By Gregory Thomas Pederson _____________________

By Gregory Thomas Pederson _____________________
LONG-TERM HYDROCLIMATIC CHANGE IN THE U.S. ROCKY MOUNTAIN REGION:
IMPLICATIONS FOR ECOSYSTEMS AND WATER RESOURCES
By
Gregory Thomas Pederson
_____________________
A Dissertation Submitted to the Faculty of the
SCHOOL OF NATURAL RESOURCES AND THE ENVIRONMENT
In Partial Fulfillment of the Requirements
For the Degree of
DOCTOR OF PHILOSOPHY
WITH A MAJOR IN NATURAL RESOURCES
In the Graduate College
THE UNIVERSITY OF ARIZONA
2010
2
THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
As members of the Dissertation Committee, we certify that we have read the dissertation
prepared by GREGORY THOMAS PEDERSON
entitled LONG-TERM HYDROCLIMATIC CHANGE IN THE U.S. ROCKY MOUNTAIN REGION:
IMPLICATIONS FOR ECOSYSTEMS AND WATER RESOURCES
and recommend that it be accepted as fulfilling the dissertation requirement for the
Degree of DOCTOR OF PHILOSOPHY
_______________________________________________________________________
Date: June 24th, 2010
_______________________________________________________________________
Date: June 24th, 2010
_______________________________________________________________________
Date: June 24th, 2010
_______________________________________________________________________
Date: June 24th, 2010
Dr. Stephen T. Gray
Dr. Connie A. Woodhouse
Dr. Julio L. Betancourt
Dr. Daniel B. Fagre
Final approval and acceptance of this dissertation is contingent upon the candidate’s
submission of the final copies of the dissertation to the Graduate College.
I hereby certify that I have read this dissertation prepared under my direction and
recommend that it be accepted as fulfilling the dissertation requirement.
_______________________________________________ Date: June 24th, 2010
Dissertation Director: Dr. Lisa J. Graumlich
3
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of requirements for an
advanced degree at the University of Arizona and is deposited in the University Library
to be made available to borrowers under rules of the Library.
Brief quotations from this dissertation are allowable without special permission, provided
that accurate acknowledgment of source is made. Requests for permission for extended
quotation from or reproduction of this manuscript in whole or in part may be granted by
the head of the major department or the Dean of the Graduate College when in his or her
judgment the proposed use of the material is in the interests of scholarship. In all other
instances, however, permission must be obtained from the author.
_________________________________
SIGNED: GREGORY THOMAS PEDERSON
4
ACKNOWLEDGEMENTS
It should come as no surprise that the work presented herein would not be possible
without the intellectual, funding, and data contributions made by members of my
graduate committee and a network of friends and collaborators. Your patience and
support has been critical in my development as a professional scientist, and has
(arguably) kept me sane through busy times. Thank You! And the U.S. Geological
Survey’s Western Mountain Initiative, the Northern Rocky Mountain Science Center, and
National Science Foundation Grants #0620793 and #0734277 for funding this work.
To my advisors and friends Lisa Graumlich and Dan Fagre, I can’t emphasize enough
how enjoyable the past decade of working with both of you has been, and I look forward
to the years ahead. You allowed me a great deal of latitude in my work, ample creative
inspiration, and perhaps more importantly you greatly expanded my horizons by
engaging me in a diversity of research projects based in some of the most spectacular
landscapes of western North America. Also, I (and society) remain indebted to you both,
since against all odds you kept me employed through some tough research funding years.
To Steve Gray, Connie Woodhouse, and Julio Betancourt: I also owe you all a great deal
of thanks as supportive committee members and friends. The span of research projects
we have worked on has underpinned my understanding of linkages between climate,
hydrology, phenology, and other ecosystem processes. Furthermore, I can’t think of a
more insightful group of people to work with, and your contribution of hard-earned treering data to this and other projects is ultimately what makes them possible.
An equally critical group of people for which many good times have been had during
some major fieldwork campaigns, and who have also contributed substantially to my
thinking on everything from snow avalanches, fire, climate, and glacier and treeline
dynamics includes: Jeremy Littell, Andy Bunn, Chris Caruso, Mike Case, Blase Reardon,
and Todd Kipfer. Your contribution of data and/or lab efforts have been critical in the
success of many of our joint research projects, but more importantly I hope we continue
to make the time to get out into the field to ski and work even though available time is
increasingly limited.
Other new and old collaborators who deserve recognition, and for which many of the
above statements also apply include: Toby Ault, Ali Macalady, Dan Griffin, Jeremy
Weiss, Scott St. George, Emma Watson, Brian Luckman, Dave Meko, Malcolm Hughes,
Jonathan Overpeck, Julia Cole, Erika Rowland, Jennifer Davidson, Dave Breshears,
Jasmine Saros, Jeffery Stone, Craig Williamson and Clint Muhlfeld.
Finally, I owe my family a good deal of thanks for all the support and encouragement
over the years. Without the influence and support I doubt if I would have found my way
into this line of work. So yes, this is your fault – and I thank you for it.
5
DEDICATION
To my family and friends for their unwavering support through this and many other of
my questionable life endeavors.
To my grandparents for sharing the wisdom of their life experiences, and for inspiring my
fascination, inquiry, and reverence for the natural world.
6
TABLE OF CONTENTS
LIST OF FIGURES .......................................................................................................... 8!
LIST OF TABLES .......................................................................................................... 14!
ABSTRACT..................................................................................................................... 16!
INTRODUCTION .......................................................................................................... 18!
PRESENT STUDY ......................................................................................................... 19!
Temperature Trends and Important Biophysical Thresholds ............................ 22!
Decadal-Scale Variability, Warming, and Shifting Seasonality in Precipitation
Patterns are Important Elements of Hydrologic Change.................................... 25!
A Millennia of Snowpack Variability and Recent Non-Stationarity Suggest a
New Era of Resource Management in Western North America......................... 26!
Future Research Directions.................................................................................... 31!
REFERENCES................................................................................................................ 34!
APPENDIX A – A CENTURY OF CLIMATE AND ECOSYSTEM CHANGE IN
WESTERN MONTANA: WHAT DO TEMPERATURE TRENDS PORTEND? .. 37!
Abstract .................................................................................................................... 38!
Introduction ............................................................................................................. 39!
Methods .................................................................................................................... 46!
Results....................................................................................................................... 49!
a. Regional Trends in Seasonal Maximum and Minimum Temperatures.......... 49!
b. Trends in Daily Tmax/Tmin Extremes and Thresholds...................................... 52!
c. Global Warming vs. Regional Trends............................................................ 56!
Discussion ................................................................................................................. 58!
a. Significance and Representativeness of Climatic Trends .............................. 58!
b. Evidence for Tipping Points and Thresholds in Climate Impacts ................. 61!
1) Alpine Glaciers are Retreating ............................................................... 62!
2) Fire and Mortality is Transforming Forests........................................... 63!
3) Rapid Changes in Freshwater Fisheries................................................. 66!
4) Ecosystem Services and Economic Impacts............................................ 68!
Conclusions .............................................................................................................. 70!
Acknowledgments.................................................................................................... 71!
References ................................................................................................................ 72!
APPENDIX B - CLIMATIC CONTROLS ON THE SNOWMELT HYDROLOGY
OF THE NORTHERN ROCKY MOUNTAINS, USA................................................ 78!
Abstract .................................................................................................................... 79!
Introduction ............................................................................................................. 80!
Data and Methods ................................................................................................... 85!
Results and Discussion ............................................................................................ 93!
7
TABLE OF CONTENTS - Continued
a. Changes in Snowpack (Snow Water Equivalent) and Streamflow................. 93!
b. Impacts of Regional Temperature and Precipitation Change on Snowpack
and Streamflow .................................................................................................. 99!
1) Mid to High-Elevation Temperature Change ......................................... 99!
2) Temperature Impacts on Snowmelt and Snowcover Duration ............. 104!
3) Changes in Surface Climate and Streamflow Timing ........................... 107!
c. Synoptic Controls on Snowpack and Streamflow Yield and Timing ............ 110!
1) Snowpack and Streamflow Yield........................................................... 110!
2) Snowmelt and Streamflow Timing ........................................................ 113!
d. Exploring the Relative Impact of Key Controls on Regional Snowpack and
Streamflow ....................................................................................................... 114!
Summary ................................................................................................................ 118!
Acknowledgements................................................................................................ 123!
Supplemental Figures............................................................................................ 123!
References .............................................................................................................. 126!
APPENDIX C - LONG-TERM SNOWPACK VARIABILITY AND CHANGE IN
THE NORTH AMERICAN CORDILLERA............................................................. 131!
Introduction to Conclusions ................................................................................. 132!
Methods .................................................................................................................. 145!
Acknowledgments.................................................................................................. 149!
Author contributions............................................................................................. 150!
Additional information ......................................................................................... 150!
Supplementary Information................................................................................. 151!
References .............................................................................................................. 156!
8
LIST OF FIGURES
Figure 1. Map of the central North American Cordillera watersheds for which tree-ring
based snowpack reconstructions were generated (left), and the Northern Rockies
watersheds for which an in-depth analysis on changes in temperature, snowpack,
streamflow, and atmospheric controls was conducted (right). ......................................... 21!
Figure 2. (a) The annual average temperature rise in the Northern Rockies and the
Northern Hemisphere. (b) The increase in number and seasonal window over which days
greater than 90°F occur. (c) The decrease in number and the first last freeze/thaw (days !
32°F) event and “extreme cold” (days ! 0°F) days of the year. ....................................... 22!
Figure 3. (a) Increasing winter, spring, and annual minimum temperatures recorded at
low-elevation valley meteorological stations and mid- to high-elevation snow telemetry
(SNOTEL) stations. The seasonal decrease in number of days minimum and maximum
temperatures drop below freezing (32°F, 0°C) across the Northern Rockies (b), and at
individual SNOTEL stations (c). ...................................................................................... 23!
Figure 4. The Boulder Glacier after 97 years of retreat in Glacier National Park, Montana,
U.S.A. ............................................................................................................................... 24!
Figure 5. (a) Co-variability between snowpack, streamflow, and North Pacific SSTs. (b,
c) Relationship between the timing of streamflow, melt-out of snowpack, number of days
without snowcover, and spring temperature and precipitation (see Appendix B for more
details)............................................................................................................................... 26!
Figure 6. (a) Long-term snowpack variability for watersheds and key hydrologic regions
along the central North American Cordillera. (b) North-South dipole in snowpack
anomalies with eras of cordillera-wide declines highlighted in gray. .............................. 27!
Figure 7. Post-1980s decline in snowpack across the central North American Cordillera.
Note, the map showing average reconstructed SWE values is not exactly equivalent to the
observational record since many individual watershed SWE reconstructions have
different end years - the earliest of which only extend to 1990. This implies the similarity
in patterns of anomalies should be noted, but the magnitudes of departure should not be
expected to be the same. ................................................................................................... 28!
Figure 8. Average seasonal and monthly temperatures of snow dominated landmasses
and snow course measurement sites across the three study regions for critical winter and
spring snow accumulation months. Average temperatures were calculated, mapped, and
estimated at each snow course site using PRISM 800m 1971-2000 climate normals, and
the average elevation of snow course monitoring sites for each of the three regions is
shown with a different color contour line. ........................................................................ 30!
9
LIST OF FIGURES - Continued
Figure A-1. Map of western Montana showing meteorological stations used in this study,
and the extensive, intact ecosystems of the region. National Parks, wilderness areas and
other protected and managed federal lands are highlighted.............................................. 41!
Figure A-2. Repeat photographs of Boulder Glacier depicting 97 years of change. The
Boulder Glacier resided in Glacier National Park, Montana, and the photos were taken
from the summit of Chapman Ridge. Morton Elrod captured the 1910 photo, and Greg
Pederson and Daniel Fagre reshot the 2007 image. See the USGS repeat photography
webpage (http://www.nrmsc.usgs.gov/repeatphoto/) for more images of glacier change in
northwestern Montana. ..................................................................................................... 45!
Figure A-3. Seasonal trends in maximum (red line) and minimum (blue line)
temperatures for western Montana. Note change in scale of temperature axis for each
season................................................................................................................................ 50!
Figure A-4. Top: Graphs showing the different rates of change in number of days-yr Tmax
" 32.2°C (gray bars) for southwestern (left) and northwestern Montana (right) from 18952006. A 5-year moving average (red line) highlights trends and variability in #days-yr "
32.2°C. Bottom: Trends in number of days (gray bars), and the first (green line) and last
(blue line) day of every summer, that temperature in western Montana equaled or
exceeded 32.2ºC................................................................................................................ 53!
Figure A-5. Average number of frost/freeze days (#days-yr Tmin !0.0°C (32.0°F)) or
extremely cold days (#days-yr Tmin !-17.8ºC (0.0ºF)) per winter year in Western Montana
(gray bars). Graphs also depict the first day of fall (blue line) and last day of spring
(green line) temperature equaled or exceeded the defined threshold. .............................. 54!
Figure A-6. Comparison of variability and trends in western Montana (blue-green) and
Northern Hemisphere (blue line) annual average temperatures. A 5-year moving average
(red line) highlights the similarity in trends and decadal variability between records. .... 57!
Figure A-7. Comparison between trends in western Montana and Northern Hemisphere
(blue line) seasonal temperatures. Red line displays the western Montana 5-year running
mean. Note change in scale of temperature axis for each season, this has implications
related to absolute temperature change............................................................................. 58!
Figure A-8. Correlations between average annual and seasonal temperature records for
the Northern Hemisphere and the nine individual Montana stations. Correlations are
calculated between the unsmoothed time series and significant values are base on p-value
! 0.05. ............................................................................................................................... 60!
10
LIST OF FIGURES - Continued
Figure B-1. Map of the Northern Rocky Mountain study region and the snow telemetry
(SNOTEL; n = 25), snow course (n = 148), stream gage (n = 14), and valley
meteorological stations (n = 37) used in this study. ......................................................... 83!
Figure B-2. Average peak snow water equivalent (SWE) calculated from 25 stations
(black line). Values are plotted as anomalies from regional SWE mean and individual
station records are plotted (gray lines). A regression line (black) shows trends in SWE,
and is bounded by 95% confidence intervals (black dashed lines) with significance (pvalue) shown. .................................................................................................................... 93!
Figure B-3. (Top) Comparison between SNOTEL peak SWE and snowcourse April 1
SWE records. (Bottom) Relationship between winter season PDO (Oct-May), total WY
stream discharge, peak WY stream discharge, and peak SWE. Note the PDO was inverted
(multiplied by -1) for ease of comparison, and all correlations are significant (p ! 0.05).
........................................................................................................................................... 95!
Figure B-4. Regional average timing of peak SWE (black dashed line) and zero SWE
(black solid line). Melt-period duration (gray bars) shows the number of days elapsed
between day of peak and zero SWE. All time series are plotted as anomalies from
regional mean, individual station records are plotted as dark gray lines, and significance
of trend (p-value) is shown though the regression line is not plotted. Strong inter-annual
and within-year station-to station variability exists for day of peak SWE, which is likely a
function of topography and other local site characteristics influencing snow accumulation
(e.g. wind) combined with the inherent spatial heterogeneity of snowfall. Inter-annual
variability in the day of zero SWE is also high, but station-to-station variability is small
since snowmelt largely temperature driven. ..................................................................... 96!
Figure B-5. Average annual number of SWE accumulation days (top), ablation days
(middle), and snow-free days (bottom) recorded at SNOTEL stations. All time series are
plotted as anomalies from regional mean, individual station records are plotted as dark
gray lines, and trends are shown with a black regression line with significance of trend
(p-value) shown ................................................................................................................ 98!
Figure B-6. Average winter (Dec-Feb; top), spring (Mar-May; middle), and annual
(bottom) minimum temperatures from SNOTEL (water year Oct-Sep) and valley MET
(calendar year Jan-Dec) stations. SNOTEL station Tmin records have been fit using a nonlinear quadratic equation due to characteristics of these time series. All trends shown are
significant (p ! 0.05) and note the y-axis temperature scale changes for each panel..... 100!
11
LIST OF FIGURES - Continued
Figure B-7. Regional average (left) and individual station records (right; grey lines)
showing number of frost days (days ! 0ºC) for Tmax (gray lines) and Tmin (black lines)
recorded at all SNOTEL stations over the winter (JFM), spring (AMJ), summer (JAS)
and fall (OND). All time series are plotted as anomalies from regional mean, and all
trends are significant* (p ! 0.001) and shown with a black regression line bracketed by
95% confidence intervals (gray dashed lines). Reduction in number of freeze/thaw days
(calculated from the trend lines) from 1983 to 2007 is printed to the right of each average
time series. ...................................................................................................................... 102!
Figure B-8. Same as Figure B-7, but for annual average number of frost days (days !
0ºC). ................................................................................................................................ 104!
Figure B-9. Correlation between number of snow-free days and average spring Tmin
(black triangles) and Tmax (gray triangles) recorded at SNOTEL stations (top left) and
valley MET stations (top right). Also plotted are, correlations between number of snowfree days and SNOTEL peak SWE (bottom left) and snowcourse April 1 SWE (bottom
right; gray squares = northern HUC, black diamonds = southern HUC). Regression lines
are plotted with corresponding 95% confidence intervals, and correlations are significant
at p ! 0.001 level............................................................................................................. 105!
Figure B-10. (Top panel) Timing of 50% (CT) and 75% cumulative discharge, day of
zero SWE, and SNOTEL average spring (MAM) Tmin . (Bottom panel) Associated trends,
variability, and step-changes in elapsed number of days between 50% and 75% stream
discharge, and average spring precipitation, Tmin, and Tmax for valley MET stations.
Correlations between 50% CT, spring Tmin, Tmax and WY day of zero SWE are significant
(p ! 0.001)....................................................................................................................... 108!
Figure B-11. Field correlations of 500mb monthly atmospheric pressure and wind
anomalies for NRM hydroclimatic variables. Maps are arranged in the following order:
A) SNOTEL peak SWE (n = 39), B) snow course April 1 SWE (n = 58), C) day of Zero
SWE (n = 39), D) day of 50% CT (n = 58), and E) number of annual snow-free days (n =
39). Plotted atmospheric circulation anomalies are significant (p ! 0.05). ................... 112!
Figure B-12. Final regression models (see run #3 in Table 4) of total annual discharge,
peak discharge and SWE anomalies. Individual stream gage records are plotted (gray
lines). Correlations between observed and modeled time series are significant (p ! 0.001).
......................................................................................................................................... 115!
Figure B-13. Idealized relationship between NRM snowpack and streamflow anomalies
with associated Pacific SSTs, atmospheric circulation, and surface feedbacks. ............ 120!
Figure B-S1. Post-1983 trends in individual SNOTEL and valley-MET station Tmin
records............................................................................................................................. 123!
12
LIST OF FIGURES - Continued
Figure B-S2. Composite 500mb monthly atmospheric pressure and wind anomaly maps
for NRM hydroclimatic variables " 1 standard deviation from the long-term mean. Maps
are arranged in the following order: A) SNOTEL peak SWE, B) snow course April 1
SWE, C) day of Zero SWE, D) day of 50% CT, and E) number of annual snow-free days.
......................................................................................................................................... 124!
Figure B-S3. Same as figure B-S2, but for composite maps associated with NRM
hydroclimatic variables ! 1 standard deviation from the long-term mean. .................... 125!
Figure C-1. Map of study area and the associated tree-ring based reconstructions of April
1 snow water equivalent (SWE) shown at multiple watershed scales. The map shows the
individual watersheds and three regions for which April 1 SWE reconstructions were
completed, the U.S. Natural Resource Conservation Service snow course sites used to
generate watershed-scale averages of observed April 1 SWE, the full set of potential
predictor chronologies, and the final set of chronologies that entered into one or more
SWE reconstruction model as a predictor. The graphs of the April 1 SWE
reconstructions are presented by region and latitude, and show the individual watershed
reconstructions of April 1 SWE (gray lines), the regional SWE average calculated from
each individual reconstruction (orange line), a 20-year cubic-smoothing spline (50%
frequency cutoff) of the regional SWE average (dark blue line), and for the Northern
Rockies and Greater Yellowstone region a cut-off date of 1376 is shown (dotted vertical
line) due to decreasing sample depth and increasing reconstruction uncertainty. The 20th
century records of observed April 1 SWE are plotted for each large region (black lines)
and smoothed with a 20-year cubic-smoothing spline to highlight decadal-scale
variability (cyan line) coherent with the snowpack reconstructions. Shaded intervals
show decadal-scale SWE anomalies mapped in Fig. 2 and Fig. 4. Lettering corresponds
to the mapped intervals. The observed and reconstructed SWE records are plotted as
anomalies from the long-term average, which was calculated sing 1400-1950 AD as a
base period. Other base periods were used to calculate the long-term average SWE
conditions yielding highly similar estimates (see Supplementary Information, Table S6).
......................................................................................................................................... 136!
Figure C-2. Decadal departures in reconstructed April 1 SWE shown for watersheds
predominately within the U.S. portion of the American Cordillera. Maps show average
SWE conditions over the following intervals previously highlighted in Fig 1: a) 14401470, b) 1511-1530, c) 1565-1600, d) 1601-1620, e) 1845-1895, and f) 1902-1932. The
mapped SWE anomalies were calculated by averaging annual conditions for each HUC 6
watershed over the time interval shown, and are plotted as anomalies from the long-term
regional mean (1400-1950 AD). The final datasets along with the ability to generate
user-defined maps of interannual- to interdecadal-scale departures in reconstructed and
observed SWE are provided at http://www.nrmsc.usgs.gov/NorthAmerSnowpack/. .... 138!
13
LIST OF FIGURES - Continued
Figure C-3. Comparison of decadal-scale variability in estimates of Upper Colorado
Basin April 1 SWE and total annual streamflow. For clarity, each reconstruction is
shown with a 20-year (dark blue line) and 50-year (red line) cubic-smoothing spline
(50% frequency cutoff), shown as departures from the long-term mean, and discussed
low flow periods are highlighted with gray bars. The reconstructed Upper Colorado
River flows21 are plotted in billions of cubic meters (BCM), and were calibrated against
the Lee’s Ferry gage record. ........................................................................................... 140!
Figure C-4. Post-1980 average April 1 SWE conditions. Maps of post-1980 average
SWE conditions (Fig. 1g) plotted as anomalies from the regional long-term mean (14001950 AD) for the observational record (left) and tree-ring based reconstructions (right).
Note, the map showing average reconstructed SWE values is not exactly equivalent to the
observational record since many individual watershed SWE reconstructions have
different end years - the earliest of which only extend to 1990. This implies the similarity
in patterns of anomalies should be noted, but the magnitudes of departure should not be
expected to be the same. ................................................................................................. 145!
Figure C-S1. Watershed-based observed and reconstructed April 1 SWE. Observed April
1 SWE for large regions/watersheds (black lines) was calculated from individual snow
course records, and are plotted alongside the tree-ring based watershed SWE
reconstructions (gray lines) and the regional estimates of average SWE conditions
(dashed orange lines; calculated by averaging individual watershed reconstructions
together). Both the observed (cyan line) and reconstructed regional averages of (dark
blue line) April 1 SWE are shown smoothed with a 20-year cubic-smoothing spline (50%
frequency cutoff) to highlight correspondence at decadal-scales. Note the spline is biased
by end effects at the start and end of any time series, and this bias is particularly evident
at the start of the observed April 1 SWE records (~1920 to 1930 AD).......................... 151!
Figure C-S2. Decadal-scale antiphasing of the N-S snowpack dipole. The 20-year splines
of the regional average snowpack anomalies are plotted for clarity and to highlight
variability at decadal scales. The shaded areas highlight the only periods of synchronous,
cordillera-wide snowpack declines. ................................................................................ 152!
Figure C-S3. Average seasonal and monthly temperatures of snow dominated landmasses
and snow course measurement sites across the three study regions for critical winter and
spring snow accumulation months. Average temperatures were calculated, mapped, and
estimated at each snow course site using PRISM (http://www.prism.oregonstate.edu/)
800m 1971-2000 climate normals, and the average elevation of snow course monitoring
sites for each of the three regions is shown with a different color contour line. ............ 153!
14
LIST OF TABLES
Table A-1. Location, elevation, and percent of record present for western Montana
meteorological stations used in this study. ....................................................................... 43!
Table A-2. Change in annual and seasonal maximum and minimum temperatures in
western Montana. Overall change and significance of trend was assessed conservatively
using ordinary least squares regression............................................................................. 51!
Table A-3. Correlation between western Montana average annual and seasonal
temperatures, and Northern Hemisphere and Global average temperature records. ........ 59!
Table B-1. SNOTEL station locations, elevation, percent of record present, and start year.
........................................................................................................................................... 86!
Table B-2. Stream gage name, location, percent of record present, and start year.......... 89!
Table B-3. Correlations between the average annual number of snow-free days, the WY
day of peak and zero SWE, and 50% CT and spring and annual Tmax and Tmin recorded at
valley (1969-2007, n = 39) and SNOTEL stations (1983-2007, n = 25)........................ 106!
Table B-4. Model diagnostics for best-fit least-squares linear regression models
constructed for peak discharge, total annual discharge, and peak SWE anomalies. ...... 116!
Table C-S1. Summary of NRCS snow course April 1 SWE sampling site elevations by
region. ............................................................................................................................. 154!
Table C-S2. Summary of April 1 SWE models by region............................................. 154!
Table C-S3. List of tree-ring chronologies that entered one or more of the 128 nested
April 1 SWE regression models (see
http://www.nrmsc.usgs.gov/NorthAmerSnowpack/tables). ........................................... 154!
Table C-S4. Tree-ring chronologies (and raw ring-width files) that entered one or more of
the 128 nested April 1 SWE regression models. All chronologies were conservatively
detrended (negative exponential, negative regression, or mean line) to preserve lowfrequency variability, and underwent r-bar variance scaling to correct for the influence of
decreasing sample depth over the early portion of the record (see
http://www.nrmsc.usgs.gov/NorthAmerSnowpack/tables). ........................................... 154!
Table C-S5. April 1 SWE nested model calibration and validation statistics (see
http://www.nrmsc.usgs.gov/NorthAmerSnowpack/tables). The “MV adjust” and “In
composite” columns indicates whether a particular reconstruction had it’s mean and
variance scaled to match the mean and variance (calculated over a common interval) of
the “best” SWE reconstruction before inclusion into the composite reconstruction of
snowpack. If the mean and variance was not adjusted, it was not included in the
composite record due to model issues. ........................................................................... 155!
15
LIST OF TABLES - Continued
Table C-S6. Different long-term April 1 SWE averages calculated using three different
base periods..................................................................................................................... 155!
16
ABSTRACT
Both natural and anthropogenic climate change are driven by forcings that interact
and result in hydroclimatic changes that alter ecosystems and natural resources at
different temporal and spatial scales. Accordingly, changes within regions (i.e.
individual points to large watersheds) may differ from patterns observed at subcontinental to global scales, thus necessitating the generation of point- to region-specific,
cross-scale hydroclimatic data to elucidate important drivers of observed changes, and
provide information at scales relevant to resource managers. Herein, we use the Northern
U.S. Rocky Mountains as a study region to explore 1) the covariability between observed
hydrologic and climatic changes, 2) the nature of changes occurring at the scale of days
to decades, and 3) the ocean-atmosphere teleconnections operating at continental- to
hemispheric-scales underlying the observed regional patterns of hydroclimatic variability.
We then expand the scope of study to include the entire central North American
Cordillera to investigate changes in winter precipitation (i.e. snowpack) spanning the last
millennia+, with a focus on the spatial and temporal coherence of events from the
medieval climatic anomaly to present. To accomplish this we utilize the full suite of
hydroclimatic observational records in conjunction with proxy records of snowpack
derived from a distributed network of tree-ring chronologies.
Results from observational records in the Northern Rockies show important
changes have occurred in the frequency and means of biophysically important
temperature thresholds, and that recent changes appear greater in magnitude at the mid-
17
to high-elevations. These changes, coupled with interannual- to interdecadal-scale
moisture variability driven by ocean-atmosphere teleconnections, are shown to be strong
controls on the timing and amount of regional snowpack and streamflow. Across the
cordillera, tree-ring based records of snowpack show that before 1950, the region
exhibited substantial inter-basin variability in snowpack, even during prolonged droughts
and pluvials, marked by a predominant north-south dipole associated with Pacific
variability. Snowpack was unusually low in the Northern Rocky Mountains for much of
the 20th century and over the entire cordillera since the 1980s; heralding a new era of
snowpack declines entrained across all major headwaters in western North America.
18
INTRODUCTION
Increasingly, land and water resource managers are faced with the daunting task
of sustaining ecosystem goods and services in the face of growing societal needs, and
rapid global environmental change1. Across the western U.S., the past century of
hydroclimatic change has resulted in greatly reduced snowpack2,3 and runoff4, the earlier
melt-out of snow and timing of peak runoff5,6, substantially lower summer base flows7,
increasing fire frequency and area burned8-10, and the ubiquitous and rapid retreat of the
regions glaciers11. These observed physical changes drive, and in turn are mirrored by,
changes documented across biological systems. For example, substantial changes have
been observed in the phenology of plants and animals12-16, the distribution of aquatic
invertebrates and cold-water fishes17, the expansion of alpine treeline and infilling of
meadows18-21, the increase in background tree-mortality rates22, and the synchrony and
size of forest area killed by recent beetle outbreaks23-25. The potential for human-induced
climate change to result in novel climate-ecosystem types26, combined with increasing
human pressures on resources are substantial enough that recent U.S. federal mandates
require management agencies to explicitly address the long-term impacts, and consequent
management options for species of special concern, habitat, and renewable resources27
(e.g. timber, water, fiber, etc.). In mountainous regions, however, the spatial scale,
patterns, and rates of change may have disproportionate impacts on different biophysical
systems (e.g. grizzly bears versus water resources), thus necessitating cross-scale
investigations of hydroclimatic change.
19
PRESENT STUDY
This dissertation seeks to explore hydroclimatic variability and change within
regions (i.e. individual points to large watersheds) to elucidate an understanding of
important drivers of observed changes, and to provide information at scales relevant to
land and water resource managers. Herein, we use the U.S. Northern Rocky Mountains
(Fig. 1; see Appendices A and B for more details) as a core study region since it’s one of
the world’s last relatively intact temperate ecosystems of high conservation and natural
resource value, and also serves as a key headwaters area for the Columbia, Saskatchewan,
and Missouri River Systems. Specifically, we explore 1) the covariability between
observed hydrologic, climatic, and ecosystem changes, 2) rates of change at different
elevations and for biophysically important thresholds, 3) the nature of changes occurring
at the scale of days to decades, and 4) the ocean-atmosphere teleconnections operating at
continental- to hemispheric-scales underlying the observed regional patterns of
hydroclimatic variability. We then expand the scope of study to include the entire central
North American Cordillera to investigate the variability and change in winter snowpack
over the last millennia+, with a focus on the spatial and temporal coherence of events
from the medieval climatic anomaly to present (see Appendix C). To accomplish this we
utilize the full suite of available hydroclimatic observational records in conjunction with
proxy records of snowpack derived from a distributed network of tree-ring chronologies.
The major implications from this research of relevance to managers, scientists,
and policy makers are summarized in the following points and further elaborated on
20
throughout the dissertation. First, important characteristics of the hydroclimatic system
(e.g. amount of snowpack, peak and total annual flows, etc.) exhibit substantial
variability and change on inter-decadal scales in addition to long-term trends. This mode
of system variability can, and has, resulted in major ecosystem and resource related shifts
that initiate rapidly and persist for many years challenging management paradigms and
techniques that may have been suitable over previous decades. Next, long-term trends in
temperature and continued warming consistent with expectations from human-induced
climate change has, and is highly likely to, continue causing important biophysical
thresholds to be crossed (e.g. 0°C frost/freeze temperature), and when combined with
surface feedback processes is resulting in more rapid warming of the higher elevations.
The combination of long-term trends and decadal-scale variability in climate makes longterm resource planning highly complex when seeking to maintain ecosystem goods and
services by setting 10 to 20+ year resource management goals since what may work over
the near-term can change rapidly (and bidirectionally) while becoming increasingly
unsustainable over the long-term. Finally, centuries-long records of snowpack variability
across the central North American Cordillera suggest a stationary north-south dipole in
high- and low-snowpack anomalies with the exception of recent post-1980s West-wide
declines and intervals of elevated warmth during the late medieval climate anomaly.
This result implies that warm temperature anomalies drive spatial and temporal
synchrony in snowpack declines, and if current temperature trends persist water resource
managers are entering a new era of management where the free storage and controlled
runoff from snowpack during the warm dry summer season becomes less available and
21
less reliable. That said, shifts in the seasonality of precipitation (i.e. specifically the
documented increases in spring precipitation across the Northern Rockies) appear to be
buffering hydrologic systems from the full magnitude of snowpack-related flow declines.
The future trend trajectory and efficacy of spring precipitation is highly uncertain,
however, and consequently unsuitable for use in long-term planning. Additional details
on these results along with summary graphics are provided over the following sections.
Figure 1. Map of the central North American Cordillera watersheds for which tree-ring
based snowpack reconstructions were generated (left), and the Northern Rockies
watersheds for which an in-depth analysis on changes in temperature, snowpack,
streamflow, and atmospheric controls was conducted (right).
22
Temperature Trends and Important Biophysical Thresholds
Warming in the Northern Rockies mirrors patterns and trends observed Globally
and across the Northern Hemisphere, with a century of records showing a two to three
times greater rise in regional extremes and seasonal averages than observed globally (Fig.
2). At low elevations, seasonal records from long-term valley stations show nighttime
minimum temperatures have risen more rapidly than daytime maximums, and that the
most rapid warming has occurred over the winter, spring, and then summer respectively.
Daily temperature records reveal extremely cold days (!-17.8°C) terminate on average 20
days earlier and decline in number, whereas extremely hot days ("32°C) show a threefold increase in number and a 24-day increase in seasonal window during which they
occur. Regionally important temperature thresholds have been exceeded, the most recent
of which include the timing and number of the 0°C freeze/thaw temperatures during
spring and fall. (see Appendix A)
Figure 2. (a) The annual average temperature rise in the Northern Rockies and the
Northern Hemisphere. (b) The increase in number and seasonal window over which
days greater than 90°F occur. (c) The decrease in number and the first last freeze/thaw
(days ! 32°F) event and “extreme cold” (days ! 0°F) days of the year.
23
More rapid warming has occurred in the higher-elevations of the Northern
Rockies with substantial seasonal and annual decreases in number of freeze/thaw days
(days ! 0°C; Fig. 3). Changes in the spring minimum temperatures correspond with
atmospheric circulation changes and surface-albedo feedbacks in March and April.
Minimum temperatures at mid- to high-elevation stations display inter-annual variability
similar in magnitude to valley stations with the exception of spring minimums. Spring
Figure 3. (a) Increasing winter, spring, and annual minimum temperatures recorded at
low-elevation valley meteorological stations and mid- to high-elevation snow
telemetry (SNOTEL) stations. The seasonal decrease in number of days minimum
and maximum temperatures drop below freezing (32°F, 0°C) across the Northern
Rockies (b), and at individual SNOTEL stations (c).
Tmin records generally exhibit approximately twice the variance of minimum
temperatures observed at valley stations, and more rapid recent warming. The high interannual variability and trend shown in spring minimum temperatures suggests changing
24
lapse rates and feedback processes associated with the earlier melt-out of snow and
consequently reduced surface albedo. The more rapid warming of the high-elevations
means forests, snowpack, and glaciers are experiencing disproportionate temperature
driven change mirrored by the recent widespread beetle kill, fire activity, and glacier
retreat. (see Appendix B)
Evidence for major hydrologic and ecosystem change in line with expectations of
warming across the Northern Rockies is exemplified by the rapid retreat of the regions
glaciers (Fig. 4). The
rapid glacier retreat,
widespread fire and beetle
activity are also good
examples of the danger in
assuming that long-term
trends in temperature will
be associated with equally
paced or simple linear
changes in ecosystems and
Figure 4. The Boulder Glacier after 97 years of retreat in
Glacier National Park, Montana, U.S.A.
their associated services.
The processes through which ecosystem changes take place are often complicated by
synergistic and non-linear relationships between variables (e.g. surface albedo
relationships and rapid surface warming), many of which are still unknown.
25
Decadal-Scale Variability, Warming, and Shifting Seasonality in Precipitation
Patterns are Important Elements of Hydrologic Change
The amount and melt-out timing of snowpack and streamflow in the Northern
Rocky Mountains is strongly controlled by warming spring temperatures, increasing
spring precipitation, and changes in winter stormtracks associated with Pacific Ocean
sea-surface-temperatures (SSTs; Fig. 5). Records of snow course April 1 SWE show the
composite time-series are good metrics of SNOTEL peak SWE, and that since 1936 a
major feature of the records is the strong variability on interannual- to decadal-scales that
coincide with changes in streamflow discharge, and North Pacific SSTs. Spring
temperatures coupled with increases in the mean and variance of spring precipitation
control the timing of snow melt-out, an increased number of snow-free days, and
observed changes in streamflow timing and discharge. Importantly, the increased spring
precipitation after 1977 appears to substantially buffer streamflow timing from what
should be considerably earlier decline in flow due to the recent decades of low snowpack
with above-average temperatures. The atmospheric controls shown to underlie observed
changes in temperature, snow melt-out, and runoff, also control multiple phenological
and ecological aspects of spring onset across western North America. This implies that
the associated snowmelt related ecosystem impacts are much broader than the Northern
Rockies study area, and if regional projections of 21st century hydroclimatic change
prove correct, major changes are in store for aquatic and terrestrial ecosystems, and
consequently land and water resource managers. (see Appendix B)
26
Figure 5. (a) Co-variability between snowpack, streamflow, and North Pacific SSTs. (b,
c) Relationship between the timing of streamflow, melt-out of snowpack, number of
days without snowcover, and spring temperature and precipitation (see Appendix B for
more details).
A Millennia of Snowpack Variability and Recent Non-Stationarity Suggest a New
Era of Resource Management in Western North America
Long-term records of snowpack variability across hydrologically important
headwaters regions of the central North American Cordillera show that for the past eight
centuries the region exhibited substantial inter-basin variability in snowpack, even during
prolonged (i.e. decadal-scale) droughts and pluvials, marked by a predominant northsouth (N-S) dipole associated with Pacific Ocean SST variability (Fig. 6). Anomalies in
Pacific SST and pressure patterns result in a N-S snowpack dipole by changing major
27
atmospheric circulation patterns, specifically the strength and position of the Aleutian
Low (i.e. the Pacific Decadal
Oscillation [PDO], see
Appendix B), which alters the
preferential positioning of the
wintertime stormtrack resulting
in a dry north and wet south or
vice versa. Apart from the N-S
dipole, snowpack was
unusually low in the Northern
Rocky Mountains relative to
the long-term mean for much of
the 20th century, and over the
entire cordillera since the
1980s; heralding a new era of
Figure 6. (a) Long-term snowpack variability for
watersheds and key hydrologic regions along the
central North American Cordillera. (b) North-South
dipole in snowpack anomalies with eras of
cordillera-wide declines highlighted in gray.
snowpack declines entrained
across all major headwaters in western North America (Fig. 7). (see Appendix C)
The early 20th century decline (~1900-1942) in snowpack across the Northern
Rockies region coincides with warm SSTs across the Gulf of Alaska and a positive PDO
phase suggesting changes in atmospheric circulation likely resulted in the southerly
displacement of the winter stormtrack and consequently a substantial portion of the
28
observed northern (southern) declines (increases) in snowpack. Additionally, the regions
relatively low-elevation landmass, snowmass, and snow course monitoring sites (~1550
m as compared to the ~2800 m of the Upper Colorado) also results in warmer average
temperatures nearer the 0°C isotherm, likely causing a greater sensitivity of the regions
snowpack to temperature change over the critical snow accumulation months (Fig. 8).
The ~3°C warmer winter and spring average temperatures estimated for the Northern
Rockies snow course sites was interpolated from meteorological station data (i.e. PRISM
800m 1971-2000 climate normals), and is consistent with the expected effects of both
Figure 7. Post-1980s decline in snowpack across the central North American
Cordillera. Note, the map showing average reconstructed SWE values is not exactly
equivalent to the observational record since many individual watershed SWE
reconstructions have different end years - the earliest of which only extend to 1990.
This implies the similarity in patterns of anomalies should be noted, but the
magnitudes of departure should not be expected to be the same.
29
elevation and latitude on temperature. Specifically, the ~1250 m difference in average
snow course elevations results in an ~ -8°C cooler temperatures across the Upper
Colorado (assuming a lapse rate of 0.65°C per100 m elevation), and the difference in
latitude (approximately 10°) causes an approximate -5°C latitudinal effect on temperature
(assuming ~0.5°C per degree latitude). Overall the net result of elevation and latitude as
calculated here supports the notion that the Northern Rockies snowmass and snow course
observation sites are approximately 3°C warmer than sites across the Upper Colorado.
This difference in average temperature across snow course sites arises from regional
differences in physiography, and reflects the fact that most the land mass and hence
hydrologic mass contained in winter snowpack is distributed across substantially lower
elevations. Plus, elevations above 2500 m in the Northern Rockies are often dangerous
for snow sampling since they are typically glaciated (i.e. steep and matterhorn shaped)
resulting in slope angles prone to frequent snow avalanche activity. The warm sampling
bias of the Northern Rockies region should be noted and accounted for in future
monitoring efforts, and research comparing trends and variability in snowpack records
across western North America. Perhaps more importantly, however, Figure 7 suggests
additional warming of several degrees centigrade would shift the majority of the snow
monitoring sites, and hence snowmass, of the Northern Rockies across the 0°C isotherm
during the core winter accumulation months, and the entire central North American
Cordillera during March and April. This would substantially enhance the melt rates of
accumulating winter snowfall, and result in more mid-winter snowmelt driven runoff
30
events, earlier and lower peak flows, and a longer period of substantially reduced summer
baseflows.
Figure 8. Average seasonal and monthly temperatures of snow dominated
landmasses and snow course measurement sites across the three study regions for
critical winter and spring snow accumulation months. Average temperatures were
calculated, mapped, and estimated at each snow course site using PRISM 800m
1971-2000 climate normals, and the average elevation of snow course monitoring
sites for each of the three regions is shown with a different color contour line.
The post-1980s cordillera-wide declines in winter snowpack have major
implications for the future of water resources across western North America, and are
unique with the exception of two periods of synchronous declines during the late
medieval climate anomaly. Each of the three periods of spatially synchronous snowpack
declines correspond with periods of elevated regional and hemispheric warmth,
potentially indicating the important role temperature plays in synchronizing low
31
snowpack conditions across the West. Past events and recent trends present challenging
scenarios for modern water management in lieu of increasing societal demands, the
prevalence of sustained decadal-scale periods of low snowpack, and long-term trends that
suggest a trajectory of declining snowpack and consequently streamflow across the West.
In terms of the management of natural resources other than water, the loss of the northsouth snowpack dipole is likely one major reason for recent spatial synchrony in
ecosystem disturbance, and suggests events like West-wide fire activity may become
commonplace over the near future.
Future Research Directions
As evidenced from the studies of high-resolution observational climate and
streamflow records herein, more intensive research is needed on local to regional
hydroclimatic and biological interactions, potential surface feedbacks, and the
atmospheric circulation changes underlying the observed trends and variability.
Additionally, to better underpin the biophysical impacts of climate change on societies
and economies, coupled socio-ecosystem impacts should be investigated through the
integration of observational studies of past climate-socio-ecosystem relationships. In
conjunction with these efforts, the capacity to monitor for new system trajectories, or
sudden shifts, in our biophysical systems should be increased for long-term planning
and/or rapid mitigation measures if land and resource management agencies ever hope to
successfully implement adaptive management strategies.
32
The proxy-based reconstructions of snowpack across the central North American
Cordillera raise many questions related to the climate dynamics underlying the observed
decadal-scale and longer-term variability, the N-S dipole in snowpack conditions, and
may also be used to validate climate models under natural forcing. Specifically, more
work is needed to disentangle the relative influence of changes in temperature and
atmospheric circulation on the observed variability and trends in snowpack. For
example, investigating why the Greater Yellowstone Region exhibits the same 20th
century decline in snowpack as the Northern Rockies region when its effective elevation
and hence average temperatures are equivalent to the Upper Colorado Basin (Fig. 6, Fig.
8). Does this observed pattern merely reflect the combination of more rapid warming
across the northern regions plus the fact that the synoptic controls on winter stormtrack
and hence snowpack are the same? Also, for consistency in the long-term monitoring of
snowpack, maintaining the current regional distribution of the historic average snow
course site temperatures (see Fig. 8) into the future is necessary to preserve a
homogeneous observational record, and needs to be accounted for in comparative studies
on the status and trends of snowpack. Finally, the long-term, spatially specific snowpack
reconstructions should also be used to assess the accuracy of coupled ocean-atmosphere
global climate models (AOGCMs) in capturing the position and variability in North
American winter storm tracks. For example, how well do AOGCMs reproduce the N-S
snowpack dipole? What’s the average latitude of the winter stormtrack, and how much
interannual to decadal-scale variability in snowpack is reproduce by the models? The
answers to these questions are relevant for assessing, and possibly improving, the spatial
33
and temporal accuracy of future forecasts of winter snowpack conditions, and
consequently has implications for small-scale detection and attribution studies.
Regardless of future efforts in carbon mitigation and sequestration, over the
course of this next century it’s highly likely we are committed to a warming climate.
Only through increased monitoring, research, and generation of spatially explicit patterns
of past and expected change, can we expect citizens, policy makers, and resource
managers to engage in well-informed decision-making.
34
REFERENCES
1
IPCC. Climate Change 2007: The Physical Science Basis. Contribution of
Working Group I to the Fourth Assessment Report of the Intergovernmental Panel
on Climate Change. 996 (Cambridge University Press, 2007).
2
Mote, P. W. Climate-Driven Variability and Trends in Mountain Snowpack in
Western North America. Journal of Climate 19, 6209-6220 (2006).
3
Pierce, D. W. et al. Attribution of Declining Western U.S. Snowpack to Human
Effects. J Climate 21, 6425-6444 (2008).
4
Hidalgo, H. G. et al. Detection and Attribution of Streamflow Timing Changes to
Climate Change in the Western United States. Journal of Climate 22, 3838-3855
(2009).
5
McCabe, G. J. & Clark, M. P. Trends and Variability in Snowmelt Runoff in the
Western United States. Journal of Hydrometeorology 6, 476-482 (2005).
6
Stewart, I. T., Cayan, D. R. & Dettinger, M. D. Changes toward earlier
streamflow timing across western North America. Journal of Climate 18, 11361155 (2005).
7
Luce, C. H. & Holden, Z. A. Declining annual streamflow distributions in the
Pacific Northwest United States, 1948-2006. Hoerling, M., J. Whitaker, A.
Kumar, and W. Wang (2001), The Midlatitude Warming During 1998-2000,
Geophys. Res. Lett., 28(5), 755-758. 36, doi:10.1029/2009gl039407 (2009).
8
Littell, J. S., McKenzie, D., Peterson, D. L. & Westerling, A. L. Climate and
wildfire area burned in western U.S. ecoprovinces, 1916 - 2003. Ecological
Applications 19, 1003-1021, doi:10.1890/07-1183.1 (2009).
9
McKenzie, D., Gedalof, Z., Peterson, D. L. & Mote, P. Climatic Change,
Wildfire, and Conservation. Conservation Biology 18, 890-902 (2004).
10
Westerling, A. L., Hidalgo, H. G., Cayan, D. R. & Swetnam, T. W. Warming and
earlier spring increase western US forest wildfire activity. Science 313, 940-943
(2006).
11
Fountain A, H. M., Jackson K, Basagic H, Nylen T, and Percy D Open File
Report 23 (USGS, 2007).
12
Caprio, J. M. Western regional phenological summary of information on
honeysuckle and lilac first bloom phase covering the period 1956-1991. 92
(Montana State Universiy, Bozeman, 1993).
35
13
Parmesan, C. & Yohe, G. A globally coherent fingerprint of climate change
impacts across natural systems. Nature 421, 37-42 (2003).
14
Root, T. L. et al. Fingerprints of global warming on wild animals and plants.
Nature 421, 57-60 (2003).
15
Schwartz, M. D. & Reiter, B. E. Changes in North American spring. Int J
Climatol 20, 929-932 (2000).
16
Ault, T. R., Macalady, A. K., Pederson, G. T., Betancourt, J. L. & Schwartz, M.
D. Northern Hemisphere Modes of Variability and the Timing of Spring Across
Western North America Quatern Int in review (in review).
17
Hauer, F. R. et al. Assessment of climate change and freshwater ecosystems of
the Rocky Mountains, USA and Canada. Hydrological Processes 11, 903-924
(1997).
18
Harsch, M. A., Hulme, P. E., McGlone, M. S. & Duncan, R. P. Are treelines
advancing? A global meta-analysis of treeline response to climate warming. Ecol
Lett 12, 1040-1049 (2009).
19
Millar, C. I., Graumlich, L. J., Delany, D. L., Westfall, R. D. & King, J. K.
Response of subalpine conifers in the Sierra Nevada, California, U.S.A., to 20thcentury warming and decadal climate variability. Arctic, Antarctic, and Alpine
Research 36, 181-200 (2004).
20
Malanson, G. P. et al. Alpine treline of western North America: Linging
organism-to-landscape dynamics. Phys Geogr 28, 378-397 (2007).
21
Elliott, G. P. & Kipfmueller, K. F. Multi-scale influence of slope aspect and
spatial pattern on ecotonal dynamics at upper treeline in the Southern Rocky
Mountains, U.S.A. Arctic, Antarctic, and Alpine Research 42, 45-56,
doi:10.1657/1938-4246-42.1.45 (2010).
22
van Mantgem, P. J. & Stephenson, N. L. Apparent climatically induced increase
of tree mortality rates in a temperate forest. Ecol Lett 10, 909-916 (2007).
23
Carroll, A. L., Taylor, S. W., Reginiere, J. & Safranyik, L. in Mountain Pine
Beetle Symposium: challenges and solutions. (eds T. L. Shore, J. E. Brooks, & J.
E. Stone) 298 (Natural Resources Canada, Canadian Forest Service, Pacific
Forrestry Center, Victoria, B.C.).
24
Logan, J. A. & Powell, J. A. Ghost forests, global warming and the mountain pine
beetle dynamics. American Entomologist 47, 160-172 (2001).
36
25
Logan, J. A., Reginiere, J. & Powell, J. A. Assessing the impacts of global
warming on forest pest dynamics. Frontiers in Ecology and the Envrionment 1,
130-137 (2003).
26
Williams, J. W. & Jackson, S. T. Novel climates, no-analog communities, and
ecological surprises. Frontiers in Ecology and the Envrionment 5, 475-482
(2007).
27
Salazar, K. in Secretarial Order No. 3289: Addressing the Impacts of Climate
Change on Apericas Water, Land, and Other Natural and Cultural Resources
(ed Department of the Interior) (Washington, D.C., 2009).
!
37
APPENDIX A – A CENTURY OF CLIMATE AND ECOSYSTEM CHANGE IN
WESTERN MONTANA: WHAT DO TEMPERATURE TRENDS PORTEND?
Authors:
1,2
Gregory T. Pederson*, 2Lisa J. Graumlich, 1Daniel B. Fagre,
3
Todd Kipfer, and 1Clint C. Muhlfeld
CLIMATIC CHANGE – Submitted February 7, 2008
Affiliations:
1
U.S. Geological Survey, Northern Rocky Mountain Science Center, 2327 University Way (STE
2), Bozeman, Montana, 597145
2
The University of Arizona, School of Natural Resources, 325 Bioscience East, Tucson, AZ
85721
3
Big Sky Institute, Montana State University, 106 AJM Johnson Hall, Bozeman, Montana 59717
*Corresponding Author: Gregory T. Pederson
Email: [email protected]
Journal Citation:
Pederson GT, LJ Graumlich, DB Fagre, T Kipfer, C Muhlfield. 2010. A century of climate and
ecosystem change in Western Montana: What4 do recent temperature trends portend? A view
from western Montana, USA. Climatic Change, 98:133-154 DOI: 10.1007/s10584-009-9642-y.
38
Abstract
The physical science linking human-induced increases in greenhouse gasses to the
warming of the global climate system is well established, but the implications of this
warming for ecosystem processes and services at regional scales is still poorly
understood. Thus, the objectives of this work were to: (1) describe rates of change in
temperature averages and extremes for western Montana, a region containing sensitive
resources and ecosystems, (2) investigate associations between Montana temperature
change to hemispheric and global temperature change, (3) provide climate analysis tools
for land and resource managers responsible for researching and maintaining renewable
resources, habitat, and threatened/endangered species, and (4) integrate our findings into
a more general assessment of climate impacts on ecosystem processes and services over
the past century. Over 100 years of daily and monthly temperature data collected in
western Montana, USA were analyzed for long-term changes in seasonal averages and
daily extremes. In particular, variability and trends in temperature above or below
ecologically and socially meaningful thresholds within this region (e.g., -17.8°C (0°F),
0°C (32°F), and 32.2°C (90°F)) are assessed. The daily temperature time series reveal
extremely cold days (!-17.8°C) terminate on average 20 days earlier and decline in
number, whereas extremely hot days ("32°C) show a three-fold increase in number and a
24-day increase in seasonal window during which they occur. Results show that
regionally important thresholds have been exceeded, the most recent of which include the
timing and number of the 0°C freeze/thaw temperatures during spring and fall. Finally,
we close with a discussion on the implications for Montana’s ecosystems. Special
39
attention is given to critical processes that respond non-linearly as temperatures exceed
critical thresholds, and have positive feedbacks that amplify the changes.
Introduction
The scientific basis for asserting that the global climate system is changing due to
human influence has been firmly established through many lines of evidence ranging
from observational studies to climate models. General circulation models have quantified
and assessed sensitivity of the Earth’s climate system to solar and greenhouse gas
forcings (e.g. Hansen et al. 2005; Lugina et al., 2005; NRC, 2005), and paleoclimatic
studies have provided important historical perspectives (e.g. Moberg et al., 2005;
Oerlemans et al., 2005; NRC, 2006). The physical science basis for detecting and
attributing climate change indicates that the recent warming of the climate system is
unequivocal, as evidenced from observations of increases in global average air and ocean
temperatures (e.g. Hansen et al. 2005; Rayner et al., 2006), widespread melting of
perennial snow and glacial ice (Dyurgerov and Meier, 2005), and a rising global average
sea level (Church and White, 2006).
The accumulated weight of the scientific evidence in line with expectation for
global warming has brought about a profound change in the level of engagement of a
wide range of stakeholders in defining potential impacts for ecosystem services and
livelihoods. This engagement provides new challenges for the scientific community to
develop robust assessments of regional to local trends and trajectories. For example,
40
federal resource managers cite the lack of site-specific information to plan for and
manage the effects of climate change as one of the key challenges they face in addressing
the impacts of climate change (GAO, 2007). More generally, site-specific information is
critical to understanding the degree to which regional patterns reflect hemispheric or
global temperature rise (Easterling et al., 2000). Accordingly, an analysis of 100+ years
of daily and monthly temperature data for western Montana, USA, is presented as a case
study with three purposes. First, high-quality, archival climate records are analyzed to
assess the nature and significance of local temperature variability and trends. Second, R
project (R Development Core Team, 2007) software developed to quantify changes in
daily temperature extremes is made freely available with modifiable script - allowing for
similar analyses to be tailored to, and conducted for, regions and species of interest.
Finally, results are used to identify and discuss the sensitivity of ecosystems and
economies in Montana to changing temperature based upon a survey of current literature.
Why investigate changes in temperature averages and extremes in western
Montana? The region presents a compelling opportunity for a case study because it
contains one of the world’s last remaining intact temperate ecosystems. At present, the
ecosystem still supports the full assemblage of native species known to have historically
inhabited the region - since the time of the first biological surveys. The coldwater
fisheries support some of the most intact, high-quality habitat remaining for native char
and Salmonids. The region also serves as home to the last remaining viable and selfsustaining mega fauna (e.g. American Bison (Bison bison)) and predator populations (e.g.
41
Grizzly Bear (Ursus arctos horribilis), Gray Wolf (Canis lupus)) in the lower 48 United
States. Aside from the biological conservation value of western Montana, the mountain
ranges serve as critical headwaters for the Columbia, Missouri, and Saskatchewan Rivers.
The importance of the hydrological, biological, and natural resources of this region are
highlighted by the extent of federally protected lands (Figure A-1), and the designation of
core areas (i.e. Glacier-Waterton International Peace Park, and Yellowstone National
Park) as United Nations Educational, Scientific, and Cultural Organization (UNESCO)
World Heritage sites. Regional warming, and specifically how impacts from global
Figure A-1. Map of western Montana showing meteorological stations used in this
study, and the extensive, intact ecosystems of the region. National Parks, wilderness
areas and other protected and managed federal lands are highlighted.
42
warming are expressed locally, represents one of the largest threats to the biological and
natural recourses contained within and around protected areas since the lands were given
special protection status.
What climatically defines and sets this region apart from other areas within the
lower 48 United States? By the standards of a vast majority of people who live outside
the Arctic, western Montana is a “cold” place. Annual average temperatures derived
from the nine western Montana station records (Figure A-1, Table A-1) show that valley
temperatures hover around 5.8°C (42.5°F). This reflects the length of the cool seasons
with fall, winter, and spring averaging 6.1°C (43.0°F), -5.0°C (23.0°F), and 5.4°C
(41.8°F) respectively. Additionally, the dry continental nature of Montana’s climate
means localized areas well suited for radiative cooling can experience extreme
temperatures exceeding -46°C (-50°F) (Western Regional Climate Center (WRCC),
2007). For example, the lowest temperatures ever recorded in the United States
(excluding Alaska) occurred on January 20, 1954 at Rodgers Pass where temperatures
fell to -57.0°C (-70.0°F; WRCC, 2007). Conversely, summers are defined by warm days
and cool nights with valley temperatures averaging 16.9°C (62.3°F). The intermountain
regions of western Montana not only have a large climatological difference in cool and
warm season temperatures, but also are prone to large and rapid variations in temperature
on extremely short time scales (i.e. weather events).
43
Table A-1. Location, elevation, and percent of record present for western
Montana meteorological stations used in this study.
Elevation
Record Present (%)
(m)
Weather Station
Latitude Longitude (Years of Record)
Bozeman MSU
1497.5
45°40’N 111°03’W
99.8
(1892-2006)
Choteau AP
1172.0
47°49’N 112°12’W
97.3
(1893-2006)
Cut Bank FAA AP
1169.8
48°36’N 112°23’W
99.0
(1903-2006)
Dillon WMCE
1593.5
45°13’N 112°39’W
99.5
(1895-2006)
Fortine 1N
914.4
48°47’N 114°54’W
99.5
(1906-2006)
Hamilton
1080.5
46°15’N 114°10’W
98.6
(1895-2005)
Helena WSO
1166.8
46°36’N 111°58’W
99.9
(1893-2006)
Kalispell WSO AP
901.3
48°18’N 114°16’W
99.9
(1896-2006)
Saint Ignatius
883.9
47°19’N 114°06’W
98.6
(1896-2006)
The high inter-seasonal and daily extremes in temperature that define this region
and its ecosystems motivated our investigation of changes in seasonal averages and daily
extremes (vs. seasonal or annual averages). For many physical processes critical to
resource management (e.g. timing of snowmelt, timing and amount of streamflow, and
forest fire activity), changes in variability, duration, and timing of extreme events within
a season may be just as important (if not more so) than changes in seasonal averages. For
example, the timing, duration, and severity of a specific temperature anomaly may cause
undesirable biological organisms (i.e. invasive) to thrive, and strain physical resources
such as water and its sources (e.g. snow and glaciers). Consecutive days of high
44
temperatures combined with earlier snowmelt and a longer dry season have also resulted
in more frequent forest fires in the western U.S. (e.g. Westerling et al., 2006). This point
is particularly acute for western Montana forests, as temperatures preceding and during
the fire season exert a significant influence on area burned, potentially more than longterm fuel accumulation (Littell et al., in press; Westerling et al., 2006; Mckenzie et al.,
2004). The same summertime temperature changes have also been documented to be
impacting western Montana’s glaciers and perennial snow and ice masses (e.g. Figure A2; Hall and Fagre, 2003; Pederson et al., 2004; Pederson et al., 2006; Watson et al.,
2008), with recent evidence for a portion of the retreat being driven by wintertime
warming (Fountain et al., 2007).
A core objective of this paper is to advance understanding of how western
Montana’s climate is changing across seasons, the degree to which these changes reflect
those observed for the Northern Hemisphere (and globally), and to discuss implications
for ecosystems and their services. We also assessed changes in variability and timing of
temperature thresholds important to biological and physical systems. Specifically, 100+
years of change in absolute temperature thresholds of -17.8°C (0°F), 0°C (32°F), and
32.2°C (90°F) are investigated due to their importance to the people and ecosystems of
western Montana. It should be noted that extremes in climate could be quantified in
terms of relative and absolute extremes. Relative extremes are expressed in terms of
percentiles or statistical distributions, whereas absolute extremes may refer to numerical
values that represent the genetic, physiological, or even cultural adaptation of plants or
45
Figure A-2. Repeat photographs of Boulder Glacier depicting 97 years of change. The
Boulder Glacier resided in Glacier National Park, Montana, and the photos were taken
from the summit of Chapman Ridge. Morton Elrod captured the 1910 photo, and Greg
Pederson and Daniel Fagre reshot the 2007 image. See the USGS repeat photography
webpage (http://www.nrmsc.usgs.gov/repeatphoto/) for more images of glacier change
in northwestern Montana.
animals to certain temperatures. In this paper absolute extremes are used because of their
documented regional importance for physical and biological processes. Further, from the
standpoint of science communication, there is much greater potential for immediate
understanding of the relevance of changes in absolute rather than relative temperatures.
The importance of a particular absolute temperature, however, will vary by region and
ecosystem. Recognizing the potential utility of producing these records for different
46
regions and processes of interest, the data resources, methods, and analyses are kept
simple, adaptable, and made available to scientists and the general public.
Methods
Temperature data were obtained from the nine meteorological stations (out of a
potential 21 stations) with the longest records of high-quality daily and monthly records
positioned within or along the Rocky Mountains of western Montana (Figure A-1, Table
A-1). The subset of stations maximized the length (100+ yrs) and completeness (>97%)
of daily and monthly measurements, and minimized potential bias from station moves
and urban heat island effects. The long-term (1895-2002) monthly and daily data used in
analyses were provided by the U.S. Historic Climatology Network (USHCN; Williams et
al., 2007c), and recent updates (2002-2006) obtained from the Western Regional Climate
Center (WRCC, 2007). Northern Hemisphere and Global temperature data were obtained
from the Global Historic Climatology Network (Lugina et al., 2005).
Utilizing USHCN monthly temperature data ensured records have been corrected
for time-of-observation biases, station moves, instrument changes, urban heat island
effects, and incompleteness of record (Karl et al., 1986; Karl and Williams, 1987; and
Karl et al., 1988). Though the USHCN daily temperature data represent the best highresolution, long-term climate records available, and are primarily free of time-ofobservation biases, care must be taken when interpreting an individual station record
since other types of bias corrections (e.g. station moves) are lacking (see Easterling et al.,
47
1999). The potential influence resulting from problems with a single station record were
minimized by producing an average regional climatology through mean-scaling followed
by record averaging (see below for details).
Missing data always present a potential problem for analyzing and averaging time
series. The U.S. HCN monthly temperature data, however, has missing values
interpolated by using a network of surrounding stations, so records are generally free of
gaps. No such data interpolation has been performed on daily temperature records; so
only stations with records greater than 97% complete (Table A-1) were used. This
ensured data gaps were minimal (usually restricted to a portion of any particular month),
and only portions of regional average time series with data from at least 5 of the 9
meteorological stations were presented in final figures.
Regional temperature time series were constructed using monthly and daily data,
and then analyzed for long-term seasonal trends, correspondence with variability and
trends in Global and Northern Hemisphere temperature, and investigation of changes in
temperature thresholds important to biological and physical systems. Individual station
maximum, minimum, and average temperature (Tmax, Tmin, and Tavg henceforth) records
were used to construct regional Tmax, Tmin, and Tavg records for all seasons by scaling the
mean and averaging the records. Mean-scaling was performed by calculating a regional
20th century mean for each temperature variable over each season, and centering each
station record on that mean before averaging records together. This method preserves the
48
temperature units in degrees Celsius (vs. dimensionless z-scores), allows for investigation
of different rates of change between Tmax and Tmin, and provides metric-scaled
temperature estimates of seasonal averages. After the seasonal and annual temperature
time series were produced, the strength of the relationships between regional, Global, and
Hemispheric records was assessed using Pearson product-moment correlation
coefficients.
Regional daily temperature data statistics were created following the analysis of
Weiss and Overpeck (2005), who documented the decreasing number of days and length
of freezing season in the Sonoran Desert. Code was written in the R project software
environment (R Development Core Team, 2007; http://www.r-project.org/) that tallied
and summed the annual number of Tmin days below -17.8°C (0°F) and 0°C (32°F) over a
winter year defined as beginning July 1 and ending June 30. Concurrently, the program
tallies and sums the number of Tmax days above 32.2°C (90°F) for a summer year
beginning January 1 and ending December 31. For each temperature threshold variable,
the program also records the first and last summer or winter day the threshold value was
met. (The daily temperature analysis code along with documentation is available from the
U.S. Geological Survey Northern Rocky Mountain Science Center web site
(http://www.nrmsc.usgs.gov/MTclimate/)). After the individual station records were
generated, a regional average was produced following the same mean-scaling techniques
described above. Principal component analysis (see Timm, 2002) was performed on the
49
individual station records to identify common variance and regional patterns of
temperature change.
Results
a. Regional Trends in Seasonal Maximum and Minimum Temperatures
Globally, instrumental evidence exists for significant decreases in Diurnal
Temperature Range (DTR) with a more rapid increase observed in Tmin relative to Tmax
(Easterling et al., 1997; Vose et al., 2005). Current thinking on the physical explanation
for the observed rapid rises in Tmin includes a combination of increased retention of longwave radiation due to higher atmospheric CO2 values, land cover changes, and/or
potential increases in cloudiness and atmospheric humidity (Easterling et al., 1997).
Though atmospheric CO2 undoubtedly plays a role, other specific mechanisms such as
increases in cloudiness and/or atmospheric humidity have not been shown to be a
significant factor in this region. In a previous regional study, Watson et al. (2008) found
a significant reduction in DTR due to more rapid increases in Tmin compared to Tmax from
a network of stations running from northwestern Montana north through Jasper National
Park, Alberta, CA. The same relationship is apparent for western Montana; however,
changes in Tmax and Tmin are different across seasons (Figure A-3, Table A-2). Annually
and seasonally all linear trends in maximum and minimum temperatures are significant (p
! 0.001). The largest decreases in DTR over the past century were due to a more rapid
rise in Tmin and have occurred in summer (DTR reduction = -0.77°C) and winter (DTR
reduction = -0.65°C). Spring Tmax appears to be rising as fast as Tmin (DTR = 0.01°C),
50
while conversely DTR over the fall season has increased (DTR increase = 0.21°C) with
Tmax rising more rapidly than Tmin.
Figure A-3. Seasonal trends in maximum (red line) and minimum (blue line)
temperatures for western Montana. Note change in scale of temperature axis for each
season.
The records presented in Figure A-3 exhibit other patterns of variability in
seasonal Tmax and Tmin, and present evidence of seasonal averages crossing the 0°C solid
to liquid phase change (i.e. freeze/thaw) threshold for water. Though freeze/thaw
terminology is typically reserved for daily, rather than seasonal, observations of
Tmin/Tmax, in the interest of keeping discussion simple it will be used interchangeably
here. In the spring, summer, and fall interannual variability in Tmax is greater than Tmin,
but correlation between records is high. During the winter, correlation between records
remains high, however, Tmin exhibits greater absolute interannual variability than Tmax.
Most relevant to water resources (via snow and ice melt) and biological processes,
51
however, are the trend lines for average spring and fall Tmin that will soon cross the 0°C
freeze/thaw threshold if the current trend continues. Though seasonal means are not
always the best surrogate for numbers of days above or below specific temperature
thresholds, the metrics typically correspond. More importantly, seasonal temperature
means do represent the average amount of ambient energy available to do work on a
system (e.g. driving phase changes in water), and thus are worth examining in detail. For
example, since the 1980s the frequency with which spring and fall Tmin were above 0°C
has increased dramatically (Figure A-3), resulting in a rapid decline in number of days
below freezing (discussed further in next section). Winter temperatures show the Tmax
trend-line crossing the 0°C threshold in the 1950s. Specific impacts from further
increases of winter Tmax, however, are unknown since variability is high and has regularly
crossed the 0°C threshold in the past. Changes in winter Tmax may, have substantial
impacts on snow depth, snow water equivalent, and cold content of valley snowpacks.
Table A-2. Change in annual and seasonal maximum and minimum
temperatures in western Montana. Overall change and significance of
trend was assessed conservatively using ordinary least squares regression.
20th Century #
# per decade
Variable
(°C)
(°C)
p-value
Annual Tmax
1.13
0.11
0.000
Winter Tmax
1.77
0.18
0.009
Spring Tmax
1.58
0.16
0.004
Sumer Tmax
1.00
0.10
0.015
Fall Tmax
1.10
0.11
0.027
Annual Tmin
1.55
0.15
0.000
Winter Tmin
2.42
0.24
0.002
Spring Tmin
1.57
0.16
0.000
Summer Tmin
1.77
0.18
0.000
Fall Tmin
0.89
0.09
0.010
52
b. Trends in Daily Tmax/Tmin Extremes and Thresholds
A major concern associated with temperature changes expected from global
warming are the potential associated changes in frequency and duration of extreme
events. Significant changes in number, duration, and variability of the first and last
calendar days-yr temperatures were " 32.2°C (90°F; Figure A-4), ! 0°C (32°F; Figure A5), and ! -17.8°C (0°F; Figure A-5) are documented below. Changes in the number of
days " 32.2°C indicate this region has experienced an approximate three-fold increase in
number of extremely hot days-yr. On average, five days-yr were " 32.2°C throughout the
early decades of the 20th century. The previous two decades, however, averaged fifteen
days-yr " 32.2°C. More conspicuous than the positive trend in days-yr "32.2°C, is the
increased frequency of years with an extreme number of days " 32.2°C. The record
setting year for number of days "32.2°C was 2003 (analysis ends in 2006), at 32 days
"32.2°C. Other years similar to 2003 in terms of numbers of days "32.2°C (i.e. "20
days) include, 2001, 2000, 1998, 1988, 1967, 1962, 1961, 1941, 1937, 1932, 1930, and
1919. The number of days-yr the 32.2°C Tmax threshold was equaled or exceeded also
exhibits pronounced sub-regional variability. Principal components analysis of
individual station data revealed different timing and rates of change for northwestern and
southwestern Montana (G. Pederson, unpublished data). Regional averages from stations
located in northwestern and southwestern Montana exhibit this relationship (Figure A-4
Top). An extremely rapid early-20th century rise (1917-1942) in average number of daysyr
" 32.2°C, and strong decadal-scale variability is evident for the northwestern region of
53
Montana. The southwestern region, however, exhibits a more linear increase in number
of days-yr " 32.2°C over the period of record (Figure A-4 Top). The rapid 1917-1942 rise
in northwestern Montana temperatures was associated with the severe dustbowl and predustbowl droughts of the Pacific Northwest (Pederson et al., 2006; Watson et al., 2004),
rapid recession of glaciers in Glacier National Park (Carrara, 1989; Key et al., 2002;
Pederson et al., 2004), as well as increased fire activity (Littell et al., in press; Pederson
et al., 2006).
Figure A-4. Top: Graphs showing the different rates of change in number of days-yr
Tmax " 32.2°C (gray bars) for southwestern (left) and northwestern Montana (right)
from 1895-2006. A 5-year moving average (red line) highlights trends and variability in
#days-yr " 32.2°C. Bottom: Trends in number of days (gray bars), and the first (green
line) and last (blue line) day of every summer, that temperature in western Montana
equaled or exceeded 32.2ºC.
With a three-fold increase in the average number of days " 32.2°C, an increase in
the seasonal window of time these temperatures can occur is expected. Figure A-4
54
(bottom) shows evidence for the temporal extension of the “hot” summer season with the
average first occurrence of temperatures " 32.2°C in the early-20th century beginning on
summer year day (YD) 189 (~July 8), and ending on average by summer YD 223
(~August 11). By the early-21st century, however, the average number of days " 32.2°C
has increased by 24 days, with the first summer YD " 32.2°C occurring on day 179
(~June 28) and the last “hot” event on summer YD 237 (~August 25).
Figure A-5. Average number of frost/freeze days (#days-yr Tmin !0.0°C (32.0°F)) or
extremely cold days (#days-yr Tmin !-17.8ºC (0.0ºF)) per winter year in Western
Montana (gray bars). Graphs also depict the first day of fall (blue line) and last day
of spring (green line) temperature equaled or exceeded the defined threshold.
55
With a demonstrated increase in number of “hot” days ("32.2°C) experienced per
year across western Montana, it follows logically that a reduction in number of “cold”
days per year should be evident. With few exceptions, western Montana meteorological
stations have experienced a decrease in annual number of freeze/thaw days (Tmin ! 0°C),
and extremely cold days (Tmin !-17.8°C). 1The average loss of number of days at or
below the freeze/thaw threshold (Tmin ! 0°C) in western Montana is approximately 16
days, declining from an average of ~186 to ~170 days-yr (Figure A-5 top). The sharpest
decline in number of freeze/thaw days has occurred within the last 20 years. The decline
in number of days !0°C corresponds with the increasing number of years that average
spring Tmin has equaled or exceeded 0°C (Figure A-3). Also associated with the decline
in average annual number of freeze/thaw days is a narrowing of the annual window over
which Tmin ! 0°C. Figure A-5 (top) shows a post-1980s trend towards a later arrival of
the first freeze/thaw day in the fall and an earlier termination of freeze/thaw events in the
spring.
The regional data document a loss in annual average number of extremely cold
days (Tmin !-17.8°C). From 1895-1980, western Montana averaged 20 days per year
where Tmin was !-17.8°C, with many “extremely cold” individual years experiencing
between 30 to 44 days below -17.8°C (Figure A-5 bottom). Recent decades (1981-2006)
have averaged 14 days per year Tmin !-17.8°C, with a reduction in frequency of years
1
In this paper, a freeze/thaw day and an extremely cold day are defined as any 24 hour period where the
average regional Tmin is ! 0°C or !-17.8°C respectively. There may be other thresholds of interest, and
these can be defined using either Tavg or Tmax, and the R program script.
56
with high numbers of extremely cold days. For example, the 1992 winter year was the
only year with close to a month (29 days) of Tmin !-17.8°C.
The onset and end of extremely cold temperatures in the region also exhibits
temporal variability. Over the length of record (1895-2006), the onset of extremely cold
temperatures (Tmin !-17.8°C) is highly variable, exhibits no trends, and typically occurs
near winter YD 158 (~December 5; Figure A-5). The last extremely cold day of the
winter season, however, has changed significantly, arriving an average of 19 days earlier.
During the early-20th century (1900-1910) extremely cold temperatures (tmin !-17.8°C)
typically ended on winter YD 248 (~March 5). Over the past decade (1996-2006) the end
of winter season’s extremely cold events has occurred on average by winter YD 228
(~February 15). The earlier termination of extreme cold events (tmin !-17.8°C)
documented here reflects the autumn/spring asymmetry in warming noted below.
c. Global Warming vs. Regional Trends
There is a high degree of regional variability in the expression of climate change
because temperature variability at regional scales results from the interaction of local(e.g. land-use and land cover change) and global-scale forcings (e.g. ocean atmosphere
interactions, changing concentrations of greenhouse gasses, volcanic events, solar
variability). For western Montana, annual and seasonal temperature variability tracks
Global and Northern Hemisphere (NH henceforth) trends on short- (interannual) and
long-term (multi-decadal and greater) scales (Figure A-6, Figure A-7). The similarity
57
Figure A-6. Comparison of variability and trends in western Montana (blue-green) and
Northern Hemisphere (blue line) annual average temperatures. A 5-year moving
average (red line) highlights the similarity in trends and decadal variability between
records.
between the interannual temperature variability implies that on a year-to-year basis, if the
NH is warm (or cold), typically western Montana is as well. The similarity between
long-term trends (or more generally low-frequency change; highlighted in Figures A-6
and A-7 using a 5-year moving average) suggests that the same large-scale forcings (e.g.
GHG forcings, solar variability, SST patterns, volcanic events) driving global
temperature change may drive a substantial portion of the observed low-frequency
change in western Montana. Correlations between western Montana, NH, and Global
temperatures range from r = 0.304 to r =0.581 and are significant (p! 0.005; Table A-3).
Similarly, although the relationships between regional and Global temperatures appear
robust, at the scale of an individual station the strength of the relationship varies (Figure
A-8). The intermediate strengths of the correlations indicate the importance of local- to
regional-scale climate forcings as compared to global-scale climate forcings (see
58
Figure A-7. Comparison between trends in western Montana and Northern
Hemisphere (blue line) seasonal temperatures. Red line displays the western Montana
5-year running mean. Note change in scale of temperature axis for each season, this
has implications related to absolute temperature change.
Discussion), but also imply results should be interpreted cautiously since the high number
of observations functionally results in lower correlation thresholds for a statistically
significant finding. Also, correlation analysis assumes a linear relationship between
variables; meaning if non-linear relationships between variables exist an overall lower
correlation coefficient will be produced.
Discussion
a. Significance and Representativeness of Climatic Trends
The long-term climatic trends in western Montana are consistent with other
studies of mountains in the temperate zone (e.g. Nogues-Bravo et al., 2006; Diaz and
Eischeid 2007; Mote 2003). Significant trends in daily and seasonal temperature
59
resemble the more rapid warming occurring at high latitudes and across heavily forested
regions (Dang et al., 2007). Western Montana has thus far experienced a +1.33°C (19002006) rise in annual average temperatures (Figure A-6), which is 1.8 times greater than
the +0.74°C (1900-2005) rise in Global temperatures (Lugina et al., 2005).
Table A-3. Correlation between western Montana average
annual and seasonal temperatures, and Northern Hemisphere and
Global average temperature records.
Northern Hemisphere
Global Tavg
Tavg Correlation
Correlation
Season
(p-value)
(p-value)
0.581
0.572
Annual
(0.000)
(0.000)
0.416
0.401
Winter (DJF)
(0.000)
(0.000)
0.414
0.430
Spring (MAM)
(0.000)
(0.000)
0.465
0.489
Summer (JJA)
(0.000)
(0.000)
0.323
0.304
Fall (SON)
(0.001)
(0.002)
Any discussion of local- to regional-scale changes in climate, however, warrants
further discussion of signal-to-noise ratios, and specifically the increasing amount of
unexplainable variability (or “noise”) that is encountered at increasingly smaller spatial
scales. For example, temperature records from a given meteorological station reflect a
mix of local- (e.g. microtopography, orography, development, instrument changes,
station moves), regional- (e.g. land-use and land-cover changes, concentration of
tropospheric aerosols), and global-scale climate forcings (e.g. ocean atmosphere
60
interactions, changing concentrations of greenhouse gasses, volcanic events, solar
variability) whereby the contribution to observed variance in climate from any particular
forcing is non-stationary through time. This relationship is reflected by the variable
correlation strengths shown for annual and seasonal average temperature records from
individual Montana meteorological stations as compared to averages for the Northern
Hemisphere (Figure A-8).
Figure A-8. Correlations between average annual and seasonal temperature records
for the Northern Hemisphere and the nine individual Montana stations. Correlations
are calculated between the unsmoothed time series and significant values are base on
p-value ! 0.05.
61
Regional climatologies reduce noise associated with local forcings thereby
providing a more representative picture of temperature trends and variability at scales
comparable to broader climate system changes, yet relevant to regional ecosystems. The
fact that western Montana tracks global and hemispheric patterns of temperature change
(Figure A-6, Figure A-7) suggests that global-scale forcings dominate, or at least are not
strongly counteracted by, local- to regional-scale climate forcings. While this
relationship is suggestive that knowledge of global or hemispheric temperature change
might provide some predictive relationship of western Montana temperatures, we caution
against such simplistic explanations. Using an ensemble of global climate models
Williams et al. (2007b) predict the Northern Rockies and consequently western Montana
will likely see the rise of a no-analog climate system configuration (see Figure A-3 in
Williams et al., 2007b). What this means for the future correspondence between western
Montana’s temperature in relation to global trends and variability is unclear. Similarly,
the relationship between global temperature trends (and variability) and trends shown for
western Montana do not constitute a formal test of attribution – and should not be
interpreted as such.
b. Evidence for Tipping Points and Thresholds in Climate Impacts
The rapid rise in temperatures has had a substantial impact on regional resources
and ecosystems, in part due to the crossing of critical daily and seasonal temperature
thresholds documented here (Figures A-3, A-4, and A-5). In the discussion below, we
describe the most sensitive elements of Montana’s ecosystems and economies as those
62
systems that 1) have critical processes that change rates as temperatures exceed critical
thresholds, and 2) have positive feedbacks that amplify the changes.
1) Alpine Glaciers are Retreating
Perhaps the most iconic impact of climate change in Montana is the retreat of
alpine glaciers within Glacier National Park, Montana ([Key et al., 2002; Hall and Fagre,
2003; Pederson et al., 2004], cf. Boulder Glacier repeat photographs; Figure A-2). The
dramatic recession of glaciers has resulted in part from the accelerated rate of ablation
(i.e. melting) that occurs with increasing average temperatures over the critical spring and
summer ablation season. Increases in ablation season average temperatures coupled with
spring and fall average Tmin crossing the freeze/thaw threshold implies both an
intensification of available energy to melt ice, and a potential lengthening of season over
which melt occurs. The analysis of daily data presented here shows a reduction in
extreme cold events and frost/freeze events, and an intensification of daily hightemperature events with a lengthening of ablation season (Figure A-4, Figure A-5).
The melting of alpine glaciers and perennial snow and ice masses, however, is not
a simple linear response to an outside climatic forcing. The initial response of alpine
glaciers to regional warming (i.e. glacier retreat) entrains a number of local climatic
positive feedbacks that amplify the physical response of the system. This is especially
true when average temperatures begin crossing important physical thresholds, such as the
0°C freeze/thaw isotherm. For example, as ice retreats and exposes more of the
surrounding bedrock and cirque walls, approximately 50% more long-wave radiation is
63
reemitted due to greatly reduced surface albedo resulting in more sensible heating of the
local environment. Thus, the area around the glacier warms non-linearly and increases
melt rates along the glacial margins, thereby accelerating the retreat process.
2) Fire and Mortality is Transforming Forests
The increasing number of hot days (days-yr "32.2°C) and the loss of frost/freeze
(days-yr !0°C) coincides with a number of observed changes in forest ecosystems. First,
the overall warming of the mountainous areas of western Montana during the winter and
spring, and the average loss of 16 days of temperatures !0°C matches the observed five
to > 20 day advance in the center-of-mass timing (CT) in streamflow for Montana
(Stewart et al., 2005), and a 30% reduction in April 1 snow water equivalent across the
Pacific Northwest since 1950 (Mote, 2003). Though changes in temperature records
from western Montana have not explicitly been linked to changes in snowpack and
streamflow timing here, Stewart et al. (2005) documents changes in temperature as a
significant driver in CT with the effects of larger synoptic controls (i.e. the Pacific
Decadal Oscillation) removed (see Figure A-11; Stewart et al. 2005). Also, a recent
detection and attribution study from Barnett et al. (2008) ascribes a majority (>60%) of
the observed changes in Western U.S. snowpack (and hence streamflow) from 1950-1999
to anthropogenic causes.
The earlier melt-out of mountain snowpacks combined with the increasing
number of days and length of season over which hot days occur has been shown by
Westerling et al. (2006) to have resulted in significant increases in fire frequency and
64
area burned both regionally and throughout the Western U.S. since the 1970s. This
study, however, lacks the long-term perspective provided by other fire/climate research
(e.g Littell et al, in press; McKenzie et al., 2004), which provide a more nuanced
perspective. The forests of western Montana, and the northern Rockies have been found
to have a fire regime strongly controlled by temperature and precipitation (i.e. water
balance [Littell et al., in press]), thus making the forests of western Montana highly
vulnerable to increased temperatures (Westerling et al., 2006). Increased temperatures,
however, are not the sole cause of frequent and large forest fire activity. Both an ignition
source (e.g. lightning, human activity) and a mechanism for rapid spread (i.e. wind
generated by storm fronts or the fires themselves) are a necessity for widespread regional
fire activity (McKenzie et al., 2004), and it is unclear how these variables may be
changing. The potential increased vulnerability to fire, however, is important since forest
fires play an important role in species structure and composition (Bond et al., 2005), and
changes in fire frequency are expected to play a role in driving biome shifts (Scholze et
al., 2006). The recent decades of increasingly large fires (Westerling et al., 2006) may
signal Montana’s forested ecosystems are currently in the process of change.
Insect outbreaks are another component of temperature-driven changes to western
forests. Increases in temperature affect bark beetle species in different ways. Warmer
temperatures – such as the winter and spring warming and loss of extreme cold days in
western Montana - may alter outbreak frequency/duration, reduce winterkills, speed up
life cycles, modify herbivory and damage rates, and lead to range expansion or
65
contraction (Carroll et al., 2003; Logan and Powell, 2001; Logan et al., 2003). From
western Montana through interior British Columbia the mountain pine beetle
(Dendroctonus ponderosae) has expanded its range to higher elevations, and farther east
then previously documented, causing widespread tree mortality (Logan et al., 2001). For
example, the hardest hit areas of the recent outbreak lie within interior British Columbia
with the Mountain Pine Beetle affecting 9,243,408 hectares (22,840,959 acres) of
lodgepole pine (Pinus contorta) as of September 2006 (B.C. Ministry of Forests, 2007).
With the Mountain Pine Beetles expansion to higher elevations, novel
species/host associations have occurred with the beetle infesting and killing whitebark
pine (Pinus albicalus). This is disconcerting since whitebark pine is considered a
‘keystone’ species due to its provision of food for more then seventeen animal species
(Arno and Hoff, 1990; and Tomback et al., 2001). With whitebark pine already
considered to be “functionally extinct over a third of its range” due to blister rust, a
changing fire regime, and now infestation from the mountain pine beetle there is serious
cause for concern regarding the future functional response of this subalpine ecosystem
(Tomback et al., 2001). Additionally, modeling efforts investigating the influence of
bioclimatic variables on the functional niche of whitebark pine within the Greater
Yellowstone Ecosystem suggest the potential for a complete loss of whitebark pine due to
projected changes in temperature and precipitation alone (Schrag et al., 2007; Bartlein et
al., 1997).
66
3) Rapid Changes in Freshwater Fisheries
Rapid ecosystem change may already be impacting the aquatic ecosystems of our
highest and coldest watersheds. Globally, over the last four decades the world has
witnessed the disappearance of over 7000 km2 of mountain glaciers (Dyurgerov, 2003).
The Greater Yellowstone and the Crown of the Continent Ecosystems (which
encompasses Glacier National Park) have lost 42% and 66% of their glacial and perennial
snow and ice cover respectively since 1900 (Fountain et al., 2007). Within Glacier
National Park the loss of glaciers has been most dramatic. Since the termination of the
Little Ice Age (ca. 1850 AD) glaciers and perennial snow/ice masses have been reduced
in number from approximately 150 to 26, which corresponds to an 83% loss in total ice
mass. With the loss of glaciers has come the loss of watersheds containing glaciers: 23
first-order watersheds historically contained glaciers at the source of their headwaters,
today only 14 watersheds contain glaciers or perennial snow and ice.
The influence of glaciers on aquatic ecosystems ranges from providing base flows
during the hot, dry summers to moderating stream temperatures. In these remnant coldwater fisheries regulation of stream temperature is of key importance in controlling the
distribution and abundance of invertebrates (Hauer et al., 1998) and fish (Keheler and
Rahel, 1996; Dunham et al., 2003). Salmonids are often considered a keystone species
for aquatic and terrestrial ecosystems, and may be an especially important indicator of
ecosystem health in the face of climate change. Salmonids provide an excellent early
warning indicator of climate warming because their body temperature is dependent on the
67
temperature of their surroundings, and they have a characteristically narrow range of
thermal tolerance. Almost all of the native inland cutthroat (Oncorhynchus clarkii spp.)
species, grayling (Thymallus arcticus) and bull trout (Salvelinus confluentus) have been
proposed for listing under the Endangered Species Act and a number are currently listed
as “threatened”. Trout, grayling, and char historically inhabited a variety of freshwater
habitats, but have declined due to habitat degradation, fragmentation, introductions of
nonnative species, and elevated water temperature. Many native salmonid populations
are restricted to small, fragmented headwater habitats, which are increasingly vulnerable
to wildfires and subsequent flooding. The bull trout (Salvelinus confluentus), for
example, is a native char to northwestern North America that requires cold, connected
and complex habitats for growth, survival, and long-term population persistence. A
warming climate, and changes to seasonal and daily temperature extremes during the
season of greatest stress (i.e. summer; see Figures A-4 and A-7), may negatively impact
existing trout populations by increasing water temperatures beyond critical physiological
thresholds. Bull trout have among the lowest mean tolerance for high water temperatres
of North American salmonids (Selong et al. 2001). Using theoretical modeling and
empirical air temperature data, Rieman et al. (2007) estimated that warming temperatures
over the distributional range of bull trout could result in the losses of 18-92% of
thermally suitable natal habitat area and 27-99% of large (>10,000-ha) habitat patches
(Rieman et al., 2007).
68
Global warming may ultimately be the greatest threat to the persistence of native
fishes because it will exacerbate current negative effects of invasive aquatic species and
habitat degradation while increasing water temperatures to unsuitable thresholds
(Williams et al. 2007a). In conjunction with losing the cooling effect of glaciers and
perennial snow and ice, stream temperature will be affected most by changes in
maximum summer temperatures and minimum winter temperatures (Keleher and Rahel
1996). Using an upper temperature threshold of 22°C (71.6°F) as a constraining variable
for a guild of cold water fish (brook trout, cutthroat trout, and brown trout), Keleher and
Rahel (1996) predicted that the length of streams occupied in Wyoming would decrease
7.5-43.3% for increases in temperature from 1 to 5°C. Changes in temperature of this
magnitude may not only decrease existing habitat by increasing air and water
temperatures, but will likely further fragment available suitable habitat, increase the risk
of catastrophic fire, change the timing and quantity of water from snowpack, increase
winter flooding in some areas, and provide habitat conditions that favor introduced
species, putting many native salmonids at high risk of extinction in the western United
States (Williams et al., 2007a).
4) Ecosystem Services and Economic Impacts
Montana’s ecotourism- and agriculture-based economy will likely experience a
mixture of positive and negative impacts as a consequence of future climate change. A
potential positive impact for ecotourism may arise from weather conditions more
amenable to people at the start and end of the traditional summer tourist season – thereby
69
increasing overall tourism numbers and length of visitation season. Conversely, the
premiere ski resort industry is likely to see a reduction in profits due to a shortening
season over which a high quantity and quality of snowpack is available for skiers
(Breiling and Charmanza, 1999). With a reduction in snowpack, and increased stream
temperatures over the spring and summer, fishing guides may expect increasingly more
frequent closures of streams and rivers due to reduced flows and increased thermal stress
on aquatic species.
Montana’s agricultural system has always had a tenuous relationship with
Montana’s climate. The predominance of dry-land farming coupled with the region’s
tendency to experience sustained drought conditions (Pederson et al., 2006) led to the
harsh human conditions and massive number of farm foreclosures during the dustbowl
and pre-dustbowl droughts of the 1920s and 30s (Murphy, 2003; Pederson et al., 2006).
In spite of this, western Montana’s highly productive and high-quality valley grasslands
have always served as valuable land for livestock production. With changes in timing of
specific chilling periods, which is likely happening as shown by decreases in winter
season cold temperatures (Figure A-5), it is expected that crop yields will decline and
more xeric conditions will prevail reducing pasture quality (Slingo et al., 2005) and
threatening Montana’s livestock industry. Economic models, coupled with climate
change predictions show agricultural regions in Montana that will be hardest hit are those
with the least resources and ability to adapt (Antle et al., 2004), which mirrors
expectations of global societal and agricultural impacts (Tol et al., 2004).
70
Conclusions
Regional analysis of trends in western Montana temperatures reveals changes that
track both interannual and multi-decadal variability exhibited in global and NH
temperature records. In all cases, however, the rise in extremes and seasonal averages
has been two to three times greater than that of the global average. Importantly, we see
substantial changes in extreme conditions, with both a loss of extremely cold days as well
as an increase in extremely hot days. Important thresholds are being exceeded, in
particular the 0°C seasonal frost/freeze value for average Tmin over the spring and fall.
Evidence for major ecosystem changes within the region is associated with these changes
in seasonal average temperatures, and the changing frequency of daily temperature
extremes and thresholds associated with physical and biological processes. A take-home
message of this analysis and review of impacts is that there is danger in assuming that
long-term trends in temperature will be associated with equally paced or simple linear
changes in ecosystems and their services. The processes by which ecosystem changes
take place are often complicated by synergistic and non-linear relationships between
variables (e.g. Nogues-Bravo et al., 2006; Scheffer et al., 2001), many of which are still
unknown. Thus, more intensive research on local and regional climate interactions with
coupled socio-ecosystems should be completed through the integration of observational
studies of past climate-socio-ecosystem relationships. The software provided herein is
intended to assist with the often-daunting task of extracting meaningful climate metrics.
In conjunction with these efforts, capacity to monitor for new system trajectories, or
sudden shifts, in our biophysical systems should be increased for long-term planning
71
and/or rapid mitigation measures. Regardless of future efforts in carbon mitigation and
sequestration, over the course of this next century we are committed to a warming
climate. Only through increased monitoring, research, and generation of spatially explicit
patterns of expected change, can we expect citizens and resource managers to engage in
well-informed decision-making.
Acknowledgments
We would like to thank C. Caruso for his early and helpful contribution to this
manuscript, along with J. Weiss, J. Overpeck, and S. Running for providing the
motivation to conduct this research. In addition, we are thankful for the insightful
comments on this manuscript from J. Littell, and three anonymous reviewers. Their time
and efforts greatly improved the writing quality and figures. A. Macalady provided both
valuable insights along with important additions to the R program for daily temperature
analysis. This work was supported by and is a product of the U.S. Geological Survey’s
Western Mountain Initiative and the Global Change Program.
72
References
Arno SF, and Hoff RJ (1990) Whitebark pine (Pinus albicaulis Engelm.). In: Silvics of
North America, Burns RM, and Honkala BH (eds). Agriculture Handbook 654, US
Department of Agriculture Forest Service, Washington, DC. 530–554 pp
Antle JM, Capalbo SM, Elliot ET, Paustian KH (2004) Adaptation, spatial heterogeneity,
and vulnerability of agricultural systems to climate change and CO2 fertilization:
An integrated assessment approach. Climatic Change 64(3):289-315
Barnett TP, Pierce DW, Hidalgo HG, Bonfils C, Santer BD, Das T, Bala G, Wood AW,
Nozawa T, Mirin AA, Cayan DR, and Dettinger MD (2008) Human-induced
changes in the hydrology of the western United States. Science Express: 2
November 2007; accepted 23 January 2008. Published online 31 January 2008;
10.1126/science.1152538
Bartlein PJ, Whitlock C, and Shafer SL (1997) Future climate in the Yellowstone
National Park region and its potential impact on vegetation. Conservation Biology
11:782–792
B.C. Minestry of Forests (2007) 2006 Aerial Overview Survey Summary. Forest
Practices Branch, Forest health conditions. URL
http://www.for.gov.bc.ca/hfp/health/overview/2006table.htm. Cited 22 January
2008.
Bond WJ, Woodward FI, and Midgley GF (2005) The global distribution of ecosystems
in a world without fire. New Phytol. 165:525-538
Breiling M, and Charmanza P (1999) the impact of global warming on winter tourism
and skiing: a regionalized model for Austrian snow conditions. Regional
Environmental Change 1(1):4-14
Carroll AL, Taylor SW, Régniére J, and Safranyik L (2003) Effects of climate change on
range expansion by the mountain pine beetle in British Columbia. In: Mountain
Pine Beetle Symposium: Challenges and Solutions. October 30-31, 2003, Kelowna,
British Columbia. Shore TL, Brooks JE, and Stone JE (eds). Natural Resources
Canada, Canadian Forest Service, Pacific Forestry Centre, Information Report BCX-399, Victoria, BC. 298 pp
Carrara PE (1989) Late Quaternary glacial and vegetative history of the Glacier National
Park region, Montana. U.S. Geol. Surv. Bull., 1902. 64 pp.
Church JA, and White NJ (2006) A 20th century acceleration in global sea-level rise.
Geophysical Research Letters 33: L01602 DOI: 10.1029/2005GL024826
73
Dang H, Gillett NP, Weaver AJ, and Zwiers FW (2007) Climate change detection over
different land surface vegetation classes. International Journal of Climatology 27:
211-220 DOI: 10.1002/joc.1397
Diaz HF, and Eischeid JK (2007) Disappearing “alpine tundra” Koppen climatic type in
the western United States. Geophysical Research Letters 34: L18707, DOI:
10.1029/2007GL031253.
Dunham, Rieman JB, and Chandler G (2003) Influences of temperature and
environmental variables on the distribution of bull trout within streams at the
southern margin of its range. North American Journal of Fisheries Management
23:894-904
Dyurgerov M (2003) Mountain and subpolar glaciers show an increase in sensitivity to
climate warming and intensification of the water cycle. Journal of Hydrology
282:164-176
Durgerov M, and Meier MF (2005) Glaciers and the changing Earth System: A 2004
snapshot. Occasional Paper 58, Institute of Arctic and Alpine Research, University
of Colorado, Boulder, CO. 118 pp
Easterling DR, Meehl GR, Parmesan C, Changnon SA, Karl TR, Mearns LO (2000)
Climate extremes: observations, modeling, and impacts. Science 289:2068-2074
Easterling DR, Karl TR, Lawrimore JR, and Del Greco SA. 1999. United States
Historical Climatology Network Daily Temperature, Precipitation, and Snow Data
for 1871-1997. ORNL/CDIAC-118, NDP-070. Carbon Dioxide Information
Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak
Ridge, Tennessee.
Easterling DR, Horton B, Jones PD, Peterson TC, Karl TR, Parker DE, Salinger MJ,
Razuvayev V, Plummer N, Jamason P, and Folland CK (1997) Maximum and
minimum temperature trends for the Globe. Science 277:364-367
Fountain A, Hoffman M, Jackson K, Basagic H, Nylen T, and Percy D (2007) Digital
Outlines and Topography of the Glaciers of the American West. USGS Open File
Report 2006-1340, 23pp
GAO: U.S. Government Accountability Office (2007) Climate change: Agencies should
develop guidance for addressing the effects on federal lands and water resources.
US government open file report GAO-07-863, 184pp. http://www.gao.gov/cgibin/getrpt?GAO-07-863. Cited 2 December 2008.
Hansen J, Nazarenko L, Ruedy R, Sato Mki, Willis J, Del Genio A, Koch D, Lacis A, Lo
K, Menon S, Novakov T, Perlwitz Ju, Russell G, Schmidt GA, and Tausnev N
74
(2005) Earth’s energy imbalance: confirmation and implications. Science 308:14311435 DOI:10.1126/science.1110252
Hall MP, and Fagre DB (2003) Modeled climate-induced glacier change in Glacier
National Park, 1850-2100. Bioscience 53(2):131-140 DOI: 10.1002/(SICI)10991085(19970630)11:8<903::AID-HYP511>3.0.CO;2-7
Hauer RF, Baron JS, Campbell DH, Fausch KD, Hostetler SW, Leavesley GH, Leavitt
PR, McKnight, DM, and Stanford JA (1998) Assessment of climate change and
freshwater ecosystems of the Rocky Mountains, USA and Canada. Hydrological
Processes 11(8):903-924
Karl TR, Diaz HF, and Kukla G (1988) Urbanization: Its detection and effect in the
United States climate record. J. Climate 1:1099-1123
Karl TR, and Williams CN Jr. (1987) An approach to adjusting climatological time series
for discontinuous inhomogeneities. J. Clim. Appl. Meteor. 26:1744-63.
Karl TR, Williams CN Jr., Young PJ, and Wendland WM (1986) A model to estimate the
time of observation bias associated with monthly mean maximum, minimum and
mean temperatures for the United States. J. Clim. Appl. Meteor. 25:145-60.
Keleher CJ, and Rahel FJ (1996) Thermal limits to salmonid distributions in the Rocky
Mountain region and potential habitat loss due to global warming: a geographic
information system (GIS) approach. Transactions of the American Fisheries Society
125:1-13
Key CH, Fagre DB, and Menicke RK (2002) Glacier retreat in Glacier National Park,
Montana. In: Satellite image atlas of glaciers of the world, glaciers of North
America—Glaciers of the Western United States. Williams RS Jr., and Ferrigno JG
(eds). United States Government Printing Office, J365–J381.
Littell JS, McKenzie D, Peterson DL, Westerling AL (2008) Climate and wildfire area
burned in Western U.S. ecoprovinces, 1916-2003. Ecological Applications (in
press)
Logan JA, and Powell JA (2001) Ghost forests, global warming and the mountain pine
beetle (Coleoptera: Scolytidae). American Entomologist, Fall:160–172
Logan JA, Régniére J, and Powell JA (2003) Assessing the impacts of global warming on
forest pest dynamics. Frontiers in Ecology and the Environment 1:130–137
Lugina KM, Groisman PYa, Vinnikov KYa, Koknaeva VV, and Speranskaya NA (2005)
Monthly surface air temperature time series area-averaged over the 30-degree
latitudinal belts of the globe, 1881-2005. In: Trends Online: A Compendium of
75
Data on Global Change. Carbon Dioxide Information Analysis Center, Oak Ridge
National Laboratory, U.S. Department of Energy, Oak Ridge, Tennessee, U.S.
http://cdiac.ornl.gov/trends/temp/lugina/data.html. Cited 22 January 2008.
McKenzie D, Gedalof Z, Peterson DL, and Mote P (2004) Climatic change, wildfire, and
conservation. Conservation Biology 18:890-902
Moberg A, Sonechkin DM, Holmgren K, Datsenko NM, and Karlen W (2005) Highly
variable Northern Hemisphere temperatures reconstructed from low- and highresolution proxy data. Nature 433:613-617
Mote PW (2003) Trends in snow water equivalent in the Pacific Northwest and their
climatic causes. Geophys. Res. Lett. 30: 1601, doi:10.1029/2003GL017258
Murphy M (2003) Hope in hard times: New Deal photographs of Montana, 1936-1942.
Montana Historical Society Press, Helena, Montana. 256 pp
Nogués-Bravo D, Arújo MB, Errea MP, and Martinez-Rica JP (2007) Exposure of global
mountain systems to climate warming during the 21st Century. Global
Environmental Change 17: 420-428. DOI: 10.1016/j.gloenvcha.2006.11.007
NRC (2005) Radiative forcing of climate change: Expanding the concept and addressing
uncertainties. The National Academies Press, Washington, D.C. 208 pp
NRC (2006) Surface temperature reconstructions for the last 2,000 years. The National
Academies Press, Washington, D.C. 145 pp
Oerlemans J (2005) Extracting a climate signal from 169 glacier records. Science
308:675-677
Pederson GT, Fagre DB, Gray ST, and Graumlich LJ (2004) Decadal-scale climate
drivers for glacial dynamics in Glacier National Park, Montana, USA. Geophys.
Res. Lett. 31, L12203, doi:10.1029/2004GL019770.
Pederson GT, Gray ST, Fagre DB, and Graumlich LJ (2006) Long-duration drought
variability and impacts on ecosystem services: A case study from Glacier National
Park, Montana. Earth Interactions 10:1-28
R Development Core Team (2007) R: A language and environment for statistical
computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3900051-07-0, URL http://www.R-project.org/. Cited 22 January 2008.
Rayner NA, Brohan P, Parker DE, Folland CK, Kennedy JJ, Vanicek M, Ansell TJ, and
Tett SFB (2006) Improved analyses of changes and uncertainties in sea surface
76
temperature measured in situ since the mid-nineteenth century: the HadSST2
dataset. Journal of Climate 19:446-469.
Rieman BE, Isaak D, Adams S, Horan D, Nagel D, Luce C, and Myers D (2007)
Anticipated climate warming effects on bull trout habitats and populations across
the interior Columba River basin. Transactions of the American Fisheries Society
136:1552-1565
Scheffer M, Carpenter S, Foley JA, Folke C, and Walker B (2001) Catastrophic shifts in
ecosystems. Nature 413: 591-596
Scholze M, Knorr W, Arnell NW, and Prentice IC (2006) A climate-change risk analysis
for world ecosystems. PNAS 103(35):13116-13120
Schrag AM, Bunn AG, and Graumlich LJ (2007) Influence of bioclimatic variables on
tree-line conifer distribution in the Greater Yellowstone Ecosystem: implications
for species of conservation concern. Journal of Biogeography, OnlineEarly Articles,
DOI: 10.1111/j.1365-2699.2007.01815.x
Selong JH, McMahon TE, Zale AV, and Barrows FT (2001) Effect of temperature on
growth and survival of bull trout, with application of an improved method for
determining thermal tolerances in fishes. Transactions of the American Fisheries
Society 130:1026-1037.
Slingo JM, Challinor AJ, Hoskins BJ, and Wheeler TR (2005) Introduction: food crops in
a changing climate. Philosophical Transactions of the Royal Society B-Biological
Sciences 360:1983-1989
Stewart IT, Cayan DR, Dettinger MD (2005) Changes toward earlier streamflow timing
across western North America. Journal of Climate 18:1136-1155
Tomback DF, Arno SF, and Keane RE (2001) The compelling case for management
intervention. In: Whitebark pine communities: ecology and restoration. Tomback
DF, Arno SF, and Keane RE (eds), Island Press, Washington DC. 3–25 pp.
Tol RSJ, Downing TE, Kuik OJ, and Smith JB (2004) Distributional aspects of climate
change impacts. Global Environmental Change 14:259-272
Timm NH (2002) Applied multivariate analysis. Springer-Verlag, New York, NY. 693
pp. ISBN: 978-0-387-95347-2.
Vose RS, Easterling DR, and Gleason B (2005) Maximum and minimum temperature
trends for the globe: An update through 2004. Geophysical Research Letters, 32,
L23822, doi:10.1029/2005GL024379.
77
Watson E, Pederson GT, Luckman BH, and Fagre DB (2008) Glacier mass balance in the
northern U.S. and Canadian Rockies: Paleo-perspectives and 20th century change.
B Orlove, Luckman BH, and Wiegandt E (eds) In: Darkening Peaks, University of
California Press, Berkley, C.A. pp 139-151
Watson E, and Luckman BH (2004) Tree-ring based reconstructions of precipitation for
the southern Canadian Cordillera. Climatic Change 65:209–241
Weiss JL, and Overpeck JT (2005) Is the Sonoran Desert losing its cool? Global Change
Biology. 11: 2065-2077. DOI 110.1111/j.1365-2486.2005.01020.x
Westerling AL, Hidalgo HG, Cayan DR, Swetnam TW (2006) Warming and early spring
increase Western U.S. forest wildfire activity. Science 313(5789):940-943. DOI
10.1126/science.1128834
Western Regional Climate Center (2007) Climate of Montana.
http://www.wrcc.dri.edu/narratives/MONTANA.htm. Cited 22 January 2008.
Williams JE, Haak AL, Neville HM, Colyer WT, and Gillespie NG (2007a) Climate
change and western trout: strategies for restoring resistance and resilience in native
populations. Wild Trout IX Proceedings. In-press.
Williams JW, Jackson ST, and Kutzbach JE (2007b) Projected distributions of novel and
disappearing climates by 2100 AD. PNAS 104(14): 5738-5742. DOI:
10.1073/pnas.0606292104.
Williams CN Jr., Menne MJ, Vose RS, and Easterling DR (2007c) United States
historical climatology network monthly temperature and precipitation data.
ORNL/CDIAC-118, NDP-019. Available on-line (monthly,
http://cdiac.ornl.gov/epubs/ndp/ushcn/usa_monthly.html; daily,
http://cdiac.ornl.gov/epubs/ndp/ushcn/usa.html) from the Carbon Dioxide
Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of
Energy, Oak Ridge, Tennessee. Cited 22 January 2008.
78
APPENDIX B - CLIMATIC CONTROLS ON THE SNOWMELT HYDROLOGY
OF THE NORTHERN ROCKY MOUNTAINS, USA
Authors:
1,2
Gregory T. Pederson*, 3Stephen T. Gray, 4Toby Ault, 5Wendy Marsh, 1Daniel B. Fagre,
6
Andrew G. Bunn, 7Connie A. Woodhouse, and 2Lisa J. Graumlich
JOURNAL OF CLIMATE - Submitted March 10, 2010
Affiliations:
1
U.S. Geological Survey, Northern Rocky Mountain Science Center, 2727 University Way (Suite
2), Bozeman, MT 59715
2
School of Natural Resources, University of Arizona, 325 Biosciences East, Tucson, AZ 85721
3
Water Resource Data System, University of Wyoming, Laramie, WY 82071
4
Geosciences Department, University of Arizona, Gould-Simpson Building, 127A, Tucson, AZ
85721
5
Big Sky Institute, Montana State University, 229 AJM Johnson Hall, Bozeman, MT 59717
6
Huxley College, Western Washington University, Bellingham, WA 98225-9181
7
School of Geography and Development, 412 Harvill Building, University of Arizona, Tucson,
Arizona 85721-0076
*Corresponding Author: Gregory T. Pederson
Email: [email protected]
Phone: 406-994-7390
79
Abstract
The Northern Rocky Mountains (NRMs) are a critical headwaters region with the
majority of water resources originating from mountain snowpack. Observations showing
declines in western U.S. snowpack have implications for water resources and biophysical
processes in high-mountain environments. This study investigates oceanic and
atmospheric controls underlying changes in timing, variability, and trends documented
across the entire hydroclimatic monitoring system within critical NRM watersheds.
Analyses were conducted using records from 25 snow telemetry (SNOTEL) stations, 148
April 1 snow course records, stream gage records from 14 relatively unimpaired rivers,
and 37 valley meteorological stations. Over the past four decades, mid-elevation
SNOTEL records show a tendency towards decreased snowpack with peak snow water
equivalent (SWE) arriving and melting out earlier. Temperature records show significant
seasonal and annual decreases in number of frost days (days ! 0°C), and changes in the
spring minimum temperatures that correspond with atmospheric circulation changes and
surface-albedo feedbacks in March and April. Spring temperatures coupled with
increases in mean and variance of spring precipitation are associated with timing of snow
melt-out, an increased number of snow-free days, and observed changes in streamflow
timing and discharge. The majority of the variability in peak and total annual snowpack
and streamflow, however, is explained by season dependent inter-annual to inter-decadal
changes in atmospheric circulation associated with Pacific Ocean sea-surfacetemperatures. Over recent decades, increased spring precipitation appears to be buffering
NRM total annual streamflow from what would otherwise be greater snow-related
80
declines in hydrologic yield. Results have important implications for ecosystems, water
resources, and long-lead forecasting capabilities.
Introduction
Much of the western U.S. and Canada is characterized by an arid to semi-arid
climate, and the majority (up to 80% or more) of surface water in this region originates as
mountain snowpack (Hamlet et al. 2007; Serreze et al. 1999; Stewart et al. 2005). Snow
serves as a natural reservoir for water that is released over the spring, summer, and fall,
thereby providing municipal and industrial water supplies to support a rapidly growing
population, as well as furnishing water for energy production, agriculture and
ecosystems. In turn, both the quantity and timing of water released from snowpack are
critical societal and environmental concerns. A growing number of studies have
demonstrated that since 1950, western North America has experienced a substantial
decline in peak snow water equivalent (SWE; Das et al. 2009; Mote et al. 2005; Pierce et
al. 2008), and subsequently a reduced and earlier snowmelt runoff (Aguado et al. 1992;
Cayan et al. 2001; Dettinger and Cayan 1995; Hidalgo et al. 2009; McCabe and Clark
2005; Rajagopalan et al. 2009; Regonda et al. 2005; Stewart et al. 2005). Accompanying
these changes is evidence for a higher proportion of precipitation falling as rain rather
than snow (Knowles et al. 2006), increasingly low base-flows during dry years (Luce and
Holden 2009), and significant increases in the percentage of total water-year discharge
occurring during the winter (Das et al. 2009; Dettinger and Cayan 1995; Stewart et al.
2005).
81
Numerous studies have identified increases in winter and spring minimum
temperatures (Tmin) linked to greenhouse gas (GHG) induced global warming, as the
primary drivers of hydrologic change in snow-dominated areas of western North America
(Barnett et al. 2008; Dettinger et al. 2004; Hamlet et al. 2007; Knowles and Cayan 2004;
Regonda et al. 2005; Stewart et al. 2005). However, the suite of observed hydroclimatic
changes within a particular region or location also incorporates complex relationships
between localized weather events, and synoptic- to hemispheric-scale forcings (e.g.
regional land-cover change to ocean-atmosphere teleconnections). For example, over
inter-annual to decadal timescales the seasonal controls on hydroclimatic variability and
change are strongly influenced by natural ocean-atmosphere interactions extending from
the Tropical and North Pacific Ocean (e.g., Cayan 1996; Cayan et al. 1998; Dettinger et
al. 1994; Rood et al. 2005), with modifying influences associated with the North Atlantic
(e.g., Enfield et al. 2001; McCabe et al. 2004). Variations in sea surface temperatures
(SSTs) substantially alter temperature and precipitation patterns across large regions (i.e.
at sub-continental scales) by modifying atmospheric circulation patterns and
consequently changing the preferential positioning of storm tracks (Cayan 1996; Cayan et
al. 1998; Dettinger et al. 1994). For the NRMs, the Pacific Decadal Oscillation (PDO;
Mantua et al. 1997) is the dominant inter-annual to decadal-scale index of Pacific Basin
sea-surface-temperature variability in this regard because it integrates high- and lowfrequency information from both the Tropical Pacific and North Atlantic - though the
specific mechanisms are still being explored (e.g. Dong et al. 2006; Enfield et al. 2001;
McCabe et al. 2008; Zhang and Delworth 2007).
82
In mountainous regions, studies seeking to understand hydroclimatic variability
and trends are further complicated by the influence of topography on climate (Das et al.
2009; Karoly and Wu 2005; Pederson et al. 2010). One common approach to detection
and attribution studies in the western U.S. has been to examine average conditions over
relatively-large spatial domains (usually many 1,000’s to 10,000’s of km2), but this
invariably masks local to regional-scale (100’s to 1,000’s of km2) variations and trends
that may be critical for water- and natural-resource management. This is particularly true
in that local to regional variability and change may often differ from the patterns that are
observed or predicted at continental to sub-continental scales (Das et al. 2009; Pederson
et al. 2010). As a result, disentangling the relative influence of secular warming and
other drivers on the amount and timing of snowmelt and related streamflow processes can
be enormously difficult. This complexity necessitates the close examination of a wide
range of observational data from multiple hydroclimatic monitoring networks.
In this study we assess historic variability and trends in the hydroclimatology of
snow-dominated watersheds in the Northern Rocky Mountains (NRMs; Figure B-1),
defined here to include the U.S. northern Rocky Mountains and the southern Canadian
Rocky Mountains. In addition, we examine the role of regional and synoptic-scale
drivers, as well as remote forcings, as controls on hydroclimatic variability and change
within these watersheds. This study is motivated by increasing pressure on state and
federal resource management agencies to incorporate this type of climate information
into sustainability- and adaptation-planning efforts (Rocky Mountain Climate Working
83
Figure B-1. Map of the Northern Rocky Mountain study region and the snow telemetry
(SNOTEL; n = 25), snow course (n = 148), stream gage (n = 14), and valley
meteorological stations (n = 37) used in this study.
Group 2009). The NRM region is particularly suitable for this type of study because it
contains a number of the longest-running, high-resolution snow observation stations in
North America, as well as a suite of long-duration, high-quality records from stream
gages and meteorological stations (MET). Understanding the drivers of hydroclimatic
variability and change in this region is critical for several reasons: 1) the NRMs
encompass the transboundary Crown of the Continent Ecosystem (Figure B-1), 2) they
form a key headwaters for three of North America’s largest river systems (i.e. the
Columbia, Missouri, and South Saskatchewan Rivers), and 3) the region also contains the
84
world’s first International Peace Park (collectively Glacier [U.S.] and Waterton Lakes
National Parks [Canada]) which is also a UNESCO World Heritage Site, and arguably
one of the most intact and functional temperate ecosystems in the world (Prato and Fagre
2007).
Based on prior research (e.g Cayan et al. 1998, Das et al. 2009) we expect that the
majority of the observed variance in NRM peak and total annual snowpack and
streamflow arise from changes in synoptic-scale drivers (e.g. PDO, ENSO, with attendant
changes in atmospheric circulation) with a minor, but significant influence of regional
surface temperatures. As for changes in the timing of hydrologic events (i.e. snowmelt,
snowcover duration, streamflow center-of-mass timing), we expect observed trends and
variability to be strongly controlled by surface temperature. In this paper we specifically
diagnose changes within, and co-variability between, climatic and hydroclimatic timeseries from individual stations and region-wide averages. In particular we examine
relations between snow-cover duration and snow water equivalent (SWE) along with
associated changes in streamflow discharge. We also detail changes in low- to midelevation temperatures (including number of frost days [! 0°C]), along with linkages to
snowmelt and streamflow timing. Finally, we examine connections between long-term
hydroclimatic changes (i.e. amount and timing) and ocean-atmosphere teleconnections.
85
Data and Methods
The study region was first divided into two sub-watersheds based on U.S.
Geological Survey watershed units (Hydrologic Unit Maps; Seaber et al. 1987), and snow
telemetry (SNOTEL), then snow course records providing the broadest spatial coverage
were selected for analyses (Figure B-1). SNOTEL stations provided the daily records of
snow water equivalent (SWE) and co-located temperature loggers used in high-resolution
assessments of changing snow dynamics, whereas April 1 SWE records from snow
course measurements provide long-term, coarse-resolution estimates of peak SWE for
extended record comparisons. SWE and temperature data were initially obtained from 37
SNOTEL stations in the U.S. and Canadian portions of the NRM study region. Although
other Canadian hydroclimatic records were used, the 12 Canadian SNOTEL records were
excluded from further analyses because long reporting intervals (" 5 days) and numerous
data gaps precluded their use in generating the daily statistics. The remaining 25 records
(Figure B-1, Table B-1) are taken from Natural Resource Conservation Service (NRCS;
http://www.wcc.nrcs.usda.gov/snow/) SNOTEL stations in western Montana. The
selected SNOTEL records contained a minimum of 15 years of daily data, the longest of
which extend back to 1969 - the networks earliest observations. Stations are located at an
average elevation of 1743 m a.s.l. (± 259 m; Table B-1), making this the highest
elevation hydroclimatic-observing network in the NRMs. It should be noted, however,
that significant land-area without any systematic monitoring exists above many of these
sites. Owing to the high reliability and statistically-rigorous infilling by NRCS (Schaefer
and Paetzold 2000), all of the selected SNOTEL stations offer serially-complete daily
86
Table B-1. SNOTEL station locations, elevation, percent of record present, and start year.
SNOTEL Site
GARVER
CREEK
GRAVE CREEK
EMERY CREEK
KRAFT CREEK
MANY
GLACIER
BISSON CREEK
HAND CREEK
POORMAN
CREEK
COPPER
BOTTOM
BANFIELD
MTN
WALDRON
DUPUYER
CREEK
PIKE CREEK
WOOD CREEK
STAHL PEAK
NOISY BASIN
SLEEPING
WOMAN
FLATTOP MTN.
NORTH FORK
JOCKO
MOUNT
LOCKHART
HAWKINS
LAKE
MOSS PEAK
BADGER PASS
COPPER CAMP
NEVADA
RIDGE
Mean
Median
Standard
deviation
Max:
Min:
Elevation
(m)
Snowpillow
Start Year
(WY)
SWE
Record
Present
(%)
Temp.
Start
Year
(WY)
Temp.
Record
Present
(%)
Latitude
Longitude
1295
1969
100
1997
100
48°58' N
115°49' W
1311
1326
1448
1976
1977
1981
100
100
100
1991
1990
1991
100
100
100
48°54' N
48°26' N
47°25' N
114°46' W
113°56' W
113°46' W
1494
1977
100
1987
100
48°47' N
113°40' W
1500
1535
1990
1977
100
100
1993
1983
100
100
47°41' N
48°18' N
113°59' W
114°50' W
1554
1969
100
1999
100
48°07' N
115°37' W
1585
1976
100
2004
100
47°03' N
112°35' W
1707
1969
100
1989
100
48°34' N
115°26' W
1707
1969
100
1990
100
47°55' N
112°47' W
1753
1984
100
1985
100
48°04' N
112°45' W
1807
1817
1838
1841
1977
1979
1976
1975
100
100
100
100
1989
1990
1990
1990
100
100
100
100
48°18' N
47°26' N
48°54' N
48°09' N
113°19' W
112°48' W
114°51' W
113°56' W
1875
1990
100
1993
99.8
47°10' N
114°20' W
1920
1970
100
1983
99.3
48°48' N
113°51' W
1929
1990
100
1990
100
47°16' N
113°45' W
1951
1969
100
1989
100
47°55' N
112°49' W
1966
1969
100
1989
100
48°58' N
115°57' W
2067
2103
2118
1986
1979
1976
100
100
100
1990
1989
2004
100
100
99.9
47°41' N
48°08' N
47°05' N
113°57' W
113°01' W
112°43' W
2140
1994
100
1995
100
46°50' N
112°30' W
1743
1807
259
2140
1295
SWE records. Daily SNOTEL temperature records began in 1983 with an
average completeness of >99%. Unlike records of SWE, SNOTEL temperature data have
87
not undergone infilling of missing data, but in all cases records contained > 99% of the
daily observations.
Daily SNOTEL SWE data were used in analyses of snowpack dynamics
throughout the water year (henceforth “WY”, which spans October 1- September 30).
For all daily SWE records the following metrics were calculated: 1) day of peak SWE; 2)
peak SWE amount; 3) number of snow accumulation and ablation days; 4) snow-free
date; 5) number of snow-free days; and 6) length of ablation period (# days between peak
and zero SWE). Snow was said to accumulate (ablate) when gains (losses) exceeded 0.5
cm for the day, and a site was considered snow-free once measurable SWE fell below
instrumentation detection limits. Daily SNOTEL temperature records were compiled into
seasonal and annual time series of minimum (Tmin) and maximum (Tmax) temperatures.
In addition, numbers of days at or below 0ºC (frost days) were tallied for winter (JFM),
spring (AMJ), summer (JAS), fall (OND) and the WY. These seasonal breakdowns were
chosen to capture the relatively late arrival of spring and summer in the NRMs.
Individual station temperature and SWE records were then checked for matching
variance, standardized (based on unit variance) against the regional mean, and combined
to form a regional-composite average (see Pederson et al. 2010).
To provide additional, long-term perspective on the SNOTEL peak SWE
observations we obtained 148 manually-measured snow course records from the NRCS
and the British Columbia and Alberta Ministry of the Environment
88
(http://a100.gov.bc.ca/pub/mss/; http://www.environment.alberta.ca/) sites (Figure B-1).
These records consist of April 1 SWE observations, and cover both major watersheds
within our study area. This dataset incorporates all records with " 30 years of April 1
SWE measurements with some records spanning the entire recording period from 1936 to
2007. Records within each watershed were normalized (converted to z-scores) and
subsequently averaged to produce a time series of mean April 1 SWE anomalies for both
watersheds. These regional April 1 SWE time series were then used as independent
verification of observed patterns in the selected SNOTEL records, as well as in analyses
of relationships with stream discharge and synoptic-scale controls.
Stream discharge data were obtained from 14 gages in both the U.S. and Canadian
portions of the study area (Figure B-1, Table B-2). The selected stream gage records are
minimally influenced by human activities (Moore et al. 2007; Rood et al. 2005), and
begin on or before 1952 with " 90% of the daily observation. The following metrics
were then calculated from the daily streamflow time series: 1) amount and timing of peak
discharge; 2) total WY discharge; 3) day of 50% cumulative discharge (center-of-mass
timing [CT]); 4) day of 75% cumulative discharge; and 5) time elapsed between dates of
50% and 75% cumulative discharge. As in previous studies (e.g. McCabe and Clark
2005), we use the WY day at which 50% of the cumulative streamflow discharge has
passed the stream gage, or CT, as our primary metric for snowmelt driven streamflow
timing. In addition to the analysis of CT we also investigate changes in WY day of 75%
cumulative discharge for evidence of changes in time elapsed between snowmelt driven
89
peak flows as they decline towards summer base flows. Individual gage time-series for
each of these variables were subsequently combined into regional-composite averages
using the same techniques that were applied to the SNOTEL and snow course records.
Stream discharge data were obtained from the U.S. Geological Survey
(http://waterdata.usgs.gov/nwis) and Environment Canada’s Hydrometric Program
(http://www.wsc.ec.gc.ca/hydat/H2O/index_e.cfm?cname=main_e.cfm).
Table B-2. Stream gage name, location, percent of record present, and start year.
ID
12358500
12355000
12370000
08NH006
08NH032
08NA002
08NB005
08NF001
08NJ013
08NK016
08NH007
05AA022
05AA023
05AD003
Stream Gage
Station Name
MF Flathead
River
Flathead at BC
Border
Swan River
Moyie River
Boundary Creek
Columbia River
at Nicholson
Columbia River
at Donald
Kootenay River
Slocan River
Elk River
Lardeau River
Castle River
Oldman River
Waterton River
State/
Province
MT
Start Year
(WY)
1940
Latitude
48°17' N
Longitude
114°00 W
Record
Present (%)
100
MT
1946
49°00' N
114°28' W
92.7
MT
ID
ID
BC
1923
1916
1930
1917
48°01' N
49°00' N
49°00' N
51°14' N
113°05' W
116°10' W
116°34' W
116°54' W
98.7
100
98.7
98.9
BC
1946
51°28' N
117°10' W
100
BC
BC
BC
BC
AB
AB
AB
1941
1915
1950
1946
1946
1950
1910
50°53' N
49°27' N
49°52' N
50°15' N
49°30' N
49°48' N
49°06' N
116°02' W
117°33' W
114°45' W
117°04' W
114°08' W
114°10' W
113°50' W
100
100
98.2
90.8
100
100
100
Monthly temperature and precipitation data were obtained from 37 meteorological
(MET) stations within the study area (Figure B-1), and were used to assess long-term,
watershed-level changes in climate, and potential elevation-related differences in
90
temperature variability and trends. We first developed regional-average precipitation
series from 37 sites, using the same techniques applied to the snow course records. Next,
we incorporate a previously developed regional time series of Tmax and Tmin from a subset
of nine stations used in a previous analysis of changes in regional temperature means and
extremes (Pederson et al. 2010). We limited our regional time-series to this subset
because the selected daily and monthly MET records are high quality, and relatively
serially complete back to 1895 (" 97% of daily observations present). Additionally, the
trends and variability captured by this subset of nine stations match those from a network
of 15 stations extending further north into the Canadian Rockies (see Watson et al. 2008).
MET data were obtained from the U.S. Historic Climatology Network (USHCN; Menne
et al. 2009; http://cdiac.ornl.gov/epubs/ndp/ushcn/newushcn.html) and the Historic
Canadian Climate Database (HCCD; Mekis and Hogg 1999;
http://www.cccma.bc.ec.gc.ca/hccd/), since these stations generally provide the longest
and most complete temperature and precipitation records available. Monthly data from
these networks have also been corrected for time-of-observation biases, station moves,
instrument changes, and urban heat island effects (Karl et al. 1988; Mekis and Hogg
1999; Menne and Williams Jr 2005; Vincent et al. 2002). All instrumental climatic data
series, derived hydroclimatic variables (e.g. center-of-mass timing, Tmin days below 0°C),
and final figures will be served from the USGS Northern Rocky Mountain Science
Center climate data and analyses website (http://www.usgs.nrmsc.gov/climate/).
91
We tested the resulting snowpack, precipitation, temperature, and streamflow time
series for statistically significant correlations and linear trends using parametric Pearson
correlation coefficients and linear bivariate least-squares regression (including quadratic
terms) coupled with analysis of variance. If the parametric assumptions of constant
variance, normal distribution (Kolmogorov-Smirnov test), and lack of autocorrelation in
residuals (Durban-Watson test) were not met, non-parametric Mann-Kendall regression
and rank correlation tests were performed and reported. Parametric regression allows for
determination of the magnitude of any significant (p < 0.05) trends, whereas nonparametric regression provides a robust approach if assumptions of parametric regression
are not met, which can be a problem with hydrologic time series (Rood et al. 2005).
To assess the impact of large-scale oceanic and atmospheric variability on
regional changes in snowpack and streamflow we calculated correlations between
regional hydroclimatic variables and: 1) time series of key Pacific Basin climate indices
(listed below); and 2) 1000-250 mb geopotential height and wind fields (500mb results
shown). Prior research has shown a strong regional influence of the PDO, and to a lesser
degree ENSO, on April 1 SWE (Cayan 1996; Cayan et al. 1998; Mote 2003; Mote et al.
2005; Pederson et al. 2004; Selkowitz et al. 2002). We revisit these relationships using
monthly SST values and winter season averages of the PDO obtained from the Joint
Institute for the Study of Atmosphere and Ocean at the University of Washington
(http://tao.atmos.washington.edu/data_sets), and the NINO 3.4 index from the National
Oceanic and Atmospheric Administration’s (NOAA) Earth System Research Laboratory
92
(ESRL; http://www.cgd.ucar.edu/cas/catalog/climind/TNI_N34/index.html). Monthly
geopotential height and wind fields were obtained from the National Center for
Environmental Prediction (NCEP) Reanalysis project (Kalnay et al. 1996).
To further assess Pacific Basin influences on regional hydroclimatic variables we
constructed best-fit least-squares linear regression models using stepwise and bestsubsets regression procedures on significantly correlated (p ! 0.05) seasonal and monthly
Pacific Basin SST indices and regional MET observations. Relative to dynamical
prediction, this method provides a simple diagnostic tool in assessing the bulk impact of
major modes of climate system variability on regional hydroclimate. To ensure highly
consistent results, predictor variables were selected based on a suite of model diagnostics
that include the adjusted R2, validation R2, Mallow’s Cp, root-mean-squared error, and a
leave-one-out cross validation (PRESS) procedure (Draper and Smith 1998). Linear
models were constructed iteratively, meaning at each step the influence of the predictor
was subtracted from the predictand, leaving a residual time series on which to regress the
next most strongly correlated predictor. The construction of final models excluded the
ENSO index because ENSO and PDO covary on inter-annual timescales due to tropical
and extra-tropical interactions (An et al. 2007; Newman et al. 2003), consequently
violating linear modeling assumptions.
93
Results and Discussion
a. Changes in Snowpack (Snow Water Equivalent) and Streamflow
Figure B-2. Average peak snow water equivalent (SWE) calculated from 25 stations
(black line). Values are plotted as anomalies from regional SWE mean and individual
station records are plotted (gray lines). A regression line (black) shows trends in SWE,
and is bounded by 95% confidence intervals (black dashed lines) with significance (pvalue) shown.
Examination of the 25 individual SNOTEL records indicates marked temporal
coherency in snowpack variability across the study area (Figure B-2). As represented by
amount of peak SWE, stations in both the northern and southern watersheds and
throughout the entire elevation range show consistently low snowpack in years such as
1977, 1995 and 2005, and high snowpack in 1972, 1974 and 1997. Likewise, patterns of
variability at individual stations tend to agree strongly with regional-average peak SWE.
94
Further analysis of regional peak SWE suggests a small, linear decline in
snowpack (p = 0.09) over the period 1969-2007 (Figure B-2). There has been ample
documentation in many parts of the western U.S. for a loss of snowpack since the mid20th century (e.g. Hamlet et al. 2007; Mote et al. 2005; Pierce et al. 2008), and our results
indicate some agreement with these previous studies. However, any trend in our study
area is embedded within pronounced inter-annual variability. Visual inspection of Figure
B-2 also suggests a possible step-change from predominantly high to low snowpack
conditions around 1976, followed by several multi-year periods with lower than average
snowpack through the mid-1990’s. Overall, a complex mix of large inter-annual
fluctuations combined with potential decadal-scale variability and a slight downward
trend characterizes peak SWE variability in the NRMs.
Records of snow course April 1 SWE compiled for both the northern and southern
NRM watersheds show the composite time-series are good metrics of SNOTEL peak
SWE, and that since 1936 a major feature of the records is the strong variability on
interannual- to decadal-scales that coincide with changes in streamflow discharge (Figure
B-3). The temporally consistent signal of regional snowpack variability is underscored
by the strong and significant (p ! 0.001) relationships between SNOTEL peak SWE and
snow course April 1 SWE, with correlations of r = 0.946 and r = 0.931 for the northern
and southern sub-watersheds, respectively. Results are supportive of previous research
showing that throughout the Rocky Mountains measurement related errors between
record types is typically constrained to twelve percent or less (Bohr and Aguado 2001).
95
Figure B-3. (Top) Comparison between SNOTEL peak SWE and snowcourse April
1 SWE records. (Bottom) Relationship between winter season PDO (Oct-May),
total WY stream discharge, peak WY stream discharge, and peak SWE. Note the
PDO was inverted (multiplied by -1) for ease of comparison, and all correlations are
significant (p ! 0.05).
Because snowpack is the primary driver of runoff in the NRMs, regional-average SWE
should also correlate strongly with streamflow for the region. The correlation between
April 1 SWE and peak and total annual discharge (Figure B-3, bottom) is in fact strong (r
= 0.687, and r = 0.756 respectively) and highly significant (p ! 0.001). We therefore
further use these records to examine the longer-term dynamics and climatic controls on
the snowmelt driven hydrology in the NRMs.
96
Returning to the short, but high-resolution SNOTEL records we examine
snowpack dynamics throughout the WY, and Figure B-4 shows significant changes in the
timing of peak and zero SWE. As in previous studies showing the progressively earlier
melt-out of snow at mid-elevations (e.g., Hamlet et al. 2007; Pierce et al. 2008), records
for the NRMs show that since 1969 SWE peaks on average nine days earlier in the WY
(p = 0.04) with an 8-day progression towards earlier melt-off (p = 0.09) as shown by the
WY day of zero SWE. Accordingly, the change in elapsed time between peak and zero
Figure B-4. Regional average timing of peak SWE (black dashed line) and zero SWE
(black solid line). Melt-period duration (gray bars) shows the number of days elapsed
between day of peak and zero SWE. All time series are plotted as anomalies from
regional mean, individual station records are plotted as dark gray lines, and significance
of trend (p-value) is shown though the regression line is not plotted. Strong inter-annual
and within-year station-to station variability exists for day of peak SWE, which is likely
a function of topography and other local site characteristics influencing snow
accumulation (e.g. wind) combined with the inherent spatial heterogeneity of snowfall.
Inter-annual variability in the day of zero SWE is also high, but station-to-station
variability is small since snowmelt largely temperature driven.
97
SWE is non-significant. The regional trend equates to a shift from peak SWE generally
occurring between April 15th and 20th in the early portion of the record (1969-1984) to
peak SWE occurring on or before April 1 in recent years.
SNOTEL records also suggest high inter-annual and intra-site variability in the
number of winter storm events and snow ablation days, with both metrics showing nonsignificant long-term linear trends (Figure B-5 top and middle). The number of SWE
accumulation days corresponds positively with regional peak SWE anomalies, and
therefore may serve as a proxy for the frequency of storm events. This suggests that over
the 1969-2007 period storm track position is a major inter-annual- to decadal-scale
control on the total amount of peak SWE received in the NRMs, but fails to account for
the long-term, negative linear trend in peak SWE (Figure B-2). The most likely
explanation of drivers behind the long-term negative trend in peak SWE is increasing
winter (DJF) and spring (MAM) temperatures (Figure B-6). Controls on ablation are
generally more complex than accumulation due to topography; with variability in the
number of ablation days through time tied to both sublimation and temperature induced
melting.
Lastly, SNOTEL records exhibit significant changes in snow-cover duration, as
shown by the increasing number of snow-free days per year over time. Figure B-5
(bottom) shows that the number of snow-free days has increased significantly (p = 0.032)
with the NRMs showing an average loss of 14 days of snowcover between 1969 and
98
Figure B-5. Average annual number of SWE accumulation days (top), ablation days
(middle), and snow-free days (bottom) recorded at SNOTEL stations. All time series
are plotted as anomalies from regional mean, individual station records are plotted as
dark gray lines, and trends are shown with a black regression line with significance
of trend (p-value) shown
2007. With the majority of the increase in number snow-free days apparent in the spring
(8-day increase), it follows that spring and early-summer temperatures should be strongly
linked to the observed trend. However, total winter snowfall, total peak SWE, and the
timing of major storm events could also play an important role in dictating the annual
number of snow-free days (see discussion below).
99
b. Impacts of Regional Temperature and Precipitation Change on Snowpack and
Streamflow
1) Mid to High-Elevation Temperature Change
Mid- to high-elevation SNOTEL stations within the NRMs show a coherent and
significant increase (p ! 0.05) in average winter, spring, and annual Tmin since the early
1980’s (Figure B-6). MET stations representing valley sites also show significant
increases in Tmin over the 20th century, though the recent trend magnitude is smaller than
at SNOTEL sites (Figure B-S1). Estimates on the magnitude of trend for the regionalcomposite time series of SNOTEL Tmin shown in Figure B-6 are based on quadratic
regression since results provide slightly more conservative estimate than obtained from
simple linear regression models. Results show, that since 1983 average Tmin at SNOTEL
stations have increased over the winter (DJF) by +3.8 ± 1.72ºC, spring (MAM) by +2.5 ±
1.23ºC, and annually by +3.5 ± 0.73ºC (Figure B-5). Note the short, highly variable
nature of these time series prevents us from accurately estimating the underlying trend.
The temperature records do not, however, span any shifts in known large-scale modes of
decadal variability, which could further amplify the trend. Hence, the greater magnitude
of mid-elevation warming likely reflects a response to both regional low-elevation
warming and elevation-dependent responses.
Minimum temperatures at SNOTEL stations display inter-annual variability
similar in magnitude to valley-MET stations with the exception of spring minimums.
Spring Tmin records generally exhibit approximately twice the variance of minimum
100
temperatures observed at valley-MET stations, and more rapid recent warming. Average
spring Tmin approaches the 0°C frost threshold during years of early snowmelt and severe
growing season drought (e.g. 1988, 1998, 2000, 2001, and 2003), with the warmest years
coming towards the end of the record. Mid-winter and average annual Tmin observations
from SNOTEL stations also show more rapid recent warming, with a pronounced
increase after 1998 that is consistent across individual station records (not shown).
Though recent warming has been rapid, winter Tmin remain $8°C below freezing. Similar
to observations at valley-MET stations, however, average annual Tmin appears to be
rapidly approaching 0°C.
Figure B-6. Average winter (Dec-Feb; top), spring (Mar-May; middle), and annual
(bottom) minimum temperatures from SNOTEL (water year Oct-Sep) and valley MET
(calendar year Jan-Dec) stations. SNOTEL station Tmin records have been fit using a
non-linear quadratic equation due to characteristics of these time series. All trends
shown are significant (p ! 0.05) and note the y-axis temperature scale changes for each
panel.
101
Changes in average seasonal lapse rates and surface albedo are two potential
drivers for the high variability and recent enhanced upward trend in observed SNOTEL
temperatures. While again recognizing the limitations of analyzing short time series, a
post-1983 comparison of trends from low-elevation MET station records against
SNOTEL observations suggests regional lapse rates may be changing (Figure B-6 and
Figure B-S1). Supportive of this finding, Diaz and Eisheid (2007) have shown
significant changes in lapse rates and more rapid warming of the higher elevations across
the western U.S. using SNOTEL and satellite based microwave sounding unit
observations. Additionally, the increase in average annual Tmin (Figure B-5) and, to some
degree winter and spring Tmin, does coincide with the major 1998 El Niño event (Enfield
2001) and a recent slowing of average global temperature rise (Thompson et al. 2008).
The high inter-annual variability and trend shown in the spring Tmin record
suggests feedback processes associated with reduced surface albedo might also be a
significant factor. Reduced surface albedo would result from the observed earlier snow
melt-out (Figure B-4) and the increasing number of snow-free days (Figure B-5, bottom).
As discussed below, the average WY day of zero SWE and number of snow-free days is
itself controlled by a number of factors ranging from total winter snowfall and variability
in the timing of end-of-season storm events due to spring atmospheric circulation
patterns; all of which undoubtedly adds to the inter-annual variability in SNOTEL
temperature records.
102
Concurrent with the increase in seasonal and annual SNOTEL Tmin, we show a
substantial decline in the number of days per year below the 0ºC threshold. Figure B-7
(left) shows a significant (p ! 0.001) reduction in regional average number of days with
Tmin ! 0ºC across winter (JFM), spring (AMJ), summer (JAS), and fall (OND) – with the
caveat that these are short records. The regional average for number days with Tmax !
0ºC also shows a significant drop over the winter and fall. Individual station records for
Figure B-7. Regional average (left) and individual station records (right; grey lines)
showing number of frost days (days ! 0ºC) for Tmax (gray lines) and Tmin (black lines)
recorded at all SNOTEL stations over the winter (JFM), spring (AMJ), summer (JAS)
and fall (OND). All time series are plotted as anomalies from regional mean, and all
trends are significant* (p ! 0.001) and shown with a black regression line bracketed by
95% confidence intervals (gray dashed lines). Reduction in number of freeze/thaw days
(calculated from the trend lines) from 1983 to 2007 is printed to the right of each
average time series.
*Except for the trends in spring and summer Tmax.
103
both Tmin and Tmax (Figure B-7, right) show strong coherency, thereby reducing the
potential for outlying years or single station records to be responsible for driving the
observed trends in loss of frost days. Over the 25-year period of record, the smallest
reductions in number of days with Tmax or Tmin ! 0ºC are seen over the winter, with this
season showing an average reduction of 7 and 4 days ! 0ºC respectively. Spring and
summer generally feature few daytime Tmax values below freezing, however, nighttime
Tmin shows a dramatic loss of days below 0°C. Since 1983, the number of spring Tmin
days ! 0°C have declined by an average of 21 days, now approaching only ~30 days
where nighttime Tmin falls below 0°C. Similarly, the summer has lost an average of 14
days where Tmin drops below 0°C, and is now rapidly approaching average conditions
whereby summer frost events rarely occur at the average elevation of the SNOTEL
stations (~1743m a.s.l.). Finally, over the fall season (OND) records show an average
reduction of 11 and 20 days with Tmin and Tmax below 0ºC, respectively.
Summed across the entire WY, we document an average loss of 53 days with Tmin
! 0ºC over this 25-yr period, and an average loss of 30 days with Tmax ! 0ºC (Figure B8). NRM valley MET stations also show a decline in number of days with Tmin and Tmax
! 0ºC (Pederson et al. 2010), but the recent rate and magnitude of change is noticeably
larger at the mid- to higher-elevations represented by these SNOTEL records. The
combined 35-day loss in Tmin frost events over the spring (AMJ) and summer (JAS)
months are perhaps the most hydrologically relevant temperature changes since relatively
warm nighttime temperatures likely reduce the cold-content and refreezing of snowpack.
104
Also, though not explicitly shown here, the documented declines in fall Tmax and Tmin
days ! 0ºC may also be increasingly limiting the early-season accumulation of snowpack.
Figure B-8. Same as Figure B-7, but for annual average number of frost days (days !
0ºC).
2) Temperature Impacts on Snowmelt and Snowcover Duration
Further exploring the role of temperature as both a driver and response to regional
snowpack dynamics, we compare 1) the regional-average number of snow-free days, and
2) the average day of peak and zero SWE, with temperature records from both SNOTEL
and valley MET stations (Table B-3, Figure B-9 top). Both spring and annual Tmax and
Tmin are significantly (p ! 0.05) correlated with the average number of snow-free days per
year (r = 0.434 to 0.757), day of peak SWE (r = -0.443 to -0.756), and day of zero SWE
(r = -0.354 to -0.820). Correlations are highest when comparing selected snowpackrelated metrics against spring temperatures recorded at SNOTEL stations relative to
records from lower-elevation MET stations. On the whole, these observations support
the potential combined impacts of recent, more rapid high-elevation warming (Diaz and
Eischeid 2007) and a positive feedback associated with mid- to higher-elevation albedo
changes. The influence of total annual SWE on snowcover duration and consequently
105
albedo is likewise captured by the strong negative correlations between the number of
snow-free days with SNOTEL Peak SWE (r = -0.689, p ! 0.001) and snow course April 1
SWE (r = -0.617 to -0.689, p ! 0.001) shown in Figure B-9 (bottom).
Figure B-9. Correlation between number of snow-free days and average spring Tmin
(black triangles) and Tmax (gray triangles) recorded at SNOTEL stations (top left) and
valley MET stations (top right). Also plotted are, correlations between number of
snow-free days and SNOTEL peak SWE (bottom left) and snowcourse April 1 SWE
(bottom right; gray squares = northern HUC, black diamonds = southern HUC).
Regression lines are plotted with corresponding 95% confidence intervals, and
correlations are significant at p ! 0.001 level.
Overall, results are supportive of previous studies (e.g. Das et al. 2009; Pierce et
al. 2008) showing that warmer temperatures (especially minimums) are an important
driver behind the increasing number of snow-free days and earlier WY day of peak and
zero SWE. Additionally, snowpack and temperature relationships at mid- to higherelevations may be complicated by the influence of low-elevation warming trends (see
106
Pederson et al. 2010) on valley snow cover (see Knowles et al. 2006). Low-elevation
warming, subsequent loss of snow cover, and accompanying changes in albedo may, in
turn, drive a positive feedback process more rapidly melting snow in the mid-elevations
and further contributing to temperature increases at the elevation of SNOTEL stations.
Rapid warming shown at SNOTEL stations would then relate, in part, to loss of mid- to
higher-elevation snow, with resulting changes in albedo reflected in station-level
temperature records.
Table B-3. Correlations between the average annual number of snow-free days, the WY
day of peak and zero SWE, and 50% CT and spring and annual Tmax and Tmin recorded at
valley (1969-2007, n = 39) and SNOTEL stations (1983-2007, n = 25).
Variable
SNOTEL Spring Tmin
SNOTEL Spring Tmax
SNOTEL Annual Tmin
SNOTEL Annual Tmax
Spring MET Tmin
Spring MET Tmax
Annual MET Tmin
Annual MET Tmax
# Snow-free
Days (p-value)
0.718
0.001
0.586
0.003
0.479
0.024
0.757
0.001
0.491
0.002
0.509
0.001
0.434
0.007
0.505
0.001
WY day
peak SWE
(p-value)
-0.750
0.001
-0.709
0.001
-0.443
0.039
-0.595
0.003
-0.615
0.001
-0.756
0.001
-0.443
0.039
-0.595
0.003
WY day
zero SWE
(p-value)
-0.815
0.001
-0.820
0.001
-0.375
0.085
-0.669
0.001
-0.639
0.001
-0.766
0.001
-0.354
0.031
-0.486
0.002
50% CT
(p-value)
-0.567
0.005
-0.681
0.001
-0.177
0.431
-0.640
0.001
-0.596
0.001
-0.710
0.001
-0.184
0.275
-0.423
0.009
107
3) Changes in Surface Climate and Streamflow Timing
Changes in spring temperatures are linked to the regional average of snow-free
days and the day of peak and zero SWE in the NRMs, with spring temperatures serving
as both a driver and response to these events. In turn, changes in the melt-out of snow
should be reflected in corresponding changes in timing of streamflow. Over the period of
record, streamflow timing is shown to be highly variable with CT arriving on or slightly
after June 1, and 75% of the hydrologic mass passing stream gages in late June to early
July (Figure B-10, top panel). Between 1925 and 1976, however, CT and 75%
cumulative discharge trended towards a later arrival date. After 1976, CT and 75%
cumulative discharge timing show a slight late-1970s and larger mid-1980s tendency
towards earlier runoff. The mid-1980s shift to earlier snowmelt runoff is supported by
previous studies showing similar shifts in timing of snowmelt runoff across the western
U.S. (e.g. McCabe and Clark 2005) that coincide with increasing spring temperatures
associated with increasing atmospheric heights, and thus an earlier onset of spring (Ault
et al. in prep.; Cayan et al. 2001; Schwartz and Reiter 2000).
NRM regional average CT correlates strongly with the average day of zero SWE
(r = 0.804, p ! 0.001), number of snow-free days (r = -0.518, p ! 0.001), and
consequently surface observations of spring Tmin and Tmax from SNOTEL and valley
MET stations (Figure B-10, Table B-3). On the whole, these observations show the
strong influence of increasing temperature in driving the earlier melt-out of regional
snowpack and earlier streamflow CT. Note also that the WY day of zero SWE and
108
streamflow CT are almost exactly the same (Figure B-10, top panel). This suggests that
SNOTEL sites in the NRMs capture the snow-driven portion of regional flows, and that
SNOTEL records are a good indicator of potential runoff contributions from these basins.
Figure B-10. (Top panel) Timing of 50% (CT) and 75% cumulative discharge, day of
zero SWE, and SNOTEL average spring (MAM) Tmin . (Bottom panel) Associated
trends, variability, and step-changes in elapsed number of days between 50% and 75%
stream discharge, and average spring precipitation, Tmin, and Tmax for valley MET
stations. Correlations between 50% CT, spring Tmin, Tmax and WY day of zero SWE
are significant (p ! 0.001).
109
Long-term changes in spring temperature and precipitation are consistent with a
late-1970s and mid-1980s progression towards earlier runoff relative to the 1920s
through mid-1970s (Figure B-10). The interaction between spring temperature and
precipitation is underscored by changes in the number of days between the 50% (i.e., CT)
and 75% discharge points, with the latter being an indicator of the transition to summer
baseflows. Specifically, after 1976/77 the mean of this time series shifts by 5 days, and
the variance almost doubles (Figure B-10, bottom). Whereas before 1976/77 (and after
1940), relatively high spring precipitation and lower spring Tmin and Tmax corresponded
with later CT and fewer days between 50% and 75% streamflow discharge. A notable
single year exception occurred in 1934 during the “Dust-Bowl” drought when spring
temperature anomalies of +4°C and low spring precipitation resulted in the earliest CT
event on record. The late-1970s and mid-1980s transition towards earlier CT, and change
in mean and variance of days between 50% and 75% stream discharge, corresponds with
an increase in mean and variance of total spring precipitation and a period of rapid and
sustained spring warming. Overall, results suggest that 1) regional spring temperatures
are the dominant driver of streamflow timing, and 2) increased variance in spring
precipitation may be contributing significantly to high variance in discharge timing since
the late-1970s.
110
c. Synoptic Controls on Snowpack and Streamflow Yield and Timing
1) Snowpack and Streamflow Yield
A large body of prior research (e.g. see Das et al. 2009 and references therein) has
demonstrated that throughout the western U.S. (e.g. Cayan 1996; Cayan et al. 1998) and
within the NRM region specifically (Moore et al. 2007; Mote 2006; Pederson et al. 2004;
Rood et al. 2005), variations in SWE and streamflow are affected by inter-annual and
inter-decadal changes in the PDO, and the interaction of the PDO with the El Niño
Southern Oscillation (ENSO; An et al. 2007; Newman et al. 2003). Using the records
compiled here, both cool season (Oct - May) or winter (Nov – Mar) PDO correlates
significantly (p ! 0.001) with SNOTEL peak SWE at r = -0.638, snow course April 1
SWE at r = -0.693, total WY discharge at r = -0.489, and peak discharge at r = -0.429
(Figure B-3). All records show potential shifts in mean occurring in 1945/46 and
1976/77, which is consistent with well-documented, inter-decadal shifts in sea surface
temperature and pressure in the North Pacific Ocean (Figure B-3, bottom). Correlations
between peak SWE, streamflow and winter ENSO are also significant (not shown),
though the respective correlation values were roughly half as strong.
Changes in North Pacific SSTs ultimately affect SWE and runoff by altering
major modes of atmospheric circulation and consequently the strength and position of the
winter westerlies (i.e. preferential storm-track position) across North America (Dettinger
et al. 1994). Correlation fields of 500mb atmospheric heights and winds over the core
winter months (JFM) show that for NRM snowpack anomalies there is a broad region of
111
strong positive correlations in the north Pacific centered over the Aleutian Islands, and
strong negative correlations poleward of 45°N (the mean wintertime position of the jet
stream) over North America (Figure B-11 a,b). This indicates high (low) snowpack
anomalies in the NRMs tend to be associated with winter-season persistence of high
(low) pressure anomalies over the north Pacific basin and low (high) pressure anomalies
over most of western North America. These winter (JFM) atmospheric circulation
patterns reflect the position, persistence, and depth of the Aleutian Low, and
consequently the preference (or not) for mid-latitude cyclones to track from the Gulf of
Alaska southeast through the Pacific Northwest and ultimately across the NRMs (Cayan
et al. 1998). Accordingly, patterns shown here are consistent with the so-called Pacific
North American pattern (PNA; Thompson and Wallace 2000), a leading mode of
Northern Hemisphere atmospheric variability that is, in part, linked to inter-annual and
decadal changes in PDO and ENSO (Quadrelli and Wallace 2004).
Strong negative correlations between snowpack and the tropical Pacific are seen
throughout the winter months (Figure B-11 a,b), which is consistent with an ENSO
forcing component. These relationships persist through March and April, even as
correlations with the North Pacific begin to weaken. The April atmospheric circulation
patterns, however, are strongest over western North America and weakened over the
tropical Pacific, suggesting regional spring weather plays an important role in dictating
total amount of WY peak SWE. Additional 500mb composites (Figure B-S2 a,b; B-S3
a,b) of atmospheric circulation patterns during years with snowpack anomalies ±1
112
Figure B-11. Field correlations of 500mb monthly atmospheric pressure and wind anomalies for NRM hydroclimatic
variables. Maps are arranged in the following order: A) SNOTEL peak SWE (n = 39), B) snow course April 1 SWE (n = 58),
C) day of Zero SWE (n = 39), D) day of 50% CT (n = 58), and E) number of annual snow-free days (n = 39). Plotted
atmospheric circulation anomalies are significant (p ! 0.05).
113
standard deviation above or below the long-term mean are supportive of the above
results: winter (JFM) ocean-atmosphere teleconnections control NRM snowpack while
spring (MA) regional atmospheric circulation anomalies (with links to the tropical
pacific) modulate snowmelt and consequently streamflow. Overall, these results
highlight the importance of different seasonal mechanisms driving the amount of coolseason precipitation the NRMs receive, with winter-season oceanic and atmospheric
teleconnections controlling the mean storm track position and inter-annual to interdecadal variations in SWE and streamflow.
2) Snowmelt and Streamflow Timing
Atmospheric circulation patterns associated with the melt-out of snowpack and
timing of streamflow (Figure B-11 c-e) suggest that CT, day of peak and zero SWE
(similar to Figure B-11c), and number of snow-free days are most strongly influenced by
spring (MAM) geopotential heights over western North America. The influence of
associated circulation patterns becomes evident in March, is strongest in April, and then
weakens by May (Figure B-11 c-e). Similar 700mb atmospheric circulation patterns
across all spring months (AMJJ) have also been associated with the mid-1980s shift
towards earlier snowmelt and runoff across the western U.S (McCabe and Clark 2005).
The April 500mb correlation fields and composite maps (Figure B-S2 c-e; B-S3 c-e) also
show atmospheric circulation anomalies of opposite sign located off the coast of western
North America near 45°N. Overall, results imply early (late) arrival of CT, or day of
zero SWE, is associated with high (low) pressure centered over western North America
114
during the spring (Figure B-S1 c-d and B-S2 c-d). The same is true for the average
number of snow-free days in the NRMs, with figure B-10e showing that springs
characterized by high (low) pressure result in higher (lower) temperatures and more
(fewer) snow-free days (Figure B-S2e and B-S3e).
The spring correlation fields and composites also show how the timing of NRM
snowmelt and the timing and quantity of streamflow are linked to the tropical (i.e. ENSO)
and extratropical (i.e. !45°N) Pacific (Figure B-11 a-e). These correlations are strongest
in months preceding snowmelt and key runoff events, and this lag suggests a modest
potential for longer-lead (i.e., seasonal) forecasting of related hydrologic processes across
the NRMs. However, substantial variation in this relationship arises due to the influence
of atmospheric circulation patterns associated with the end of winter and the transition to
spring-like circulation patterns centered over western North America. These respective
circulation patterns relate directly to the potential for Arctic air intrusions and moisture
delivery from different source areas (e.g., mid-latitude Pacific or Gulf of Mexico), and
thereby add substantially to variability observed in the timing and amount of snowmelt
and streamflow.
d. Exploring the Relative Impact of Key Controls on Regional Snowpack and Streamflow
We have demonstrated that the amount and timing of snowpack and streamflow in
the NRMs is linked to 1) wintertime conditions in the Pacific basin via controls on the
strength and position of the winter westerlies and, 2) springtime (MAM) changes in
115
atmospheric circulation and their influence on regional temperature and precipitation
anomalies. We now explore the relative influence of these factors using regression-based
techniques. Though there is covariability between these drivers, the observed
hydroclimatic impacts arise from distinct, seasonally dependent processes (Figure B-11),
thus allowing for this type of simple modeling approach.
Figure B-12. Final regression models (see run #3 in Table 4) of total annual
discharge, peak discharge and SWE anomalies. Individual stream gage records are
plotted (gray lines). Correlations between observed and modeled time series are
significant (p " 0.001).
Results show that peak SWE, peak discharge, and total annual discharge can all
be adequately modeled using the following three predictors: 1) winter PDO, 2) spring
temperature (specifically Tmax), and 3) spring precipitation (Figure B-12, Table B-4). We
116
have excluded ENSO from these analyses due to interactions with PDO and its influence
on regional spring temperatures and precipitation. Thus, an ENSO forcing component is
implicit in the results shown here. In all models, winter PDO (Oct-May or Nov-Mar)
entered as the primary predictor variable, explaining 40.1% of the variability in peak
SWE, 18.4% in peak streamflow, and 23.9% of total annual streamflow (Table B-4). The
reduced explanatory power of the PDO in capturing variability in peak flows likely
reflects the greater importance of spring weather events, such as rapid warming and
precipitation variability (including rain-on-snow events), in driving hydrograph
anomalies. The next best-fit variable for snowpack and streamflow was spring (MAM)
Table B-4. Model diagnostics for best-fit least-squares linear regression models
constructed for peak discharge, total annual discharge, and peak SWE anomalies.
Run
Model
R2
R2adj
R2validation
p-value !
1
Peak SWE = 62.6 - 10.68 (Oct-May PDO)
40.6
39.0
34.7
0.001
2
Peak SWE = 106 - 8.50 (Oct-May PDO) 3.32 (Spring Tmax)
51.7
48.9
43.8
0.001
3
Peak SWE = 96.5 - 10.0 (Oct-May PDO) 2.67 (Spring Tmax) + 5.28 (MAM Pcp)
55.9
51.9
45.8
0.001
1
Peak Discharge = - 0.0475 - (0.313) OctMay PDO
Peak Discharge = 1.90 - 0.219 (Oct-May
PDO) - 0.154 (Spring Tmax)
18.4
17.4
14.7
0.001
31.2
29.4
26.4
0.001
Peak Discharge = 1.49 - 0.267 (Oct-May
PDO) - 0.122 (Spring Tmax) + 0.313
(MAM pcp)
Total Discharge = - 0.0766 - 0.453 (NovMar PDO)
Total Discharge = 1.81 - 0.381 (Nov-Mar
PDO) - 0.149 (Spring Tmax)
39.1
36.7
33.6
0.001
23.9
23.0
20.2
0.001
30.8
29.0
25.0
0.001
44.6
42.4
38.5
0.001
2
3
1
2
3
Total Discharge = 1.15 - 0.446 (Nov-Mar
PDO) - 0.0962 (Spring Tmax) + 0.555
(MAM Pcp)
117
Tmax (or Tmin, either substitutes), which explained an additional 6.9% to 12.8% of
the observed snowpack and streamflow variability (Table B-4). Models for peak SWE
and peak stream discharge show a proportionately greater amount of variance explained
by spring Tmax as compared to models for total annual flows. This may be due to the
sensitivity of snowpack and peak streamflow to the timing and magnitude of spring
warming, whereas total annual flows integrate the combined influence of recent increases
in spring precipitation and temperature.
The addition of spring (MAM) precipitation noticeably improves models of peak
and annual streamflow, while only marginally contributing to models of peak SWE
(Table B-4). The relatively small role of spring precipitation in controlling peak SWE
may be due in part to the cumulative negative impacts of increasing temperatures on
snowpack. Additionally, the inverse trends in NRM snowpack (Figure B-3 top) and
spring precipitation (Figure B-10 bottom) may be capturing a shift in the regional rain to
snow ratio, as previously shown for much of the western U.S. (Knowles et al. 2006).
Likewise, the relatively large proportion of variance explained by spring precipitation in
the models of peak and total WY discharge may indicate the increasing influence of
precipitation outside the winter months on hydrologic yield and variance in the NRMs.
For the post-1977 period in particular, increases in the amount and variance of spring
precipitation may be buffering against changes in the amount and timing of hydrologic
flows during years of low snowpack (Figure B-11, Figure B-12). Conversely, low spring
precipitation would exacerbate these impacts. The implication is that as snowpack melts
118
out of the mountains progressively earlier, summer season streamflow amount and timing
has become increasingly dependent on spring precipitation to make up for snow-related
deficits.
Summary
Snowpack in the NRMs serves as a primary source of runoff and hydrologic
storage for North American river systems, as well as a key driver of regional biophysical
processes. In this work we first sought to characterize local to regional variations and
trends in temperature, precipitation, snowpack, and streamflow using observations from a
comprehensive hydroclimatic-monitoring network. We then explored the linkages
among these variables, and examined the role of ocean-atmosphere teleconnections,
along with other potential large-scale drivers of snowmelt hydrology. Results related to
the characterization of the relevant local to regional hydroclimate provide important
baseline information on historical variability and change. When combined with our
detailed examination of linkages among hydroclimatic variables and potential largerscale forcings, this study may in turn, inform water resource and ecosystem management,
and help improve regional forecasting and prediction in the face of climate change.
Records from SNOTEL stations show that from 1969 to 2007 the mid-elevations
of the NRMs exhibited a tendency towards decreased regional snowpack, with peak SWE
and melt-out arriving an average of 8-days earlier accompanied by an average of 14
fewer days of seasonal snow-cover (Figure B-2 and Figure B-4). Though winter (JFM)
119
atmospheric circulation patterns and numbers of winter storm events are tightly linked to
peak SWE (Figure B-5 and Figure B-11a,b), a lack of significant long-term trend in these
two drivers suggests that other factors must account for snowpack decline. Mid-elevation
temperature records extending back to 1983 show significant seasonal and annual
decreases in numbers of frost days (days " 0°C), and highly variable but increasing
minimum temperatures (Figure B-6, B-7, and B-8). Consistent with expectations, the
observed eight-day earlier melt-out of snow and 14 additional snow-free days since 1969
are found to covary strongly with streamflow timing (CT) and spring temperatures
(Figure B-10).
Strong inter-annual to inter-decadal variations are shown for the duration of
record for NRM April 1 snowpack and peak and total annual streamflow records (Figure
B-3). A substantial proportion (18-41%) of this variability in snowpack and streamflow
is associated with winter (JFM) Pacific Basin SST anomalies (i.e. PDO and ENSO) due
to the influence on atmospheric circulation patterns controlling the preferential
positioning of winter storm tracks (Figure B-11 a,b; and Figure B-13 a). Winters with
high (low) snowpack in the NRMs tend to be associated with – (+) PDO conditions, a
weakened (strengthened) Aleutian Low, and low (high) pressure centered poleward of
45°N across western North America (Figure B-13 a-b). During years of high snowpack,
for example, the tendency is for mid-latitude cyclones to track from the Gulf of Alaska
southeast through the Pacific Northwest and into the NRMs. The relatively persistent
low-pressure anomaly centered over western North America is also conducive to more
120
frequent Arctic-air outbreaks resulting in colder winter temperatures. Conditions in the
tropical Pacific are also an important driver of snowpack and streamflow at inter-annual
time-scales, and the influence of related tropical Pacific atmospheric circulation
anomalies persists well into spring (Figure B-11 a-b).
Figure B-13. Idealized relationship between NRM snowpack and streamflow
anomalies with associated Pacific SSTs, atmospheric circulation, and surface
feedbacks.
121
Changes in spring (MAM) temperatures and precipitation are associated with
changes in regional atmospheric circulation, and are shown to strongly influence the
timing of NRM streamflow (Figure B-13 b-d). Specifically, high (low) pressure
anomalies centered over western North America correspond with increased (decreased)
spring temperatures and consequently the increasing (decreasing) number of snow-free
days, early (late) arrival of snow melt-out, and streamflow CT. Thus, spring atmospheric
circulation changes can, in turn, initiate surface feedbacks that contribute to surface
temperature and hydrograph anomalies (Figure B-13 c-d). The timing of CT for streams
within the NRMs exhibits high variability, with a tendency towards earlier CT dates
beginning in the late-1970s and mid-1980s (Figure B-10). Increasing spring precipitation
after 1977 appears to substantially buffer streamflow timing from substantially early
arrival during what was otherwise a period of low snowpack with above-average
temperatures. The more substantial mid-1980s shift towards earlier CT, is supportive of
findings by McCabe and Clark (2005), and appears to be driven by spring warming
associated with greater March through May geopotential heights over western North
America (Figure B-11d and B-13a-b). Cayan et al. (2001) and Ault et al. (in prep) also
show that similar March and April atmospheric circulation patterns control multiple
phenological and ecological aspects of spring onset across western North America. This,
in turn, suggests phenological and ecological changes in the NRMs are likely driven by
the same spring atmospheric circulation changes and consequently associated with
changes in streamflow timing and the duration of snow-cover.
122
Overall, our results agree with past research showing the influence of warming
temperatures on earlier snowmelt and runoff, along with decreasing snowpack and
streamflow (e.g. Barnett et al. 2008; McCabe and Clark 2005). However, as summarized
in Figure 13 we also show the influence of different seasonally dependent oceanatmosphere teleconnections and atmospheric circulation patterns in driving snowpack and
streamflow dynamics. Likewise this research also points to potential surface-albedo
feedbacks that interact with broad-scale controls on snowpack and runoff (Figure B-13 cd); however, more work here is required to disentangle the magnitude of potential
influence such surface feedbacks may have. Under future warming scenarios, snowpack
throughout the NRMs and other mid-latitude regions is expected to decline, with melt-out
and runoff coming progressively earlier (Adam et al. 2009; Hamlet et al. 2001). Our
results imply that if spring precipitation continues to increase in the NRMs, this could
offset snow-related declines in streamflow amounts and changes in the timing of
snowmelt and runoff to some degree. However, if spring precipitation was to remain at
current levels or decline for any reason (e.g. drought or shifting spring storm tracks), we
expect a significant amplification of trends toward earlier CT, reduced peak and annual
streamflow, and lower (but highly variable) summer base-flows. These expectations are
consistent with other regional and global projections of hydroclimatic change over the
21st century (Adam et al. 2009; Barnett et al. 2004; Hamlet et al. 2001; IPCC 2007), and
have important implications for aquatic and terrestrial ecosystems.
123
Acknowledgements
This research has benefited from the helpful comments of J. Weiss, and J.
Betancourt. NSF grant 0620793 and the USGS Western Mountain Initiative supported
this research. Any use of trade, product, or firm names is for descriptive purposes only
and does not imply endorsement by the U.S. Government.
Supplemental Figures
Figure B-S1. Post-1983 trends in individual SNOTEL and valley-MET station Tmin
records.
124
Figure B-S2. Composite 500mb monthly atmospheric pressure and wind anomaly maps for NRM hydroclimatic variables ≥ 1
standard deviation from the long-term mean. Maps are arranged in the following order: A) SNOTEL peak SWE, B) snow
course April 1 SWE, C) day of Zero SWE, D) day of 50% CT, and E) number of annual snow-free days.
125
Figure B-S3. Same as figure B-S2, but for composite maps associated with NRM hydroclimatic variables ≤ 1 standard
deviation from the long-term mean.
126
References
Adam, J. C., A. F. Hamlet, and D. P. Lettenmaier, 2009: Implications of global climate
change for snowmelt hydrology in the 21st century. Hydro. Proc., 23, 962-972.
Aguado, E., D. Cayan, L. Riddle, and M. Roos, 1992: Climatic Fluctuations and the
Timing of West Coast Streamflow. J. Climate, 5, 1468-1483.
An, S.-I., J.-S. Kug, A. Timmermann, I.-S. Kang, and O. Timm, 2007: The Influence of
ENSO on the Generation of Decadal Variability in the North Pacific. J. Climate, 20,
667-680.
Ault, T. R., A. K. Macalady, G. T. Pederson, J. L. Betancourt, and M. D. Schwartz, in
prep: Northern Hemisphere Modes of Variability and the Timing of Spring Across
Western North America J. Climate, in prep.
Barnett, T., R. Malone, W. Pennell, D. Stammer, B. Semtner, and W. Washington, 2004:
The Effects of Climate Change on Water Resources in the West: Introduction and
Overview. Climatic Change, 62, 1-11.
Barnett, T. P., and Coauthors, 2008: Human-induced changes in the hydrology of the
western United States. Science, 319, 1080-1083.
Bohr, G. S. and E. Aguado, 2001: Use of April 1 SWE measurements as estimates of
peak seasonal snowpack and total cold-season precipitation. Water Resour. Res.,
37, 51-60.
Cayan, D. R., 1996: Interannual climate variability and snowpack in the western United
States. J. Climate, 9, 928-948.
Cayan, D. R., M. D. Dettinger, H. F. Diaz, and N. E. Graham, 1998: Decadal Variability
of Precipitation over Western North America. J. Climate, 11, 3148-3166.
Cayan, D. R., S. A. Kammerdiener, M. D. Dettinger, J. M. Caprio, and D. H. Peterson,
2001: Changes in the onset of spring in the western United States. Bull. Amer.
Meteor. Soc., 82, 399-415.
Das, T., and Coauthors, 2009: Structure and Detectability of Trends in Hydrological
Measures over the Western United States. J. Hydrometeorology, 10, 871-892.
Dettinger, M. D., D. Cayan, and G. J. McCabe, 1994: Decadal trends in runoff over the
western U.S. and links to persistent North Pacific sea-surface-temperature and
atmospheric-circulation patterns. 18th Annual NOAA Climate Diagnostics
Workshop, 240-243.
127
Dettinger, M. D. and D. R. Cayan, 1995: Large-Scale Atmospheric Forcing of Recent
Trends toward Early Snowmelt Runoff in California. J. Climate, 8, 606-623.
Dettinger, M. D., D. R. Cayan, M. K. Meyer, and A. E. Jeton, 2004: Simulated
Hydrologic Responses to Climate Variations and Change in the Merced, Carson,
and American River Basins, Sierra Nevada, California, 1900–2099. Climatic
Change, 62, 283-317.
Diaz, H. F. and J. K. Eischeid, 2007: Disappearing "Alpine tundra" Koppen climatic type
in the western United States. Geophys. Res. Lett., 34.
Dong, B., R. T. Sutton, and A. A. Scaife, 2006: Multidecadal modulation of El Nino
Southern Oscillation (ENSO) variance by Atlantic Ocean sea surface temperatures.
Geophys. Res. Lett., 33.
Draper, N. R. and H. Smith, 1998: Applied Regression Analysis. 3 ed. Wiley.
Enfield, D. B., 2001: Evolution and historical perspective of the 1997/1998 El
NinoSouthern Oscillation event. Bull. Marine Sci., 69, 7-25.
Enfield, D. B., A. M. Mestas-Nunez, and P. J. Trimble, 2001: The Atlantic multidecadal
oscillation and its relation to rainfall and river flows in the continental US.
Geophys. Res. Lett., 28, 2077-2080.
Rocky Mountain Climate Working Group, R. M. C. W., 2009: Rocky Mountain Climate
Protocol: Climate monitoring in the Greater Yellowstone and Rocky Mountain
Inventory and Montioring Networks., National Park Service, Fort Collins, CO.
Natural Resource Report, NPS/XXXX/NRR-20XX/XXX.
Hamlet, A. F., D. Fluharty, D. P. Lettenmaier, N. J. Mantua, E. L. Miles, P. W. Mote, and
L. Whitely-Binder, 2001: Effects of climate change on water resources in the
Pacific Northwest: Impacts and policy implications. S. Preparatory White Paper for
Climate and Water Policy Meeting, Washington, July 2001., Ed., Climate Impacts
Group, University of Washington, 16.
Hamlet, A. F., P. W. Mote, M. P. Clark, and D. P. Lettenmaier, 2007: Twentieth-century
trends in runoff, evapotranspiration, and soil moisture in the western United States.
J. Climate, 20, 1468-1486.
Hidalgo, H. G., and Coauthors, 2009: Detection and Attribution of Streamflow Timing
Changes to Climate Change in the Western United States. J. Climate, 22, 38383855.
IPCC, 2007: Climate Change 2007: The Physical Science Basis. Contribution of Working
Group I to the Fourth Assessment Report of the Intergovernmental Panel on
128
Climate Change. Eds. Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis,
K.B. Averyt, M. Tignor and H.L. Miller. Cambridge University Press, United
Kingdom and New York, NY, USA. pp. 996.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull.
Amer. Meteor. Soc., 77, 437-471.
Karl, T. R., H. F. Diaz, and G. Kukla, 1988: Urbanization: its detection and effect in the
United States climate record. J. Climate, 1, 1099-1123.
Karoly, D. J. and Q. Wu, 2005: Detection of Regional Surface Temperature Trends. J.
Climate, 18, 4337-4343.
Knowles, N. and D. R. Cayan, 2004: Elevational Dependence of Projected Hydrologic
Changes in the San Francisco Estuary and Watershed. Climatic Change, 62, 319336.
Knowles, N., M. D. Dettinger, and D. R. Cayan, 2006: Trends in snowfall versus rainfall
in the Western United States. J. Climate, 19, 4545-4559.
Luce, C. H. and Z. A. Holden, 2009: Declining annual streamflow distributions in the
Pacific Northwest United States, 1948-2006. Geophys. Res. Lett., 36.
Mantua, N. J., S. R. Hare, Y. Zhang, J. M. Wallace, and R. C. Francis, 1997: A Pacific
interdecadal climate oscillation with impacts on salmon production. Bull. Amer.
Meteor. Soc., 78, 1069-1079.
McCabe, G. J., M. A. Palecki, and J. L. Betancourt, 2004: Pacific and Atlantic Ocean
influences on multidecadal drought frequency in the United States. Proc. Nat. Acad.
Sci., 101, 4136-4141.
McCabe, G. J. and M. P. Clark, 2005: Trends and Variability in Snowmelt Runoff in the
Western United States. J. Hydrometeorology, 6, 476-482.
Mccabe, G. J., J. L. Betancourt, S. T. Gray, M. A. Palecki, and H. G. Hidalgo, 2008:
Associations of multi-decadal sea-surface temperature variability with US drought.
Quat. Int., 188, 31-40.
Mekis, E. and W. D. Hogg, 1999: Rehabilitation And Analysis Of Canadian Daily
Precipitation Time Series. Atmosphere-Ocean, 37, 53-85.
Menne, M. J. and C. N. Williams Jr, 2005: Detection of undocumented changepoints
using multiple test statistics and composite reference series. J. Climate, 18, 42714286.
129
Menne, M. J., C. N. Williams Jr, and R. S. Vose, 2009: The United States Historical
Climatology Network Monthly Temperature Data - Version 2. Bull. Amer. Meteor.
Soc., 90, 993-1007.
Moore, J. N., J. T. Harper, and M. C. Greenwood, 2007: Significance of trends toward
earlier snowmelt runoff, Columbia and Missouri Basin headwaters, western United
States. Geophys. Res. Lett., 34.
Mote, P. W., 2003: Trends in snow water equivalent in the Pacific Northwest and their
climatic causes. Geophys. Res. Lett., 30.
Mote, P. W., A. F. Hamlet, M. P. Clark, and D. P. Lettenmaier, 2005: Declining
mountain snowpack in western north America. Bull. Ame. Meteor. Soc., 86, 39-+.
Mote, P. W., 2006: Climate-Driven Variability and Trends in Mountain Snowpack in
Western North America. J. Climate, 19, 6209-6220.
Newman, M., G. P. Compo, and M. A. Alexander, 2003: ENSO-Forced Variability of the
Pacific Decadal Oscillation. J. Climate, 16, 3853-3857.
Pederson, G., L. Graumlich, D. Fagre, T. Kipfer, and C. Muhlfeld, 2010: A century of
climate and ecosystem change in Western Montana: what do temperature trends
portend? Climatic Change, 98, 133-154.
Pederson, G. T., D. B. Fagre, S. T. Gray, and L. J. Graumlich, 2004: Decadal-scale
climate drivers for glacial dynamics in Glacier National Park, Montana, USA.
Geophys. Res. Lett., 31.
Pierce, D. W., and Coauthors, 2008: Attribution of Declining Western U.S. Snowpack to
Human Effects. J. Climate, 21, 6425-6444.
Prato, T. and D. Fagre, Eds., 2007: Sustaining Rocky Mountain Landscapes: Science,
Policy, and Management for the Crown of the Continent Ecosystem. 1 ed.
Resources for the Future Press, 321 pp.
Quadrelli, R. and J. M. Wallace, 2004: A Simplified Linear Framework for Interpreting
Patterns of Northern Hemisphere Wintertime Climate Variability. J. Climate, 17,
3728-3744.
Rajagopalan, B., K. Nowak, J. Prairie, M. Hoerling, B. Harding, J. Barsugli, A. Ray, and
B. Udall, 2009: Water supply risk on the Colorado River: Can management
mitigate? Water Resour. Res., 45, 1-7.
Regonda, S. K., B. Rajagopalan, M. Clark, and J. Pitlick, 2005: Seasonal cycle shifts in
hydroclimatology over the western United States. J. Climate, 18, 372-384.
130
Rood, S. B., G. M. Samuelson, J. K. Weber, and K. A. Wywrot, 2005: Twentieth-century
decline in streamflows from the hydrographic apex of North America. J. Hydrol.,
306, 215-233.
Schaefer, G. L., and R. F. Paetzold, 2000: SNOTEL (SNOpack TELemetry) and SCAN
(Soil Climate Analyses Network. Presented at Automated Weather Stations for
Applications in Agriculture and Water Resources Management: Current Use and
Future Perspectives, Lincoln, NB, March 2000, pp. 7. Document URL:
http://www.wcc.nrcs.usda.gov/publications/factsheets.html, last accessed February
27, 2010.
Schwartz, M. D. and B. E. Reiter, 2000: Changes in North American spring. Int. J. Clim.,
20, 929-932.
Seaber, P. R., F. P. Kapinos, and G. L. Knapp, 1987: Hydrologic Unit Maps. USGS
Water Resources Division, Ed., United States Government Printing Office, 63.
Selkowitz, D. J., D. B. Fagre, and B. A. Reardon, 2002: Interannual variations in
snowpack in the crown of the continent ecosystem. Hydrol. Processes, 16, 36513665.
Serreze, M., M. Clark, R. Armstrong, D. Mcginnis, and R. Pulwarty, 1999:
Characteristics of the Western United States Snowpack from Snowpack Telemetry
(SNOTEL) Data. Water Resour. Res., 35, 2145-2160.
Stewart, I. T., D. R. Cayan, and M. D. Dettinger, 2005: Changes toward earlier
streamflow timing across western North America. J. Climate, 18, 1136-1155.
Thompson, D. W. J. and J. M. Wallace, 2000: Annular Modes in the Extratropical
Circulation. Part I: Month-to-Month Variability. J. Climate, 13, 1000-1016.
Thompson, D. W. J., J. J. Kennedy, J. M. Wallace, and P. D. Jones, 2008: A large
discontinuity in the mid-twentieth century in observed global-mean surface
temperature. Nature, 453, 646-649.
Vincent, L. A., B. R. Zhang, B. R. Bonsal, and W. D. Hogg, 2002: Homogenization of
daily temperatures over Canada. J. Climate, 15, 1322-1344.
Watson, E., G. Pederson, B. Luckman, and D. Fagre, 2008: Glacier mass balance in the
northern U.S. and Canadian Rockies: Paleo-perspectives and 20th century change.
Darkening Peaks, L. B. B Orlove, and Wiegandt E, Ed., University of California
Press, 139-151.
Zhang, R. and T. L. Delworth, 2007: Impact of the Atlantic Multidecadal Oscillation on
North Pacific climate variability. Geophy. Res. Lett., 34.
131
APPENDIX C - LONG-TERM SNOWPACK VARIABILITY AND CHANGE IN
THE NORTH AMERICAN CORDILLERA
Authors:
1,2,3
Gregory T. Pederson*, 4Stephen T. Gray, 3,5Connie A. Woodhouse, 6Julio L. Betancourt,
1
Daniel B. Fagre, 7Jeremy S. Littell, 8Emma Watson, 9Brian H. Luckman,
and 2,3Lisa J. Graumlich
NATURE GEOSCIENCE (Letters)– To be submitted August, 2010
Affiliations:
1
U.S. Geological Survey, Northern Rocky Mountain Science Center, 2727 University Way (Suite
2), Bozeman, MT 59715
2
School of Natural Resources, University of Arizona, 325 Biosciences East, Tucson, AZ 85721
3
Laboratory of Tree-Ring Research, University of Arizona, 105 W. Stadium, Tucson, AZ 85721
4
Water Resource Data System, University of Wyoming, Laramie, WY 82071
5
School of Geography and Development, 412 Harvill Building, University of Arizona, Tucson,
Arizona 85721-0076
6
U.S. Geological Survey, National Research Program, Water Resources Division, Tucson, AZ
85719
7
Climate Impacts Group, University of Washington, Box 354235, Seattle, WA 98195-4235
8
Environment Canada, Toronto, Ontario, Canada
9
Department of Geography, University of Western Ontario, London, Ontario, Canada N6A 5C2
*Corresponding Author: Gregory T. Pederson
Email: [email protected]
132
Introduction to Conclusions
In mountainous regions, snowpack controls the amount and timing of streamflow
runoff1,2 and influences myriad aquatic and terrestrial ecosystem processes3-6. Within the
North American Cordillera, for example, 60-80% of the runoff for the Columbia,
Missouri, and Colorado Rivers originates as snowpack1,7, and serves as the primary water
source for ~72 million people8. In this region, snowpack declined noticeably since 19509,
climate models attribute most of this decline to human activities2, and continued
reductions in snowpack are expected through the 21st century10. The significance of these
changes can only be understood in the context of natural, long-term snowpack variability
across time and space, and in its climatology. Here we use a distributed network of treering chronologies to produce a series of nested snowpack reconstructions spanning most
of the last millennium for key headwaters regions in the cordillera. Before 1950, the
region exhibited substantial inter-basin variability in snowpack, even during prolonged
droughts and pluvials, marked by a predominant north-south dipole associated with
Pacific variability11,12. Snowpack was unusually low in the Northern Rocky Mountains
for much of the 20th century and over the entire cordillera since the 1980s; heralding a
new era of snowpack declines entrained across all major headwaters in western North
America2,13.
Snowpack in the high mountains of western North America (NA) serves as
hydrologic storage that is gradually released as runoff over the warm dry summer
months. Both the timing and amount of water released is critical for hydroelectric
133
generation, manufacturing, irrigation, and water resource management. The distribution
of snowpack and timing of snowmelt influences average stream temperature and flows2,3,
the distribution and spawning of fishes3, plant and animal phenology5, the probability and
size of forest fires6, and the distribution and mass of glaciers14. Recognizing snowpack’s
relevance, the U.S. Natural Resource Conservation Service15 first began monthly manual
measurements of SWE along transects (known as snow courses) in the 1920s with
comprehensive coverage attained by ~1950. The winter season accumulation of
snowpack has traditionally been measured in terms of depth and mass of a standing
column of water – commonly referred to as snow water equivalency (SWE). SWE
measurements made on April 1 generally capture the annual peak accumulation of
snowpack13,16,17, and in conjunction with automated daily SWE measurements from snow
telemetry stations, inform seasonal streamflow forecasting and water resource planning.
The accurate measurement of SWE within any small watershed, however, is
exceedingly difficult. Once on the ground, snow is a dynamic medium that changes
continuously as a function of energy fluxes (e.g. temperature, solar radiation, wind, vapor
pressure deficit) and delivery of moisture, both dictated by atmospheric pressure and
circulation7,17. The interaction between synoptic and local conditions can make
individual snow course records of SWE highly variable and often times more reflective
of local conditions. The aggregation of individual site observations up to large regions or
watersheds, however, distributes the types of environments sampled and measurement
errors to capture a common regional snowpack signal2,13,16,17. Many studies have shown
134
that snowpack declined across individual sites (except the highest elevations) and all
regions since 1950, and is associated with increases in winter and spring minimum
temperatures attributed mostly to human activities2,13. Continued warming across the
West discourages a stationarity assumption-based water management strategy18, and will
challenge current storage and allocation strategies due to the combination of declining
supplies and increasing demands. Long-term reconstructions of snowpack are needed to
understand the non-stationary nature and full range of snowpack variability at all scales,
to validate climate models under natural forcing, and plan for climate change impacts in
western North America.
To date, there has been no systematic effort to reconstruct snowpack across
multiple watershed scales. Tree-rings have a long history of use in reconstructing
drought19,20, streamflow21,22, and temperature23,24 , which integrate similar moisture and
energy fluxes as snowpack. Like local snowpack measurements, tree growth
incorporates both local idiosyncrasies and shared climatic signals that are minimized and
amplified, respectively, by generating a suite of reconstructions with a network of
chronologies, then averaging individual watershed reconstructions of snowpack across
large regions. Successful reconstructions of snowpack have been generated for specific
watersheds within the Upper Colorado River basin25,26, indicating potential for much
broader application.
135
Our approach was to create annual snowpack reconstructions spanning the last
500-1000+ years at the level of medium-size (~39,070 ± !26,181 km2; HUC 6)
watersheds for key headwaters of the Columbia, Missouri, and Colorado River Basins,
thereby providing information on patterns and processes at relatively fine to large spatial
scales (Fig. C-1). For calibration of the tree-ring based reconstructions, temporally
contiguous, watershed-level snowpack datasets were constructed by standardizing
individual April 1 SWE records to unit deviation then averaging records together for each
watershed. The April 1 SWE records for three major headwaters regions are
representative of peak snow accumulation16, and distributed along an elevational transect
with the Northern Rockies containing the lowest elevation sites on average (~1550 ±
!424m) and the Upper Colorado basin the highest (~2,807 ± !311m; see Supplementary
Information, Table C-S1). Recent tree-ring collections and existing chronologies were
standardized using conservative detrending methods (negative exponential or straight line
fit), combined into site chronologies using a robust weighted mean, and variancestabilized to account for possible trends in variance due to changing sample depth27.
Standard and pre-whitened (autocorrelation removed) versions of site chronologies were
produced and combined into a geographically referenced master database then screened
for significant (p " 0.05) relationships with watershed-level snowpack records.
We generated a suite of high-quality snowpack reconstructions using a spatially
and temporally nested, stepwise regression strategy similar to that employed to
reconstruct flow records in the Upper Colorado River21,22. Reconstructions were first
136
Figure C-1. Map of study area and the associated tree-ring based reconstructions of
April 1 snow water equivalent (SWE) shown at multiple watershed scales. The map
shows the individual watersheds and three regions for which April 1 SWE
reconstructions were completed, the U.S. Natural Resource Conservation Service
snow course sites used to generate watershed-scale averages of observed April 1 SWE,
the full set of potential predictor chronologies, and the final set of chronologies that
entered into one or more SWE reconstruction model as a predictor. The graphs of the
April 1 SWE reconstructions are presented by region and latitude, and show the
individual watershed reconstructions of April 1 SWE (gray lines), the regional SWE
average calculated from each individual reconstruction (orange line), a 20-year cubicsmoothing spline (50% frequency cutoff) of the regional SWE average (dark blue
line), and for the Northern Rockies and Greater Yellowstone region a cut-off date of
1376 is shown (dotted vertical line) due to decreasing sample depth and increasing
reconstruction uncertainty. The 20th century records of observed April 1 SWE are
plotted for each large region (black lines) and smoothed with a 20-year cubicsmoothing spline to highlight decadal-scale variability (cyan line) coherent with the
snowpack reconstructions. Shaded intervals show decadal-scale SWE anomalies
mapped in Fig. 2 and Fig. 4. Lettering corresponds to the mapped intervals. The
observed and reconstructed SWE records are plotted as anomalies from the long-term
average, which was calculated sing 1400-1950 AD as a base period. Other base
periods were used to calculate the long-term average SWE conditions yielding highly
similar estimates (see Supplementary Information, Table S6).
attempted using a “watershed limited” approach taking only the chronologies located
137
within 50 km from each watershed as a potential predictor. If well-fit models were not
obtained, which was common, the search area was increased to the region, and finally the
“full pool” of potential predictors. To prevent model overfitting, the entry of predictors
was halted when it resulted in a reduction of either the root mean squared error (RMSEv)
or R2 validation (aka reduction of error RE) statistic. Once a “best” snowpack model was
obtained, the second stage of the nested regression procedure generated successively
longer reconstructions, albeit with diminished skill, using subsets of predictor
chronologies filtered by progressively earlier termination dates. To remove spurious
trends in the statistical properties of each successively longer reconstruction the mean and
variance were scaled to match the long-term mean and variance of the “best”
reconstruction. Records were then combined into a single contiguous snowpack
reconstruction for each watershed. This process was repeated for larger watershed-scales
resulting in a total of 128 models of generally good, but variable, quality that integrate
information from 66 chronologies (see Supplementary Information, Tables C-S2 – C-S5).
A comparison of the final 27 composite HUC 6 watershed snowpack reconstructions
against the average regional observed April 1 SWE records show the models skillfully
capture both interannual and decadal-scale variability thereby providing high-quality
estimates of historic snowpack variability (see Supplementary Information, Fig. C-S1).
The long-term, watershed-based reconstructions of April 1 SWE are arrayed
along a latitudinal and elevational gradient spanning the north-south (N-S) moisture
138
Figure C-2. Decadal departures in reconstructed April 1 SWE shown for watersheds
predominately within the U.S. portion of the American Cordillera. Maps show average
SWE conditions over the following intervals previously highlighted in Fig 1: a) 14401470, b) 1511-1530, c) 1565-1600, d) 1601-1620, e) 1845-1895, and f) 1902-1932. The
mapped SWE anomalies were calculated by averaging annual conditions for each HUC 6
watershed over the time interval shown, and are plotted as anomalies from the long-term
regional mean (1400-1950 AD). The final datasets along with the ability to generate
user-defined maps of interannual- to interdecadal-scale departures in reconstructed and
observed SWE are provided at http://www.nrmsc.usgs.gov/NorthAmerSnowpack/.
dipole, and show distinctly different modes of variability (Fig. C-1). The Greater
Yellowstone and Northern Rocky Mountain watersheds (i.e. northern cordillera) exhibit
similar decadal-scale and longer-term phasing of snowpack anomalies typically
antiphased with the Upper Colorado River Basin (Fig. C-S2). During the 1450s (~1440-
139
1470 AD; Fig. C-1a, C-2a), and 1550s drought 19,20 (~1550-1600 AD; Fig. C-1c, C-2c)
sustained low snowpack conditions were centered over the Upper Colorado River basin
and coincided with sustained and severe low flows21 (Fig. C-3). Over the same intervals,
northern cordilleran watersheds generally experienced average to above average
snowpack. Severe low snowpack conditions across the northern cordillera prevailed from
~1511-1530 AD when the Upper Colorado experienced average to high snowpack
conditions (Fig. C-1b, C-2b) and above average streamflow (Fig. C-3). With the
exception of ~1303-1330 AD, no analogous periods of severe, low snowpack are
apparent in the northern cordilleran watersheds until they were met and exceeded in the
early 20th century (~1900-1942 AD; Fig. C-1f, C-2f). From 1650 to the 1890s generally
high snowpack conditions prevailed in the northern cordillera coinciding with the
maximum Holocene advance of glaciers28, and a period of reduced west-wide fire
synchronicity with regional quiescence in fire activity29. Two notable decadal-scale
departures of high snowpack coincided with cool summer temperatures23 and
consequently the major intervals of Little Ice Age (LIA) glacier advance and moraine
formation in the northern cordillera28; one centered on the 1700s, and the other from
1845-1895 (Fig. C-1e, C-2e). The highly variable and generally opposing decadal
snowpack anomalies in the Upper Colorado (Fig. C-2e, C-S2) may, in conjunction with
regional summer temperatures24, explain the lack of evidence for a substantial LIA
glacial advance over the southern cordillera.
140
Further comparison of Upper Colorado Basin snowpack and streamflow records
suggests periods of increased sublimation (and/or groundwater recharge, vegetation
uptake, etc.) due to higher regional temperatures and reduced hydrologic effectiveness of
snowpack. For example, during the 1950s (~1950-56 AD) and medieval drought21
(1118-1179 AD), average to slightly below average snowpack conditions are apparent
over the majority of both low-flow intervals (Fig. C-3). The low-flow conditions are
more severe than suggested by snowpack conditions alone, and coincide with two periods
Figure C-3. Comparison of decadal-scale variability in estimates of Upper Colorado
Basin April 1 SWE and total annual streamflow. For clarity, each reconstruction is
shown with a 20-year (dark blue line) and 50-year (red line) cubic-smoothing spline
(50% frequency cutoff), shown as departures from the long-term mean, and discussed
low flow periods are highlighted with gray bars. The reconstructed Upper Colorado
River flows21 are plotted in billions of cubic meters (BCM), and were calibrated against
the Lee’s Ferry gage record.
141
of elevated regional and hemispheric temperatures24,30. The combined influence of
warmer temperatures and severe decadal-scale winter drought is suggested by the
extreme low-flow interval of the medieval drought (~1143-1155 A.D.). Due to reduced
sample depth, increasing model uncertainty, and potential overlap between predictor
datasets, we cannot rule out that these relationships don’t arise by chance. The same
negative interaction between temperature and cool season precipitation, however, has
been demonstrated for the 1950s and the recent 2000-2009 drought30 suggesting the
physical mechanism is consistent with observations, and may portend future snowtemperature-runoff relationships across the NA Cordillera.
From individual watersheds to broader regions, snowpack reconstructions show a
generally stationary N-S antiphasing of decadal moisture anomalies maintained over the
length of record (Fig. C-S2, C-2a-f). The N-S dipole suggests sustained departures in the
average position of the wintertime stormtrack consistent with tropical and extratropical
Pacific Basin related forcings11,12. Though the major N-S dipole pattern of snowpack
variability is maintained across broad regions, the watershed-level records show
substantial interbasin variability exists within regions. Several notable exceptions to the
N-S dipole configuration did occur however. Immediately following the 1550s drought
(~1601-1620 AD) sustained high snowpack conditions across the NA cordillera are
evident (Fig. C-1d, C-S2), with the highest estimated snowpack levels centered over the
southern watersheds, perhaps suggesting a southerly stormtrack with highly meridional
flows. Cordilleran-wide periods of low snowpack shown for the 1350s, 1400s, and post-
142
1980s era (Fig. C-S2) correspond with late medieval era and recent regional and
hemispheric periods of anomalous warmth24,30. The N-S dipole pattern of snowpack
variability and potential role of temperature in synchronizing snow declines shown here
should be considered for use in accuracy assessments of coupled ocean-atmosphere
global climate models (AOGCMs) ability to resolve the average position and temporal
variability in NA winter storm tracks. The reproducibility of these patterns by AOGCMs
is relevant for assessing, and possibly improving, the spatial and temporal accuracy of
future forecasts of winter snowpack conditions, and also has implications for small-scale
detection and attribution studies2,31,32.
The early 20th century decline (~1900-1942 AD) in snowpack observed across
the northern cordillera coincides with warm SSTs across the Gulf of Alaska33 (i.e. a
positive Pacific Decadal Oscillation34 phase) suggesting changes in atmospheric
circulation likely resulted in the southerly displacement of the winter stormtrack and
consequently a substantial portion of the observed northern (southern) declines
(increases) in snowpack (Fig. C-1f, C-2f). Additionally, the regions relatively lowelevation landmass, snowmass, and snow course monitoring sites (~1550 m as compared
to the ~2800 m of the Upper Colorado) also results in warmer average temperatures
nearer the 0°C isotherm, likely causing a greater sensitivity of the regions snowpack to
temperature change over the critical snow accumulation months (Fig. C-S3). The ~3°C
warmer winter and spring average temperatures estimated for the Northern Rockies snow
course sites was interpolated from meteorological station data (i.e. PRISM 800m 1971-
143
2000 climate normals), and is consistent with the expected effects of both elevation and
latitude on temperature. The warm sampling bias of the Northern Rockies region should
be noted and accounted for in future monitoring efforts, and research comparing trends
and variability in snowpack records across western North America. Perhaps more
importantly, however, Figure C-S3 suggests additional warming of several degrees
centigrade would shift the majority of the snow monitoring sites, and hence snowmass, of
the Northern Rockies across the 0°C isotherm during the core winter accumulation
months, and the entire central North American Cordillera during March and April. This
would substantially enhance the melt rates of accumulating winter snowfall, and likely
result in more mid-winter snowmelt driven runoff events, earlier and lower peak flows,
and a longer period of substantially reduced summer baseflows.
Substantial 20th century declines in snowpack are shown across all northern
cordillera watersheds (Fig. C-1), with synchronous declines evident over the entire NA
cordillera since 1980 (Fig. C-1g, C-4, and C-S2). The reduced hydrologic effectiveness
of snowpack in dictating runoff, and the breakdown of the N-S snowpack dipole during
the medieval periods of elevated warmth may 1) provide insight into mechanisms behind
the synchronous declines in snowpack conditions since the 1980s, and 2) serve as a
conservative analog for likely future hydroclimatic conditions across the NA Cordillera.
Additionally, the high LIA and anomalously-low 20th century snowpack conditions
observed across the northern cordillera suggests a potentially greater sensitivity to
regional long-term temperature change due to the relatively low-elevation of the land and
144
snow masses, and the consequently ~3°C higher average temperatures. The substantial
early 20th century snowpack decline, however, was likely predominately driven by
anomalous high pressure and warmth over the Gulf of Alaska33 and consequently a
preference for a more southern stormtrack position. Evidence for southern winter
stormtrack over the early 20th century is shown by the anomalously high streamflow and
snowpack conditions across the Upper Colorado, and when combined with the higher
cooler elevations of the southern cordillera may have, until recently, served to buffer
snowpack from substantial temperature driven declines. High-resolution physical climate
models with well resolved atmospheric circulation patterns, however, are required to
disentangle the range of potential mechanisms. Since the early-1980s the NA cordillerawide decline in snowpack (Fig. C-4) is relatively unique within the context of the longterm records, consistent with previous studies2,13, and heralds a new era of accounting for
non-stationarity in water resource management18.
145
Figure C-4. Post-1980 average April 1 SWE conditions. Maps of post-1980 average
SWE conditions (Fig. 1g) plotted as anomalies from the regional long-term mean (14001950 AD) for the observational record (left) and tree-ring based reconstructions (right).
Note, the map showing average reconstructed SWE values is not exactly equivalent to the
observational record since many individual watershed SWE reconstructions have
different end years - the earliest of which only extend to 1990. This implies the similarity
in patterns of anomalies should be noted, but the magnitudes of departure should not be
expected to be the same.
Methods
We produced two sets of watershed-level snowpack reconstructions using
standard and pre-whitened (i.e. residual) chronologies respectively, however, no mixing
of standard and pre-whitened chronologies was allowed in model construction. The
models presented herein, are constructed solely from the standard chronologies since
146
common interval analysis (1936-1990) indicates the autoregressive (AR) coefficients, and
coherence of decadal-scale anomalies and longer-term trends in snowpack closely match
that of the observational records. Reconstructions produced using residual chronologies
result in over-whitened snowpack reconstructions compared to the observational SWE
records, particularly in the northern cordilleran watersheds, consequently over dampening
decadal-scale and longer-term variability.
All potential predictor chronologies used in the snowpack reconstructions are
from recent collections of the co-authors, and a subset of records contributed to the
World Data Center for Paleoclimatology’s International Tree-Ring Data Bank (ITRDB,
http://www.ncdc.noaa.gov/paleo/treering.html). Chronologies from the ITRDB were
screened to ensure they were not collected for use in ecological disturbance studies (e.g.
pandora moth outbreaks), and that they extended at least to 1980. All chronologies were
screened for significant (p " 0.05) relationships with watershed-level April 1 SWE at year
t and t ± 1. Growth year relationships (t) with winter season April 1 SWE were most
common, with few events of significant lag/lead correlations, so only year t relationships
were allowed to enter as potential model predictors. Snowpack models constructed with
chronolgies extending to 1990 or greater were preferred even if model calibration and
validation statistics over shorter intervals indicated greater model skill. With the
exception of a few cases, all models extended to and beyond the 1990s. This ensured the
tree-ring chronologies accurately captured the recent few decadades of snowpack
declines, and therefore exhibit the capacity to capture any potential periods of historic
147
snowpack reductions that equaled and potentially exceeded recent observed declines in
April 1 SWE.
The nested, multiple linear regression models of April 1 SWE were constructed
using a forward-backward stepwise regression procedure that allowed screened predictor
chronologies to enter at # = 0.05, with a removal # = 0.10. Model strength was
summarized by the adjusted R2 and F level of the regression equation, and potential
problems with multicolinearity of predictors detected using the Variance Inflation Factor
(VIF) statistic. Model validation was conducted using a leave-one-out (PRESS) crossvalidation procedure. Data subsetting into different calibration and validation intervals
was not performed due to the relatively short length of the April 1 SWE records, and the
substantial increase in uncertainty of watershed-level snowpack conditions pre-1950 due
to the rapid decline in number of available snow course records. As an additional quality
control check on the final HUC 6 watershed models of April 1 SWE, both step-wise
regressions for larger watersheds and regions were performed (see Supplementary
Information, Table S5) along with principal components analysis (PCA) based
regressions (not shown). These larger watershed and regional April 1 SWE models were
then compared against the individual HUC 6 watershed-level April 1 SWE
reconstructions for commonalty in high- and low-frequency variance. Only the Lower
Snake HUC 6 watershed reconstruction of April 1 SWE (in the Northern Rockies/ Upper
Columbia region) indicates a potential problem with capturing the shared regional lowfrequency signal in snowpack variability over the 1750-1900 interval. Since this is one
148
of the lowest elevation watersheds with the greatest potential for a strong temperature
influence on snowpack, the model was not rejected, but caution in interpretation and use
of data is urged. Otherwise, all other reconstructions cross-validated well (see
Supplementary Information, Table C-S5) with the occasional exception of the DurbanWatson (DW) statistic indicating the potential for problems with autocorrelation in the
residuals. Visual inspection of the residual plots and sensitivity tests showed the
potential problem to be minor and typically driven by a series of years in the early and
lower-quality portion of the observed SWE record. More specific information on the
methodologies employed here, along with the relevant citations can be found in
Woodhouse et al.22 and Meko et al.21.
An assessment of the N-S snowpack dipole shown in Fig. C-S2 was performed
using PCA (varimax rotated and unrotated, not shown) on the correlation matrix of the
individual HUC 6 watershed April 1 SWE reconstructions over a common interval (15001990). Results support our (longer-tem) interpretation of Fig. C-S2, splitting the southern
and northern cordilleran watersheds onto the first and second components respectively,
explaining 54% of the total variance, with time-series plots of the component scores
showing stationarity in the N-S snowpack dipole. An additional 14.6% of the variance is
explained by the third principal component, which indicates the northern watersheds of
the Upper Colorado region (i.e. the Webber, Green, White-Yampa, and the North and
South Platte) contain variance orthogonal to both the northern or southern cordillera. The
variance of the individual watershed reconstructions, and the time-series plot of the
149
component scores, suggest these watersheds in particular are a dynamic boundary
between the N-S dipole. For example, at the height of the LIA these watersheds mirror
average snowpack conditions in the northern cordillera (Fig. C-2e), whereas during the
1450s and 1550s drought (Fig C-2a,c) conditions are more similar to those shown across
the southern cordillera.
All the snowpack reconstructions, and tree-ring chronologies used to generate
them, are available online at the World Data Center for Paleoclimatology in Boulder,
Colorado, U.S.A. (http://www.ncdc.noaa.gov/paleo/paleo.html), and from the U.S.
Geological Survey Northern Rocky Mountain Science Center in Bozeman, Montana,
U.S.A. (http://www.nrmsc.usgs.gov/NorthAmerSnowpack/). Also available online from
the USGS is a web-mapping tool that allows for the generation of user-defined snowpack
anomaly maps and animations.
Acknowledgments
We are grateful for all the data screening and analysis provided by S. Laursen,
and assistance in the construction of web-mapping tools in Arc10 provided by S. Moore,
B. Ralston, and L. Saunders of ESRI, and S. Carrithers and L. Clampitt of the USGS
Northern Rocky Mountain Science Center. A special thanks to contributors to the
International Tree Ring Databank, and the U.S. Natural Resource Conservation Service
for making your April 1 SWE records available. This research was financially supported
in part by the U.S. Geological Survey’s Western Mountain Initiative, and National
150
Science Foundation Grants #0620793 and #0734277. Any use of trade, product, or firm
names is for descriptive purposes only and does not imply endorsement by the U.S.
Government.
Author contributions
G.T.P, S.T.G, C.A.W., and L.J.G planned the project, contributed data, designed
and participated in the data analyses and writing of the paper. G.T.P and S.T.G.
conducted all the analyses. J.L.B. and D.B.F. contributed substantially to the analysis
design and writing of the paper. J.S.L, E.W., and B.H.L provided critical northern
cordilleran tree-ring chronologies and contributed to the writing of the paper.
Additional information
Supplementary Information accompanies this paper on
www.nature.com/naturegeoscience. Reprints and permissions information is available
online at http://npg.nature.com/reprintsandpermissions. Correspondence and requests for
materials should be addressed to G.T.P.
151
Supplementary Information
Figure C-S1. Watershed-based observed and reconstructed April 1 SWE. Observed April
1 SWE for large regions/watersheds (black lines) was calculated from individual snow
course records, and are plotted alongside the tree-ring based watershed SWE
reconstructions (gray lines) and the regional estimates of average SWE conditions
(dashed orange lines; calculated by averaging individual watershed reconstructions
together). Both the observed (cyan line) and reconstructed regional averages of (dark
blue line) April 1 SWE are shown smoothed with a 20-year cubic-smoothing spline (50%
frequency cutoff) to highlight correspondence at decadal-scales. Note the spline is biased
by end effects at the start and end of any time series, and this bias is particularly evident
at the start of the observed April 1 SWE records (~1920 to 1930 AD).
152
Figure C-S2. Decadal-scale antiphasing of the N-S snowpack dipole. The 20-year splines
of the regional average snowpack anomalies are plotted for clarity and to highlight
variability at decadal scales. The shaded areas highlight the only periods of synchronous,
cordillera-wide snowpack declines.
153
Figure C-S3. Average seasonal and monthly temperatures of snow dominated landmasses
and snow course measurement sites across the three study regions for critical winter and
spring snow accumulation months. Average temperatures were calculated, mapped, and
estimated at each snow course site using PRISM (http://www.prism.oregonstate.edu/)
800m 1971-2000 climate normals, and the average elevation of snow course monitoring
sites for each of the three regions is shown with a different color contour line.
154
Table C-S1. Summary of NRCS snow course April 1 SWE sampling site elevations by
region.
Mean Median
Max
Min
Region
No. ele. (m) ele. (m) ele. (m) ele. (m) ! (m)
Northern Rockies
373
1544
1532
2516
345
424
Greater Yellowstone
279
2307
2288
3093
1708
291
Upper Colorado
305
2807
2837
3538
1678
311
Table C-S2. Summary of April 1 SWE models by region.
Regions
# Models
# Chronologies
Northern Rockies and Greater
Yellowstone (HUC6)
38
28
Northern Rockies and Greater
Yellowstone (Large Watersheds)
12
13
Upper Colorado (HUC6)
52
34
Upper Colorado (Large
Watershed)
26
22
128
66*
Total:
*Total number of chronologies used in all models. Replicate
appearances by individual chronologies have been removed.
Table C-S3. List of tree-ring chronologies that entered one or more of the 128 nested
April 1 SWE regression models (see
http://www.nrmsc.usgs.gov/NorthAmerSnowpack/tables).
Table C-S4. Tree-ring chronologies (and raw ring-width files) that entered one or more of
the 128 nested April 1 SWE regression models. All chronologies were conservatively
detrended (negative exponential, negative regression, or mean line) to preserve lowfrequency variability, and underwent r-bar variance scaling to correct for the influence of
decreasing sample depth over the early portion of the record (see
http://www.nrmsc.usgs.gov/NorthAmerSnowpack/tables).
155
Table C-S5. April 1 SWE nested model calibration and validation statistics (see
http://www.nrmsc.usgs.gov/NorthAmerSnowpack/tables). The “MV adjust” and “In
composite” columns indicates whether a particular reconstruction had it’s mean and
variance scaled to match the mean and variance (calculated over a common interval) of
the “best” SWE reconstruction before inclusion into the composite reconstruction of
snowpack. If the mean and variance was not adjusted, it was not included in the
composite record due to model issues.
Table C-S6. Different long-term April 1 SWE averages calculated using three different
base periods.
1400-1950 1400-Present Full Record
Region
(z-scores)
(z-scores)
(z-scores)
Northern Rocky
Mountains
0.463
0.431
0.432
Greater Yellowstone
0.369
0.346
0.350
Upper Colorado
0.074
0.070
0.076
156
References
1
Bales, R. C. et al. Mountain hydrology of the western United States. Water Resour
Res 42 (2006).
2
Barnett, T. P. et al. Human-induced changes in the hydrology of the western United
States. Science 319, 1080-1083 (2008).
3
Hauer, F. R. et al. Assessment of climate change and freshwater ecosystems of the
Rocky Mountains, USA and Canada. Hydrological Processes 11, 903-924 (1997).
4
Littell, J. S., McKenzie, D., Peterson, D. L. & Westerling, A. L. Climate and
wildfire area burned in western U.S. ecoprovinces, 1916 - 2003. Ecological
Applications 19, 1003-1021, doi:10.1890/07-1183.1 (2009).
5
Schwartz, M. D., AHAS, R. & AASA, A. Onset of spring starting earlier across the
Northern Hemisphere. Global Change Biology 12, 343-351 (2006).
6
Westerling, A. L., Hidalgo, H. G., Cayan, D. R. & Swetnam, T. W. Warming and
earlier spring increase western US forest wildfire activity. Science 313, 940-943
(2006).
7
Serreze, M., Clark, M., Armstrong, R., McGinnis, D. & Pulwarty, R. Characteristics
of the Western United States Snowpack from Snowpack Telemetry (SNOTEL)
Data. Water Resour Res 35, 2745-2160 (1999).
8
Bureau, U. S. C. U.S. Western Region 2009 Population Estimates,
<http://www.census.gov/> (2010).
9
Mote, P. W., Hamlet, A. F., Clark, M. P. & Lettenmaier, D. P. Declining mountain
snowpack in western north America. B Am Meteorol Soc 86, 39-+ (2005).
10
IPCC. Climate Change 2007: The Physical Science Basis. Contribution of Working
Group I to the Fourth Assessment Report of the Intergovernmental Panel on
Climate Change. 996 (Cambridge University Press, 2007).
11
Woodhouse, C. A., Russell, J. L. & Cook, E. R. Two Modes of North American
Drought from Instrumental and Paleoclimatic Data. Journal of Climate 22, 43364347, doi:10.1175/2009JCLI2705.1 (2009).
12
Brown, D. P. & Comrie, A. C. A winter precipitation 'dipole' in the western United
States associated with multidecadal ENSO variability. Geophys Res Lett 31 (2004).
13
Mote, P. W. Climate-Driven Variability and Trends in Mountain Snowpack in
Western North America. Journal of Climate 19, 6209-6220 (2006).
157
14
Fountain A, H. M., Jackson K, Basagic H, Nylen T, and Percy D Open File Report
23 (USGS, 2007).
15
NRCS. Snowcourse April 1 snow water equivalent records,
<http://www.wcc.nrcs.usda.gov/snowcourse/> (2010).
16
Bohr, G. S. & Aguado, E. Use of April 1 SWE measurements as estimates of peak
seasonal snowpack and total cold-season precipitation. Water Resour Res 37, 51-60
(2001).
17
Pederson, G. T. et al. Climatic controls on the snowmelt hydrology of the Northern
Rocky Moutnains. Journal of Climate (In review).
18
Milly, P. C. D. et al. Climate Change: Stationarity Is Dead: Whither Water
Management? Science 319, 573-574, doi:10.1126/science.1151915 (2008).
19
Cook, E. R., Woodhouse, C. A., Eakin, C. M., Meko, D. M. & Stahle, D. W. Longterm aridity changes in the western United States. Science 306, 1015-1018 (2004).
20
Stahle, D. W., Fye, F. K., Cook, E. R. & Griffin, R. D. Tree-ring reconstructed
megadroughts over North America since AD 1300. Climatic Change 83, 133-149
(2007).
21
Meko, D. M. et al. Medieval drought in the upper Colorado River Basin. Geophys
Res Lett 34, doi:10.1029/2007GL029988 (2007).
22
Woodhouse, C. A., Gray, S. T. & Meko, D. M. Updated streamflow reconstructions
for the Upper Colorado River Basin. Water Resour Res 42, WO5415,
doi:10.1029/2005wr004455 (2006).
23
Luckman, B. H. & Wilson, R. J. S. Summer temperatures in the Canadian Rockies
during the last millennium: a revised record. Clim Dynam 24, 131-144 (2005).
24
Salzer, M. W. & Kipfmuller, K. F. Reconstructed Temperature and Precipitation on
a Millennial Timescale from Tree-Rings in the Southern Colorado Plateau, U.S.A.
Climatic Change 70, 465-487, doi:10.1007/s10584-005-5922-3 (2005).
25
Timilsena, J. & Piechota, T. Regionalization and reconstruction of snow water
equivalent in the upper Colorado River basin. Journal of Hydrology 352, 94-106
(2008).
26
Woodhouse, C. A. A 431-Yr Reconstruction of Western Colorado Snowpack from
Tree Rings. Journal of Climate 16, 1551-1561, doi:10.1175/1520-0442(2003)016
(2003).
158
27
Cook, E. R., Briffa, K., Shiyatov, S. & Mazepa, V. in Methods of
Dendrochonology: Applications in the Environmental Sciences (eds E.R. Cook &
L.A. Kairiukstis) 104-123 (Springer, 1990).
28
Luckman, B. H. The Little Ice Age in the Canadian Rockies. Geomorphology 32,
357-384 (2000).
29
Kitzberger, T., Brown, P. M., Heyerdahl, E. K., Swetnam, T. W. & Veblen, T. T.
Contingent Pacific-Atlantic Ocean influence on multicentury wildfire synchrony
over western North America. P Natl Acad Sci USA 104, 543-548,
doi:10.1073/pnas.0606078104 (2007).
30
Woodhouse, C. A., Meko, D. M., MacDonald, G. M., Stahle, D. W. & Cook, E. R.
A 1,200-year perspective of 21st century drought in southwestern North America. P
Natl Acad Sci USA, early edition, doi:10.1073/pnas.0911197107 (2010).
31
Das, T. et al. Structure and Detectability of Trends in Hydrological Measures over
the Western United States. Journal of Hydrometeorology 10, 871-892 (2009).
32
Pierce, D. W. et al. Attribution of Declining Western U.S. Snowpack to Human
Effects. Journal of Climate 21, 6425-6444 (2008).
33
Wilson, R., Wiles, G., D’Arrigo, R. & Zweck, C. Cycles and shifts: 1,300 years of
multi-decadal temperature variability in the Gulf of Alaska. Clim Dynam 28, 425440 (2007).
34
Mantua, N. J., Hare, S. R., Zhang, Y., Wallace, J. M. & Francis, R. C. A Pacific
interdecadal climate oscillation with impacts on salmon production. B Am Meteorol
Soc 78, 1069-1079 (1997).
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement