Global Sea Ice Concentration Reprocessing Product User Manual

Ocean & Sea Ice SAF
Global Sea Ice Concentration
Reprocessing
Product User Manual
Product OSI-409
Document version: 1.3
Data set version: 1.1
October 2011
Steinar Eastwood, met.no
Kristian Rune Larsen, DMI
Thomas Lavergne, met.no
Esben Nielsen, DMI
Rasmus Tonboe, DMI
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
Documentation Change Record
Document Data
Software Date
Change Description
version
set
version
version
V 1.0
V 1.0
V 4.0
14.01.2010 First version.
V 1.1
V 1.0
V 4.0
30.03.2010 Updates after DRI review.
V 1.2
V 1.1
V 4.0
23.06.2011 The uncertainties have been re-calculated
with a corrected and re-implemented
algorithm. Added 2008 -10.2009.
V 1.3
V 1.1
V 4.0
21.10.2011 Updated after DRI2 review
The software version number gives the corresponding version of the OSI SAF High Latitude
software chain which was used to produce the reprocessing data set.
OSI SAF
1
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
CONTENTS
1.
Introduction...................................................................................................................... 4
1.1 OSI SAF overview...................................................................................................... 4
1.2 Sea ice concentration reprocessing project................................................................ 4
1.3 Ownership and copyright of data................................................................................ 4
1.4 Acknowledgment........................................................................................................ 4
1.5 Glossary..................................................................................................................... 5
1.6 Reference Documents................................................................................................ 5
2. Input data......................................................................................................................... 6
2.1 SMMR........................................................................................................................ 6
2.2 SSM/I......................................................................................................................... 6
2.3 ECMWF NWP............................................................................................................ 7
3. Algorithms........................................................................................................................ 9
3.1 The ice concentration algorithm.................................................................................. 9
3.1.1 Selection of algorithm.......................................................................................... 9
3.1.2 The hybrid ice concentration algorithm................................................................ 9
3.1.3 Tb correction for water vapor and open water surface roughness variability......10
3.2 Dynamical tie-points................................................................................................. 10
3.3 Sea ice concentration uncertainties.......................................................................... 11
3.3.1 Error sources and definitions............................................................................. 12
3.3.2 Instrument errors............................................................................................... 12
3.3.3 Algorithm and tie-point uncertainties.................................................................. 12
3.3.4 Representativeness error.................................................................................. 12
3.3.5 Geo-location error.............................................................................................. 12
3.4 The sea ice concentration uncertainty algorithm....................................................... 13
4. Processing scheme........................................................................................................ 15
4.1 L1/L2 processing...................................................................................................... 15
4.1.1 Converting antenna temperature to brightness temperature.............................. 15
4.1.2 Selection of data................................................................................................ 15
4.1.3 Adding NWP data.............................................................................................. 16
4.1.4 RTM correction and ice concentration calculations............................................ 16
4.1.5 Uncertainty calculations..................................................................................... 16
4.1.6 Dynamical tie-points.......................................................................................... 16
4.2 L3 processing........................................................................................................... 16
4.2.1 Daily gridding..................................................................................................... 16
4.2.2 Coastal correction.............................................................................................. 17
4.2.3 Climatological maximum extent masking........................................................... 17
4.2.4 T2m check......................................................................................................... 17
4.3 L4 processing........................................................................................................... 18
4.3.1 Applying masks and coastal correction.............................................................. 18
4.3.2 Gap filling by interpolation................................................................................. 18
4.3.3 Final formatting.................................................................................................. 19
5. Product description......................................................................................................... 20
5.1 Product specification................................................................................................ 20
5.1.1 Sea ice concentration........................................................................................ 20
5.1.2 Uncertainty estimate.......................................................................................... 20
5.1.3 Processing flag.................................................................................................. 20
5.2 Grid specification...................................................................................................... 21
5.3 Meta data specification............................................................................................. 21
5.4 File naming convention............................................................................................. 23
5.5 Product availability.................................................................................................... 23
5.6 Delivery of updates................................................................................................... 23
6. References..................................................................................................................... 24
7. Appendix A : Example of NetCDF format....................................................................... 26
OSI SAF
2
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
8. Appendix B: Examples of monthly climatological maximum extent masks..................... 29
9. Appendix C: Missing dates............................................................................................. 30
10. Appendix D: Metadata List for EUMETSAT Data Center.............................................. 31
OSI SAF
3
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
1. Introduction
1.1 OSI SAF overview
The Ocean and Sea Ice Satellite Application Facility, OSI SAF, is a EUMETSAT project that
started in 1997. The OSI SAF is a part of the EUMETSAT distributed ground segment for
production of operational near real time value added satellite products. The OSI SAF delivers
a range of air-sea interface products, namely: sea ice characteristics, sea surface
temperature, radiative fluxes and wind. The sea ice products are sea ice concentration, sea
ice edge, sea ice type and sea ice drift.
The OSI SAF project is managed by CMS, Meteo-France. The sea ice products are
produced at the OSI SAF High Latitude processing facility under the responsibility of the
Norwegian Meteorological Institute, operated jointly by the Norwegian and Danish
Meteorological Institutes.
1.2 Sea ice concentration reprocessing project
Since the start of the operational production of sea ice products in 2002 the growing user
group has brought more focus on expanding the available data set. It was therefore decided
to reprocess historical passive microwave data to extend the OSI SAF sea ice data set. This
effort was started in 2006 as a part time EUMETSAT visiting scientist activity in collaboration
with the UK Met Office. The goal was to reprocess the SSM/I data record. A collaboration
was also established with NSIDC to include the SMMR data record in the project, and an
EUMETSAT visiting scientist project was set up for this task. These two Visiting Scientist
projects initiated the OSI SAF reprocessing and produced a first version of reprocessing data
set based on the SSM/I data. Later further improvements have been implemented before the
current version was finished.
This reprocessing product fulfills the requirement OSI-PRD-PRO-205 in the OSI SAF Product
Requirement Document [RD.1]. No validation results are presented in this document. The
validation results are presented in a separate validation report [RD.2].
1.3 Data set version
This product user manual describes version 1.1 of the OSI SAF reprocess sea ice
concentration data set, which was released in October 2011.
1.4 Ownership and copyright of data
The OSI SAF reprocessed sea ice concentration data set has been produced under
responsibility of Norwegian Meteorological Institute and Danish Meteorological Institute. The
ownership and copyrights of the data set belongs to EUMETSAT. The data set is distributed
freely, but EUMETSAT must be acknowledged when using the data. EUMETSAT's copyright
credit must be shown by displaying the words "copyright (year) EUMETSAT" on each of the
products used.
1.5 Acknowledgment
We would like to thank John Stark (previously at the UK Met Office) for his effort as a visiting
scientist under the OSI SAF IOP-SG06-VS01 Visiting Scientist project. His efforts are
documented in Stark (2008).
We would also like to thank Walt Meier and Jeff Smith at NSIDC for their cooperation and
effort during the OSI SAF CDOP-SG01-VS03 Visiting Scientist project. Their work is
documentation in Meier (2008).
OSI SAF
4
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
1.6 Glossary
ASCII
CMS
CDL
CDOP
CF
DMI
DMSP
EASE
ECMWF
ERA
FTP
EUMETSAT
GCMD
met.no
NASA
NetCDF
NSIDC
NWP
OSI SAF
RTM
RSS
SAR
SMMR
SSM/I
TUD
American Standard Code for Information Interchange
Centre de Météorologie Spatiale
network Common data form Description Language
Continuous Development and Operations Phase (OSI SAF project)
Climate and Forecast (Metadata Conventions)
Danish Meteorological Institute
Defense Meteorological Satellite Program
Equal-Area Scalable Earth
European Centre for Medium range Weather Forecast
ECMWF Re-Analysis
File Transfer Protocol
European Organisation for the Exploitation of Meteorological Satellites
Global Change Master Directory
Norwegian Meteorological Institute
National Aeronautics and Space Administration
Network Common Data Form
National Snow and Ice Data Center
Numerical Weather Prediction
Ocean and Sea Ice Satellite Application Facility
Radiative Transfer Model
Remote Sensing Systems
Synthetic Aperture Radar
Scanning Multichannel Microwave Radiometer
Special Sensor Microwave/Imager
Technical University of Denmark
1.7 Reference Documents
[RD.1] OSI SAF CDOP Product Requirement Document, v1.2.
[RD.2] OSI SAF Sea Ice Concentration Reprocessing Product Validation Report.
[RD.3] OSI SAF IOP-SG06-VS01 report (Stark, 2008).
[RD.4] OSI SAF CDOP-SG01-VS03 report (Meier, 2008).
[RD.5] Adaption of SAFOSI sea ice processing system to southern hemisphere (Toudal,
2006).
OSI SAF
5
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
2. Input data
This chapter describes the main input data that have been used for the OSI SAF sea ice
concentration reprocessing.
2.1 SMMR
The Scanning Multichannel Microwave Radiometer (SMMR) instrument on board the Nimbus
7 satellite operated from October 1978 to August 1987 (Gloersen et al., 1992). The
instrument was operated only every second day due to power supply limitations. The
instrument had 10 channels from the six Dicke radiometers at five frequencies (6.6, 10.7,
18.0, 21.0, 37.0 GHz) and vertical and horizontal polarization. The scanning across track was
ensured by tilting the reflector from side to side while maintaining constant incidence angle
on the ground of about 50°. The scan track on the ground formed a 780 km wide arc in front
of the satellite (Gloersen and Barath, 1977). Because of the satellite orbit inclination and
swath width there is no coverage poleward of 84°. The SMMR instrument is further
described in http://nsidc.org/data/docs/daac/smmr_instrument.gd.html and the brightness
temperature data and the initial formatting of the data are described in Meier (2008).
Frequency
(GHz)
Polarizations
6.6
Field of view
Along-track
Cross-track
H,V
148 km
95 km
10.7
H,V
91 km
59 km
18.0
H,V
55 km
41 km
21.0
H,V
46 km
30 km
37.0
H,V
27 km
18 km
Table 1: Characteristics of the Nimbus 7 SMMR channels (Gloersen and Barath,
1977).
2.2 SSM/I
The Special Sensor Microwave/Imager (SSM/I) sensors on bord the Defence Meteorological
Satellite Program (DMSP) started its record with the F08 satellite on 9. July 1987 shortly
before the SMMR ceased to operate on 20. August 1987. The different SSM/I instrument
records are summarised in table 2. The SSM/I is a total power radiometer with a conical scan
measuring the upwelling radiation from the Earth at a constant incidence angle of about
50deg at 7 different channels. The channels are summarised in table 3. The swath width is
about 1400km.
The Special Sensor Microwave/Imager (SSM/I) data set used for this reprocessing was
purchased by EUMETSAT from Remote Sensing Systems (RSS) and covers the whole
period of available satellites with SSMI/I instruments from 1987 to 2009. We have used the
version 6 of the RSS SSM/I data set. The different satellites and covered periods are listed in
Table 2.
The SSM/I data were received as antenna temperatures from RSS. These antenna
temperatures were converted to brightness temperatures (Tb's) using a software provided by
RSS, as described in chapter 4.1.1.
The SSM/I instrument have five low frequency channels similar to SMMR. In addition, two
higher frequency channels with twice the sampling rate are available on the SSM/I. The
characteristics of these channels are listed in Table 3. The 85GHz channels had a
OSI SAF
6
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
malfunction on F08, so they are only useful starting with the F11 satellite. The F10 has not
been used because they are more noisy than F11 and were not needed for a complete
coverage.
Satellite
Period covered
F08
Jul 1987 – Dec 1991
F11
Dec 1991 – May 2000
F13
Dec 1995 – Oct 2009
F14
May 1997 – Aug 2008
F15
Dec 1999 – Oct 2009
Table 2: The different satellite missions carrying the SSM/I instrument and the
periods they cover.
Frequency
(GHz)
Polarizations
19.35
Footprint size
Along-track
Cross-track
H,V
69 km
43 km
22.235
V
50 km
40 km
37.0
H,V
37 km
28 km
85.5
H,V
15 km
13 km
Table 3: Characteristics of the different SSM/I channels (from Wentz, 1991).
The RSS SSM/I version 6 data set incorporates geolocation correction, sensor calibration
and quality control procedures, as well as inter calibration between the different satellites
from overlapping periods. These procedures are documented in the RSS SSM/I User's
Manuals (Wentz, 1991; Wentz, 1993; Wentz, 2006).
The RSS SSM/I data set is constrained with a license for use and distribution, and the
brightness temperatures on swath format used in the OSI SAF reprocessing product can
therefore not be distributed.
2.3 ECMWF NWP
The brightness temperatures (Tb) are corrected explicitly for wind roughening over open
water and water vapor in the atmosphere prior to the calculation of ice concentration. The
correction is using a radiative transfer model function (RTM) and NWP data. The model
function used for correction of the SMMR and SSM/I Tbs using ECMWF NWP input is
denoted by (Wentz, 1983):
Tb= f Ts ,U * , V , L ,Ta ,
(Eq. 1)
where Ts is the physical surface temperature, U* is the sea surface wind friction velocity, V is
the integrated atmospheric water vapor column, L is the atmospheric liquid water column,
and Ta is the surface (2 m) air temperature. Over areas with both ice and water the influence
of open water roughness on the Tb's and the ice emissivity is scaled linearly with the ice
OSI SAF
7
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
concentration. The emissivity of ice is given by standard tie-point emissivities. The correction
procedure is described in detail in Andersen et al. (2006B). The NWP model grid points are
co-located with the satellite swath data in time and space and a correction to the Tb's using
Eq. 1 is applied. ECMWF ERA 40 data are used for the period from 1978 to 2002, and
ECMWF data from the operational model are used from 2002 onwards. A description of the
ERA 40 data archive and the reprocessing can be found in Kålberg et al. (2004). These two
NWP model data sets are not completely consistent, however the inconsistencies are
expected to be small and the residual error is included in the error estimate.
The representation of atmospheric liquid water column (L) in the NWP data is not suitable to
use for Tb correction (see Andersen et al., 2006B). The Tb's are therefore not corrected for
the influence of L. Assuming neutral atmospheric conditions, the wind speed at 10 m, given
by the NWP model, is converted to the surface friction velocity using the factor 0.047. The
other NWP variables are used directly.
OSI SAF
8
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
3. Algorithms
This chapter gives a scientific description of the algorithms that have been used in the
reprocessing, and how the reprocessing setup differs from the operational OSI SAF
algorithm.
3.1 The ice concentration algorithm
A set of three different ice concentration algorithms were applied to reprocess the satellite
microwave radiometer data record from 1978 to 2007. The three ice concentration algorithms
are the TUD (Technical University of Denmark) using the high resolution and high frequency
channels on the SSM/I from 1991 and onwards (Andersen et al., 2006B), the Bootstrap
algorithm in frequency mode (Comiso, 1986; Comiso et al., 1997), and the Bristol algorithm
(Smith, 1996). The latter two are used in combination as a hybrid algorithm, which uses the
near 18 and 37 GHz channels from the SMMR and SSM/I instruments since 1978.
3.1.1 Selection of algorithm
When selecting an ice concentration algorithm it is important to ensure low sensitivity to the
error sources, including variability in atmospheric emission and surface emission. It is
particularly important to find low sensitivity to the parameters which are not properly
corrected for by (using auxiliary data), such as cloud liquid water in the atmosphere and for
ice surface emissivity variability. For climate time series it is important to find an algorithm
using the 19 and 37 GHz channels which can be used as far back in time as possible,
including the SMMR period from 1978 to 1987 and the first SSM/I (F8) from 1987 to 1991.
The SMMR did not have near 90 GHz radiometers and the 85 GHz radiometer on F8 was not
functional. The 85 GHz radiometer was only used for internal use during the SSM/I period.
The analysis of atmospheric sensitivity in Andersen et al. (2006B) showed that the Bootstrap
frequency mode algorithm (Comiso, 1986) had the lowest sensitivity to atmospheric noise
over open water. Furthermore, the comparison to high resolution SAR imagery in Andersen
et al. (2007) revealed that among the algorithms using the low frequency channels (18 and
37 GHz), the Bristol algorithm (Smith, 1986) had the lowest sensitivity to ice surface
emissivity variability. In addition this algorithm had a low sensitivity to atmospheric emission
in particular at high ice concentrations.
Consequently, a hybrid algorithm has been established as a linear combination of two of the
tested algorithms, the Bristol algorithm and the Bootstrap frequency mode algorithm. To
ensure an optimum performance over both marginal and consolidated ice, and to retain the
virtues of each algorithm, the Bristol algorithm is given little weight at low concentrations,
while the opposite is the case over high ice concentrations.
3.1.2 The hybrid ice concentration algorithm
The Bootstrap algorithm (Comiso, 1986) is based on the observation of linear clustering of
ice Tb's in scatter plots of T 37v vs T19v whereas open water Tb's cluster around a single point.
It assumes only two surface types: ice and open water, taking into account the variability of
both to optimize the detection of small sea ice concentrations. The linear relationship yields
the following simple formulation for the total ice concentration, Ct:
Ct=Tb−TbW /Tb I TbW  ,
(Eq. 2)
where Tb is the measured brightness temperature, TbW is the open water tie point, and TbI is
the ice tie point (see section 3.2).
OSI SAF
9
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
The Bristol algorithm (Smith, 1996) is conceptually similar to the Bootstrap algorithm. In a
three-dimensional scatter plot spanned by T19v, T37v and T37h the ice Tb's tend to lie in a plane.
The only difference to the Bootstrap algorithm is that instead of viewing the data in the T 19v,
T37v space, the Bristol algorithm views the data perpendicular to the plane in which the data
lies, i.e. in a transformed coordinate system:
1. axis :T 37v 1.045T37h 0.525T19v ,
(Eq. 3a)
2. axis : 0.9164T19v −T 37v 0.4965T 37h .
(Eq. 3b)
The remaining analysis is identical to the Bootstrap algorithm.
The Bootstrap algorithm is used over open water and the Bristol algorithm is used over ice.
At intermediate concentrations up to 40% the ice concentration is an average weighted
linearly between the two algorithms, as shown in Eq. 3c and 3d. This hybrid algorithm is the
OSI SAF sea ice concentration algorithm.
iceconc=1−wc∗conc bristol wc∗conc bootstrap ,
(Eq. 3c)
wc=∣t−conc bootstrap ∣t−concbootstrap / 2∗t ,
(Eq. 3d)
where t is the threshold of 40%.
3.1.3 Tb correction for water vapor and open water surface roughness variability
Using the model function presented in 2.3, the Tb's are corrected for the influence of water
vapor in the atmosphere and open water surface roughness caused by wind shear. The
model function is a semi empirical radiative transfer ocean model describing the Tb as a
function of sea surface temperature, surface wind friction velocity, total atmospheric water
vapor, total cloud liquid water and surface air temperature. The model function used for
SMMR processing is described in Wentz (1983) and the one used for SSM/I processing is
described in Wentz (1997). The correction procedure is described in Andersen et al. (2006B).
At intermediate ice concentrations the surface emission term is a linear combination of ice
emissivity derived from tie-point signatures and the open water emissivity derived from the
model. The total ice concentration is solved by iteration.
3.2 Dynamical tie-points
Tie-points are typical signatures of 100% ice and open water which are used in the ice
concentration algorithms as a reference. The tie-points are derived by selecting Tb's from
regions of known open water and 100% ice. Usually these tie-points are static in time and
space, but they can be adjusted to follow the seasonally changing signatures of ice and open
water as it is currently done in the operational OSI SAF ice concentration processing. Static
tie-points are prone to be affected by sensor drift, inter sensor calibration differences and
climatic trends in surface and atmospheric emission. The data must therefore be carefully
calibrated before computing the ice concentrations. Here we use dynamic tie-points, a
method that minimizes these unwanted effects, with or without prior calibration.
During winter, in the consolidated pack ice well within the ice edge, the ice concentration is
very near 100 % (Andersen et al., 2007). This has been established using high resolution
SAR data, ship observations and by comparing the estimates from different ice concentration
algorithms. The apparent fluctuations in the derived ice concentration in the near 100% ice
regime are primarily attributed to snow/ice surface emissivity variability around the tie-point
signature and only secondarily to actual ice concentration fluctuations. In the marginal ice
OSI SAF
10
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
zone the atmospheric emission may be significant. The fluctuations due to atmospheric and
surface emission are systematic. In fact, different algorithms with different sensitivity to
atmospheric and surface emission compute very different trends in sea ice area on seasonal
and decadal time scales (Andersen et al., 2007). This means that not only do the sea ice
area have a climatic trend, but the atmospheric and surface constituents affecting the
microwave emission are also changing. For example, different wind patterns, water vapor
and liquid water concentrations in the atmosphere, snow depth, fraction of perennial ice etc.
In an attempt to compensate for the influence of these unwanted trends the tie-points are
derived dynamically using a monthly mean around the reprocessing point in time. It is
assumed that ice concentrations from the NASA Team algorithm above 95 % are in fact near
100 % ice and that the mean value of these data points can be used to derive the ice tiepoint (for SMMR NASA team tie-points from Gloersen et al. (1992) were used and for SSM/I
tie-points from Andersen (1999) were used). The NASA Team ice concentration is the initial
guess before the iteration and the OSI SAF ice concentration does not depend on the NASA
Team ice concentration. The analysis of SAR data in Andersen et al. (2007) from the central
arctic showed that during winter there is more than 99% ice cover. During strong ice drift
divergence and during the summer there may situations where this is not the case. However,
during one month of tie-point data collection we are sure to have captured the situations with
near 100% ice cover. The standard deviation of the tie-point is included in the total ice
concentration error estimate which is the justification for this assumption.
Geographically, the selection of data for the ice tie-point is limited by excluding regions
poleward of 84°. This is due to the limited coverage of the SMMR instrument, and the same
constraint has been applied to the SSM/I data for consistency. Regions of open water were
used for selecting the water tie-point data (53N to 75N and 65S to 80S). There is no attempt
to compensate explicitly for sensor drift or inter-sensor calibration differences between the
seven different sensors used in the analysis. The dynamical tie-point method is in principle
compensating for these problems in a consistent manner.
3.3 Sea ice concentration uncertainties
Uncertainty estimates are needed when the ice concentration data are compared to other
data sets or when the ice concentrations are assimilated into numerical models. The
uncertainties are computed for each data point and are not interpolated or filled into regions
with no data. The uncertainty is given in terms of one standard deviation and quantifies the
confidence we have in a certain data point. The uncertainty scales with the weighting given
each observation when assimilated in models. The overall accuracy of the SMMR total ice
concentrations are estimated to be ±7% (Gloersen et al., 1992). The mean accuracy of
some of the more common algorithms, used to compute ice concentration from SSM/I data,
such as NASA Team and Bootstrap are reported to be 1-6 % in winter (Andersen et al.,
2006A).
The polar atmosphere is generally transparent for microwave radiation in between the
sounding channels called the atmospheric windows near 18, 36, 90, and 150 GHz. For
typical polar atmospheric states the down-welling emission at the surface is about 5-15 K at
18 GHz, 20-40 K at 36 GHz, 30-100 K at 90 GHz. For comparison, the sea ice surface
emission is typically 150-260 K. When computing the ice concentration using the
atmospheric window channels, the atmospheric emission and scattering is an error source.
The tie-points are typical ice and water signatures representative on a hemispheric scale.
Deviations from the typical surface emission signatures result in ice concentration
uncertainties. Both the SMMR and the SSM/I instruments have large foot-prints on the
ground, and the algorithms with the lowest sensitivity to both atmospheric and surface
emissivity variability use Tb's at different frequencies with different foot-print size.
Representing these large foot-prints on a finer, predefined grid results in a
representativeness error. In addition there is the geo-location error, sensor noise, drift, and
sea ice variability over the sampling period.
OSI SAF
11
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
3.3.1 Error sources and definitions
The errors described in the following sections are generally independent. The tie-point
uncertainty, including residual atmospheric noise, sensor noise and ice surface emissivity
variability, is derived from measurements. The representativeness error is simulated.
3.3.2 Instrument errors
Random instrument noise results in Tb and ice concentration uncertainties. The SSM/I ice
concentration uncertainty due to sensor noise is less than 2.6 %. The SSM/I instrument noise
results in a Bristol algorithm ice concentration uncertainty of 1.4 %, 1.7 % for the Bootstrap in
frequency mode (Andersen et al., 2006A).
Systematic sensor drift is in particular an important issue for ice concentration algorithms
using static tie-points. Here we use dynamical tie-points intended for alleviating problems
with sensor drift, inter-sensor calibration and climatic trends in ice surface emissivity and
atmospheric emission, i.e. this method minimizes the uncertainties caused by sensor drift.
3.3.3 Algorithm and tie-point uncertainties
Both the water surface and ice surface emissivity variabilities result in ice concentration
uncertainties. Emission and scattering in the atmosphere also affects the Tb's and the
computed ice concentrations. Different algorithms have different sensitivities to these surface
and atmospheric parameters (Andersen et al., 2006B). Further, both the atmospheric and
surface parameters affecting the ice concentration estimates have climatic trends (Andersen
et al., 2007). To minimize the uncertainties due to these two parameters, the Tb's are
corrected using NWP data for atmospheric humidity and open water roughness in this
reprocessing. The dynamical tie points minimize uncertainty due to the climatic trends in the
atmosphere and on the ice surface on a hemispheric scale while regional trends may still
exist. The remaining tie-point uncertainties are estimated from the spatial ice concentration
variability in regions with open water or 100% ice.
3.3.4 Representativeness error
Foot-print sizes for the channels used for ice concentration mapping range from about 50 km
for the 19 GHz channels to about 12 km for the 85 GHz channels. Foot-prints of uneven size
are combined in the algorithms when computing the ice concentration. The foot-print ice
concentration is represented on a predefined grid. The ice concentration data are normally
represented on a finer grid (typically 12.5 or 25 km) than the sensor resolution (12 to 50 km).
This is sometimes called smearing. The combination of foot-prints of uneven size in the ice
concentration algorithm results in an additional smearing effect. This we call the foot-print
mismatch error. The smearing and the foot-print mismatch error can not be estimated
separately. However, the combined error can be estimated if all other error sources and the
ice cover reference are known a priori. It can also be simulated using high resolution ice
concentration reference data and a model for the satellite measurement foot-print patterns. In
this reprocessing we use a model simulating the smearing uncertainties for different
algorithms and ice concentrations (see Appendix E).
3.3.5 Geo-location error
A geo-location error occurs when the satellite is not exactly oriented. Simulations show that
because the foot-prints are large compared to the typical geo-location errors (about 5km), the
geo-location error is small, and therefore neglected here.
OSI SAF
12
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
3.4 The sea ice concentration uncertainty algorithm
The detailed OSI SAF ice concentration algorithm is described in Appendix E. Below is
shown examples of the derived uncertainty for an SSMI product.
Figure 1: Ice concentration from one day in September 2005
derived from SSMI data. The white areas are areas where
data are not processed since climatologically they are always
open water (climatology described in section 4.2.3).
Figure 2: Total uncertainty (standard deviation) of the product
shown in Figure 1.
OSI SAF
13
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
Figure 3: Total, algorithm and smear uncertainty as function of ice concentration for
the product shown in Figure 1.
OSI SAF
14
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
4. Processing scheme
This chapter describes all the processing steps in the OSI SAF sea ice concentration
reprocessing scheme. The processing steps can be divided in three main steps; Level 1/2,
Level 3 and Level 4. An overview of these three steps are shown in Figure 4, and the
technical details are presented in more detail in the next three sections. The science behind
some of these steps are discussed in Chapter 3. SMMR and SSM/I data are processed
separately and not merged in the overlapping period.
Figure 4: The three main processing elements in the ice concentration
processing chain.
4.1 L1/L2 processing
This main step contains all processing done on the original swath data, without any gridding
or averaging.
4.1.1 Converting antenna temperature to brightness temperature
The antenna temperatures measured by the radiometer are different from the brightness
temperatures, due to antenna pattern effects such as antenna side-lobe contributions and
cross polarization coupling (Njoku et al. 1980; Njoku, 1980). The SSM/I data provided by the
RSS are converted from antenna temperatures to brightness temperatures using the method
described in Wentz (1991).
4.1.2 Selection of data
The SMMR and SSM/I data sets provide almost global coverage (see also chapter 2.1 and
2.2). The exception is a circular area around each pole which is not covered due to the
inclination of the satellite orbit and the limited swath width. For SMMR there are no data
available poleward of 84°, and for SSM/I this limit is at 87°.
To reduce the amount of data to be processed the global data sets were filtered by selecting
only data from areas where ice is ever likely to occur. For this purpose a sea ice maximum
extend climatology from NSIDC has been used. Because this open water mask was close to
the ice edge in certain places the open water mask has been moved 100 km further away
from the ice edge. All land data more than 100km from the coast are skipped.
OSI SAF
15
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
4.1.3 Adding NWP data
NWP data from ECMWF (see also Chapter 2.3) were interpolated in time and space to the
position of each set of low resolution brightness temperatures. The following parameters
were collected:
●
surface temperature at 2 m height
●
total column water vapor
●
total column liquid water
●
wind speed and wind direction at 10m height
Total column liquid water is not used in the processing, but is kept for reference.
4.1.4 RTM correction and ice concentration calculations
The NWP model grid points are co-located with the satellite swath Tb's in time and space,
and a correction for water vapor and open water roughness, using Eq. 1, is applied. Over
fractions of both ice and water the influence of open water roughness on the Tb's and the ice
emissivity is scaled linearly with the ice concentration using iteration. The emissivity of ice is
given by standard tie-point emissivities. The correction procedure is described in detail in
Andersen et al. (2006B).
The ice concentration is calculated as part of the correction procedure. This gives ice
concentration estimates both for uncorrected and corrected brightness temperatures. The
uncorrected ice concentration estimates are only kept for reference. More details are given in
Chapter 2.3 and 3.1.
4.1.5 Uncertainty calculations
The representativeness error is computed as a function of ice concentration for each
algorithm and sensor using a model. The remaining error sources (including the instrument
noise, residual atmospheric noise, surface emissivity noise) are estimated in combination
using the hemispheric variability of the measurements. The total uncertainty is the sum of the
partial uncertainties squared as described in Eq. 4. The smearing uncertainty is zero for open
water and for 100 % ice. More details is given in Chapter 3.3 and Appendix E.
4.1.6 Dynamical tie-points
The L1/L2 processing is done in two steps. The two last steps are done twice, as shown in
Figure 4. This is necessary for the generation of the dynamical tie-points. First, all data must
be organized, a first tie-point set is generated based on uncorrected brightness temperatures
and corrected brightness temperatures are calculated. Secondly, a new tie-point set is
generated using the corrected brightness temperatures. With these tie-points the final ice
concentration calculations are made. More details are given in Chapter 3.2.
4.2 L3 processing
This main step contains the gridding of the swath data to daily fields, and calculation of
corrections and masking fields without applying these.
4.2.1 Daily gridding
The daily gridding searches for all observation within 24 hours, centered on 12:00 UTC, and
grids these to the final output grid. The observations within one grid cell are averaged, using
a weight defined as:
weight
OSI SAF
=1.0− d
i s
t
/ i n
f lr
ad
16

∗
0.3
,
(Eq. 5)
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
where dist is the distance between the observation and center of grid point, and inflrad is the
radius of influence. The influence radius depends on the sampling radius for the channel. For
low resolution channels (19, 22 and 37 GHz) 18 km is used and for 85GHz 9 km is used. The
85GHz is kept for internal use, and no ice concentration product based on this sensor is
provided.
The gridding is done for all areas with data coverage, including the coastal zone. A gridded
field is made for all the ice concentration estimates, based on both uncorrected and
corrected brightness temperatures, and for the algorithms present in the L2 files. A similar
gridding/averaging is applied for the uncertainty estimates.
In the averaging, observations from multiple satellite missions are available in overlapping
periods. During overlaps in the SSM/I period, observations from different satellites are
averaged. The dynamical tie-point approach handles the possible inter satellite differences,
which is shown in the validation report [RD.2]. Data during the overlap period between
SMMR and SSM/I data are not mixed, as these instruments are more different.
4.2.2 Coastal correction
Observations in the coastal zone are a mixture between land and water/sea ice. Land has
similar signatures to sea ice, and the algorithms therefore overestimate the sea ice
concentration in these areas. To correct for these so-called land spill-over effects, an extra
coastal correction step has been implemented. The method implemented is described in
detail in Cavalieri et al. (1999). In short, this method first calculates monthly average ice
concentration for all the months in a selected year and then finds the minimum ice
concentration from these averages. This minimum is then used to correct the ice
concentration values in the coastal zone if adjacent non-coastal grid points are ice free. The
minimum monthly average ice concentration fields were calculated using data from 1985 for
SMMR and 1992 for SSM/I. Separate fields were calculated for all algorithms processed.
This processing step generates an additional field with the coastal correction to be added to
the concentration field, without applying it.
4.2.3 Climatological maximum extent masking
To mask out erroneous ice outside areas where sea ice is ever likely to occur, a monthly
maximum extent climatology has been used. This climatology has been produced by NSIDC
using SMMR and SSM/I monthly averaged ice concentrations and finding the maximum
extent for each month between 1979 and 2007. A zone of 100 km has been added to the
maximum extent NSIDC maps to assure that the masks are surely outside the areas were
sea ice is ever likely to occur. Examples of these masks are given in Appendix B. More
details about these monthly climatological maximum fields (or ocean mask as called by
NSIDC) are available from NSIDC:
http://nsidc.org/data/smmr_ssmi_ancillary/ocean_masks.html.
This processing step collects the respective monthly mask and adds it to the file as a
separate layer.
4.2.4 T2m check
A quality check using the NWP T2m (air temperature at 2 meters) field is used in the
processing. The T2m NWP values that have been interpolated in time and space to each
observation in the add NWP step, are gridded and averaged to the product grid similar as for
the ice concentrations. This field is then added as a separate layer for later use. The test
applying the T2m value is based on experience from the operational OSI SAF chain, and
shows useful for removing gross errors far from the sea ice edge. In the reprocessing data
set it is not used for modifying the nominal value, just mark questionable data.
OSI SAF
17
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
4.3 L4 processing
This final main step contains filling of some areas with missing data by interpolation and
applying masks and corrections to present the final ice concentration product. The
processing flag variable (section 5.1.3) is computed during these steps.
4.3.1 Applying masks and coastal correction
A sea ice concentration analysis step is performed. It deals with applying the various
masking steps, the coastal correction and the T2m check (as described in sections 4.2.2,
4.2.3 and 4.2.4) are applied.
The nominal values of the ice concentrations and uncertainties are modified (except the T2m
test), and a variable holding the processing flags is created. The flags are described in
section 5.1.3.
4.3.2 Gap filling by interpolation
For easing the use of the reprocessing data set, it was decided that some level of spatial
interpolation should be performed for reducing the occurrence of gaps. Only missing data are
interpolated. Interpolated data points are clearly marked in the product file (section 5.1.3) so
that users can choose to discard them and only ingest retrievals that rely on satellite signal.
Data gaps can occur in several forms, such as missing scan lines, missing orbits and polar
observation hole. While spatial interpolation might be efficient in filling small gaps (e.g. one or
two missing scan lines), it necessarily blurs the sea ice concentration features. This effect
becomes overwhelming when large areas are missing. To overcome this issue, yet
implementing a general approach for all cases, the ice concentration estimates from the
previous and next daily products are used in the interpolation as well. In the case of SSM/I, it
means that interpolation on a given date D uses pixels from 3 data files: D-1, D and D+1.
The interpolated value at grid cell (i,j) for day D is given by:
D−1
D1
D1
D
D
X iD, j=K⋅w Di , −1
j ⋅X i , j w i , j ⋅X i , j ∑ W  k , l ; i , j X k , l  (Eq. 6)
k,l
where X is the sea ice concentration value and K is a normalizing factor given by:
D−1
D1
D
w i , j wi , j ∑ W k , l ; i , j=1/ K .
k,l
From Eq. 6, it is clear that the spatial interpolation from neighbors of cell (i,j) only uses values
from date D, while the temporal interpolation is only concerned with the value from the exact
(i,j) cell but from dates D-1 and D+1. This strategy ensures that the interpolation will be
efficient in the two following extreme scenarios. In a region where we never have satellite
observations (e.g. the polar observation hole in the Northern Hemisphere), the spatial
interpolation term will be the only contribution. Conversely, in the case of several missing
swath on date D only (nominal coverage on D-1 and D+1), the interpolated values will be
computed from the previous and next dates, taking advantage of the persistence of sea ice
concentration over such a short period. The interpolation for intermediate cases (when both
spatial and temporal neighbors exist) is a compromise of those extreme situations.
In Eq. 6, the weighting parameters are computed as follows:
wiD, j =1/ Di , j 2⋅ 2 N max 1
 k , l ; i , j 2 ,
D
D 2
W  k ,l ; i , j=1/ k ,l  ×exp −0.5⋅

Ri , j
OSI SAF
18
(Eq. 7)
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
where  is the standard deviation associated to each ice concentration estimate (section
3.3),  is the distance between a given (k,l) neighbor and cell (i,j) and R is an auto-correlation
radius. The spatial interpolation weight is thus based on an isotropic gaussian shape, and
almost all (>99.9%) of the interpolation weight is concentrated inside a [-3R;+3R] x [-3R;+3R]
km2 area, which translates into a [-Nmax;+Nmax] x [-Nmax;+Nmax] grid cells square area. It was
found that a spatially varying radius R was needed for optimal gap filling and the value R =
latitude of (i,j) (in degrees) was taken.
In the case of SMMR which was operated every second day, the interpolation is performed
with D-2 and D+2 instead of D-1 and D+1.
4.3.3 Final formatting
This step is only concerned with the final formatting of the data sets into a CF compliant
NetCDF file. This implies selection of a limited number of variables (ice concentration,
uncertainties, processing flags, latitude, longitude) for the selected algorithm and extraction
to a new file where global attributes are added. The format of this product file is described in
the next chapter.
Since SMMR and SSM/I are processed separately, there is a period with overlapping
products in July and August 1987. The quality is better for the SSM/I and should be preferred
during the overlapping period.
OSI SAF
19
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
5. Product description
This chapter gives a description of the product specification, meta data, data format and
product availability.
5.1 Product specification
The product consists of three major fields:
●
sea ice concentration
●
uncertainty estimate
●
processing flag
The definitions of these three fields are given in the sections below. These fields are given in
the same grid.
5.1.1 Sea ice concentration
Sea ice concentration indicates the areal fraction of a given grid point covered by ice. It is
given as a real number in percentage, with a range from 0-100%.
5.1.2 Uncertainty estimate
An estimate of the uncertainty of each sea ice concentration value is given in a separate
field. The uncertainty is given as standard deviation in percentage, with at range from 0100%.
5.1.3 Processing flag
The processing flag contains information about the processing steps that have influenced the
ice concentration value. It is coded as a signed character, with one flag value for each ice
concentration value. The different values are described in Table 4. These flags are not
additive, so all the values are exclusive (e.g. coastal correction and gap filling can not occur
for the same data point).
Value
Quality indication Definition
0
Nominal
1
Nominal value
questionable
2
10
11
12
101
Nominal value, but T2m check indicates possibility of false
ice due to high surface air temperature (> +10°C).
Nominal value, but quality questionable because of lake ice.
Coastal correction has been applied.
Nominal value
changed
Climatological maximum extent mask has been applied
inside data area.
As value 11, but outside of data coverage area.
Gap filling has been applied.
13
100
Nominal ice concentration value given.
Missing value
Missing value due to land.
Missing value due to missing data.
Table 4: Definition of sea ice concentration processing flags.
OSI SAF
20
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
5.2 Grid specification
The sea ice concentration product is available on two projections and grids, each with one
product for each hemisphere. The projections used are a Lambert Azimuthal Equal Area
projection with grid a resolution of about 12.5 km, and a Polar Stereographic projection with a
grid resolution of 10.0km. The Lambert grid is also called the EASE grid, and it is used by
NSIDC for several of their sea ice products. More documentation about the EASE grid can
be found on their web site: http://nsidc.org/data/ease/.
The details of the grid definitions are given in Table 5 and Table 6 below. Maps of the areas
are shown in Figure 5 and Figure 6. Projection definitions in the form of PROJ-4 initialization
strings are also given (see [http://www.remotesensing.org/proj] for details).
Projection:
Lambert Azimuthal Equal Area
Resolution:
12.5337625 km
Size:
849 columns, 849 lines
Central Meridian:
0°
Radius of Earth:
Spherical: 6371228.0 m
PROJ-4 string:
NH: +proj=laea +R=6371228.0 +lat_0=90 +lon_0=0
SH: +proj=laea +R=6371228.0 +lat_0=-90 +lon_0=0
Table 5: Geographical definition for the EASE 12.5 km grid, Northern and Southern
Hemisphere.
Projection:
Polar Stereographic
Resolution:
10.0 km
Size:
NH: 860 columns, 1120 lines
SH: 790 columns, 830 lines
Central Meridian:
-45°
Radius of Earth:
Elliptical: a=6378273.0m , b=6356889.44891 m
PROJ-4 string:
NH: +proj=stere +a=6378273 +b=6356889.44891 +lat_0=90 +lat_ts=70 +lon_0=-45
SH: +proj=stere +a=6378273 +b=6356889.44891 +lat_0=-90 +lat_ts=-70 +lon_0=0
Table 6: Geographical definition for the Polar Stereographic 10.0 km grid, Northern
and Southern Hemisphere.
5.3 Meta data specification
The meta data included in the product file are given as NetCDF attributes to the variables
and to the file (Global Attributes). Attributes associated to the variables are those required by
the CF convention. The Global Attributes have been selected for matching requirements from
the International Polar Year projects DAMOCLES and iAOOS-Norway (http://dokipy.met.no).
GCMD and IMO keywords were also selected. An ASCII version of the NetCDF header, as
given by ncdump, is given in Appendix A.
OSI SAF
21
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
Figure 5: Area covered by the EASE 12.5 km grids for Northern (left) and Southern
(right) hemispheres.
Figure 6: Area covered by the Polar Stereographic 10.0 km grids for Northern (left)
and Southern (right) hemispheres.
OSI SAF
22
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
5.4 File naming convention
The NetCDF product file have the following naming convention:
ice_conc_<area>_<proj>-<gridRes>_reproc_<date12>.nc ,
where:
<area>
:
<nh> (Northern Hemisphere) and <sh> (Southern Hemisphere)
<proj>
:
<ease> for the equal area Lambert azimuthal projection, <polstere>
for the Polar Stereographic projection.
<gridRes>
:
spatial resolution of the grid (<125> for 12.5 km, <100> for 100 km)
<date12>
:
central date of the analysis <YYYYMMDD1200>, e.g. 199112021200.
5.5 Product availability
The OSI SAF sea ice concentration reprocessing data set covers the period from 26.10.1978
to 23.10.2009. Some dates are missing due to lack of satellite data. These dates are listed in
Appendix C.
The data set is distributed freely through the OSI SAF Sea Ice FTP server, available at this
address:
ftp://saf.met.no/reprocessed/ice/conc/v1p1/
The data are organized in year and month directories. The SMMR data covers the period
from from 26.10.1978 to 09.07.1987 and the SSM/I data covers the period from 10.07.1987
to 23.10.2009. The overlapping SMMR data (11.07.1987 – 20.08.1987) are made available
for reference, and kept in the directory called “overlapping_smmr”.
5.6 Delivery of updates
The current version covers the period from 1978 to 2009. A continuation of this reprocessing
effort is planned. A service with monthly updates is expected in the OSI SAF CDOP-2 phase
at end of 2012.
OSI SAF
23
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
6. References
Andersen, S., L. Toudal Pedersen, G. Heygster, R. Tonboe, and L Kaleschke,
Intercomparison of passive microwave sea ice concentration retrievals over the high
concentration Arctic sea ice. Journal of Geophysical Research 112, C08004,
doi10.1029/2006JC003543, 2007.
Andersen, S., R. T. Tonboe and L. Kaleschke. Satellite thermal microwave sea ice
concentration algorithm comparison. Arctic Sea Ice Thickness: Past, Present and Future,
edited by Wadhams and Amanatidis. Climate Change and Natural Hazards Series 10,
EUR 22416, 2006A.
Andersen, S., R. Tonboe, S. Kern, and H. Schyberg. Improved retrieval of sea ice total
concentration from spaceborne passive microwave observations using Numerical
Weather Prediction model fields: An intercomparison of nine algorithms. Remote Sensing
of Environment104, 374-392, 2006B.
Cavalieri, D.J., C.L. Parkinson, P. Gloersen, J.C. Comiso, and H.J. Zwally. Deriving longterm time series of sea ice cover from satellite passive-microwave multisensor data sets.
Journal of Geophysical Research 104(C7), 15803-15814, 1999.
Comiso J.C, D.J. Cavalieri, C.L. Parkinson, and P. Gloersen. Passive microwave algorithms
for sea ice concentration: A comparison of two techniques. Remote Sensing of
Environment 60, 357-384, 1997.
Comiso J.C. Characteristics of arctic winter sea ice from satellite multispectral microwave
observations. Journal of Geophysical Research 91(C1), 975-994, 1986.
Gloersen, P., and F. T. Barath. A scanning multichannel microwave radiometer for Nimbus-G
and SeaSat-A. IEEE Journal of Oceanic Engineering OE-2(2), 172-178, 1977.
Gloersen, P., W. J. Campbell, D. J. Cavalieri, J. C. Comiso, C. L. Parkinson, H. J. Zwally.
Arctic and Antarctic sea ice, 1978-1987: satellite passive-microwave observations and
analysis. NASA SP-511, Washington D. C., 1992.
Kållberg, P., A. Simmons, S. Uppala, and M. Fuentes. The ERA-40 archive. ERA-40 Project
Report Series, ECMWF, Reading, 2004.
Meier, W. Scanning Multichannel Microwave radiometer (SMMR) reprocessing for
EUMETSAT. OSI SAF Visiting Scientist Report. 9 pages, 2008.
Njoku, E. G. Antenna pattern correction procedures for the scanning multichannel microwave
radiometer (SSMR). Boundary Layer Meteorology 18, 78-98, 1980.
Njoku, E. G., E. J. Christensen, and R. E. Cofield. The Seasat scanning multichannel
microwave radiometer (SMMR): Antenna corrections - development and implementation.
IEEE Journal of Oceanic Engineering OE-5(2), 125-137, 1980.
Smith, D. M. Extraction of winter sea ice concentration in the Greenland and Barents Seas
from SSM/I data. International Journal of Remote Sensing 17(13), 2625-2646, 1996.
Stark, J. (2008) Sea ice reanalysis using the OSI SAF sea ice processing system. OSI SAF
Visiting Scientist Report. 39 pages.
Toudal, L. (2006) Adaptation of SAFOSI sea ice processing system to southern hemisphere.
OSISAF Visiting Scientist Report.
Wentz, F. J. A model function for ocean microwave brightness temperatures. Journal of
Geophysical Research 88(C3), 1892-1908, 1983.
Wentz, F. J. A well-calibrated ocean algorithm for SSM/I. Journal of Geophysical Research
102(C4), 8703-8718, 1997.
Wentz, F. J. User's Manual, SSM/I Antenna Temperature Tapes, Revision 1. RSS Technical
Report 120191, 1991.
OSI SAF
24
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
Wentz, F. J. User's Manual, SSM/I Antenna Temperature Tapes, Revision 2. RSS Technical
Report 120193, 1993.
Wentz, F. J. User's Manual, SSM/I Antenna Temperature, Version 6. RSS Technical Memo
082806, 2006.
OSI SAF
25
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
7. Appendix A : Example of NetCDF format
Below is given an example of the NetCDF header of an OSI SAF sea ice concentration
reprocessing file (CDL format).
netcdf ice_conc_nh_ease-125_reproc_199112021200 {
dimensions:
time = 1 ;
nv = 2 ;
xc = 849 ;
yc = 849 ;
variables:
int Lambert_Azimuthal_Grid ;
Lambert_Azimuthal_Grid:grid_mapping_name =
"lambert_azimuthal_equal_area" ;
Lambert_Azimuthal_Grid:longitude_of_projection_origin = 0.f ;
Lambert_Azimuthal_Grid:latitude_of_projection_origin = 90.f ;
Lambert_Azimuthal_Grid:false_easting = 0.f ;
Lambert_Azimuthal_Grid:false_northing = 0.f ;
Lambert_Azimuthal_Grid:earth_radius = 6371228.f ;
Lambert_Azimuthal_Grid:semi_major_axis = 6371228.f ;
Lambert_Azimuthal_Grid:proj4_string = "+proj=laea +R=6371228.0
+lat_0=90 +lon_0=0" ;
double time(time) ;
time:long_name = "reference time of product" ;
time:standard_name = "time" ;
time:units = "seconds since 1978-01-01 00:00:00" ;
time:axis = "T" ;
time:bounds = "time_bnds" ;
double time_bnds(time, nv) ;
time_bnds:units = "seconds since 1978-01-01 00:00:00" ;
double xc(xc) ;
xc:axis = "X" ;
xc:units = "km" ;
xc:long_name = "x coordinate of projection (eastings)" ;
xc:standard_name = "projection_x_coordinate" ;
xc:grid_spacing = "12.5334 km" ;
double yc(yc) ;
yc:axis = "Y" ;
yc:units = "km" ;
yc:long_name = "y coordinate of projection (northings)" ;
yc:standard_name = "projection_y_coordinate" ;
yc:grid_spacing = "12.5334 km" ;
float lat(yc, xc) ;
lat:long_name = "latitude coordinate" ;
lat:standard_name = "latitude" ;
lat:units = "degrees_north" ;
float lon(yc, xc) ;
lon:long_name = "longitude coordinate" ;
lon:standard_name = "longitude" ;
lon:units = "degrees_east" ;
short ice_conc(time, yc, xc) ;
ice_conc:long_name = "concentration of sea ice" ;
ice_conc:standard_name = "sea_ice_area_fraction" ;
ice_conc:units = "%" ;
ice_conc:_FillValue = -32767s ;
ice_conc:valid_min = 0s ;
ice_conc:valid_max = 10000s ;
ice_conc:grid_mapping = "Lambert_Azimuthal_Grid" ;
ice_conc:coordinates = "lat lon" ;
ice_conc:scale_factor = 0.01f ;
short standard_error(time, yc, xc) ;
standard_error:long_name = "total uncertainty (one standard deviation)
of concentration of sea ice" ;
standard_error:standard_name = "sea_ice_area_fraction
standard_error" ;
standard_error:units = "%" ;
standard_error:_FillValue = -32767s ;
OSI SAF
26
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
standard_error:valid_min = 0s ;
standard_error:valid_max = 10000s ;
standard_error:grid_mapping = "Lambert_Azimuthal_Grid" ;
standard_error:coordinates = "lat lon" ;
standard_error:scale_factor = 0.01f ;
short algorithm_standard_error(time, yc, xc) ;
algorithm_standard_error:long_name = "algorithm uncertainty (one
standard deviation) of concentration of sea ice" ;
algorithm_standard_error:standard_name = "sea_ice_area_fraction
standard_error" ;
algorithm_standard_error:units = "%" ;
algorithm_standard_error:_FillValue = -32767s ;
algorithm_standard_error:valid_min = 0s ;
algorithm_standard_error:valid_max = 10000s ;
algorithm_standard_error:grid_mapping = "Lambert_Azimuthal_Grid" ;
algorithm_standard_error:coordinates = "lat lon" ;
algorithm_standard_error:scale_factor = 0.01f ;
short smearing_standard_error(time, yc, xc) ;
smearing_standard_error:long_name = "smearing uncertainty (one
standard deviation) of concentration of sea ice" ;
smearing_standard_error:standard_name = "sea_ice_area_fraction
standard_error" ;
smearing_standard_error:units = "%" ;
smearing_standard_error:_FillValue = -32767s ;
smearing_standard_error:valid_min = 0s ;
smearing_standard_error:valid_max = 10000s ;
smearing_standard_error:grid_mapping = "Lambert_Azimuthal_Grid" ;
smearing_standard_error:coordinates = "lat lon" ;
smearing_standard_error:scale_factor = 0.01f ;
byte status_flag(time, yc, xc) ;
status_flag:long_name = "status flag for sea ice concentration
retrieval" ;
status_flag:standard_name = "sea_ice_area_fraction status_flag" ;
status_flag:_FillValue = -128b ;
status_flag:valid_min = 0b ;
status_flag:valid_max = 101b ;
status_flag:grid_mapping = "Lambert_Azimuthal_Grid" ;
status_flag:coordinates = "lat lon" ;
status_flag:flag_values = 0b, 1b, 2b, 10b, 11b, 12b, 13b, 100b, 101b ;
status_flag:flag_meanings = "nominal t2m lake coastcorr maxclim_change
maxclim_set interpolated land missing" ;
status_flag:flag_descriptions = "\n",
" 0 -> nominal value from algorithm used\n",
" 1 -> t2m check indicates possibly false ice\n",
" 2 -> over lake caused possibly less accurate\n",
" 10 -> value changed by coast correction method\n",
" 11 -> value changed by applying maximum climatology\n",
" 12 -> missing value set by applying maximum climatology\n",
" 13 -> value set by applying interpolation\n",
"100 -> missing value due to over land\n",
"101 -> missing value due to missing data" ;
// global attributes:
:title = "OSI SAF Global Reprocessed Sea Ice Concentration" ;
:product_id = "OSI-409" ;
:product_name = "osi_saf_ice_conc_reproc" ;
:product_status = "offline" ;
:abstract = "The reprocessing of sea ice concentration is obtained
from\n",
"passive microwave satellite data over the polar regions.
It\n",
"is based on atmospherically corretected signal and an
optimal\n",
"sea ice concentration algorithm. This product is
available\n",
"for free from the EUMETSAT Ocean and Sea Ice
Satellite\n",
"Application Facility (OSI SAF)." ;
:topiccategory = "Oceans ClimatologyMeteorologyAtmosphere" ;
:keywords = "Sea Ice Concentration,Sea
Ice,Oceanography,Meteorology,Climate,Remote Sensing" ;
OSI SAF
27
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
:gcmd_keywords = "Cryosphere > Sea Ice > Sea Ice Concentration\n",
"Ocean > Sea Ice > Sea Ice Concentration\n",
"Geographic Region > Northern Hemipshere\n",
"Vertical Location > Sea Surface\n",
"EUMETSAT/OSISAF > Satellite Application Facility on Ocean and
Sea Ice, European Organisation for the Exploitation of Meteorological Satellites" ;
:northernmost_latitude = 90.f ;
:southernmost_latitude = 17.71581f ;
:easternmost_longitude = 180.f ;
:westernmost_longitude = -180.f ;
:activity_type = "Space borne instrument" ;
:area = "Northern Hemisphere" ;
:start_date = "1991-12-02 00:00:00" ;
:stop_date = "1991-12-03 00:00:00" ;
:project_name = "EUMETSAT OSI SAF" ;
:institution = "EUMETSAT OSI SAF" ;
:PI_name = "Rasmus Tonboe and Steinar Eastwood" ;
:contact = "osisaf-manager@met.no" ;
:distribution_statement = "Free" ;
:references = "Global Sea Ice Concentration Reprocessing Product User
Manual, Tonboe and Eastwood (editors), v1.1, October 2011\n",
"http://saf.met.no\n",
"http://www.osi-saf.org" ;
:history = "2010-01-13 creation" ;
:product_version = "1.0" ;
:software_version = "4.0" ;
:netcdf_version = "3.6.3" ;
:Conventions = "CF-1.4" ;
}
OSI SAF
28
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
8. Appendix
B:
Examples
of
monthly
climatological maximum extent masks
Figure 7: Climatological maximum sea ice
extent during March (upper) and September
(lower)
OSI SAF
29
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
9. Appendix C: Missing dates
The reprocessing data set covers the period from 26.10.1978 to 23.10.2009. During the
SMMR period only every second day is available. The SMMR data have been used until
11.07.1987. SMMR data are missing during November and December 1980, in addition to
shorter periods. Table 7 below lists the dates with no product due to lack of satellite data,
except the expected missing SMMR days (every second day).
Year
Missing dates
SMMR
1979
21/5-27/5
1980
4/1-10/1, 27/2-4/3, 16/3-21/3, 9/4-15/4, 1/1131/12
1981
27/2-5/3
1982
14/7-16/7, 30/7-1/8, 3/8-5/8, 15/8-17/8
1984
12/8-24/8
1985
22/9-28/9
1986
29/3-3/4, 5/4-9/4, 21/5-25/5, 27/5-6/6, 8/6-16/6,
8/12-10/12, 16/12-18/12
1987
3/1-15/1, 7/4-9/4
SSM/I
1987
25/08-26/08, 06/10-07/10, 03/12-31/12
1988
01/01-21/01, 06/05-08/05, 23/09, 25/12-27/12
1989
07/06, 23/07-24/07, 23/10
1990
13/08, 25/08-26/08, 21/10-22/10, 26/10-28/10,
22/12-26/12
1992
18/06
1993
04/01
1994
20/07, 20/11-21/11
2000
01/12
Table 7: Dates with no reprocessing product due to lack of satellite data. SMMR
(25.10.1978-20.08.1987) was operated every second day and the table shows the
periods with missing SMMR data for more than two days.
OSI SAF
30
Version 1.3 October 2011
SAF/OSI/CDOP/met.no/TEC/MA/138
Global Sea Ice Concentration Reprocessing PUM
10. Appendix D: Metadata List for EUMETSAT
Data Center
The EUMETSAT Data Center metadata parameters [RD.6] applicable to the reprocessed
OSI SAF Sea Ice Concentration product are listed in Table 8. These metadata parameters
are available in XML formatted files, one for each product file. These XML metadata files are
available through the EUMETSAT Data Center.
Short Name
Attribute Name
Notes
AAAR
Geographic Area
'NH' or 'SH'
AARF
Archive Facility
UMARF
Instrument ID
'SSM/I' or 'SMMR'
APAS
Product Actual Size
In bytes
APNM
Product Type
OSICOGBRE
ASTI
Satellite ID
Nimbus, F-08, F-10, F-11,F-13,
F-14 or F-15
AVBA
Base Algorithm Version
AVPA
Product Algorithm Version
GDMD
Disposition Mode
O = Operational
GGTP
Granule Type
DP = Data Product
GNFV
Native Product Format Version
GORT
Orbit Type
LEO
GPLV
Processing Level
O3
GPMD
Processing Mode
O = Offline
PPRC
Processing Center
OSNMI
PPST
Processing End Date and Time
QQOV
Overall Quality Flag
AIID
'OK' or 'NOK'
SNIT
Reference Time
SSBT
Sensing Start Date and Time
SSST
Sensing End Date and Time
Table 8: EUMETSAT Data Center metadata parameters applicable to the
reprocessed OSI SAF Sea Ice Concentration product.
OSI SAF
31
Version 1.3 October 2011
Figure 8: Ice concentration September 2005 derived from one day of SSMI
data
11
Appendix E: Ice concentration uncertainties
The uncertainties are calculated based on all the swath pixels contributing
to a level 3 ice concentration product (see figure 8).
11.1
Uncertainty implementation
We make the assumption the total uncertainty σtot can be written as
2
2
2
+ σsmear
σtot
= σalgo
(1)
where σalgo is the inherent uncertainty of the concentration algorithm, σsmear
is the uncertainty due to resampling to a grid where the sensor footprint covers more than one pixel.
11.1.1
Algorithm uncertainty
Ice concentration can be interpreted as a superposition of water and ice
iceconc = (1 − α(ic)) · water + α(ic) · ice
(2)
where ic is the ice concentration calculated by the algorithm.
The functional dependency between α and the calculated ice concentration
ic is described by:
E-1
ic ≤ 0
0 < ic < 1
ic ≥ 1
α=0
α = ic
α=1
which can be written as
α(ic) = Π0,1 (ic)ic + H(ic − 1)
(3)
where Πa,b (x) is the Boxcar function and H(x) the Heaviside step function.
Using equation 2 and assuming the uncertainty for the ice and water part
is independant this leads to a total algorithmic uncertainty as
q
2
2
σalgo (α(ic)) = (1 − α(ic))2 σwater
+ α2 (ic)σice
(4)
where σwater = σ(IC(Popenwater )) and open water is determined by a monthly
varying ocean mask, IC is the functional mapping of the ice concentration
algorithm and PC denotes the set of swath pixels for all swaths (used calculating the daily product) selected on the condition C.
σice = σ(IC(Pocean, nasateam>0.95 )) e.g. the standard deviation of the calculated ice concentration of those pixels clear of the coast having a NASA
Team concentration > 95%.
An example of the algorithm uncertainty is shown in figure 9
Figure 9: Algorithm uncertainty of the ice concentration product shown in
figure 8
E-2
Figure 10: MODIS 1 km channel 1 image
11.1.2
Smear uncertainty
The smear uncertainty is the error due to the sensor footprint covering more
than one pixel in the level3 product grid.
The error is calculated taking a cloud free 1 km MODIS image and assigning
ice concentrations to all pixels based on the channel 1 brightness. For each
pixel the corresponding brightness temperature is calculated for all relevant
microwave channels based on standard tiepoints (Comiso et al. 1997). Using channel specific sensor footprints for weighting the ice concentration is
calculated from the 1km brightness temperature image in the specified final
resolution i.e. 10, 12, 25 and 50 km. This ice concentration is compared
to the MODIS ice concentration regridded to the same resolution. The
standard deviation of the difference between these sets of ice concentration
values is the standard deviation of the smeared points.
The standard deviation of smeared points is denoted as
σsmear (sensor, resolution, algorithm) (abbreviated to σsmear (s, r, a)). As
the ice concentration approaches zero or 1 the spatial variation of the ice
cover diminishes thus the smearing has less effect. This is reflected in the
E-3
Figure 11: Resulting calculated OSI SAF ice concentration regridded to 1
km resolution
scaling:
σss (s, r, a, ic) =
Π0,σwater (ic)
ic
σwater
1 − ic
+ Πσwater ,1−σice (ic) + Π1−σice ,1 (ic)
σice
σsmear (s, r, a)
(5)
The use of scaling thresholds σwater and 1 − σice are somewhat arbitrarily
selected.
An example of the smear uncertainty is shown in figure 12
11.1.3
OSI SAF algorithm
The OSI SAF ice concentration is calculated as a linear combination of the
Comiso and Bristol algorithm which means the total smear uncertainty as
icosi (iccomiso , icbristol ) = (1 − β(iccomiso )) iccomiso + β(iccomiso )icbristol (6)
where the scaling factor β(iccomiso ) is
β(iccomiso ) = Π0,τ (iccomiso )
iccomiso
+ H(iccomiso − τ )
τ
E-4
(7)
Figure 12: Smear uncertainty of the ice concentration product shown in
figure 8
and τ a threshold value set to 0.4
11.1.4
OSI SAF uncertainty
Using equation 1 the total uncertainty is calculated as
q
2
2
σtotal−osi (s, r, osi, ic) = σalgo−osi
(ic) + σss−osi
(s, r, osi, ic)
(8)
An example of the smear uncertainty is shown in figure 14
11.2
Reported uncertainties
The reported uncertainties in the OSI SAF ice concentration product is the
total uncertainty σtotal−osi (s, r, osi, ic) (equation 8), the algorithmic uncertainty σalgo−osi (ic) (equation 4) and the smear uncertainty σss−osi (s, r, osi, ic)
(equation 5).
E-5
Figure 13: Total uncertainty of the ice concentration product shown in figure
8
Figure 14: Uncertainty components as a function of ice concentration for
the ice concentration product shown in figure 8
E-6