Nutrient variations in boreal and subarctic sea fluxes

Nutrient variations in boreal and subarctic sea fluxes
Linköping University Postprint
Nutrient variations in boreal and subarctic
Swedish rivers: Landscape control of land–
sea fluxes
Christoph Humborg, Erik Smedberg, Sven Blomqvist, Carl-Magnus Mörth, Jenni Brink, Lars
Rahm, Åsa Danielsson and Jörgen Sahlberg
N.B.: When citing this work, cite the original article.
Original publication:
Christoph Humborg, Erik Smedberg, Sven Blomqvist, Carl-Magnus Mörth, Jenni Brink, Lars
Rahm, Åsa Danielsson and Jörgen Sahlberg, Nutrient variations in boreal and subarctic
Swedish rivers: Landscape control of land–sea fluxes, 2004, Limnology & Oceanology, (49),
5, 1871-1883.
http://aslo.org/lo/toc/vol_49/issue_5/1871.pdf.
Copyright: American Society of Limnology and Oceanology, http://www.aslo.org/lo/
Postprint available free at:
Linköping University E-Press: http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-12551
Limnol. Oceanogr., 49(5), 2004, 1871–1883
q 2004, by the American Society of Limnology and Oceanography, Inc.
Nutrient variations in boreal and subarctic Swedish rivers: Landscape control of land–
sea fluxes
Christoph Humborg,1 Erik Smedberg, and Sven Blomqvist
Department of Systems Ecology, Stockholm University, SE-106 91 Stockholm, Sweden
Carl-Magnus Mörth and Jenni Brink
Department of Geology and Geochemistry, Stockholm University, SE-106 91 Stockholm, Sweden
Lars Rahm and Åsa Danielsson
Department of Water and Environmental Studies, Linköping University, SE-581 83 Linköping, Sweden
Jörgen Sahlberg
Swedish Meteorological and Hydrological Institute, SE-601 76 Norrköping, Sweden
Abstract
We examined the hypothesis that the extent of vegetation cover governs the fluxes of nutrients from boreal and
subarctic river catchments to the sea. Fluxes of total organic carbon (TOC) and dissolved inorganic nitrogen,
phosphorus, and dissolved silicate (DIN, DIP, and DSi, respectively) are described from 19 river catchments and
subcatchments (ranging in size from 34 to 40,000 km 2) in northern Sweden with a detailed analysis of the rivers
Luleälven and Kalixälven. Fluxes of TOC, DIP, and DSi increase by an order of magnitude with increasing proportion of forest and wetland area, whereas DIN did not follow this pattern but remained constantly low. Principal
component analysis on landscape variables showed the importance of almost all land cover and soil type variables
associated with vegetation, periglacial environment, soil and bedrock with slow weathering rates, boundary of upper
tree line, and percentage of lake area. A cluster analysis of the principal components showed that the river systems
could be separated into mountainous headwaters and forest and wetland catchments. This clustering was also valid
in relation to river chemistry (TOC, DIP, and DSi) and was confirmed with a redundancy analysis, including river
chemistry and principal components as environmental variables. The first axis explains 89% of the variance in river
chemistry and almost 100% of the variance in the relation between river chemistry and landscape variables. These
results suggest that vegetation change during interglacial periods is likely to have had a major effect on inputs of
TOC, DIP, and DSi into the past ocean.
Changes in vegetation cover on land have altered biogeochemical cycles by affecting both silicate weathering rates
1
Present address: Institute of Applied Environmental Research,
Stockholm University, SE-106 91 Stockholm, Sweden ([email protected]).
Acknowledgments
We are grateful to Vattenfall AB (especially Hans Lindmark) for
generous field support. The staff of the Tarfala Research Station and
Nils and Olof Sarri at Nikkaluokta kindly assisted us with the field
work. Water samples of TOC and nutrients (DIN, DIP) were analyzed by Anders Sjösten, Karin Wallman, and Tomas Thillman at
the Department of Systems Ecology, Stockholm University. We are
also grateful to Curt Broman for assistance with the bedrock classification scheme and to Ragnar Elmgren for his generous help with
linguistic improvements of the manuscript. Two anonymous reviewers greatly improved the manuscript with their thoughtful and very
helpful comments.
The Swedish Natural Science Research Council (NFR), the
Swedish Research Council (VR), the Swedish Research Council for
the Environment, Agricultural Sciences and Spatial Planning (FORMAS), and the European Commission (R&D priority Sustainable
Marine Ecosystems contract EVK3-CT-2002-00069) funded the
project. This is a contribution of the Scientific Committee on Problems of the Environment (SCOPE) project on ‘‘Land–Ocean Nutrient Fluxes: The Silica Cycle.’’
and carbon sequestration on land (Schwartzman and Volk
1989; Berner 1992). How these altered weathering patterns
have been manifested in land–sea fluxes of nutrients is still
not clear. Natural changes in river runoff of nutrients might
have influenced the primary production regime of the entire
ocean environment, and thus the CO2 concentration of the
atmosphere. Both dissolved silicate (DSi) and dissolved inorganic nitrogen (DIN) inputs into the contemporary ocean
are estimated to be between 5 and 6 Tmol yr21 (Tréguer et
al. 1995; Jaffe 2000). Thus, the Si : N ratio of total annual
input (i.e., loads from rivers and the atmosphere, including
N fixation) corresponds to the molar uptake demand ratio of
diatoms, which is ;1 (Brzezinski 1985). An improved understanding of changes in inputs of DSi will help explain
the variation of diatom production over geological timescales. About 90% of the DSi input to the global ocean is
estimated to come from rivers. However, in global biogeochemical budgets and models (Tréguer et al. 1995; Tréguer
and Pondaven 2000), river inputs to the ocean are still considered as having been constant during the late Quaternary.
Instead, it was argued that increased input of aeolian silicate
(Tréguer and Pondaven 2000) or fertilizing iron (Falkowski
1997) to the ocean have been responsible for an increased
CO2 uptake via the biological pump during glacial periods.
Today, arctic rivers contribute roughly a sixth (5,500 km3)
1871
1872
Humborg et al.
of the global annual water discharge to the ocean (Shiklomanov et al. 2000). However, during deglaciation periods,
these contributions should have been much more significant
because of the massive release of melting water over extended areas from about 708N down to 408N (Andersen and
Borns 1997). A prominent example is the St. Lawrence River, which is supposed to have influenced the entire North
Atlantic during the Younger Dryas (Broecker et al. 1989).
Still, the biogeochemical significance of glacial periodicity
to such river systems is unclear.
Most studies on land indicate an increase of chemical
weathering and, thus, of DSi fluxes with an increase in vegetation cover (Berner and Berner 1996), suggesting that the
highest weathering rates can be expected during an interglacial period. On the other hand, from analyses of marine sedimentary paleoceanographic records, it has been inferred that
during glacial periods, DSi fluxes might have been twice as
high as today because of an increase in physical weathering
(Froelich et al. 1992).
Vegetation might be a crucial force in the downstream
change of nutrients and major elements in boreal watersheds,
as indicated by a positive correlation between the two river
chemistry variables of total organic carbon (TOC) and DSi
found during winter base flow in a recent study of some
headwater lakes and streams of northern Sweden (Humborg
et al. 2002). Previous field studies in small tropical watersheds (Cochran and Berner 1996; Alexandre et al. 1997;
Oliva et al. 1999), and especially in weathering limited environments (sensu Drever 1997), indicate that vascular plant
vegetation appears to be a major factor affecting the silica
weathering rate (Drever and Zobrist 1992; Drever 1994; Anderson et al. 2000). In contrast, a negative correlation between vegetation cover, dissolved organic carbon (DOC),
and weathering products has been reported by Engstrom et
al. (2000) for small lakes in a recently deglaciated terrain in
Alaska. However, most of these studies investigating the effect of vegetation on weathering are small in scale (i.e., deduced from soil profile analyses or from investigations in
watersheds ,100 km 2), and differences in weathering patterns might be explained by local geochemical and hydrological conditions. Thus, there is still a lack of large-scale
analyses of how river chemistry is related to landscape variables in boreal and subarctic watersheds that have undergone repeated changes in vegetation cover during glacial cycles (Kohfeld and Harrison 2000) and that could potentially
have a major impact on land–sea nutrient fluxes globally.
Moreover, an evaluation of potential causal mechanisms regulating river chemistry is also missing. The areas of our
study catchments ranged from 34 to 40,000 km 2, allowing
us to test whether an effect of vegetation cover on weathering rates can be found over various spatial catchment
scales and is, therefore, also displayed in nutrient land–sea
fluxes. We use physical geographic, hydrological, and biogeochemical information from northern Swedish river catchments to study how land–sea fluxes of nutrients from boreal
and subarctic rivers are regulated.
Materials and methods
Investigated area—Our study encompasses the northern
Swedish river systems of Torneälven, Kalixälven, Råneäl-
ven, Luleälven, Piteälven, Skellefteälven, and Umeälven,
which are ideal case studies with sufficient background data
for investigating the significance of vegetation cover on nutrient land–sea fluxes. All studied rivers drain into the northern Gulf of Bothnia (Baltic Sea) and have their headwaters
in the mountainous area close to the Norwegian border, except Råneälven, which originates in the eastern forested region (Fig. 1A). The investigated catchments are located between 648 and 698N, several originating above the Arctic
Circle. The whole area is sparsely inhabited and can, apart
from regulation of the rivers Luleälven, Skellefteälven, and
Umeälven, be characterized as relatively unperturbed. In the
Scandinavian mountain range (the Scandes), the climate is
typically subarctic, with continuous frost from mid-October
to May (Ångström 1974). Westerly winds prevail and deliver
1,000 to 2,000 mm yr21 of precipitation to the headwater
areas (Carlsson and Sanner 1994). Cold climate favors boreal biomes (taiga, tundra) as typical also for the large Siberian and Canadian rivers (Kohfeld and Harrison 2000).
The spatial vegetation gradient found in our investigated watersheds—from essentially unvegetated rock at the top elevations, through alpine pasture and deciduous brushwood/
short deciduous forest, to coniferous forest and mires in the
river catchments—represents dominating biotopes of environments deglaciated in the Holocene. The upper mountainous headwater areas (summits below 2,200 m above sea level) are sparsely vegetated or even barren (Table 1), generally
with thinner soil thickness than in the forested landscape
downstream. The tree line is found at 650–800 m above sea
level. Two of the rivers, Kalixälven and Luleälven (Fig. 1B),
have been studied in greater detail.
Sampling strategy—Since 1970, the rivers of the present
study have been routinely monitored (monthly, n 5 360) for
biogeochemical variables at their mouths by the Swedish
University of Agricultural Sciences (SLU, Department of
Environmental Assessment). Our own field samplings were
carried out in 1999–2001 in the rivers Kalixälven (n 5 8)
and Luleälven (n 5 9), where several subcatchments, including streams as well as inlets and outlets of natural lakes
and reservoirs, were investigated in detail. The water discharge of these high-latitude rivers is dominated by the
spring flood in late May to early July, when roughly half of
the annual water discharge occurs. We therefore used discharge-weighted mean concentrations (Table 1) of the various constituents to evaluate empirical relationships between
physical geographical data and concentration records of dissolved constituents. We chose to sample early winter conditions in November and December and late winter conditions in March, April, and May. We also sampled the spring
flood from snowmelt in late May to early July. Summer conditions were investigated during August and September. The
sampled catchments, streams, rivers, lakes, and reservoirs
are indicated in Fig. 1A,B. The mire-draining stream Muddusätno (Fig. 1B) was sampled monthly (n 5 192) by the
SLU.
Analytical methods—We sampled water directly at the water surface using syringes that were prewashed in 0.1 M
hydrochloric acid before sampling. The water samples were
Landscape control of nutrient fluxes
1873
Fig. 1. (A) Catchment areas and land cover characteristics of the investigated major river systems in northern Sweden. (B) Detail of the headwater area of the rivers Kalixälven and Luleälven,
showing location of sampling sites, subcatchment areas, and land cover characteristics.
filtered immediately through prewashed cellulose membrane
filters (0.45 mm, Milliporet), and the filtrates were collected
in carefully cleaned polypropylene and polyethylene tubes
and bottles. All containers used for collecting TOC and nutrient samples were prewashed with ultraclean water, and for
inductively coupled plasma–optical emission spectrometry
(ICP-OES) analysis, they were acid washed. The water samples for ICP analysis were immediately preserved by adding
1 ml suprapur concentrated HNO3 to each 100 ml of sample.
All syringes, tubes, and bottles were precleaned with ultraclean water (ELGAt systems, Maxima Analytica and Spectrum RO1).
The concentration of DIN and dissolved inorganic phosphorus (DIP) was determined colorimetrically by means of
flow injection analysis (Lachatt; see Ranger 1993) or segmented flow analysis (Flow Solutiont IV, Alpkem Corpo-
1874
Humborg et al.
Table 1. Catchment areas, land cover, water discharge, and discharge-weighted annual mean concentration of total organic carbon (TOC)
and dissolved inorganic nutrients (DIN, DIP, DSi; b.d.l., below detection limit) in 19 river systems of northern Sweden. Runoff and river
chemistry data from monthly measurements monitored at the river mouths by the Swedish University of Agricultural Sciences, Uppsala,
Sweden.
River
Torneälven§
Kalixälven§
Tarfalajåkka
Ladtjojåkka 1
Ladtjojåkka 2
Råneälven§
Luleälven§
Upmasjåkka
Sårkajåkka
Valtajåkka
Ritsem hydropower plant
Vuojatätno
Suorva dam
Sjaunjaätno
Porjus hydropower plant
Muddusätno\
Piteälven§
Skellefteälven§
Umeälven§
Nutrient (mmol L21)
Catchment
area
(km 2)
Deciduous
UnLake
vegConif- brushWetand Snow etated Open erous wood/
land stream and ice rock land forest forest Runoff
(%)
(%)* (%)* (%)
(%)
(%)† (%)‡ (km3)
TOC
DIN
DIP
DSi
39,865
17,994
34
200
298
4,176
25,095
306
305
147
973
2,832
5,603
849
9,828
450
11,221
11,671
26,684
10.3
19.3
0.0
0.3
0.2
26.5
8.7
0.8
0.0
2.0
0.1
0.1
0.2
27.7
4.6
38.9
8.9
9.4
7.9
540
438
29
—
—
579
203
88
47
82
43
39
71
360
—
—
320
321
325
3.9
6.7
1.8
4.5
1.9
3.0
3.1
1.9
1.6
1.4
2.6
2.6
2.3
2.1
1.0
3.4
3.6
3.4
3.5
0.16
0.15
0.19
0.02
0.02
0.12
0.07
0.02
0.02
b.d.l.
b.d.l.
b.d.l.
b.d.l.
0.02
0.02
0.08
0.09
0.06
0.07
86.5
82.2
22.5
53.6
66.2
97.2
39.3
17.0
9.7
30.6
12.0
15.7
13.2
73.2
22.4
78.4
66.3
39.9
46.8
4.4
2.9
2.0
2.8
2.9
2.8
7.8
16.1
3.6
0.5
9.3
9.4
11.8
4.0
11.0
2.1
6.3
11.8
6.4
1.5
1.4
47.5
17.6
15.0
0.0
5.9
11.4
31.2
11.5
29.0
16.2
17.3
0.0
10.5
0.0
1.3
1.4
1.5
0.3
0.2
2.3
2.5
3.3
0.0
0.4
0.3
0.5
0.4
0.8
0.4
0.7
0.0
0.6
0.0
0.8
0.4
0.4
27.0
21.2
50.2
70.6
66.1
14.1
41.5
61.5
61.4
70.8
60.8
72.9
67.0
39.2
59.4
6.9
32.8
34.1
42.0
41.5
39.5
0.0
0.7
1.4
46.6
28.4
1.8
0.2
1.1
0.0
0.2
0.5
21.7
8.8
49.5
42.3
34.4
32.1
15.0
15.2
0.0
5.6
11.1
9.9
7.2
8.1
3.1
13.6
0.0
0.7
2.6
7.4
5.1
2.6
7.4
8.2
9.5
13.89
10.28
0.02
0.12
0.17
1.47
17.50
0.31
0.36
0.07
1.26
3.43
4.86
0.38
8.19
0.19
5.93
5.86
15.22
* Recorded by satellite at the end of June/early July 1999.
† Mainly Norway spruce (Picea abies) and Scots pine (Pinus sylvestris).
‡ Mainly short deciduous forest/deciduous brushwood of birch (Betula spp.) and willow (Salix spp.).
§ Runoff and river chemistry data from monthly measurements monitored at the river mouths by the SLU between 1970–2000.
\ Runoff and river chemistry data from monthly measurements monitored at the river mouths by the SLU between 1988–2000.
ration). DSi was determined by ICP-OES (Varian Vistat Pro
Ax). TOC in water samples was determined by a high-temperature combustion technique (Shimadzut TOC-5000).
Runoff estimates and discharge-weighted mean concentrations—To be able to calculate discharge-weighted TOC
and nutrient concentrations, we estimated the water discharge of the various subcatchments of the rivers Kalixälven
and Luleälven (Fig. 1B) using a semidistributed conceptual
runoff model, the HBV model (Lindström et al. 1997). Here,
semidistributed means that a basin can be separated into a
number of subbasins and that each one is distributed according to altitude and type of vegetation in the catchment
area. Each subbasin can then be further divided into a number of elevation zones, depending on the height variation of
the subbasin. The elevation zone was in turn divided into
three vegetation classes: forest, open land, and lakes.
The standard routine of the HBV model is based on a
degree-day approach involving ambient air temperature as
well as water-holding capacity of snow; the latter results in
varying delays in the runoff. Primary input data are daily
mean values of precipitation, air temperature, and evapotranspiration rate in each subbasin, respectively. The model
performance was measured over both a calibration period
and a validation period of independent data. The main criterion of the model performance is the R 2 value, calculated
according to Nash and Sutclife (1970). The R 2 value was
0.927 for the calibration and 0.890 for the validation periods.
The difference in R 2 values reflects a more variable measured flow during the validation period. The high validation
R 2 value shows that the HBV model predicts runoff well.
Land cover, soil type, and bedrock—Calculations of percent land cover and of soil and bedrock types for each drainage area were performed by ARC Viewt 8.1 (ESRI). The
drainage basin boundaries were obtained from the Swedish
Meteorological and Hydrological Institute. Table 1 summarizes land cover, and Tables 2, 3 soil and bedrock types of
the investigated catchments and subcatchments, respectively.
The data used for land cover calculations were compiled
from satellite images, with a spatial resolution of 150 3 150
m, and various reference data sets provided by Metria Miljöanalyst. Soil and bedrock types have been compiled from
geological maps of the Swedish Geological Survey.
Slight differences in lake area percentages between Table
1 and Tables 2, 3 arose because the lake areas given in the
land cover and soil type maps were obtained by different
methods (i.e., by means of satellite information recorded at
the end of June/early July in 1999 [Table 1] and by airplane
photography and field mapping of consecutive years [Table
2, 3], respectively). Moreover, different filling stages further
resulted in various surface areas of the lakes/reservoirs and
Landscape control of nutrient fluxes
1875
Table 2. Soil types (and percent area of glacier and lake and stream) in the investigated catchments.
River
Torneälven§
Kalixälven§
Tarfalajåkka
Ladtjojåkka 1
Ladtjojåkka 2
Råneälven
Luleälven
Upmasjåkka
Sårkajåkka
Valtajåkka
Ritsem hydropower plant
Vuojatätno
Suorva dam
Sjaunjaätno
Porjus hydropower plant
Muddusätno
Piteälven
Skellefteälven
Umeälven
GlacioTill below Till above
Thin soils fluvial Sand and Silt and
Lake and
Catchment tree line tree line
and bare sediment gravel
clay
Glacier stream
area (km 2)
(%)
(%)
Peat (%) rock (%)
(%)
(%)
(%)
(%)
(%)
39,865
17,994
34
200
298
4,176
25,095
306
305
147
973
2,832
5,603
849
9,828
450
11,221
11,671
26,684
34.9
45.6
0.0
1.7
5.3
52.2
35.4
2.5
1.9
14.2
0.0
0.9
1.6
47.7
16.8
40.8
54.9
54.4
59.2
11.8
7.4
30.4
41.6
48.7
0.0
12.6
19.0
0.3
4.7
24.9
24.2
20.7
0.6
19.5
0.0
5.5
5.2
12.2
streams studied (cf. Tables 1–3). ‘‘Snow and ice’’ (Table 1)
indicates high-altitude areas of surface draining from snowmelt during summer, whereas ‘‘Glacier’’ (Table 2) indicates
strictly glaciated areas, only. Note, that ‘‘Unvegetated rock’’
(Table 1) indicates truly hard rock outcrops, whereas ‘‘Thin
soils and bare rock’’ (Table 2) includes soils up to 100 cm
thick as well.
The bedrock is usually covered by till (Fredén 1994),
22.7
23.3
0.0
0.0
0.0
29.0
10.9
0.0
0.0
0.8
0.0
0.2
0.1
35.3
6.7
42.0
7.5
10.5
7.5
12.5
9.6
43.9
44.7
36.4
5.7
25.3
61.7
97.6
67.7
55.7
57.8
59.4
6.5
40.0
4.6
16.6
15.0
10.3
13.1
9.7
8.4
1.9
1.7
8.3
5.6
0.0
0.0
9.5
2.2
3.2
2.2
9.8
4.2
12.7
7.3
4.0
4.3
1.7
2.4
0.0
5.9
5.1
2.5
2.1
0.0
0.0
2.6
0.0
1.2
0.7
0.0
0.6
0.0
2.7
1.8
3.1
0.4
0.6
0.0
0.0
0.0
1.4
1.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.0
0.0
1.3
1.5
0.7
0.1
0.2
17.3
4.2
2.8
0.0
0.7
0.0
0.0
0.0
0.0
1.5
0.9
0.0
0.6
0.0
0.2
0.0
0.0
2.7
1.2
0.0
0.0
0.0
0.8
6.3
16.7
0.2
0.5
17.2
11.1
14.3
0.0
11.5
0.0
4.1
7.5
2.7
sometimes several tens of meters thick. In all, seven soil
types were recognized (Table 2). The dominant soil type, till
(Table 2; see ‘‘Till below tree line’’ and ‘‘Till above tree
line,’’ respectively) reflects the regional bedrock composition, which is dominated by minerals like quartz, plagioclase, microcline, micas, and amphiboles.
Bedrock types (Table 3) were grouped according to weathering properties: carbonate rocks and carbonate-rich shale
Table 3. Bedrock types in the investigated catchments. The bedrock types were classified according to their chemical influence on water
chemistry (see text).
River
Torneälven§
Kalixälven§
Tarfalajåkka
Ladtjojåkka 1
Ladtjojåkka 2
Råneälven
Luleälven
Upmasjåkka
Sårkajåkka
Valtajåkka
Ritsem hydropower plant
Vuojatätno
Suorva dam
Sjaunjaätno
Porjus hydropower plant
Muddusätno
Piteälven
Skellefteälven
Umeälven
Catchment
area (km 2)
Shale
(%)
Carbonate-rich
shale
(%)
39,865
17,994
34
200
298
4,176
25,095
306
305
147
973
2,832
5,603
849
9,828
450
11,221
11,671
26,684
11.6
12.3
0.0
0.0
0.0
12.8
7.0
24.8
38.3
2.4
24.9
19.4
17.8
0.3
10.2
0.6
9.4
14.9
23.2
0.0
0.0
0.0
0.0
0.0
0.0
2.9
11.3
15.7
20.8
9.1
16.8
13.3
0.0
7.7
0.0
0.1
0.1
2.6
Carbonate
rock
(%)
Sandstone
(%)
Quartzite
(%)
Gneiss
(%)
0.9
0.1
0.0
0.0
0.0
0.1
1.4
0.0
0.0
0.0
2.0
7.2
4.0
0.0
2.4
0.0
1.0
0.5
0.9
0.4
0.1
0.0
0.0
0.0
0.9
3.2
0.0
0.0
0.0
0.0
0.0
0.0
0.0
1.4
0.0
2.4
1.7
0.1
3.4
1.6
6.1
10.1
7.6
0.0
5.3
2.6
1.7
3.3
15.1
3.9
4.9
0.0
4.3
0.0
9.2
10.1
9.7
23.0
3.2
23.4
41.2
39.7
0.0
11.4
44.2
35.0
64.6
7.9
12.4
25.0
0.0
18.0
0.0
5.0
6.9
14.6
Granite
and acid Lake and
Alkaline volcanic stream
rock (%) rock (%)
(%)
14.6
17.6
61.5
40.9
37.9
13.1
14.2
0.4
6.4
0.0
2.0
28.2
16.8
6.6
15.9
21.9
8.0
7.6
6.5
43.4
63.8
9.0
7.8
14.8
72.3
48.3
0.0
2.6
8.4
21.8
1.1
3.9
93.1
28.6
77.5
60.9
50.6
39.7
2.7
1.2
0.0
0.0
0.0
0.8
6.3
16.7
0.2
0.5
17.2
11.1
14.3
0.0
11.5
0.0
4.1
7.5
2.7
1876
Humborg et al.
Table 4. Results from PCA on landscape variables of 19 river catchments in northern Sweden.
Explained variance and loadings for each component and respective landscape variable (Tables 1–
3) are reported. Loadings mainly controlling the component (.z0.5z) are marked in bold.
PC1
Explained variance (%)
PC2
PC3
PC4
PC5
43.2
17.8
11.6
9.3
4.6
Land cover variable
Wetland
Lake and stream
Snow and ice
Unvegetated rock
Open land
Coniferous forest
Deciduous forest
20.940
0.317
0.635
0.435
0.891
20.847
20.218
20.149
20.358
0.440
0.832
0.113
20.206
20.264
20.127
0.355
20.177
0.020
20.233
0.346
0.030
20.003
20.170
20.576
0.218
0.060
0.157
0.735
20.104
0.586
20.114
20.104
0.249
20.153
20.427
Soil type variable
Till below treeline
Till above treeline
Peat
Thin soils and bare rock
Glaciofluvial sediment
Sand and gravel
Glacier
Silt and clay
20.757
0.529
20.952
0.870
20.788
0.128
0.153
20.348
20.273
0.721
20.149
20.142
0.040
0.369
0.796
20.211
0.439
2.050
20.155
20.309
20.210
0.198
20.050
0.819
0.233
0.231
0.020
20.282
0.040
0.831
20.411
0.137
20.162
0.282
20.104
20.010
20.312
20.104
20.240
20.050
0.412
0.640
0.184
20.184
0.436
0.759
0.020
20.617
20.498
20.080
20.119
0.313
0.020
0.929
0.157
20.408
20.080
0.782
0.563
20.363
20.150
20.341
20.165
20.144
0.040
0.090
0.305
20.109
0.150
0.175
0.862
0.020
0.196
20.342
0.010
20.924
20.134
0.204
0.080
20.107
Bedrock type variable
Shale
Carbonate-rich shale
Carbonate rock
Sandstone
Quartzite
Gneiss
Alkaline rock
Granite and acid volcanic
rock
(high weathering rate); shale and alkaline rocks (intermediate weathering rate); gneiss, granite, and acid volcanic rock
and sandstone (low weathering rate); and quartzite (very low
weathering rate). Sandstone and quartzite were chosen as
two separate classes because of the higher content of feldspar (plagioclase and microcline) in the sandstone (Björklund 1989) and the higher weathering resistance of the
quartzite. Some of the rock types can also contain weathering-resistant clay minerals, especially the shales, which
might affect weathering. However, the weathering rates of
clay minerals are much less known than those of primary
silicate minerals.
Statistical analysis—We used principal components analysis (PCA) and cluster analysis to distinguish landscape
characteristics (land cover, types of soil and bedrock) of the
river catchments. In addition, multiple linear regression and
redundancy analysis (RDA) were performed to statistically
test the relationship between river chemistry (TOC and dissolved inorganic nutrients) and landscape variables.
PCA is a common multivariate technique used for data
reduction. It ordinates a number of correlated variables along
orthogonal principal components, in which the principal
components describe the variation within the original data
in descending order. In practice, this commonly means that
only the first few principal components are needed to de-
scribe the major part of the original variation. Each of these
principal components is a weighted linear combination of all
variables, in which the weight of a variable is called loading.
These loadings indicate which variables are primarily responsible for describing the principal components (i.e., those
with largest positive or negative loadings have a larger influence than those with small ones; Table 4). Principal scores
for each river catchment were calculated as the sum of individual variable loadings (Table 4) multiplied by the area
percentage of the corresponding variables in the river catchments (Tables 1–3). Thus, a principal score was calculated
for each river catchment and principal component.
Cluster analysis was used to divide river catchments into
clusters on the basis of their similarity in terms of all landscape variables. For this study, a hierarchical method, Ward’s
minimum variance clustering algorithm (Ward 1963), was
used. At each iterative step, the clustering routine joins the
two clusters that minimize the variance within the cluster
and simultaneously maximize the variance between clusters.
Clustering was performed on the resulting principal components, similar to that conducted for other environmental
data (Danielsson et al. 1999).
We used two independent approaches to statistically test
the role of vegetation in regulating river chemistry: multiple
linear regression and RDA. Multicollinearity among environmental variables is a common problem in these types of
Landscape control of nutrient fluxes
analyses and could cause unstable regression and canonical
coefficients. With the use of the orthogonal principal components of the landscape variables, this problem is avoided.
Multiple linear regression analysis offers the opportunity to
test the linear relationship between a single river chemistry
variable (TOC, DSi, DIP, and DIN) and the principal components. RDA is an ordination method applied for multivariate classification, which analyzes linear relationships between several river chemistry variables and principal
components (derived from a PCA on landscape variables)
simultaneously. Linearity between these variables was tested
prior to RDA with the use of detrended canonical correspondence analysis.
The significance of each RDA ordination axis is statistically tested by unrestricted Monte Carlo permutations with
forward selection to identify which principal components explained significant (P , 0.05) amounts of variation in river
chemistry (Ter Braak 1986). The final result is presented in
an ordination diagram in which the main pattern of variation
in river chemistry and principal components (derived from
a PCA on landscape variables) are presented as arrows, together with the distribution of river catchments. Rivers found
near the head of an arrow are strongly and positively correlated with that variable (and vice versa). All analyses were
made in CANOCO (vers. 4.5; Ter Braak 1990). Some DIP
values below the detection limit (see Table 1) were replaced
by half their detection limit (0.008 mmol L21) in order to
permit their use in the statistical analysis.
Results
PCA and cluster analysis on landscape variables—The
landscape characteristics were highly correlated, with 86.5%
of the variance explained along the first five principal components (PCs). Table 4 reports the loadings for the various
landscape variables within the five PCs. PC1, which describes 43% of the data variance, is characterized by high
loadings of peat, wetland, granite and acid volcanic rock,
coniferous forest, glaciofluvial sediment, and till below the
tree line, but low loadings of snow and ice area, carbonaterich shale, gneiss, thin soils and bare rock, and open land
area (Table 4). Thus, almost all land cover and soil type
variables associated with vegetation are linked to PC1, and
a high negative score for this component corresponds to
catchments with high cover of vegetation. The second principal component (PC2), which describes 18% of the data
variance, is described by high loadings of alkaline rock, unvegetated rock, glacier, and till above the tree line, as is
typical for periglacial environments in the area. PC3, which
describes 12% of the data variance, is described by high
loadings of silt and clay, sandstone, and quartzite (i.e., dominated by soil and bedrock types that have low weathering
rates). Sand and gravel and deciduous brushwood/forest are
the major constituents of PC4, which describes 9% of the
data variance, is interpreted as typical for the upper tree line
in northern Sweden that consists of short birch (Betula pubescens tortuosa). Finally, PC5, which describes 5% of the
data variance, holds high loadings for carbonate rock and
lake and stream areas.
1877
Fig. 2. Dendrogram from Ward’s minimum variance clustering
with the use of the principal scores of individual river catchments
from the five principal components (based on landscape variables)
as input variables. The river catchments can be divided in two major
groups: mountainous headwaters (four clusters) and forest and wetland catchments (two clusters). The main characteristics are: 2, glacier; 3, alpine area and subalpine birch forest; 5, alpine area (open
highland); 6, large lakes/reservoirs. Both forest and wetland clusters
have high vegetation cover, whereas cluster 1 is richer in area of
wetland and coniferous forest than cluster 4.
The cluster analyses of the five principal components distributed the 19 river catchments into six groups (Fig. 2),
which can be further summarized into two major groups:
mountainous headwaters versus forest and wetland catchments. The river headwaters have in common that most of
their catchments are located above the tree line (i.e., ,15%
of the catchment area is covered by forest; Table 1). Figure
3 shows the score distribution of the five principal components for headwaters, as well as forest and wetland catchments, respectively. Statistically, the catchments of headwaters, as well as forest and wetland catchments, only
differed significantly in terms of land cover and soil type
variables associated with vegetation (PC1).
Within the mountainous headwaters, four clusters appeared (Fig. 2). In the first, the headwater of river Kalixälven, with its uppermost glacier stream, Tarfalajåkka, was
found (cluster 2). This catchment showed a very high PC2
score and low scores for PC4 and PC5 (i.e., an alpine environment with glaciers and low percentage of lake area).
The watercourses Ladtjojåkka 1 and 2, the two downstream
stations of the Kalixälven headwater, are located below the
tree line. Their catchments were high in scores PC2 and PC4
(i.e., consisted both of alpine environments and subalpine
birch forest; cluster 3). The three minor streams discharging
into the northwestern part of the Akka reservoir (Upmasjåkka, Sårkajåkka, and Valtajåkka) formed cluster 5, with a
highly positive PC1 score and highly negative PC2 score.
They all represent open highlands (i.e., alpine environments
with low and little vegetation but, in contrast to clusters 2
and 3, with low percentage of snow- and ice-covered areas).
Cluster 6 consists of the major inputs to the Akka reservoir
(Vuojatätno and Ritsem hydropower plant) and the outlets
of the major reservoirs (Suorva dam and Porjus hydropower
plant). These catchments are mainly characterized by highly
1878
Humborg et al.
Fig. 3. Box plot of scores for each principal component (PC1–
PC5) separated into the two main groups: mountainous headwaters
(,15% forest) and forest and wetland catchments. The boxes indicate the 25th and 75th percentiles, the bold line the median value,
and whiskers the largest and smallest observations, respectively.
One outlier (circle) is excluded.
positive PC5 scores, namely, headwater catchments with a
high percentage of lake/reservoir area.
The wetland catchments (Sjuanjaätno and Muddusätno;
first-order tributaries) are most similar to the relatively unperturbed river systems Torneälven, Kalixälven, and Råneälven (cluster 1) and are characterized by a high percentage of vegetation (negative PC1 scores). The rivers of
Luleälven, Piteälven, Skellefteälven, and Umeälven form the
second group of forest and wetland catchments (cluster 4),
with slightly negative scores for PC1, related to less vegetation cover (Table 1) compared with cluster 1, but high
scores of PC3 from certain occurrence of silica-rich soils and
bedrock (Tables 2, 3) with low weathering rates.
River chemistry classified by landscape clusters—A clear
relationship appeared when relating the six clusters (based
on the landscape variables identified with PCA) to our chemistry data set. The various clusters reappear clearly in the
positive relationships between DIP/DSi (Fig. 4A), DIP/TOC
(Fig. 4B), and DSi/TOC (Fig. 4C). Highest concentrations
of TOC, DIP, and DSi were found in the unperturbed forest
and wetland catchments (cluster 1), with lower concentra→
Fig. 4. Relationships of long-term river water chemistry variables (discharge-weighted annual mean concentrations). (A) DIP
versus DSi, (B) DIP versus TOC, and (C) DSi versus TOC. River
catchments are classified according to cluster identity of Fig. 2;
cluster 3 is not shown in Fig. 4B,C because of the lack of TOC
measurements. The main characteristics of the headwater clusters
are: 2, glacier; 3, alpine area and subalpine birch forest; 5, alpine
area (open highland); 6, large lakes/reservoirs. Both forest and wetland clusters have high vegetation cover, whereas cluster 1 is richer
in area of wetland and coniferous forest than cluster 4. hpp, hydropower plant.
Landscape control of nutrient fluxes
1879
tions in the rivers of cluster 4, and lowest concentrations in
all mountainous headwaters (clusters 2, 3, 5, and 6). TOC,
DIP, and DSi concentrations (mean 6 SD, mmol L21) were
significantly (Kruskal–Wallis test, a 5 0.05) lower in headwaters (TOC 5 57 6 23, DIP 5 0.021 6 0.005, DSi 5 26
6 19) than in the forest and wetland catchments (TOC 5
580 6 126, DIP 5 0.092 6 0.043, DSi 5 68 6 21).
The DIN concentration pattern (not shown) differed from
that of the other nutrients. Whereas TOC, DIP, and DSi concentrations ranged over about one order of magnitude from
the headwaters to the river mouth, DIN concentrations remained rather low, about 2.1 6 1.0 mmol L21, regardless of
whether the lotic waters originated from catchments dominated by barren ground, forest, or wetland (Table 1). Consequently, DIN concentrations did not change between the
clusters, but remained fairly constant.
River chemistry versus landscape variables by multiple
linear regression analysis and RDA—Given that the first
five principal components explained such a large proportion
(87%) of the variation in landscape variables, the relationships of the individual river chemistry variables to the principal components were further analyzed. Ordinary stepwise
regression (with the use of partial F-test with criteria entry
p 5 0.05 and removal p 5 0.10) and with principal components as independent variables and TOC, DIP, DSi, and
DIN as dependent variables, resulted in the following statistically significant (a 5 0.05) relationships.
TOC 5 221.2 2 192.2 · PC1 2 65.6
· PC2
DIP 5 0.05 2 0.03 · PC1
DSi 5 45.9 2 22.6 · PC1 1 13.4
· PC4 2 7.3 · PC5
(r 2adj 5 0.80)
(r 2adj 5 0.43)
(r 2adj 5 0.87)
For DIN, none of the principal components were significant.
As for TOC, DIP, and DSi, the first component was significant throughout. In other words, the higher the percentage
of land cover and soil type variables associated with vegetation, the higher the concentrations of TOC, DIP, and DSi.
For DSi, PC4 and PC5 also make a significant contribution,
with a higher percentage of deciduous brushwood/forest or
of sand and gravel relating to higher DSi concentration,
whereas percentages of both lake and carbonate rock are
linked to lower concentrations. TOC was also negatively associated with the periglacial environment.
RDA was used to statistically determine the relationship
between the set of river chemistry variables and the set of
principal components. DIN was excluded because in the
multiple linear regression analysis, it was not found to correlate to any of the landscape variables studied. Therefore,
the following results are restricted to TOC, DIP, and DSi as
river chemistry variables. Only PC1 explained significant (p
5 0.002) amounts of variation in river chemistry. All in all,
the principal components explained 89% of the variance in
river chemistry, and 74% was related to PC1.
The angles between the arrows of the river chemistry variables (TOC, DIP, and DSi) in the ordination diagram are all
small (Fig. 5), indicating high intercorrelation. Both river
chemistry variables and river catchments of cluster 1 are
Fig. 5. Redundancy analysis (RDA) ordination diagram. River
chemistry (TOC, DIP, and DSi) and landscape variables (PC1–PC5)
are shown as arrows. The river catchments are presented according
to cluster identity of Fig. 2. The river catchments are: 1, Torneälven;
2, Kalixälven; 3, Tarfalajåkka; 4, Råneälven; 5, Luleälven; 6, Upmasjåkka; 7, Sårkkajåkka; 8, Valtajåkka; 9, Ritsem hydroelectric
power plant; 10, Vuojatätno; 11, Suorva dam; 12, Sjaunjaätno; 13,
Piteälven; 14, Skellefteälven; 15, Umeälven.
highly, negatively correlated to PC1. This is in accord with
the results reported above; the river catchments with high
percentages of vegetation were negatively related to PC1
(Table 4) and showed highest concentrations of TOC, DIP,
and DSi. The rivers of clusters 5 and 6 are found on the
negative side of the first axis, together with the arrow of
PC1 because both are related to snow and ice, open land,
unvegetated rock, till above the tree line and thin soils (Table
4). The arrow of PC4 is located at the positive site of the
first axis, indicating that deciduous brushwood/forest also
might be related to river chemistry variables. The first axis
also explains 99.9% of the variance in the relation between
river chemistry and landscape variables.
Discussion
Our study of entire landscapes in northern Sweden suggests that the effects of vegetation cover on weathering processes also are found at larger scales. This is indicated by
the clusters of landscape variables related to TOC, DIP, and
DSi concentrations, irrespective of catchment scale (Figs.
4A–C, 5). Thus, the presence of coniferous forest and wetland/peat, in combination with glaciofluvial sediments and
till below the tree line overlying granite and acid volcanic
bedrocks (Table 4), coincides with high TOC, DIP, and DSi
concentrations in the boreal and subarctic river catchments
studied. In contrast, the DIN concentrations were not significantly correlated to landscape characteristics. Possibly, this
1880
Humborg et al.
DIN might emanate from atmospheric deposition and biological (diazotrophic) fixation, whereas TOC, DIP, and DSi
contribute to or result from weathering processes. A major
implication from this study is that changes in landscape characteristics during glacial cycles might have affected land–
sea fluxes of TOC, DIP, and DSi significantly.
Vegetation control versus temperature and physical
weathering—The evaluation of potential causal mechanisms
regulating river chemistry in boreal and subarctic watersheds
revealed a clear response. PC1 is negatively associated with
landscape variables that are typical for deglaciated environments (i.e., mainly by coniferous trees and wetlands and soil
types of peat, glaciofluvial sediments, and till below the tree
line). The river systems with the lowest scores for PC1 had
the highest TOC, DIP, and DSi concentrations (cluster 1; see
Fig. 2; Table 1). The RDA analysis demonstrated a significant relation between river chemistry and landscape variables (Fig. 5), in which all land cover and soil type variables
associated with vegetation (PC1 and PC4; see Table 4) contribute strongly in describing the variation in TOC, DIP, and
DSi.
A large effect of temperature on river chemistry can be
ruled out because the temperature differences between the
mountainous headwaters and the river mouths are quite
small. In the Luleälven River, for example, the mean annual
temperature is 20.18C in the headwaters (Ritsem), 20.38C
at a station located in the middle of the catchment (Porjus),
and 12.58C at the river mouth. The weathering rate increases
by about 6% for each degree C as calculated for plagioclase
(by the Arrhenius equation and an activation energy of 10
kcal mol 21; see Lasaga 1998). Thus, a catchment-integrated
increase in weathering in the investigation area because of
temperature is probably ,10%. Moreover, landscape characteristics associated with periglacial environments (i.e.,
snow and ice, unvegetated rock, thin soils and bare rock, as
well as till above tree line) appear to have little influence on
biogeochemical dynamics, in contrast to what has been previously suggested (Froelich et al. 1992).
Low concentrations of TOC and dissolved inorganic nutrients that resemble that of rain and proglacial meltwater
(Humborg et al. 2002; Thillman 2003) were found both in
pristine systems, such as Tarfalajåkka (cluster 2; see Fig. 2)
and the headwaters of the Akka reservoir (cluster 5; see Fig.
2), as well as in the major inputs to and outlets from the
major reservoirs of the Luleälven headwaters (cluster 6; see
Fig. 2). In all these, water has had essentially no contact
with vegetated soils. In the uppermost headwater of the Kalixälven River, sparsely vegetated rock matter or alpine pastures with a thin soil layer dominate the landscape. In the
Luleälven River system, 95% of the water running into the
Akka reservoir is from areas above the tree line. Thus, the
biogeochemistry of the water systems above the tree line is
spatially and temporally uniform, although freshly weathered
rock matter is continuously produced, for instance, by glaciers. When deciduous brushwood/forest appears, concentrations of dissolved constituents increase, as demonstrated in
the Ladtjojåkka 1 and 2 streams (cluster 3; see Fig. 2), which
showed higher DSi concentrations than the other headwaters,
presumably as a result of a combination of weathered soils
and the appearance of more vegetation.
Damming strongly affects the land–sea fluxes of TOC and
nutrients (Humborg et al. 2002). In the mountainous headwater of the Luleälven River, damming has inundated the
river valley, causing major losses of vegetated soil. In total,
15 major dams are located along the Luleälven River. Hard
rock fragments and organic matter in the littoral zone of
these reservoirs are easily washed away. In reaches between
the reservoirs, underground headrace channeling of water in
combination with a reduction of water level fluctuations has
further decreased soil–water contact and, consequently, lowered the weathering rates. Therefore, the river biogeochemistry of the Luleälven River is uniform in space (Table 1)
and time and resembles that of periglacial systems with
sparse vegetation and low weathering regimes (Humborg et
al. 2002). The somewhat higher nutrient concentration at the
river mouth can be explained by tributaries draining wetlands (i.e., Sjaunjaätno and Muddusätno; Table 1). However,
the rivers of Piteälven, Skellefteälven, Luleälven, and
Umeälven form a distinct group of their own (cluster 4; see
Fig. 2). In fact, they have low scores for all components
except PC3 (Table 4), described by high loadings for silt and
clay, sandstone, and quartzite. However, it is not very likely
that the presence of the latter soil and bedrock types (see
area percentages in Tables 2, 3) will affect river biogeochemistry. More important, the transformations of wetlands
and forests into lake (reservoir) areas have certainly resulted
in reduced scores for PC1 (Table 4) in the three regulated
rivers of Luleälven, Skellefteälven, and Umeälven.
TOC as a large-scale proxy for organic degradation and
chemical weathering—Our multivariate statistical analyses
on landscape variables and river biogeochemistry integrate
small-scale interactions between bedrock, soil, pore water,
and biota (from microbes to vascular plants) and provide
evidence that they also are significant on a larger scale in
weathering-limited environments, such as the studied boreal
and subarctic watersheds. On the basis of laboratory studies,
it is well known that chemical weathering rates of silicate
minerals in acid solutions are related to temperature and pH
(Brantley and Stillings 1996; Lasaga 1998), as well as the
presence of organic (ligand/acid) compounds (Amrhein and
Suarez 1988). It has also been argued that the concentration
of such organic compounds under normal soil conditions
seems too low to significantly increase the weathering rate
(Drever and Stillings 1997) and that weathering products
even decrease with increasing vegetation and soil cover
(Engstrom et al. 2000). However, the latter study was performed in an area rich in carbonate, where calcium fluxes
decreased when the formerly exposed (easily weathered)
bedrock was covered with soil and vegetation. This contrasts
with our findings and previous field studies in deglaciated
terrain dominated by bedrock rich in feldspar and quartz
covered with till (Drever and Zobrist 1992; Anderson et al.
2000) that show that vascular plant vegetation appears to be
a major factor in increasing the weathering rate of silica and
phosphorus.
Among plausible chemical mechanisms by which vascular
plants might enhance the weathering rate, three main pro-
Landscape control of nutrient fluxes
cesses are discussed in the literature: root exudation of organic acids (Grayston et al. 1997), the activity of ectomycorrhizal fungi (van Breemen et al. 2000; Landeweert et al.
2001), and associated mineral dissolution by bacteria (Bennett et al. 2001). In high-latitude river systems, TOC often
consists of .80% DOC (cf. Wetzel 2001) and can be regarded as a proxy for various organic acids and ligands
formed. The positive correlation between TOC and both DIP
and DSi is striking (Fig. 4B,C). That DOC is a function of
forest cover has also been reported for subarctic lakes in
Canada (Pienitz and Vincent 2000). Intriguingly, the area of
deciduous brushwood/forest in our studied catchments (Table 1; PC4 in Table 4) appeared less significant for river
biogeochemistry than that of coniferous forest or wetland
(PC1; see Table 4). Moreover, in our investigated area, the
biomass per unit area of deciduous brushwood/forest (i.e.,
mainly birch, Betula spp., and willow, Salix spp.) is often
less than that in wetland or coniferous forest, the latter dominated by Norway spruce (Picea abies) and Scots pine (Pinus sylvestris). In support of our findings, a long-term field
study in France has reported that coniferous species enhance
the weathering rate on feldspar, compared with broad-leaved
deciduous trees (Augusto et al. 2000).
Northern Swedish rivers as model systems for high-latitude areas—Mean global concentrations of TOC, DIN, DIP,
and DSi for recent unperturbed subarctic and boreal rivers
are estimated at 750, 6, 0.1, and 110 mmol L21, respectively
(Meybeck 1979, 1982). The investigated Swedish rivers, and
especially the unperturbed forest and wetland catchments
(cluster 1), represent these conditions reasonably well (Table
1). Most riverine nutrient inputs to the Arctic Ocean come
from Russia, and an extensive data set has recently been
intercalibrated (Holmes et al. 2001). DIN values reported
from the three largest Siberian rivers, Ob, Lena and Yenisey,
respectively (Holmes et al. 2000), show low concentrations
comparable to the studied rivers of Northern Sweden (Table
1). Time series of DIP from the rivers Lena and Yenisey
indicate concentrations between 0.1 and 0.3 mmol L21, respectively (Holmes et al. 2000, 2001). Concentrations of
TOC and DSi in the major Siberian rivers likewise resemble
our records (Table 1; Fig. 4). Mean TOC and DSi concentrations of major Siberian rivers, including the Lena, Ob,
Yenesey, and Onega, are 615 and 95 mmol L21, respectively
(Gordeev 2000). Thus, the fragmented data sets of major
arctic Siberian rivers indicate that their nutrient concentrations and dynamics are similar to those presented here for
the northern Swedish rivers. Also, a recent study in the Mackenzie River Basin, Canada, reported mean DOC and DSi
concentrations of 1,088 and 78 mmol L21, respectively, and
positive relationships between DOC and DSi concentrations,
as well as DOC concentrations and weathering rate estimates
(Millot et al. 2003).
Similar patterns in river chemistry appear even in the
Southern Hemisphere. Stream chemistry data from unpolluted primary forests in temperate South America exhibiting
a broad range of environmental factors that influence ecosystem nutrient cycles showed a remarkably consistent pattern of low nitrogen loss from all forests. As in our study,
stream nitrate concentrations were exceedingly low, and dis-
1881
solved organic nitrogen was the main fraction of nitrogen
loss from these forests (Perakis and Hedin 2002). Assuming
our findings to be general for high-latitude areas, we suggest
that changes in type and cover of vegetation from glacial to
interglacial periods might increase TOC, DIP, and DSi fluxes
to the sea by an order of magnitude, in contrast to DIN
fluxes, which might have remained constantly low.
The potential effect of such variations in river chemistry
is challenging, for instance, when considering the melting of
the North American Laurentide Ice Sheet in the Younger
Dryas. During this period, the river discharge of the St.
Lawrence increased to 2 3 105 m3 s21 (Clark et al. 2001),
and simultaneously the vegetation along the ice edge
changed from polar desert or tundra to coniferous and broadleaved forest (Kohfeld and Harrison 2000). Given that much
of this water had passed through the forested landscapes, this
single river could then have transported about 0.6 Tmol DSi
into the North Atlantic each year, which is more than 10%
of the total present annual supply to the ocean (Tréguer et
al. 1995). In contrast, DIN and DIP concentrations in highlatitude rivers were presumably low throughout glacial cycles, and even lower than in the recent Arctic Ocean, North
Atlantic, and North Pacific (Levitus et al. 1993). Thus, much
of the nitrogen supply to the northern marine environments
might have come from biological nitrogen fixation in the
warmer areas of the ocean, which has been assumed to be
highest in glacial periods (Falkowski 1997). During glacial
periods, the low DSi supply to the northern seas presumably
slowed the biological pump (Dugdale et al. 1995), whereas
during periods of active deglaciation, the DSi supply to the
northern marine environments has probably been maximal.
The biological pump is most sensitive to diatom production,
and hence DSi variations. It can be estimated from global
ocean carbon and DSi budgets (Tréguer et al. 1995; Falkowski et al. 1998) that diatom sedimentation is responsible
for more than half of the carbon export production of the
contemporary ocean.
The biogeochemistry of high-latitude rivers is greatly influenced by vegetation cover and soil types such as peat.
Thus, vegetation changes during the deglaciation process are
likely to have had major effects on river biogeochemistry
and land–sea fluxes. For instance, inputs of TOC and DSi
to the ocean from formerly glaciated areas could have increased by an order of magnitude after deglaciation, judging
from the spatial vegetation gradient in our study and vegetation change between glacial and interglacial periods. This
would have had crucial effects on coastal environments
(Humborg et al. 1997, 2000; Ittekkot et al. 2000) and possibly even in midlatitude areas of the North Atlantic, where
DSi at present is often found in low concentrations after the
spring bloom (Levitus et al. 1993). A lower input might even
lead to Si limitation in the North Atlantic or elsewhere as
reported for some of the high nitrate–low chlorophyll areas
in contemporary upwelling Pacific water, where DSi set the
upper limit on the total possible biological utilization of dissolved inorganic carbon (Dugdale and Wilkerson 1998). Finally, because global warming is believed to be especially
pronounced at high latitudes in the Northern Hemisphere
(IPCC 2001), a change in structure and cover of vegetation
could quite rapidly alter the biogeochemistry of river catch-
1882
Humborg et al.
ments and land–ocean interactions along the coasts of the
Arctic Ocean.
References
ALEXANDRE, A., J. D. MEUNIER, F. COLIN, AND J. M. KOUD. 1997.
Plant impact on the biogeochemical cycle of silicon and related
weathering processes. Geochim. Cosmochim. Acta 61: 677–
682.
AMRHEIN, C., AND D. L. SUAREZ. 1988. The use of a surface complexation model to describe the kinetics of ligand-promoted
dissolution of anorthite. Geochim. Cosmochim. Acta 52: 2785–
2793.
ANDERSEN, B. G., AND H. W. BORNS. 1997. The Ice Age world.
Scandinavian University Press.
ANDERSON, S. P., J. I. DREVER, C. D. FROST, AND P. HOLDEN. 2000.
Chemical weathering in the foreland of a retreating glacier.
Geochim. Cosmochim. Acta 64: 1173–1189.
ÅNGSTRÖM, A. 1974. Sveriges klimat. Generalstabens Litografiska
Anstalts Förlag.
AUGUSTO, L., M. P. TURPAULT, AND J. RANGER. 2000. Impact of
forest tree species on feldspar weathering rates. Geoderma 96:
215–237.
BENNETT, P. C., J. R. ROGERS, AND W. J. CHOI. 2001. Silicates,
silicate weathering, and microbial ecology. Geomicrobiol. J.
18: 3–19.
BERNER, E. K., AND R. A. BERNER. 1996. Global environment: Water, air and geochemical cycles. Prentice Hall.
BERNER, R. A. 1992. Weathering, plants, and the long-term carboncycle. Geochim. Cosmochim. Acta 56: 3225–3231.
BJÖRKLUND, L. 1989. Geology of the Akkajaure-Tysfjord-Lofoten
traverse, N. Scandinavian Caledonides. Department of Geology, Gothenburg University.
BRANTLEY, S. L., AND L. STILLINGS. 1996. Feldspar dissolution at
25 degrees C and low pH. Am. J. Sci. 296: 101–127.
BROECKER, W. S., AND OTHERS. 1989. Routing of meltwater from
the Laurentide ice-sheet during the Younger Dryas cold episode. Nature 341: 318–321.
BRZEZINSKI, M. A. 1985. The Si : C : N ratio of marine diatoms:
Interspecific variability and the effect of some environmental
variables. J. Phycol. 21: 347–357.
CARLSSON, B., AND H. SANNER. 1994. Influence of river regulation
on runoff to the Gulf of Bothnia—the Gulf of Bothnia Year
1991. Swedish Meteorological and Hydrological Institute, Rep.
Hydrol. 9.
CLARK, P. U., S. J. MARSHALL, G. K. C. CLARKE, S. W. HOSTETLER,
J. M. LICCIARDI, AND J. T. TELLER. 2001. Freshwater forcing
of abrupt climate change during the last glaciation. Science
293: 283–287.
COCHRAN, M. F., AND R. A. BERNER. 1996. Promotion of chemical
weathering by higher plants: Field observations on Hawaiian
basalts. Chem. Geol. 132: 71–77.
DANIELSSON, Å., I. CATO, R. CARMAN, AND L. RAHM. 1999. Spatial
clustering of metals in sediments of Skagerrak/Kattegat. Appl.
Geochem. 14: 689–706.
DREVER, J. I. 1994. The effect of land plants on weathering rates
of silicate minerals. Geochim. Cosmochim. Acta 58: 2325–
2332.
. 1997. The geochemistry of natural waters. Prentice Hall.
, AND L. L. STILLINGS. 1997. The role of organic acids in
mineral weathering. Colloid. Surface A 120: 167–181.
, AND J. ZOBRIST. 1992. Chemical-weathering of silicate
rocks as a function of elevation in the southern Swiss Alps.
Geochim. Cosmochim. Acta 56: 3209–3216.
DUGDALE, R. C., F. P. WILKERSON, AND H. J. MINAS. 1995. The
role of a silicate pump in driving new production. Deep-Sea
Research I 42: 697–719.
AND
. 1998. Silicate regulation of new production
in the equatorial Pacific upwelling. Nature 391: 270–273.
ENGSTROM, D. R., S. C. FRITZ, J. E. ALMENDINER, AND S. JUGGINS.
2000. Chemical and biological trends during lake evolution in
recently deglaciated terrain. Nature 408: 161–166.
FALKOWSKI, P. G. 1997. Evolution of the nitrogen cycle and its
influence on the biological sequestration of CO2 in the ocean.
Nature 387: 272–275.
, R. T. BARBER, AND V. SMETACEK. 1998. Biogeochemical
controls and feedbacks on ocean primary production. Science
281: 200–206.
FREDÉN, C. 1994. Quaternary deposits, p. 104–119. In C. Fredén
[ed.], Geology, national atlas of Sweden. Almqvist and Wiksell.
FROELICH, P. N., V. BLANC, R. A. MORLOCK, S. N. CHILLRUD, W.
DUNSTAN, A. UDOMKIT, AND T.-H. PENG. 1992. River fluxes
of dissolved silica to the oceans were higher during glacials:
Ge/Si in diatoms, rivers, and oceans. Paleooceanography 7:
739–767.
GORDEEV, V. V. 2000. River input of water, sediment, major ions,
nutrients and trace metals from Siberian territory to the Arctic
Ocean, p. 297–322. In E. L. Lewis [ed.], The freshwater budget
of the Arctic Ocean. Kluwer.
GRAYSTON, S. J., D. VAUGHAN, AND D. JONES. 1997. Rhizosphere
carbon flow in trees, in comparison with annual plants: The
importance of root exudation and its impact on microbial activity and nutrient availability. Appl. Soil Ecol. 5: 29–56.
HOLMES, R. M., AND OTHERS. 2000. Flux of nutrients from Siberian
rivers to the Arctic Ocean: Can we establish a baseline against
which to judge future changes? Water Resour. Res. 36: 2309–
2320.
, AND OTHERS. 2001. Nutrient chemistry of the Ob’ and Yenisey Rivers, Siberia: Results from June 2000 expedition and
evaluation of long-term data sets. Mar. Chem. 75: 219–227.
HUMBORG, C., V. ITTEKKOT, A. COCIASU, AND B. VON BODUNGEN.
1997. Effect of Danube River dam on Black Sea biogeochemistry and ecosystem structure. Nature 386: 385–388.
, D. J. CONLEY, L. RAHM, F. WULFF, A. COCIASU, AND V.
ITTEKKOT. 2000. Silicon retention in river basins: Far-reaching
effects on biogeochemistry and aquatic food webs in coastal
marine environments. Ambio 29: 45–50.
, AND OTHERS. 2002. Hydrological alterations with river
damming in northern Sweden: Implications for weathering and
river biogeochemistry. Glob. Biogeochem. Cycles 16. [doi:
10.1029/2000GB001369]
IPCC (INTERGOVERNMENTAL PANEL ON CLIMATE CHANGE). 2001.
Climate change 2001: The scientific basis. Contribution of
working group I to the third assessment report of the Intergovernmental Panel on Climate Change. J. T. Houghton, Y. Ding,
D. J. Griggs, M. Nouger, P. J. van der Linden, X. Dai, K.
Maskell, and C. A. Johnson [eds.]. Cambridge University
Press.
ITTEKKOT, V., C. HUMBORG, AND P. SCHÄFER. 2000. Hydrological
alterations and marine biogeochemistry: A silicate issue?
BioScience 50: 776–782.
JAFFE, D. A. 2000. The nitrogen cycle, p. 322–342. In M. C. Jacobsen, R. J. Charlson, H. Rodhe, and G. H. Orians [eds.],
Earth system science. Academic.
KOHFELD, K. E., AND S. P. HARRISON. 2000. How well can we
simulate past climates? Evaluating the models using global palaeoenvironmental datasets. Quat. Sci. Rev. 19: 321–346.
LANDEWEERT, R., E. HOFFLAND, R. D. FINLAY, T. W. KUYPER, AND
N. VAN BREEMEN. 2001. Linking plants to rocks: Ectomycor-
Landscape control of nutrient fluxes
rhizal fungi mobilize nutrients from minerals. Trends Ecol.
Evol. 16: 248–254.
LASAGA, A. 1998. Kinetic theory in the earth science. Princeton
Univ. Press.
LEVITUS, S., M. E. CONKRIGHT, J. L. REID, R. G. NAJJAR, AND A.
MANTYLA. 1993. Distribution of nitrate, phosphate and silicate
in the world oceans. Prog. Oceanogr. 31: 245–273.
LINDSTRÖM, G., B. JOHANSSON, M. PERSSON, M. GARDELIN, AND
S. BERGSTRÖM. 1997. Development and test of the distributed
HBV-96 hydrological model. J. Hydrol. 201: 272–288.
MEYBECK, M. 1979. Pathways of major elements from land to
ocean through rivers, p. 18–30. In J.-M. Martin, J. D. Burton,
and D. Eisma [eds.], River inputs to ocean systems. UNEP IOC
SCOR United Nations.
. 1982. Carbon, nitrogen, and phosphorus transport by world
rivers. Am. J. Sci. 282: 401-450.
MILLOT, R., J. GAILLARDET, B. DUPRE, AND C. J. ALLEGRE. 2003.
Northern latitude chemical weathering rates: Clues from the
Mackenzie River Basin, Canada. Geochim. Cosmochim. Acta
67: 1305–1329.
NASH, J. E., AND J. V. SUTCLIFE. 1970. River flow forecasting
through conceptual models. Part I: A discussion of principles.
J. Hydrol. 10: 282–290.
OLIVA, P., AND OTHERS. 1999. The effect of organic matter on chemical weathering: Study of a small tropical watershed: NsimiZoetele site, Cameroon. Geochim. Cosmochim. Acta 63:
4013–4035.
PERAKIS, S. S., AND L. O. HEDIN. 2002. Nitrogen loss from unpolluted South American forests mainly via dissolved organic
compounds. Nature 415: 416–419.
PIENITZ, R., AND W. F. VINCENT. 2000. Effect of climate change
relative to ozone depletion on UV exposure in subarctic lakes.
Nature 404: 484–487.
1883
RANGER, C. 1993. Determination of low level nutrients in environmental water samples using flow injection analysis. Int. Environ. Technol. 3(4): 21–23.
SCHWARTZMAN, D. W., AND T. VOLK. 1989. Biotic enhancement of
weathering and the habitality of Earth. Nature 340: 457–459.
SHIKLOMANOV, I. A., R. B. SHIKLOMANOV, B. J. LAMMERS, B. J.
PETERSON, AND C. J. VÖRÖSMARTY. 2000. The dynamics of
river water inflow to the Artcic Ocean, p. 281–296. In E. L.
Lewis [ed.], The freshwater budget of the Arctic Ocean. Kluwer.
TER BRAAK, C. F. J. 1986. Canonical correspondence analysis: A
new eigenvector technique for multivariate direct gradient analysis. Ecology 67: 1167–1179.
. 1990. CANOCO: A Fortran program for canonical community ordination. Microcomputer Power.
THILLMAN, T. 2003. Geochemical characteristics of proglacial meltwater from Storglaciären, Kebnekaise massif, Sweden. Department of Geology and Geochemistry, Stockholm Univ.
TRÉGUER, P., AND P. PONDAVEN. 2000. Global change silica control
of carbon dioxide. Nature 406: 358–359.
, D. M. NELSON, A. J. VAN BENNEKOM, D. J. DEMASTER,
A. LEYNAERT, AND B. QUEGUINER. 1995. The silica balance in
the world ocean—A reestimate. Science 268: 375–379.
VAN BREEMEN, N., R. FINLAY, U. LUNDSTROM, A. G. JONGMANS,
R. GIESLER, AND M. OLSSON. 2000. Mycorrhizal weathering:
A true case of mineral plant nutrition? Biogeochemistry 49:
53–67.
WARD, J. H. 1963. Hierarchical grouping to optimize an objective
function. J. Am. Stat. Assoc. 58: 236–244.
WETZEL, R. G. 2001. Limnology. Academic.
Received: 26 August 2003
Amended: 19 March 2004
Accepted: 9 April 2004
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

Related manuals

Download PDF

advertisement