Global Change Biology Global Change Biology (2014) 20, 1225–1237, doi: 10.1111/gcb.12491
Almost 50 years of monitoring shows that climate, not forestry, controls long-term organic carbon fluxes in a large boreal watershed € 1 , M A R T Y N N F U T T E R 2 and P I R K K O K O R T E L A I N E N 1 AHTI LEPISTO 1 Finnish Environment Institute SYKE, P.O. Box 140, FI-00251 Helsinki, Finland, 2Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences SLU, P.O. Box 7050, SE-75007 Uppsala, Sweden
Abstract Here, we use a unique long-term data set on total organic carbon (TOC) fluxes, its climatic drivers and effects of land management from a large boreal watershed in northern Finland. TOC and runoff have been monitored at several sites in the Simojoki watershed (3160 km2) since the early 1960s. Annual TOC fluxes have increased significantly together with increased inter-annual variability. Acid deposition in the area has been low and has not significantly influenced losses of TOC. Forest management, including ditching and clear felling, had a minor influence on TOC fluxes – seasonal and long-term patterns in TOC were controlled primarily by changes in soil frost, seasonal precipitation, drought, and runoff. Deeper soil frost led to lower spring TOC concentrations in the river. Summer TOC concentrations were positively correlated with precipitation and soil moisture not temperature. There is some indication that drought conditions led to elevated TOC concentrations and fluxes in subsequent years (1998–2000). A sensitivity analysis of the INCA-C model results showed the importance of landscape position, land-use type, and soil temperature as controls of modeled TOC concentrations. Model predictions were not sensitive to forest management. Our results are contradictory to some earlier plot-scale and small catchment studies that have shown more profound forest management impacts on TOC fluxes. This shows the importance of scale when assessing the mechanisms controlling TOC fluxes and concentrations. The results highlight the value of long-term multiple data sets to better understand ecosystem response to land management, climate change and extremes in northern ecosystems. Keywords: drought, forestry, frost, INCA-C, modeling, organic carbon, TOC flux, watershed Received 7 June 2013 and accepted 2 November 2013
Introduction The widespread occurrence of increasing organic carbon concentrations across Europe and North America indicates regional causes and drivers, the relative importance of which are still under discussion. Various explanatory factors have been proposed, e.g., changes in temperature causing enhanced decomposition of organic soils, changes in hydrology leading to shifts in flow paths, decreased acid deposition, elevated atmospheric CO2 levels, or changes in land use and management practices. Increased temperature associated with climate change, contributing to the enhanced decomposition of peat soils and release of total organic carbon (TOC), was proposed by Freeman et al. (2001). Elevated atmospheric carbon dioxide levels and consequent stimulation of primary production may also be expected to result in increased levels of TOC (Freeman et al., 2004). Further, possible reasons for increased TOC concentrations include changes in hydrological Correspondence: Dr A. Lepist€ o, tel. +358 400 148824, fax +358 9 5490 2390, e-mail: [email protected]
© 2013 John Wiley & Sons Ltd
regimes due to increasing precipitation (Hongve et al., 2004), causing increasing flow volumes and consequent changes in flow paths (Tranvik & Jansson, 2002; Hejzlar et al., 2003) and the frequency of severe droughts (Worrall et al., 2004). Changes in land-use and management practices may in most cases be expected to result in increased decomposition of soil organic matter and release of TOC. Mitchell and McDonald (1995) found that areas of greatest peat drainage density were the most important sources of TOC. In contrast, peatland drainage decreased total inorganic carbon (TIC) and TOC concentrations in the long term, but this did not affect carbon export from small boreal catchments (Rantakari et al., 2010). Studies in Swedish headwater catchments have shown elevated TOC concentrations and fluxes after clear-felling (Schelker et al., 2012), but comparable studies in Finland have shown limited effects on TOC (Palviainen et al., 2013). In the Fennoscandic forest, land management practices are typically not sufficiently widespread to explain long-term TOC trends, because the effect on TOC of climatic variability is often large (e.g., Kortelainen et al., 1997). 1225
€ et al. 1226 A . L E P I S T O Decreasing mineral acid deposition may have resulted in an increase in organic acidity, resulting in increased TOC concentrations (e.g., Evans et al., 2005; Monteith et al., 2007; Haaland et al., 2010). It has been argued that increased TOC concentrations in 522 remote lakes and streams were due to a decrease in mineral acid and base cation deposition (Monteith et al., 2007). Recently, Oni et al. (2013) ascribed some of the change in DOC at a long-term monitoring site in northern Sweden to the subtle effects of declines in acid deposition. Sarkkola et al. (2009) found that hydrometeorological factors were more important than declining acid deposition in predicting TOC trends in boreal headwater streams in eastern Finland. However, the most important mechanisms behind the increasing TOC trends remain a subject of discussion (e.g., Roulet & Moore, 2006), and only a few studies have dealt with trends in organic carbon fluxes as compared with trends in concentrations. Longer wet or drought periods may significantly contribute to leaching. Wet periods may have strong regional effects, i.e., higher groundwater tables leading to closer-to-surface flow paths, and higher TOC leaching. It may not be changes in the amount of runoff but rather its source, i.e., as flow paths shift new, richer sources of TOC are accessed (Worrall & Burt, 2007). Consequently, droughts could augment TOC production. Changes in the relationship between flow and TOC were observed after a severe drought that persisted even through more minor droughts (Worrall & Burt, 2004). The snow pack in winter insulates soil from belowfreezing air temperatures. This facilitates a significant amount of biological activity (Groffman et al., 2001; Campbell et al., 2005). Northern forest ecosystems typically experience a deep and persistent snow pack, but future changes in climate may alter snow pack dynamics and thus winter soil temperatures. Snow removal experiments indicate an inverse relationship between snow depth and soil frost (e.g., Groffman et al., 2001; Decker et al., 2003). However, analyses of historical data suggest that warmer winters could result in either an increase or decrease in the frequency of soil frost (Lindstr€ om et al., 2002; Henry, 2008). Models that incorporate soil frost dynamics predict that warmer winters may result in fewer days with soil frost (Ven€ al€ainen et al., 2001; Campbell et al., 2010), although mid-winter soil frost may become more common (Ven€ al€ ainen et al., 2001). Because changes in winter climate are likely to alter belowground temperature regimes, it is important to understand how changes in soil frost dynamics could impact forest processes such as C and N cycling (Reinmann et al., 2012). Most of organic C in boreal surface waters is dissolved as opposed to particulate: in Finnish rivers,
where both total and filtered samples have been analyzed, particulate organic C fractions (POC) have been very minor and an average of 94% was found in dissolved form (Mattsson et al., 2005). The dominance of dissolved organic C and N fractions suggests very low retention of C and N in fast-flowing northern rivers such as Simojoki (Lepist€ o et al., 2006). In the boreal zone, in northern Finland, typical climate features are long winters (5–7 months) with continuous snow cover, deep soil frost, and ice cover over rivers and lakes. A snowmelt–induced spring flood in late April–May dominates the annual hydrological pattern. Smaller flow peaks occur in autumn due to rainfall. Most TOC leaching occurs during these high flow periods (Lepist€ o & Kortelainen, 2011). The Simojoki watershed experiences low, declining sulfate deposition and limited human impacts apart from forestry. Process-based and conceptual models are often used to check whether we understand the controls on surface water DOC dynamics (Futter et al., 2007; Ledesma et al., 2012; Oni et al., 2012). The Integrated Catchments model for Carbon (INCA-C) has been applied to headwater catchments in Fennoscandia (Futter & de Wit, 2008; Futter et al., 2008, 2009, 2011) and to larger, temperate catchments in Sweden (Ledesma et al., 2012) and Canada (Oni et al., 2012). In this paper, long-term changes (>40 years) in organic carbon concentrations and fluxes, and the factors behind these changes were studied on the boreal Simojoki river system. We investigated the impact of soil frost, temperature, and drought on organic C concentrations, using a monthly data set of possible drivers for the study period 1971–2008. To assess the possible impacts of land management, climate, landscape position, and the potential importance of drought, the integrated catchment model INCA-C was applied to the river system to simulate changing C concentrations and fluxes all through the way from the uppermost lakedominated reach down to the river outflow.
Material and methods
Site description, forest management, hydrological monitoring, and water sampling The Simojoki river (watershed area 3160 km2) (Fig. 1) discharges to the Gulf of Bothnia of the Baltic Sea. It is a salmon river in near-natural state. Mineral soil forests (39%), and peatlands and peatland forests (53%) together dominate in the watershed. The dominant human impact is forest management including drainage and felling: of the total P load, forestry covers 8770 kg yr1 of the total of 18 860 kg yr1, i.e., 47%. Of the total N load, forestry covers 89 000 kg yr1 of the o et al., 1995). total of 230 500 kg yr1, i.e., 39% (Perkki€
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
3 Fertilizing (%) Clear-cutting (%) Supplementary drainage (%)
2
Forest drainage (%)
1
0 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008
Agriculture is practiced only in 2.7% of the watershed area, and point sources (municipal wastewaters) have negligible contribution to the total nutrient load (2.0% N, 2.4% P) (Perkki€ o et al., 1995; Rankinen et al., 2002; Lepist€ o et al., 2004). The total length of the river between the outlet of Simoj€arvi, a large lake in the upper reaches of the watershed, and the sea is 193 km. In the Simojoki watershed, forest management was most intensive in the early 1970s. At that time, annual forestry activities covered 2.0–2.5% of the area, decreasing to 0.8–1.0% annually during the late 1980s–1990s (Fig. 2, data based on the Finnish Statistical Yearbooks of Forestry by METLA). The most evident change has been in the spatial intensity of firsttime peatland drainage for forest production, which has been replaced by supplementary drainage works. Forest fertilization (N–P–K) was almost discontinued in the 1990s but has increased in recent years. The percentage of cut area varied from year to year with no major long-term change. Cut areas are scattered throughout the river basin, with relatively minor annual treatment areas (0.2–0.6% in 1968–2009). Intensive forestry in the 1970s may have caused some increase in DOC concentrations at the river outlet (Lepist€ o et al., 2008). Simojoki has two discharge gauging stations that are used for calibration and validation of the hydrological part of the model; one is located at the river outlet, the other upstream at the Hosionkoski rapids (Fig. 1). Discharge has been monitored on a daily basis since 1965 at the river outlet. The annual average discharge at the outlet (1971–2000) was 40.4 m3 s1, i.e., annual runoff was 403 mm. Water samples have been analyzed from nine locations along Simojoki during the study period 1962–2008 by a regional environment centre. High variability is found in sampling periods and frequency between the reaches. At the outlet, 4–5 samples annually in 1962–1981 and 10–25 samples annually in 1982–2008 have been analyzed. These nine locations correspond to the reaches used in the INCA-C model calibration (Fig. 1).
Forest treatments (% of watershed)
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1227
Fig. 2 Annual forestry treatments (fertilizing, clear-cutting, forest drainage and supplementary drainage) in 1968–2009, as percentages of the watershed area.
Chemical oxygen demand (CODMn) was analyzed by titrimetric determination with KMnO4 during the whole observation period 1962–2008. TOC was estimated from CODMn concentrations (TOC = 0.675*CODMn + 1.94; Kortelainen, 1993), because CODMn measurements were available for the whole period. This equation is also valid in the Simojoki case, where the correlation coefficient between measured TOC and COD concentrations was high, i.e., 0.88 (R€ aike et al., 2012). In the Simojoki river, differences between TOC and DOC are very small (Mattsson et al., 2005; T. Mattsson, unpublished). The patterns of TOC that are analyzed in this study can hence directly be compared and discussed in light of literature values and patterns of DOC.
Seasonal data sets of climate, hydrology, and TOC A monthly climate data set was assembled for the study period of 1971–2008, from which seasonal average values were
Fig. 1 Map of the Simojoki watershed showing the subcatchment areas, reaches 1–9, and the locations of water quality and discharge monitoring sites. © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
€ et al. 1228 A . L E P I S T O calculated. The data set covers meteorological drivers including temperature, precipitation, and frost depth. Frost depth during 1971–2008 was available from the groundwater and soil frost data bank Povet by SYKE. Frost is typically measured in forest sites one to five times a month, totaling approximately 20 measurements during the whole winter period. The station Ylitornio R1301, closest to the Simojoki watershed, was used here. Temperature and precipitation observations used were areal weighted values based on meteorological stations operated by the Finnish Meteorological Institute in the region. The daily temperature and precipitation values were also inputs to national scale watershed simulation and forecasting system (WSFS), which has been widely used in operational hydrology in Finland since 1990 (Vehvil€ainen, 1994). Daily hydrologically effective rainfall (HER; mm) and soil moisture deficit (SMD; mm) were simulated by the WSFS model and converted into monthly values. Observed runoff for the study period 1971–2008 was obtained from the discharge station close to the Simojoki river outlet (Fig. 1). Annual or monthly TOC fluxes were estimated by interpolating observed TOC concentrations linearly to estimate daily concentrations, which were then multiplied by daily flow and summed.
INCA-C model description The Integrated Catchments model for Carbon (INCA-C) is described in detail by Futter et al. (2007, 2009). The INCA-C model describes the major factors and processes controlling TOC in surface waters. There are four components to INCA-C: (i) a GIS interface used to define subcatchment boundaries and identify the areas of different land cover classes; (ii) an external rainfall-runoff model used to calculate HER and SMD; (iii) a land phase hydro-chemical model simulating material fluxes through the soil column and transformations between chemical stocks; and (iv) an in-stream model simulating the transformations in the aquatic phase. The INCA-C model is calibrated against measured time series of surface water TOC concentrations and flow. Briefly, INCA-C simulates the hydrological fluxes within a catchment or subcatchments and the masses and fluxes of solid organic (SOC), total (dissolved) organic (TOC) and dissolved inorganic carbon (DIC) in soils and surface waters (Futter et al., 2007). HER is the amount of water arriving at the soil surface as precipitation or snow melt that will eventually contribute to runoff. It may flow directly to the surface water or percolate into the upper soil box. Water in the upper soil box may flow to the stream or percolate to the lower soil box. All water in the lower soil box eventually runs off to the surface water. The breakdown of litter and root material contributes SOC and TOC in the upper soil box. Sorption and desorption processes control the transformation of organic carbon between TOC and SOC, and mineralization controls the transformation of SOC and TOC to DIC, part of which is lost to the atmosphere through degassing. TOC and DIC are transported advectively by water movement from the upper to lower soil boxes and from the soil to surface waters. Within INCA-C, all carbon transformations in soils and surface waters are modeled as a series of first-order differential
equations. All in-soil rate coefficients are dependent on soil temperature and moisture (Eqn 1). Soil temperature is simulated using observed air temperature and a model developed by Rankinen et al. (2004). The effect of soil temperature on the rate coefficients in INCA-C is simulated using a Q10 type model. The rate coefficient for the effect of soil moisture is a linear function of soil moisture content. The processes operate at a maximum rate when the simulated SMD is equal to zero. Processes cease when the SMD is greater than SMDMax, a calibrated threshold representing the maximum SMD at which carbon processing may occur. SMDMax minðSMD, SMDMax Þ ð1Þ m ¼ QðTSoil 20Þ SMDMax Changes in the mass of organic carbon in a soil box are simulated as a combination of soil temperature and moisture effects (m), and the mass of organic carbon sorbed (kSTOC, kg C d1) and desorbed (kDSOC, kg C d1) (Eqn 2). The base rates of the desorption (kD) and sorption (kS) coefficients are estimated during model calibration. dTOC ¼ mðkD SOC ks TOCÞ dt
ð2Þ
In INCA-C, TOC can be mineralized in the open water by both microbial and photolytic processes. In the simulations presented here, only microbial mineralization is assumed. The simulated rate of microbial mineralization of TOC is linearly dependent on water temperature (TW). Water temperature is assumed to be equal to air temperature so long as the air temperature is above a user-specified threshold; otherwise, water temperature is set to the threshold value. The model has been further modified here to simulate the so-called ‘enzymic latch’ mechanism (Worrall et al., 2004). The ‘enzymic latch’ mechanism has been proposed as a mechanism to explain the increase in DOC concentrations observed following severe drought in UK peatlands. This mechanism is simulated in INCA-C as an increase in the rate of SOC to DOC transformation when the SMD crosses a critical threshold representative of severe drought conditions. The rate increase due to drought declines with time, returning to the baseline value after a user-specified number of years.
INCA-C model inputs and its calibration to Simojoki Hydrological input to the INCA-C model was calculated using the above WSFS model. Daily values of temperature (°C), HER, and SMD were calculated for the study period 1962– 2008, and taken as input to the INCA-C model. Land cover in the catchment was divided into six classes representing (i) mineral forests; (ii) cut mineral forests; (iii) undrained peatlands; (iv) peatland forests; (v) peatlands subject to ditching; and (vi) agriculture. Land use proportions were estimated for each of the 9 reaches based on values in Rankinen et al. (2002), forest management time-series in Fig. 2, and data sets based on the Finnish Statistical Yearbooks of Forestry by METLA. Effects of forest harvest and peatland drainage were assumed to last for 10 years, it was estimated that 20% of the peatlands
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1229 in the catchment were drained in 1962 and that by 2009, about 60% of peatlands had been subject to primary drainage. Until the end of 1991, about 1030 km2 of the whole watershed area was estimated to have been drained, which is in agreement with Perkki€ o et al. (1995) who report that in 1991 over 30% of watershed (i.e., about 60% of the peatlands and peatland forests) had been drained. INCA-C model parameterization was based on values for forests and peatlands from the ValkeaKotinen reference catchment in southern Finland (Futter et al., 2008, 2009). The effects of forest harvest on soil temperature were simulated by adjusting the thermal conductivity parameters of the soil-temperature submodel in INCA-C (Rankinen et al., 2004) so that the simulated soil warming matched that observed by Schelker et al. (2012). A manual calibration was performed which attempted to maximize the Nash–Sutcliffe (NS) statistics (Nash & Sutcliffe, 1970). The results of this calibration were used to define initial parameter ranges for a sensitivity and uncertainty analysis based on Markov Chain Monte Carlo (MCMC) principles. Initial parameter ranges for terrestrial and in-stream process rates were defined as 20% of the value obtained in the manual calibration. In-stream flow/velocity parameters and a temperature-dependent TOC mineralization rate were allowed to vary. Land-phase parameters related to sorption desorption rates, soil moisture thresholds, soil temperature response, and postdrought increase in TOC production were also allowed to vary. The MCMC analysis (described in Futter et al., 2013) attempted to give equal weight to simulations of flow, TOC from the lake outlet and TOC in the main stem of the river. Briefly, the MCMC analysis assumes equifinality in the high-dimensional parameter space sampled. A posterior distribution of credible parameter values is derived from the best performing parameter set in each of an ensemble of model runs. The acceptance criteria for new parameter sets in the MCMC analysis was based on weighted NS statistics for TOC in the lake outflow (reach 1, weight = 8), TOC in the main river (reaches 2–9, weight = 1 for each reach) and flow (reaches 5 and 9, weight = 4 for each reach). This was performed so as to ensure that model calibration would be sensitive to parameters related to in-lake and terrestrial processes, and so that a balance would be achieved between fitting to flow and TOC concentration. The MCMC analysis comprised 100 model chains, each consisting of 300 model runs. The parameter set associated with the best performing model run from each chain was retained for sensitivity and uncertainty analysis. Parameter sensitivity was assessed by using the Kolmogorov–Smirnov test to compare the distribution of parameter values in the best performing ensemble of model runs to a rectangular distribution.
Results
but decreased somewhat during the 2000s (2001–2008), to 17 200 t yr1. The average annual runoff was 27% higher during the 1990s (477 mm) than during the 1980s (375 mm), and therefore, runoff accounts for a major part of the increase in the TOC outputs but not all. Figure 3 also highlights increasing temperatures: in the latter part of the study period (1986–2008) average temperature was clearly higher, 1.2 °C, as compared with the first half (0.2 °C in 1962–1985). Year-to-year hydrological variability is very high and impacts considerably on TOC fluxes (Fig. 3). A dry period with low runoff and low regional groundwater table (1994–1997; average runoff 369 mm) was followed by a considerably wetter period (1998–2000; average runoff 618 mm) with high groundwater tables, high TOC concentration peaks and large leaching fluxes. High water tables contribute more surface runoff from the TOC-rich soil layers, and more runoff from riparian areas to the stream than low groundwater tables.
Soil frost and TOC concentration in the subsequent spring Spring TOC concentrations at the river outlet (reach 9) during the spring snow melt period in 1971–2008 correlated with the maximum soil frost during the preceding winter (R2 = 0.37) (Fig. 4). Average spring concentrations were clearly higher in the four mildest winters (14.3 mg l1) with a maximum frost of only 40–70 cm, as compared with four deepest frost winters with 110120 cm (9.9 mg l1). This means that also spring TOC fluxes after a mild winter are considerably higher, over double: 9600–12100 tonnes in April–May after mild ‘low-frost’ winters vs. 3700–6200 tonnes after deep-frost winters. There was practically no relationship (R2 = 0.10) between maximum soil frost and the amount of subsequent spring runoff. This means that deep-frost winters (110–120 cm) were followed by only average spring runoff, probably without any substantial amounts of surface runoff and overland flow over frozen soils. The soil frost correlated considerably better to TOC concentrations than to amounts of runoff which means that high spring flows may occur after very variable winters (low- to medium-frost conditions), while high concentrations of TOC during spring snowmelt only occur after low-frost winters.
Long-term TOC fluxes and hydrological variability Annual TOC fluxes at the river outlet displayed a statistically significant increase over the study period (Mann Kendall statistic, MK = 2.3, P = 0.01, Fig. 3). The average TOC flux increased by 38% during the 1990s, from 14 800 t yr1 in the 1980s to 20 500 t yr1 in the 1990s © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
Summer precipitation, drought, and TOC Summer precipitation (June–August) was an important driver of summer SMD (Fig. 5b), which in turn explained most (R2 = 0.66) of the variation in summer TOC concentrations at the river outlet (reach 9) (Fig. 5a).
€ et al. 1230 A . L E P I S T O
3 2 1 0 –1
1980
1982
1984
1986
1988
1990
1992
1980
1982
1984
1986
1988
1990
1992
2008
1978 1978
2006
1976 1976
2004
1974 1974
2002
1972 1972
2000
1970 1970
1998
1968 1968
1996
1966 1966
1994
1964 1964
Soil moisture deficit (mm)
–3
1962
–2
1962
Air temperature (°C)
4
0 –5 –10 –15 –20 –25 –30 –35 –40 –45 800
Runoff (mm)
700 600 500 400 300 200 100 0
30 000 25 000 20 000 15 000 10 000
2008
2006
2004
2002
2000
1998
0
1996
5000 1994
TOC flux (tonnes yr–1)
35 000
Fig. 3 Annual average air temperature (°C), soil moisture deficit (mm), annual runoff (mm), and annual total organic carbon (TOC) flux (tonnes yr1) in the Simojoki watershed in 1962–2008 (for runoff and TOC flux, data from 1966 are available).
Summer precipitation and TOC concentrations were closely linked (R2 = 0.51) (Fig. 6), with an average increase in TOC from 7.5 mg l1 in the driest summer to 15.5 mg l1 in the wettest summer. Summer TOC fluxes in the four wettest summers were over 5-fold higher: 5700–12800 tonnes (average 9400) vs. 900– 2300 tonnes (average 1700) in the four driest summers, respectively. In contrast, summer temperature was not correlated with either summer TOC concentrations (R2 = 0.08) or summer precipitation (R2 = 0.04). Close links between summer precipitation and TOC concentrations (Fig. 6) probably reflect closer-to-surface
flow paths and higher areal groundwater tables, with higher TOC concentrations in those surface soil layers, which together allow strong C fluxes to the river system. These responses to higher water volumes are rapid, while higher soil temperatures may increase TOC fluxes but with longer delays.
INCA-C calibration results Model performance statistics from the ensemble of best performing INCA-C calibrations are shown in Table 1. For all reaches, model simulations were able to achieve © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1231 20
TOC (mg L–1) in summer
Spring TOC concnentration (mg L–1)
20
15
10
5
0
0
20
40
60
80
100
120
140
Precipitation (mm) in summer 0
20
40
60
80
100
120
140
20
Fig. 6 Summer total organic carbon (TOC) concentrations vs. precipitation of that period (1971–2008).
Table 1 Model performance statistics [R2 and Nash–Sutcliffe (NS)] for the INCA-C calibrations from uppermost reach 1 down to reach 9 (river outlet)
(a)
R2
15
Reach
10 y = 0.0022x2 + 0.32x + 21.19 R² = 0.66
5 0 –100
–80
–60
–40
–20
0
SMD (mm) in summer 0 –10 –20 –30 –40 –50 –60 –70 –80 –90 –100
y = 0.074x + 6.72 R² = 0.51
5 0
Fig. 4 Spring total organic carbon (TOC) concentrations vs. deepest monthly frost depth of the preceding winter (1971–2008).
TOC (mg L–1) in summer
10
y = –0.0007x2 + 0.054x + 13.26 R² = 0.37
Maximum soil frost (cm)
SMD (mm) in summer
15
(b)
y = –0.0037x2 + 1.09x –99.36 R² = 0.74 0
20
40
60
80
100
120
140
Summer precipitation (mm) Fig. 5 (a) Summer total organic carbon (TOC) concentrations vs. soil moisture deficit (SMD mm) of that summer (1971–2008), together with (b) relationship between precipitation and SMD.
NS statistics >0 for both TOC and flow. Observed and simulated TOC values for the Simoj€ arvi outlet (reach 1), Portimoj€ arvi (reach 5), and the river outlet (reach 9) are shown in Fig. 7. The INCA-C calibration was able to reproduce the markedly different seasonal patterns of TOC at Simoj€ arvi vs. the other reaches. In most cases, the INCA-C calibration was able to reproduce the range of observed values; however, it tended to miss some of © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
TOC 1 2 3 4 5 6 7 8 9 Flow 5 9
NS
Min
Average
Max
Min
Average
Max
0.10 0.36 0.35 0.62 0.44 0.46 0.46 0.44 0.38
0.13 0.39 0.38 0.64 0.47 0.49 0.49 0.47 0.42
0.14 0.43 0.40 0.66 0.50 0.51 0.51 0.49 0.44
0.78 0.54 0.07 0.25 0.17 0.15 0.26 0.26 0.09
0.00 0.15 0.05 0.36 0.30 0.27 0.38 0.37 0.07
0.08 0.08 0.10 0.43 0.39 0.31 0.46 0.44 0.15
0.83 0.80
0.85 0.82
0.86 0.83
0.81 0.63
0.83 0.81
0.84 0.83
the high autumn concentrations observed at the outflow (Fig. 7). INCA-C estimated fluxes from individual reaches are shown in Fig. 8. Note the large difference between Simoj€arvi (reach 1) and the downstream reaches. This difference is a result of simulated in-lake mineralization of TOC in Simoj€arvi. The difference in areal fluxes between reaches below the lake is a function of the proportion of peat forest land cover in the subcatchment. By combining the reach specific fluxes with the respective land use proportions, it was possible to estimate land use specific fluxes of 3.9, 7.7, 8.2, 8.0, 8.3, and 1.0 g TOC m2 yr1 for mineral forest, cut mineral forest, undrained peatland, peatland forest, peat ditching, and agriculture, respectively. Using these values, an average flux of 5.9 g TOC m2 yr1 is estimated for Simoj€arvi. Comparing this with the INCA-C estimated flux of 3.1 g TOC m2 yr1 at the Simoj€arvi outlet
€ et al. 1232 A . L E P I S T O 30
Reach 1
25
TOC (mg L–1)
20 15 10 5 0 –5 30
Reach 5
25
TOC (mg L–1)
20 15 10 5 0 –5 30
Reach 9
25
TOC (mg L–1)
20 15 10 5 0
2008
2006
2004
2000
2002
1998
1996
1994
1990
1992
1986
1988
1982
1984
1980
1978
1974
1976
1970
1972
1968
1966
1962
1964
–5
Fig. 7 Observed and modeled total organic carbon (TOC) concentrations at reaches 1 (upstream), 5 (mid-catchment) and 9 (close to outlet) during 1962–2008. The thick line shows the average value from the ensemble of 100 best performing INCA-C model runs, while the lower thin lines show the deviations of the maximum and minimum ensemble values from the ensemble average.
indicates that ca. 50% of the simulated TOC inputs from the surrounding catchment are removed by in-lake processing in Simoj€ arvi. There was relatively little difference in model performance for downstream reaches (Fig. 8). This suggests that TOC dynamics in these reaches, with their shorter residence times, can be credibly simulated on the basis of terrestrial processes alone.
Posterior distributions of model parameters only had significant Kolmogorov–Smirnov statistics for the flow velocity ‘a’ and ‘b’ terms and DOC mineralization rate in reach 1 and the temperature response and DOC to SOC sorption rate in the peatland forest land cover type. The model was most sensitive to estimated rates of TOC mineralization in reach 1, which was © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1233
TOC (g m–2 yr–1)
was related to organic matter increases, but flow emerged as an even more important driver of OM variability (Erlandsson et al., 2008). Warming can affect DOC export in different ways, depending on whether it is accompanied by increased or decreased precipitation; the variation is related to hydrology as well as to biological productivity, and to how productivity is balanced by decomposition (Tranvik & Jansson, 2002).
Impacts of soil frost
Reach
Fig. 8 Modeled total organic carbon (TOC) fluxes (g m2 yr1) from uppermost reach 1 down to reach 9 (river outlet) during 1962–2008. Fluxes were modeled using the INCA-C model.
dominated by Simoj€ arvi, which were controlled by both the water residence times (flow ‘a’ and ‘b’ values) and the mineralization rate. Model performance was not sensitive to any parameter in the agriculture land cover type, probably due to its relatively minor contribution to total land use in the Simojoki catchment. Somewhat surprisingly, model results were not sensitive to any parameters in the forest land cover type, including the drought-related ‘enzymic latch’ parameters.
Discussion
Trends in TOC flux as a result of climate Surface water organic carbon concentrations are changing: increasing trends in concentrations and fluxes into large watercourses have been reported across Europe (Erlandsson et al., 2008; Lepist€ o et al., 2008; Klavins et al., 2012; P€ arn & Mander, 2012). In this study, annual TOC fluxes at Simojoki watershed displayed a statistically significant increase over the study period. Respectively, P€ arn and Mander (2012) found significant trends in TOC in five of 11 rivers in Estonia. They conclude that the main factors in the increase in organic carbon export are longer drought periods driven by climate change, magnified by drainage works. R€ aike et al. (2012) studied changes in DOC fluxes in the period 1975–2010 using data from 29 Finnish river basins. They found increasing normalized DOC fluxes in 8 river basins, decreasing in 7 basins, and no trends in 14 basins. In 28 large Swedish catchments, decline in anthropogenic sulfate deposition © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
The depth of soil frost had a statistically significant effect on TOC concentration in the subsequent spring (Fig. 4). Laboratory studies on Norwegian forest and heathland soils have shown that the impacts of soil frost on TOC are complex and highly variable (e.g., Austnes & Vestgarden, 2008; Austnes et al., 2008). Results from laboratory experiments using soils from a temperate hardwood forest indicate that severe soil frost (15 °C) resulted in lower losses of C compared with unfrozen soils during snow melt, while no significant impact of mild soil frost (0.5 °C) was detected (Reinmann et al., 2012). This suggests that severity of soil frost plays an important role in potentially reducing C losses during spring snow melt. These results are comparable with our results at the large watershed scale. In laboratory experiments with German forest soils, there was on the contrary an increase in TOC in response to freezing of organic soil horizons at temperatures of 8 and 13 °C, but not at 3 °C, a much more common field temperature (Hentschel et al., 2008). It is likely that the extent to which C and N cycling during snow melt is altered in response to changes in winter climate depend on both the presence of soil frost and the temperature to which soils freeze (Reinmann et al., 2012). The few short-term field-scale manipulation studies performed to examine the relationship between soil frost and TOC in soil solution have reported no or inconclusive results (Fitzhugh et al., 2001; Austnes et al., 2008). In northern Sweden, contrary to our results, it was found that longer and colder winter results in higher TOC concentrations in a very small boreal headwater stream during the subsequent snow melts (Haei et al., 2010). In very small systems, riparian zones probably play a more important role than in the larger river basin scale studied here. Field studies in Norway noted increases in TOC in response to mild freezing (Austnes et al., 2008; Kaste et al., 2008). In a field study in Germany, no relationship between TOC and soil frost was observed (Hentschel et al., 2009). Clearly, there is a need for further research on the factors regulating TOC response to soil freezing disturbance (Groffman et al., 2011).
€ et al. 1234 A . L E P I S T O The extent of soil frost development ultimately depends on the integrated effect of air temperature and the extent and duration of the snow cover. Furthermore, development and timing of snow cover is of importance for the depth of frost, and further to C leaching processes. In mild, low-frost winters also in the boreal zone, higher amounts of water flow in surface zone layers where the TOC concentrations are higher, contributing to higher TOC losses in the subsequent spring (Fig. 4). In their recent review of freeze-thaw events and C and N losses from soils, Matzner and Borken (2008) state that future milder winter climate with fewer periods of soil frost may result in greater losses of C from soils that are presently influenced by extended frost periods. The probability of mild winters with less soil frost probably increases in a changing climate, in northern ecosystems (e.g., Denman et al., 2007). These changes might increase spring C fluxes, through the concentration increase, in those cases where snow melt flow peaks remain the same, or occur somewhat earlier in spring.
Precipitation, temperature, and drought effects Summer precipitation was an important driver for summer SMD, which in turn explained most of the variation in summer TOC concentrations (Fig. 5). Also, summer precipitation and TOC concentrations were closely linked. This is relevant as Rantakari and Kortelainen (2005) showed that open water precipitation and CO2 fluxes from a set of large Finnish lakes were very closely correlated, whereas temperature and CO2 were not linked to each other. Although the overall mechanisms behind TOC and CO2 patterns are dissimilar, both seem to be closely linked to summer precipitation in boreal Finnish lakes and rivers, whereas links to temperature are minor. Further, global hotspots of stream and river CO2 evasion were demonstrated in regions where precipitation was high (Raymond et al., 2013). Isotopic compositions of DOC and CO2 in boreal Finnish headwater streams have been shown to be very similar, suggesting a predominantly single and consistent C source, i.e., decomposition of organic matter (Billett et al., 2012). There are several indirect lines of evidence to support the conclusion that droughts can be major drivers of TOC release. These include the following: (i) change in the relationship between flow and TOC after a severe drought; (ii) soil respiration and TOC production become decoupled after a severe drought; (iii) stepwise changes in the flux of TOC are observed after droughts. However, direct evidence for a drought effect on TOC concentrations is lacking (Worrall & Burt, 2007). In
Simojoki study, we found some indication of stepwise changes in the TOC flux after droughts, particularly when comparing wet period of 1998–2000 to earlier drought period of 1994–1997 (Fig. 3), but INCA-C model results were not sensitive to any parameters in the forest land cover type, including drought-related parameters. In Poland, the autumn 2000 drought was one of the most severe in many years, resulting in about 30% decrease in TOC concentrations. Average TOC and water colour were two times higher in the wetter autumn than in dry period (Zielinski et al., 2009). In a Swedish catchment, stepwise increases in DOC concentration did follow some, but not all, summer droughts; the most notable changes occurred following a sequence of dry summers (Jennings et al., 2010).
Importance of scale – forestry effects, landscape scale controls, and INCA-C Forest management activities can either increase or decrease TOC concentrations and fluxes. Numerous studies of headwater catchments in Fennoscandia have shown elevated fluxes and concentrations following final felling (Nieminen, 2004; Kreutzweiser et al., 2008; Schelker et al., 2012) while Palviainen et al. (2013) did not observe any significant increases. Piirainen et al. (2002) caution that final felling effects on TOC flux may be negligible at larger spatial scales. Our results support this latter conclusion and are consistent with the observations of Palviainen et al. (2013). Any increase in TOC associated with final felling may be offset by the decrease in TOC export that has been observed following ditch maintenance (Nieminen et al., 2010). While forest management activities have potentially severe and undesirable consequences at the local scale, these effects are not readily apparent at the landscape scale. Landscape type exerts the dominant control on TOC export in boreal rivers. As noted by R€aike et al. (2012), upland forest and wetlands have quite different TOC exports. In addition, more than half of the TOC exported from boreal catchments may be consumed by in-lake processes (Tranvik et al., 2009). These landscape level controls were reproduced in the INCA-C simulations presented here. The areal flux of TOC from the uppermost lake dominated subcatchment was almost half that of downstream subcatchments due to in-lake losses of TOC (Fig. 8). The INCA-C model (Futter et al., 2007) has been designed to simulate the effects of landscape type, climate, and land management on soil and surface water TOC concentrations and fluxes and has been successfully applied in numerous boreal headwater catchments. Our empirical and modeling results by INCA-C suggest that climate change driven patterns in runoff, © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1235 soil moisture, and temperature were more important than temporal patterns of land management in controlling surface water TOC concentrations. The absence of any clear forest management effect in the large-scale simulations presented here shows the importance of scale when addressing the impacts of forest management activities on surface water quality (Futter et al., 2010). The large changes in TOC flux and concentration which are sometimes observed in headwater catchments following final felling (Schelker et al., 2012; Palviainen et al., 2013) may be impossible to detect in larger rivers. Due to the relatively slow rates of tree growth and long forest rotation periods, forest management activities affect a relatively small amount of the total catchment at any point in time. The INCA-C simulated fluxes from peatlands were relatively similar (8.0– 8.3 g TOC m2 yr1), regardless of the management activities. The higher TOC exports from upland forests following final felling (7.7 vs. 3.9 g TOC m2 yr1) are consistent with empirical studies (Schelker et al., 2012) but the relatively small (
Almost 50 years of monitoring shows that climate, not forestry, controls long-term organic carbon fluxes in a large boreal watershed € 1 , M A R T Y N N F U T T E R 2 and P I R K K O K O R T E L A I N E N 1 AHTI LEPISTO 1 Finnish Environment Institute SYKE, P.O. Box 140, FI-00251 Helsinki, Finland, 2Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences SLU, P.O. Box 7050, SE-75007 Uppsala, Sweden
Abstract Here, we use a unique long-term data set on total organic carbon (TOC) fluxes, its climatic drivers and effects of land management from a large boreal watershed in northern Finland. TOC and runoff have been monitored at several sites in the Simojoki watershed (3160 km2) since the early 1960s. Annual TOC fluxes have increased significantly together with increased inter-annual variability. Acid deposition in the area has been low and has not significantly influenced losses of TOC. Forest management, including ditching and clear felling, had a minor influence on TOC fluxes – seasonal and long-term patterns in TOC were controlled primarily by changes in soil frost, seasonal precipitation, drought, and runoff. Deeper soil frost led to lower spring TOC concentrations in the river. Summer TOC concentrations were positively correlated with precipitation and soil moisture not temperature. There is some indication that drought conditions led to elevated TOC concentrations and fluxes in subsequent years (1998–2000). A sensitivity analysis of the INCA-C model results showed the importance of landscape position, land-use type, and soil temperature as controls of modeled TOC concentrations. Model predictions were not sensitive to forest management. Our results are contradictory to some earlier plot-scale and small catchment studies that have shown more profound forest management impacts on TOC fluxes. This shows the importance of scale when assessing the mechanisms controlling TOC fluxes and concentrations. The results highlight the value of long-term multiple data sets to better understand ecosystem response to land management, climate change and extremes in northern ecosystems. Keywords: drought, forestry, frost, INCA-C, modeling, organic carbon, TOC flux, watershed Received 7 June 2013 and accepted 2 November 2013
Introduction The widespread occurrence of increasing organic carbon concentrations across Europe and North America indicates regional causes and drivers, the relative importance of which are still under discussion. Various explanatory factors have been proposed, e.g., changes in temperature causing enhanced decomposition of organic soils, changes in hydrology leading to shifts in flow paths, decreased acid deposition, elevated atmospheric CO2 levels, or changes in land use and management practices. Increased temperature associated with climate change, contributing to the enhanced decomposition of peat soils and release of total organic carbon (TOC), was proposed by Freeman et al. (2001). Elevated atmospheric carbon dioxide levels and consequent stimulation of primary production may also be expected to result in increased levels of TOC (Freeman et al., 2004). Further, possible reasons for increased TOC concentrations include changes in hydrological Correspondence: Dr A. Lepist€ o, tel. +358 400 148824, fax +358 9 5490 2390, e-mail: [email protected]
© 2013 John Wiley & Sons Ltd
regimes due to increasing precipitation (Hongve et al., 2004), causing increasing flow volumes and consequent changes in flow paths (Tranvik & Jansson, 2002; Hejzlar et al., 2003) and the frequency of severe droughts (Worrall et al., 2004). Changes in land-use and management practices may in most cases be expected to result in increased decomposition of soil organic matter and release of TOC. Mitchell and McDonald (1995) found that areas of greatest peat drainage density were the most important sources of TOC. In contrast, peatland drainage decreased total inorganic carbon (TIC) and TOC concentrations in the long term, but this did not affect carbon export from small boreal catchments (Rantakari et al., 2010). Studies in Swedish headwater catchments have shown elevated TOC concentrations and fluxes after clear-felling (Schelker et al., 2012), but comparable studies in Finland have shown limited effects on TOC (Palviainen et al., 2013). In the Fennoscandic forest, land management practices are typically not sufficiently widespread to explain long-term TOC trends, because the effect on TOC of climatic variability is often large (e.g., Kortelainen et al., 1997). 1225
€ et al. 1226 A . L E P I S T O Decreasing mineral acid deposition may have resulted in an increase in organic acidity, resulting in increased TOC concentrations (e.g., Evans et al., 2005; Monteith et al., 2007; Haaland et al., 2010). It has been argued that increased TOC concentrations in 522 remote lakes and streams were due to a decrease in mineral acid and base cation deposition (Monteith et al., 2007). Recently, Oni et al. (2013) ascribed some of the change in DOC at a long-term monitoring site in northern Sweden to the subtle effects of declines in acid deposition. Sarkkola et al. (2009) found that hydrometeorological factors were more important than declining acid deposition in predicting TOC trends in boreal headwater streams in eastern Finland. However, the most important mechanisms behind the increasing TOC trends remain a subject of discussion (e.g., Roulet & Moore, 2006), and only a few studies have dealt with trends in organic carbon fluxes as compared with trends in concentrations. Longer wet or drought periods may significantly contribute to leaching. Wet periods may have strong regional effects, i.e., higher groundwater tables leading to closer-to-surface flow paths, and higher TOC leaching. It may not be changes in the amount of runoff but rather its source, i.e., as flow paths shift new, richer sources of TOC are accessed (Worrall & Burt, 2007). Consequently, droughts could augment TOC production. Changes in the relationship between flow and TOC were observed after a severe drought that persisted even through more minor droughts (Worrall & Burt, 2004). The snow pack in winter insulates soil from belowfreezing air temperatures. This facilitates a significant amount of biological activity (Groffman et al., 2001; Campbell et al., 2005). Northern forest ecosystems typically experience a deep and persistent snow pack, but future changes in climate may alter snow pack dynamics and thus winter soil temperatures. Snow removal experiments indicate an inverse relationship between snow depth and soil frost (e.g., Groffman et al., 2001; Decker et al., 2003). However, analyses of historical data suggest that warmer winters could result in either an increase or decrease in the frequency of soil frost (Lindstr€ om et al., 2002; Henry, 2008). Models that incorporate soil frost dynamics predict that warmer winters may result in fewer days with soil frost (Ven€ al€ainen et al., 2001; Campbell et al., 2010), although mid-winter soil frost may become more common (Ven€ al€ ainen et al., 2001). Because changes in winter climate are likely to alter belowground temperature regimes, it is important to understand how changes in soil frost dynamics could impact forest processes such as C and N cycling (Reinmann et al., 2012). Most of organic C in boreal surface waters is dissolved as opposed to particulate: in Finnish rivers,
where both total and filtered samples have been analyzed, particulate organic C fractions (POC) have been very minor and an average of 94% was found in dissolved form (Mattsson et al., 2005). The dominance of dissolved organic C and N fractions suggests very low retention of C and N in fast-flowing northern rivers such as Simojoki (Lepist€ o et al., 2006). In the boreal zone, in northern Finland, typical climate features are long winters (5–7 months) with continuous snow cover, deep soil frost, and ice cover over rivers and lakes. A snowmelt–induced spring flood in late April–May dominates the annual hydrological pattern. Smaller flow peaks occur in autumn due to rainfall. Most TOC leaching occurs during these high flow periods (Lepist€ o & Kortelainen, 2011). The Simojoki watershed experiences low, declining sulfate deposition and limited human impacts apart from forestry. Process-based and conceptual models are often used to check whether we understand the controls on surface water DOC dynamics (Futter et al., 2007; Ledesma et al., 2012; Oni et al., 2012). The Integrated Catchments model for Carbon (INCA-C) has been applied to headwater catchments in Fennoscandia (Futter & de Wit, 2008; Futter et al., 2008, 2009, 2011) and to larger, temperate catchments in Sweden (Ledesma et al., 2012) and Canada (Oni et al., 2012). In this paper, long-term changes (>40 years) in organic carbon concentrations and fluxes, and the factors behind these changes were studied on the boreal Simojoki river system. We investigated the impact of soil frost, temperature, and drought on organic C concentrations, using a monthly data set of possible drivers for the study period 1971–2008. To assess the possible impacts of land management, climate, landscape position, and the potential importance of drought, the integrated catchment model INCA-C was applied to the river system to simulate changing C concentrations and fluxes all through the way from the uppermost lakedominated reach down to the river outflow.
Material and methods
Site description, forest management, hydrological monitoring, and water sampling The Simojoki river (watershed area 3160 km2) (Fig. 1) discharges to the Gulf of Bothnia of the Baltic Sea. It is a salmon river in near-natural state. Mineral soil forests (39%), and peatlands and peatland forests (53%) together dominate in the watershed. The dominant human impact is forest management including drainage and felling: of the total P load, forestry covers 8770 kg yr1 of the total of 18 860 kg yr1, i.e., 47%. Of the total N load, forestry covers 89 000 kg yr1 of the o et al., 1995). total of 230 500 kg yr1, i.e., 39% (Perkki€
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
3 Fertilizing (%) Clear-cutting (%) Supplementary drainage (%)
2
Forest drainage (%)
1
0 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008
Agriculture is practiced only in 2.7% of the watershed area, and point sources (municipal wastewaters) have negligible contribution to the total nutrient load (2.0% N, 2.4% P) (Perkki€ o et al., 1995; Rankinen et al., 2002; Lepist€ o et al., 2004). The total length of the river between the outlet of Simoj€arvi, a large lake in the upper reaches of the watershed, and the sea is 193 km. In the Simojoki watershed, forest management was most intensive in the early 1970s. At that time, annual forestry activities covered 2.0–2.5% of the area, decreasing to 0.8–1.0% annually during the late 1980s–1990s (Fig. 2, data based on the Finnish Statistical Yearbooks of Forestry by METLA). The most evident change has been in the spatial intensity of firsttime peatland drainage for forest production, which has been replaced by supplementary drainage works. Forest fertilization (N–P–K) was almost discontinued in the 1990s but has increased in recent years. The percentage of cut area varied from year to year with no major long-term change. Cut areas are scattered throughout the river basin, with relatively minor annual treatment areas (0.2–0.6% in 1968–2009). Intensive forestry in the 1970s may have caused some increase in DOC concentrations at the river outlet (Lepist€ o et al., 2008). Simojoki has two discharge gauging stations that are used for calibration and validation of the hydrological part of the model; one is located at the river outlet, the other upstream at the Hosionkoski rapids (Fig. 1). Discharge has been monitored on a daily basis since 1965 at the river outlet. The annual average discharge at the outlet (1971–2000) was 40.4 m3 s1, i.e., annual runoff was 403 mm. Water samples have been analyzed from nine locations along Simojoki during the study period 1962–2008 by a regional environment centre. High variability is found in sampling periods and frequency between the reaches. At the outlet, 4–5 samples annually in 1962–1981 and 10–25 samples annually in 1982–2008 have been analyzed. These nine locations correspond to the reaches used in the INCA-C model calibration (Fig. 1).
Forest treatments (% of watershed)
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1227
Fig. 2 Annual forestry treatments (fertilizing, clear-cutting, forest drainage and supplementary drainage) in 1968–2009, as percentages of the watershed area.
Chemical oxygen demand (CODMn) was analyzed by titrimetric determination with KMnO4 during the whole observation period 1962–2008. TOC was estimated from CODMn concentrations (TOC = 0.675*CODMn + 1.94; Kortelainen, 1993), because CODMn measurements were available for the whole period. This equation is also valid in the Simojoki case, where the correlation coefficient between measured TOC and COD concentrations was high, i.e., 0.88 (R€ aike et al., 2012). In the Simojoki river, differences between TOC and DOC are very small (Mattsson et al., 2005; T. Mattsson, unpublished). The patterns of TOC that are analyzed in this study can hence directly be compared and discussed in light of literature values and patterns of DOC.
Seasonal data sets of climate, hydrology, and TOC A monthly climate data set was assembled for the study period of 1971–2008, from which seasonal average values were
Fig. 1 Map of the Simojoki watershed showing the subcatchment areas, reaches 1–9, and the locations of water quality and discharge monitoring sites. © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
€ et al. 1228 A . L E P I S T O calculated. The data set covers meteorological drivers including temperature, precipitation, and frost depth. Frost depth during 1971–2008 was available from the groundwater and soil frost data bank Povet by SYKE. Frost is typically measured in forest sites one to five times a month, totaling approximately 20 measurements during the whole winter period. The station Ylitornio R1301, closest to the Simojoki watershed, was used here. Temperature and precipitation observations used were areal weighted values based on meteorological stations operated by the Finnish Meteorological Institute in the region. The daily temperature and precipitation values were also inputs to national scale watershed simulation and forecasting system (WSFS), which has been widely used in operational hydrology in Finland since 1990 (Vehvil€ainen, 1994). Daily hydrologically effective rainfall (HER; mm) and soil moisture deficit (SMD; mm) were simulated by the WSFS model and converted into monthly values. Observed runoff for the study period 1971–2008 was obtained from the discharge station close to the Simojoki river outlet (Fig. 1). Annual or monthly TOC fluxes were estimated by interpolating observed TOC concentrations linearly to estimate daily concentrations, which were then multiplied by daily flow and summed.
INCA-C model description The Integrated Catchments model for Carbon (INCA-C) is described in detail by Futter et al. (2007, 2009). The INCA-C model describes the major factors and processes controlling TOC in surface waters. There are four components to INCA-C: (i) a GIS interface used to define subcatchment boundaries and identify the areas of different land cover classes; (ii) an external rainfall-runoff model used to calculate HER and SMD; (iii) a land phase hydro-chemical model simulating material fluxes through the soil column and transformations between chemical stocks; and (iv) an in-stream model simulating the transformations in the aquatic phase. The INCA-C model is calibrated against measured time series of surface water TOC concentrations and flow. Briefly, INCA-C simulates the hydrological fluxes within a catchment or subcatchments and the masses and fluxes of solid organic (SOC), total (dissolved) organic (TOC) and dissolved inorganic carbon (DIC) in soils and surface waters (Futter et al., 2007). HER is the amount of water arriving at the soil surface as precipitation or snow melt that will eventually contribute to runoff. It may flow directly to the surface water or percolate into the upper soil box. Water in the upper soil box may flow to the stream or percolate to the lower soil box. All water in the lower soil box eventually runs off to the surface water. The breakdown of litter and root material contributes SOC and TOC in the upper soil box. Sorption and desorption processes control the transformation of organic carbon between TOC and SOC, and mineralization controls the transformation of SOC and TOC to DIC, part of which is lost to the atmosphere through degassing. TOC and DIC are transported advectively by water movement from the upper to lower soil boxes and from the soil to surface waters. Within INCA-C, all carbon transformations in soils and surface waters are modeled as a series of first-order differential
equations. All in-soil rate coefficients are dependent on soil temperature and moisture (Eqn 1). Soil temperature is simulated using observed air temperature and a model developed by Rankinen et al. (2004). The effect of soil temperature on the rate coefficients in INCA-C is simulated using a Q10 type model. The rate coefficient for the effect of soil moisture is a linear function of soil moisture content. The processes operate at a maximum rate when the simulated SMD is equal to zero. Processes cease when the SMD is greater than SMDMax, a calibrated threshold representing the maximum SMD at which carbon processing may occur. SMDMax minðSMD, SMDMax Þ ð1Þ m ¼ QðTSoil 20Þ SMDMax Changes in the mass of organic carbon in a soil box are simulated as a combination of soil temperature and moisture effects (m), and the mass of organic carbon sorbed (kSTOC, kg C d1) and desorbed (kDSOC, kg C d1) (Eqn 2). The base rates of the desorption (kD) and sorption (kS) coefficients are estimated during model calibration. dTOC ¼ mðkD SOC ks TOCÞ dt
ð2Þ
In INCA-C, TOC can be mineralized in the open water by both microbial and photolytic processes. In the simulations presented here, only microbial mineralization is assumed. The simulated rate of microbial mineralization of TOC is linearly dependent on water temperature (TW). Water temperature is assumed to be equal to air temperature so long as the air temperature is above a user-specified threshold; otherwise, water temperature is set to the threshold value. The model has been further modified here to simulate the so-called ‘enzymic latch’ mechanism (Worrall et al., 2004). The ‘enzymic latch’ mechanism has been proposed as a mechanism to explain the increase in DOC concentrations observed following severe drought in UK peatlands. This mechanism is simulated in INCA-C as an increase in the rate of SOC to DOC transformation when the SMD crosses a critical threshold representative of severe drought conditions. The rate increase due to drought declines with time, returning to the baseline value after a user-specified number of years.
INCA-C model inputs and its calibration to Simojoki Hydrological input to the INCA-C model was calculated using the above WSFS model. Daily values of temperature (°C), HER, and SMD were calculated for the study period 1962– 2008, and taken as input to the INCA-C model. Land cover in the catchment was divided into six classes representing (i) mineral forests; (ii) cut mineral forests; (iii) undrained peatlands; (iv) peatland forests; (v) peatlands subject to ditching; and (vi) agriculture. Land use proportions were estimated for each of the 9 reaches based on values in Rankinen et al. (2002), forest management time-series in Fig. 2, and data sets based on the Finnish Statistical Yearbooks of Forestry by METLA. Effects of forest harvest and peatland drainage were assumed to last for 10 years, it was estimated that 20% of the peatlands
© 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1229 in the catchment were drained in 1962 and that by 2009, about 60% of peatlands had been subject to primary drainage. Until the end of 1991, about 1030 km2 of the whole watershed area was estimated to have been drained, which is in agreement with Perkki€ o et al. (1995) who report that in 1991 over 30% of watershed (i.e., about 60% of the peatlands and peatland forests) had been drained. INCA-C model parameterization was based on values for forests and peatlands from the ValkeaKotinen reference catchment in southern Finland (Futter et al., 2008, 2009). The effects of forest harvest on soil temperature were simulated by adjusting the thermal conductivity parameters of the soil-temperature submodel in INCA-C (Rankinen et al., 2004) so that the simulated soil warming matched that observed by Schelker et al. (2012). A manual calibration was performed which attempted to maximize the Nash–Sutcliffe (NS) statistics (Nash & Sutcliffe, 1970). The results of this calibration were used to define initial parameter ranges for a sensitivity and uncertainty analysis based on Markov Chain Monte Carlo (MCMC) principles. Initial parameter ranges for terrestrial and in-stream process rates were defined as 20% of the value obtained in the manual calibration. In-stream flow/velocity parameters and a temperature-dependent TOC mineralization rate were allowed to vary. Land-phase parameters related to sorption desorption rates, soil moisture thresholds, soil temperature response, and postdrought increase in TOC production were also allowed to vary. The MCMC analysis (described in Futter et al., 2013) attempted to give equal weight to simulations of flow, TOC from the lake outlet and TOC in the main stem of the river. Briefly, the MCMC analysis assumes equifinality in the high-dimensional parameter space sampled. A posterior distribution of credible parameter values is derived from the best performing parameter set in each of an ensemble of model runs. The acceptance criteria for new parameter sets in the MCMC analysis was based on weighted NS statistics for TOC in the lake outflow (reach 1, weight = 8), TOC in the main river (reaches 2–9, weight = 1 for each reach) and flow (reaches 5 and 9, weight = 4 for each reach). This was performed so as to ensure that model calibration would be sensitive to parameters related to in-lake and terrestrial processes, and so that a balance would be achieved between fitting to flow and TOC concentration. The MCMC analysis comprised 100 model chains, each consisting of 300 model runs. The parameter set associated with the best performing model run from each chain was retained for sensitivity and uncertainty analysis. Parameter sensitivity was assessed by using the Kolmogorov–Smirnov test to compare the distribution of parameter values in the best performing ensemble of model runs to a rectangular distribution.
Results
but decreased somewhat during the 2000s (2001–2008), to 17 200 t yr1. The average annual runoff was 27% higher during the 1990s (477 mm) than during the 1980s (375 mm), and therefore, runoff accounts for a major part of the increase in the TOC outputs but not all. Figure 3 also highlights increasing temperatures: in the latter part of the study period (1986–2008) average temperature was clearly higher, 1.2 °C, as compared with the first half (0.2 °C in 1962–1985). Year-to-year hydrological variability is very high and impacts considerably on TOC fluxes (Fig. 3). A dry period with low runoff and low regional groundwater table (1994–1997; average runoff 369 mm) was followed by a considerably wetter period (1998–2000; average runoff 618 mm) with high groundwater tables, high TOC concentration peaks and large leaching fluxes. High water tables contribute more surface runoff from the TOC-rich soil layers, and more runoff from riparian areas to the stream than low groundwater tables.
Soil frost and TOC concentration in the subsequent spring Spring TOC concentrations at the river outlet (reach 9) during the spring snow melt period in 1971–2008 correlated with the maximum soil frost during the preceding winter (R2 = 0.37) (Fig. 4). Average spring concentrations were clearly higher in the four mildest winters (14.3 mg l1) with a maximum frost of only 40–70 cm, as compared with four deepest frost winters with 110120 cm (9.9 mg l1). This means that also spring TOC fluxes after a mild winter are considerably higher, over double: 9600–12100 tonnes in April–May after mild ‘low-frost’ winters vs. 3700–6200 tonnes after deep-frost winters. There was practically no relationship (R2 = 0.10) between maximum soil frost and the amount of subsequent spring runoff. This means that deep-frost winters (110–120 cm) were followed by only average spring runoff, probably without any substantial amounts of surface runoff and overland flow over frozen soils. The soil frost correlated considerably better to TOC concentrations than to amounts of runoff which means that high spring flows may occur after very variable winters (low- to medium-frost conditions), while high concentrations of TOC during spring snowmelt only occur after low-frost winters.
Long-term TOC fluxes and hydrological variability Annual TOC fluxes at the river outlet displayed a statistically significant increase over the study period (Mann Kendall statistic, MK = 2.3, P = 0.01, Fig. 3). The average TOC flux increased by 38% during the 1990s, from 14 800 t yr1 in the 1980s to 20 500 t yr1 in the 1990s © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
Summer precipitation, drought, and TOC Summer precipitation (June–August) was an important driver of summer SMD (Fig. 5b), which in turn explained most (R2 = 0.66) of the variation in summer TOC concentrations at the river outlet (reach 9) (Fig. 5a).
€ et al. 1230 A . L E P I S T O
3 2 1 0 –1
1980
1982
1984
1986
1988
1990
1992
1980
1982
1984
1986
1988
1990
1992
2008
1978 1978
2006
1976 1976
2004
1974 1974
2002
1972 1972
2000
1970 1970
1998
1968 1968
1996
1966 1966
1994
1964 1964
Soil moisture deficit (mm)
–3
1962
–2
1962
Air temperature (°C)
4
0 –5 –10 –15 –20 –25 –30 –35 –40 –45 800
Runoff (mm)
700 600 500 400 300 200 100 0
30 000 25 000 20 000 15 000 10 000
2008
2006
2004
2002
2000
1998
0
1996
5000 1994
TOC flux (tonnes yr–1)
35 000
Fig. 3 Annual average air temperature (°C), soil moisture deficit (mm), annual runoff (mm), and annual total organic carbon (TOC) flux (tonnes yr1) in the Simojoki watershed in 1962–2008 (for runoff and TOC flux, data from 1966 are available).
Summer precipitation and TOC concentrations were closely linked (R2 = 0.51) (Fig. 6), with an average increase in TOC from 7.5 mg l1 in the driest summer to 15.5 mg l1 in the wettest summer. Summer TOC fluxes in the four wettest summers were over 5-fold higher: 5700–12800 tonnes (average 9400) vs. 900– 2300 tonnes (average 1700) in the four driest summers, respectively. In contrast, summer temperature was not correlated with either summer TOC concentrations (R2 = 0.08) or summer precipitation (R2 = 0.04). Close links between summer precipitation and TOC concentrations (Fig. 6) probably reflect closer-to-surface
flow paths and higher areal groundwater tables, with higher TOC concentrations in those surface soil layers, which together allow strong C fluxes to the river system. These responses to higher water volumes are rapid, while higher soil temperatures may increase TOC fluxes but with longer delays.
INCA-C calibration results Model performance statistics from the ensemble of best performing INCA-C calibrations are shown in Table 1. For all reaches, model simulations were able to achieve © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1231 20
TOC (mg L–1) in summer
Spring TOC concnentration (mg L–1)
20
15
10
5
0
0
20
40
60
80
100
120
140
Precipitation (mm) in summer 0
20
40
60
80
100
120
140
20
Fig. 6 Summer total organic carbon (TOC) concentrations vs. precipitation of that period (1971–2008).
Table 1 Model performance statistics [R2 and Nash–Sutcliffe (NS)] for the INCA-C calibrations from uppermost reach 1 down to reach 9 (river outlet)
(a)
R2
15
Reach
10 y = 0.0022x2 + 0.32x + 21.19 R² = 0.66
5 0 –100
–80
–60
–40
–20
0
SMD (mm) in summer 0 –10 –20 –30 –40 –50 –60 –70 –80 –90 –100
y = 0.074x + 6.72 R² = 0.51
5 0
Fig. 4 Spring total organic carbon (TOC) concentrations vs. deepest monthly frost depth of the preceding winter (1971–2008).
TOC (mg L–1) in summer
10
y = –0.0007x2 + 0.054x + 13.26 R² = 0.37
Maximum soil frost (cm)
SMD (mm) in summer
15
(b)
y = –0.0037x2 + 1.09x –99.36 R² = 0.74 0
20
40
60
80
100
120
140
Summer precipitation (mm) Fig. 5 (a) Summer total organic carbon (TOC) concentrations vs. soil moisture deficit (SMD mm) of that summer (1971–2008), together with (b) relationship between precipitation and SMD.
NS statistics >0 for both TOC and flow. Observed and simulated TOC values for the Simoj€ arvi outlet (reach 1), Portimoj€ arvi (reach 5), and the river outlet (reach 9) are shown in Fig. 7. The INCA-C calibration was able to reproduce the markedly different seasonal patterns of TOC at Simoj€ arvi vs. the other reaches. In most cases, the INCA-C calibration was able to reproduce the range of observed values; however, it tended to miss some of © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
TOC 1 2 3 4 5 6 7 8 9 Flow 5 9
NS
Min
Average
Max
Min
Average
Max
0.10 0.36 0.35 0.62 0.44 0.46 0.46 0.44 0.38
0.13 0.39 0.38 0.64 0.47 0.49 0.49 0.47 0.42
0.14 0.43 0.40 0.66 0.50 0.51 0.51 0.49 0.44
0.78 0.54 0.07 0.25 0.17 0.15 0.26 0.26 0.09
0.00 0.15 0.05 0.36 0.30 0.27 0.38 0.37 0.07
0.08 0.08 0.10 0.43 0.39 0.31 0.46 0.44 0.15
0.83 0.80
0.85 0.82
0.86 0.83
0.81 0.63
0.83 0.81
0.84 0.83
the high autumn concentrations observed at the outflow (Fig. 7). INCA-C estimated fluxes from individual reaches are shown in Fig. 8. Note the large difference between Simoj€arvi (reach 1) and the downstream reaches. This difference is a result of simulated in-lake mineralization of TOC in Simoj€arvi. The difference in areal fluxes between reaches below the lake is a function of the proportion of peat forest land cover in the subcatchment. By combining the reach specific fluxes with the respective land use proportions, it was possible to estimate land use specific fluxes of 3.9, 7.7, 8.2, 8.0, 8.3, and 1.0 g TOC m2 yr1 for mineral forest, cut mineral forest, undrained peatland, peatland forest, peat ditching, and agriculture, respectively. Using these values, an average flux of 5.9 g TOC m2 yr1 is estimated for Simoj€arvi. Comparing this with the INCA-C estimated flux of 3.1 g TOC m2 yr1 at the Simoj€arvi outlet
€ et al. 1232 A . L E P I S T O 30
Reach 1
25
TOC (mg L–1)
20 15 10 5 0 –5 30
Reach 5
25
TOC (mg L–1)
20 15 10 5 0 –5 30
Reach 9
25
TOC (mg L–1)
20 15 10 5 0
2008
2006
2004
2000
2002
1998
1996
1994
1990
1992
1986
1988
1982
1984
1980
1978
1974
1976
1970
1972
1968
1966
1962
1964
–5
Fig. 7 Observed and modeled total organic carbon (TOC) concentrations at reaches 1 (upstream), 5 (mid-catchment) and 9 (close to outlet) during 1962–2008. The thick line shows the average value from the ensemble of 100 best performing INCA-C model runs, while the lower thin lines show the deviations of the maximum and minimum ensemble values from the ensemble average.
indicates that ca. 50% of the simulated TOC inputs from the surrounding catchment are removed by in-lake processing in Simoj€ arvi. There was relatively little difference in model performance for downstream reaches (Fig. 8). This suggests that TOC dynamics in these reaches, with their shorter residence times, can be credibly simulated on the basis of terrestrial processes alone.
Posterior distributions of model parameters only had significant Kolmogorov–Smirnov statistics for the flow velocity ‘a’ and ‘b’ terms and DOC mineralization rate in reach 1 and the temperature response and DOC to SOC sorption rate in the peatland forest land cover type. The model was most sensitive to estimated rates of TOC mineralization in reach 1, which was © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1233
TOC (g m–2 yr–1)
was related to organic matter increases, but flow emerged as an even more important driver of OM variability (Erlandsson et al., 2008). Warming can affect DOC export in different ways, depending on whether it is accompanied by increased or decreased precipitation; the variation is related to hydrology as well as to biological productivity, and to how productivity is balanced by decomposition (Tranvik & Jansson, 2002).
Impacts of soil frost
Reach
Fig. 8 Modeled total organic carbon (TOC) fluxes (g m2 yr1) from uppermost reach 1 down to reach 9 (river outlet) during 1962–2008. Fluxes were modeled using the INCA-C model.
dominated by Simoj€ arvi, which were controlled by both the water residence times (flow ‘a’ and ‘b’ values) and the mineralization rate. Model performance was not sensitive to any parameter in the agriculture land cover type, probably due to its relatively minor contribution to total land use in the Simojoki catchment. Somewhat surprisingly, model results were not sensitive to any parameters in the forest land cover type, including the drought-related ‘enzymic latch’ parameters.
Discussion
Trends in TOC flux as a result of climate Surface water organic carbon concentrations are changing: increasing trends in concentrations and fluxes into large watercourses have been reported across Europe (Erlandsson et al., 2008; Lepist€ o et al., 2008; Klavins et al., 2012; P€ arn & Mander, 2012). In this study, annual TOC fluxes at Simojoki watershed displayed a statistically significant increase over the study period. Respectively, P€ arn and Mander (2012) found significant trends in TOC in five of 11 rivers in Estonia. They conclude that the main factors in the increase in organic carbon export are longer drought periods driven by climate change, magnified by drainage works. R€ aike et al. (2012) studied changes in DOC fluxes in the period 1975–2010 using data from 29 Finnish river basins. They found increasing normalized DOC fluxes in 8 river basins, decreasing in 7 basins, and no trends in 14 basins. In 28 large Swedish catchments, decline in anthropogenic sulfate deposition © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
The depth of soil frost had a statistically significant effect on TOC concentration in the subsequent spring (Fig. 4). Laboratory studies on Norwegian forest and heathland soils have shown that the impacts of soil frost on TOC are complex and highly variable (e.g., Austnes & Vestgarden, 2008; Austnes et al., 2008). Results from laboratory experiments using soils from a temperate hardwood forest indicate that severe soil frost (15 °C) resulted in lower losses of C compared with unfrozen soils during snow melt, while no significant impact of mild soil frost (0.5 °C) was detected (Reinmann et al., 2012). This suggests that severity of soil frost plays an important role in potentially reducing C losses during spring snow melt. These results are comparable with our results at the large watershed scale. In laboratory experiments with German forest soils, there was on the contrary an increase in TOC in response to freezing of organic soil horizons at temperatures of 8 and 13 °C, but not at 3 °C, a much more common field temperature (Hentschel et al., 2008). It is likely that the extent to which C and N cycling during snow melt is altered in response to changes in winter climate depend on both the presence of soil frost and the temperature to which soils freeze (Reinmann et al., 2012). The few short-term field-scale manipulation studies performed to examine the relationship between soil frost and TOC in soil solution have reported no or inconclusive results (Fitzhugh et al., 2001; Austnes et al., 2008). In northern Sweden, contrary to our results, it was found that longer and colder winter results in higher TOC concentrations in a very small boreal headwater stream during the subsequent snow melts (Haei et al., 2010). In very small systems, riparian zones probably play a more important role than in the larger river basin scale studied here. Field studies in Norway noted increases in TOC in response to mild freezing (Austnes et al., 2008; Kaste et al., 2008). In a field study in Germany, no relationship between TOC and soil frost was observed (Hentschel et al., 2009). Clearly, there is a need for further research on the factors regulating TOC response to soil freezing disturbance (Groffman et al., 2011).
€ et al. 1234 A . L E P I S T O The extent of soil frost development ultimately depends on the integrated effect of air temperature and the extent and duration of the snow cover. Furthermore, development and timing of snow cover is of importance for the depth of frost, and further to C leaching processes. In mild, low-frost winters also in the boreal zone, higher amounts of water flow in surface zone layers where the TOC concentrations are higher, contributing to higher TOC losses in the subsequent spring (Fig. 4). In their recent review of freeze-thaw events and C and N losses from soils, Matzner and Borken (2008) state that future milder winter climate with fewer periods of soil frost may result in greater losses of C from soils that are presently influenced by extended frost periods. The probability of mild winters with less soil frost probably increases in a changing climate, in northern ecosystems (e.g., Denman et al., 2007). These changes might increase spring C fluxes, through the concentration increase, in those cases where snow melt flow peaks remain the same, or occur somewhat earlier in spring.
Precipitation, temperature, and drought effects Summer precipitation was an important driver for summer SMD, which in turn explained most of the variation in summer TOC concentrations (Fig. 5). Also, summer precipitation and TOC concentrations were closely linked. This is relevant as Rantakari and Kortelainen (2005) showed that open water precipitation and CO2 fluxes from a set of large Finnish lakes were very closely correlated, whereas temperature and CO2 were not linked to each other. Although the overall mechanisms behind TOC and CO2 patterns are dissimilar, both seem to be closely linked to summer precipitation in boreal Finnish lakes and rivers, whereas links to temperature are minor. Further, global hotspots of stream and river CO2 evasion were demonstrated in regions where precipitation was high (Raymond et al., 2013). Isotopic compositions of DOC and CO2 in boreal Finnish headwater streams have been shown to be very similar, suggesting a predominantly single and consistent C source, i.e., decomposition of organic matter (Billett et al., 2012). There are several indirect lines of evidence to support the conclusion that droughts can be major drivers of TOC release. These include the following: (i) change in the relationship between flow and TOC after a severe drought; (ii) soil respiration and TOC production become decoupled after a severe drought; (iii) stepwise changes in the flux of TOC are observed after droughts. However, direct evidence for a drought effect on TOC concentrations is lacking (Worrall & Burt, 2007). In
Simojoki study, we found some indication of stepwise changes in the TOC flux after droughts, particularly when comparing wet period of 1998–2000 to earlier drought period of 1994–1997 (Fig. 3), but INCA-C model results were not sensitive to any parameters in the forest land cover type, including drought-related parameters. In Poland, the autumn 2000 drought was one of the most severe in many years, resulting in about 30% decrease in TOC concentrations. Average TOC and water colour were two times higher in the wetter autumn than in dry period (Zielinski et al., 2009). In a Swedish catchment, stepwise increases in DOC concentration did follow some, but not all, summer droughts; the most notable changes occurred following a sequence of dry summers (Jennings et al., 2010).
Importance of scale – forestry effects, landscape scale controls, and INCA-C Forest management activities can either increase or decrease TOC concentrations and fluxes. Numerous studies of headwater catchments in Fennoscandia have shown elevated fluxes and concentrations following final felling (Nieminen, 2004; Kreutzweiser et al., 2008; Schelker et al., 2012) while Palviainen et al. (2013) did not observe any significant increases. Piirainen et al. (2002) caution that final felling effects on TOC flux may be negligible at larger spatial scales. Our results support this latter conclusion and are consistent with the observations of Palviainen et al. (2013). Any increase in TOC associated with final felling may be offset by the decrease in TOC export that has been observed following ditch maintenance (Nieminen et al., 2010). While forest management activities have potentially severe and undesirable consequences at the local scale, these effects are not readily apparent at the landscape scale. Landscape type exerts the dominant control on TOC export in boreal rivers. As noted by R€aike et al. (2012), upland forest and wetlands have quite different TOC exports. In addition, more than half of the TOC exported from boreal catchments may be consumed by in-lake processes (Tranvik et al., 2009). These landscape level controls were reproduced in the INCA-C simulations presented here. The areal flux of TOC from the uppermost lake dominated subcatchment was almost half that of downstream subcatchments due to in-lake losses of TOC (Fig. 8). The INCA-C model (Futter et al., 2007) has been designed to simulate the effects of landscape type, climate, and land management on soil and surface water TOC concentrations and fluxes and has been successfully applied in numerous boreal headwater catchments. Our empirical and modeling results by INCA-C suggest that climate change driven patterns in runoff, © 2013 John Wiley & Sons Ltd, Global Change Biology, 20, 1225–1237
L O N G - T E R M O R G A N I C C A R B O N F L U X E S I N A B O R E A L W A T E R S H E D 1235 soil moisture, and temperature were more important than temporal patterns of land management in controlling surface water TOC concentrations. The absence of any clear forest management effect in the large-scale simulations presented here shows the importance of scale when addressing the impacts of forest management activities on surface water quality (Futter et al., 2010). The large changes in TOC flux and concentration which are sometimes observed in headwater catchments following final felling (Schelker et al., 2012; Palviainen et al., 2013) may be impossible to detect in larger rivers. Due to the relatively slow rates of tree growth and long forest rotation periods, forest management activities affect a relatively small amount of the total catchment at any point in time. The INCA-C simulated fluxes from peatlands were relatively similar (8.0– 8.3 g TOC m2 yr1), regardless of the management activities. The higher TOC exports from upland forests following final felling (7.7 vs. 3.9 g TOC m2 yr1) are consistent with empirical studies (Schelker et al., 2012) but the relatively small (
