Science of the Total Environment 408 (2009) 92–101

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r.com/ locate /sc i totenv

Trends in hydrometeorological conditions and stream water organic carbon in borealforested catchments

Sakari Sarkkola a,⁎, Harri Koivusalo b, Ari Laurén b, Pirkko Kortelainen c, Tuija Mattsson c, Marjo Palviainen d,Sirpa Piirainen b, Mike Starr d, Leena Finér b

a Finnish Forest Research Institute, Vantaa Research Unit, P.O. Box 18, FI-01301 Vantaa, Finlandb Finnish Forest Research Institute, Joensuu Research Unit, P.O. Box 68, FI-80101 Joensuu, Finlandc Finnish Environment Institute, P.O. Box 140, FI-00251 Helsinki, Finlandd University of Helsinki, Department of Forest Ecology, P.O. Box 27, FI-00014 University of Helsinki, Finland

⁎ Corresponding author. Tel.: +358 10 211 2312; faxE-mail address: sakari.sarkkola@metla.fi (S. Sarkkola

0048-9697/$ – see front matter © 2009 Elsevier B.V. Aldoi:10.1016/j.scitotenv.2009.09.008

a b s t r a c t

a r t i c l e i n f o

Article history:Received 6 February 2009Received in revised form 19 August 2009Accepted 14 September 2009Available online 12 October 2009

Keywords:CatchmentClimate changeDissolved organic carbonExportForestRunoffPeatland

Temporal trends in stream water total organic carbon (TOC) concentration and export were studied in 8forested headwater catchments situated in eastern Finland. The Seasonal Kendall test was conducted toidentify the trends and a mixed model regression analysis was used to describe how catchmentcharacteristics and hydrometeorological variables (e.g. precipitation, air and stream water temperatures,and atmospheric deposition) related to the variation in the concentration and export of stream water TOC.The 8 catchments varied in size from 29 to 494 ha and in the proportion of peatland they contained, from 8to 70%. Runoff and TOC concentration were monitored for 15–29 years (1979–2006). Trends and variation inTOC levels were analysed from annual and seasonal time series. Mean annual TOC concentration increasedsignificantly in seven of the eight catchments. The trends were the strongest in spring and most apparentduring the last decade of the study period. The slopes of the trends were generally smaller than the variationin TOC concentration between years and seasons and between catchments. The annual TOC export showedno clear trends and values were largely determined by the temporal variability in runoff. Annual runoffshowed a decreasing trend in two of the eight catchments. Mean annual air and stream water temperaturesshowed increasing trends, most clearly seen in the summer and autumn series. According to our modelingresults, stream water temperature, precipitation and peatland percentage were the most important variablesexplaining annual and most seasonal TOC concentrations. The atmospheric deposition of SO4, NH4, and NO3

decreased significantly over the study period, but no significant link with TOC concentration was found.Precipitation was the main hydrometeorological driver of the TOC export. We concluded that stream waterTOC concentrations and exports are mainly driven by catchment characteristics and hydrometeorologicalfactors rather than trends in atmospheric acid deposition.

: +358 10 211 2202.).

l rights reserved.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

Natural organic matter (NOM) in surface waters has importantimplications for water quality and management, transportation oftrace metals, natural acidity, recovery from acid deposition, andcarbon cycling. NOM is usually measured as total organic carbon(TOC) if unfiltered or as dissolved organic carbon (DOC) if filtered, thedifference being particulate organic carbon. In Nordic surface waters,DOC has been found to constitute more than 90% of TOC (Mattssonet al., 2005; Kortelainen et al., 2006). DOC in streamwater from borealforested catchments is primarily allochthonous, originating from thedecomposition of organic matter and the metabolic activities of livingorganisms (Hudson et al., 2003), and net primary production in lakesand rivers is low (Jonsson et al., 2001).

During the last 10–15 years, DOC and TOC concentrations in lakesand rivers across northern Europe and North America have beenobserved to be increasing (e.g. Freeman et al., 2001a,b; Worrall et al.,2004a,b;Worrall and Burt, 2007; Eikebrokk et al., 2004; Vuorenmaa etal., 2006; Monteith et al., 2007). However, the cause of these increasesremains partly unclear.

Warm and wet years have been frequent during the period ofincreased DOC and TOC concentrations, indicating either increaseddecomposition of soil organic matter (SOM), increased transport orboth. Future climate change scenarios for northern Europe predictincreasing mean air temperatures, particularly for the winter andspring seasons, increasing winter precipitation and decreasingsummer precipitation (Alcamo et al., 2007). Such changes can beexpected to alter decomposition rates of SOM and the magnitude andtiming of the spring snowmelt and runoff, which dominates theannual water cycle in the boreal zone. Under such conditionswintertime runoff will increase and since a greater proportion of the

93S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

precipitation will fall as rain rather than snow, the snowmelt runoffpeak in spring will become smaller and occur earlier than before(Bates et al., 2008). Several studies have found a correlation betweenthe observed increase in lake and river water DOC concentrations andthe simultaneous increase in precipitation or air temperature(Freeman et al., 2001a,b; Evans et al., 2002; Clark et al., 2005; Worrallet al., 2004b; Ellis and Freeman, 2007; Erlandsson et al., 2008).

It has also been suggested that the pattern of DOC in lakes andstreams is controlled by the deposition of strong acid anions,particularly SO4

2− and marine Cl− (Kalbitz et al., 2000; Clark et al.,2005). However, the acid anion chloride in inland areas is of minorimportance compared to the coastal areas, where Cl− can be thedominant anion associated with high sea salt deposition (Monteithet al., 2007). Recent studies link the increase in the river and lakewater DOC and TOC concentrations to the decrease in the amount ofacid atmospheric deposition (Vuorenmaa et al., 2006; Monteith et al.,2007; Erlandsson et al., 2008). Changes in land use and forestmanagement may also be involved. Management practices such asclear-cutting and ditching temporarily enhance the release andtransport of TOC to surface waters (Kortelainen and Saukkonen,1998; Nieminen, 2004). Surface water TOC concentration andcatchment TOC export have been found to be related to a number ofcatchment characteristics. The proportion of peatlands within thecatchment area has been found to be particularly important (Arvola etal., 2004; Laudon et al., 2004; Mattsson et al., 2005; Kortelainen et al.,2006; Ågren et al., 2007, 2008). During low flow periods runofforiginates fromwater stored in catchment soils, and the streamwaterDOC concentration and export are controlled by the extent ofpeatlands within the catchment (Laudon et al., 2004; Ågren et al.,2007, 2008; Buffam et al., 2007). During spring the snow cover meltsover frozen soil and generates surface runoff that is not in contactwith the peat and has a low DOC concentration (Laudon et al., 2004).While these conditions prevail stream water DOC export is morecontrolled by other catchment characteristics, such as the catchmentsize, rather than by the peatland proportion (Ågren et al., 2007, 2008).

Trends in lake, river, and stream water DOC concentrations aremainly studied in terrestrial catchments with heterogeneous land-useandmanagement. Thus it is difficult to distinguish the role of co-existingfactors behind the trends. Although increasing trends in lake, river, andstream water DOC concentrations have been widely observed, thereasons for the trends are still partly unclear and even controversial.

This study focuses on time series from 8 intensively monitoredforested first-order headwater catchments located in eastern Finland,where the proportion of DOC from TOC in stream water is 93–98%(Kortelainen et al., 2006). The objectives were: 1) to analyze howannual and seasonal stream water TOC concentrations and exportsfrom boreal forested catchments have changed during the pastdecades, 2) to identify relationships between the trends of TOC level,the hydrometeorological conditions, and the amount of sulphate

Table 1General information on the studied forested catchments.

Catchment Location coordinates N; E Monitoring period Area ha Elevat

Murtopuro 63°46′;28°28′ 1979–2006 494 225

Liuhapuro 63°47′;28°29′ 1979–2006 165 225Suopuro 63°53′;28°40′ 1979–2006 113 200Kivipuro 63°52′;28°65′ 1979–2006 54 200

Välipuro 63°52′;28°65′ 1979–2006 86 200Porkkavaara 63°52′;29°10′ 1992–2006 72 182Kangaslampi 63°52′;28°57′ 1992–2006 29 184Korsukorpi 63°53′;28°40′ 1992–2006 69 198

a Above mean sea level.b Of the total land area of the catchment.

sulphur (SO4–S), ammonium (NH4–), nitrate– (NO3–), and totalnitrogen (N) depositions during the years 1979–2006, and 3) to assessthe role of hydrometeorological conditions and catchment character-istics in controlling TOC concentration and export.

2. Material and methods

2.1. Study areas

The 8 forested catchments included in this study are all locatednear to each other in eastern Finland and therefore have similarclimate and landscape characteristics (Table 1). The long term (1979–2006) mean annual precipitation in the area is 600 mm, 40% of whichfalls as snow. The temperature sum of the growing season (thresholdvalue +5 °C) is 1100 dd.

The catchments are underlain by mainly gneiss, granite, andgranodiorite bedrock and covered with stony sandy till with less than2% clay. The upland soils are iron podzols (Haplic Podzols) withtransitions to peaty podzols downslope and fibric histosols (peat) inthe lowest parts of the catchments. Most of the peat layers (78%) are<2 m thick. In this study paludified mineral soil sites with a con-tinuous peat layer were considered peatlands.

According to the Finnish site type classification system (Cajander,1949), the upland forests are mainly medium-rich Vaccinium-Myrtil-lus and poor Empetrum-Myrtillus types, and both minerotrophic andnutrient-poor ombrotrophic peatland types are found. The forests aredominated by Scots pine (Pinus sylvestris L.) and Norway spruce (Piceaabies L. Karst.), and mixed with Silver and Downy birch (Betulapendula Roth. and B. pubescens Ehrh. The peatlands are mostlyforested although the stand density varies considerably and treelessor poorly stocked mires are common in the Suopuro and Korsukorpicatchments.

While 3 of the 8 study catchments had not been subject tomanagement, 5 had and have varying amounts of clear-cutting, sitepreparation and peatland drainage present (Table 1). According toearlier studies (e.g. Uusivuori et al., 2008) the impact of forestmanagement on water quality can only be detected if the treatmentarea exceeds about 10% of catchment area and then approximately forthe first ten years after treatment. For two of ourmanaged catchments(Välipuro and Korsukorpi), the managed area was less than 10% andtherefore we consider these data sets not to be influenced bymanagement. In the case of the other 3 managed catchments(Murtopuro, Suopuro and Kivipuro), we excluded the first ten yearsof data after a management practice larger than 10% of the total area.

The establishment of the study catchments and the effect of forestmanagement practices on water quality and runoff have beenreported in more detail in e.g. Ahtiainen (1988), Seuna (1988),Finér et al. (1997), Ahtiainen and Huttunen (1999), Mattsson et al.(2006), and Laurén et al. (2009).

ion amsla m Peatland%b Treatment methods and year Cut% Drained%

50 Clear cutting 1983 58Soil preparation 1986Clear cutting 2002 4Drainage 1983 40

48 No management70 Drainage 1983 1332 Clear cutting 1983 56

Soil preparation 198656 Clear cutting 2001 916 No management9 No management

56 Clear cutting 2000 6

Table 2The average, minimum, and maximum values of TOC concentration, TOC export,hydrometeorological variables, and catchment characteristics of the studied catch-ments in eastern Finland.

Variable Average Max Min

TOCMean TOC concentration (mg l−1yr−1)a 24.8 69 8.4TOC export (kg ha−1yr−1)a 81.3 190.2 9.6

Hydrometeorological variablesRunoff (mm yr−1) 359 589 160Precipitation (mm yr−1) 612 746 479Mean air temperature (°C)a 1.53 3.3 −1.3Snowfall (mm yr−1) 241 339 182Mean snow depth (cm day−1)b 32 57 9Annual temperature sum (dd) 1163 1362 908Stream water temperature (°C)a 4.3 7.7 2.12SO4–S deposition (kg ha−1yr−1)a 3.9 6.9 1.5TOC deposition (kg ha−1yr−1)a 9.3 12.7 6.5NO3–N deposition (kg ha−1yr−1)a 1.8 2.8 1.1NH4–N deposition (kg ha−1yr−1)a 2.4 4.4 0.7Total N deposition (kg ha−1yr−1)a 5.4 8.8 2.8

Catchment characteristicsCatchment area (ha) 132 494 29Total length of main stream (m) 1275 3700 450Total length of ditches (m) 10,756 75,700 450Length of ditches/catchment area (m ha−1) 35.4 153.3 2.5Nutrient rich sites (%)c 45.0 95 20Peatland (%)c 42.0 70.0 8.0Drained area (%)c 6.6 40.0 0.0Mean tree stand volume (m3ha−1) 97 210 16Tree biomass (1000 kg ha−1) 90.1 300.0 13.1Peat volume (m3ha−1) 6428 14,724 1338Humus volume (m3ha−1) 279 454 216Humus biomass (1000 kg ha−1) 19.4 31.1 14.8Total mass of organic material (1000 kg ha−1) 882.2 1816.6 213.2Broad leaf tree stands (%)c 12 20 5

The number of study catchments is 8. The catchment and hydrometeorologicalvariables were applied as explanatory variables (x in Eq. (2)) in the models of TOCconcentration and export (for the models see Table 4).

a Annual value.b In spring during snow covered period.c Of the catchment area.

94 S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

2.2. Runoff and stream water TOC concentrations

The water level in the outlet stream was recorded at a streamgauging station and daily time series of runoff calculated using thestage–discharge relationship. Prior to 1995, water samples wereusually taken 1–5 times per month andmore frequently in springtime(up to 6–10 times per month). After 1995, the sampling was restrictedto the spring and autumn periods and no samples were collectedduring wintertime (December–February). In 2000, water samples atPorkkavaara, Korsukorpi and Kangaslampi were collected only inOctober and November. Stream water temperature was also recordedin the runoff gauging station at the time of water sampling.

The TOC concentrations were determined at the laboratory of theNorth Carelia Regional Environment Centre, using UV-persulphateoxidation followed by IR (Infra-Red) gas measurements or a high-temperature oxidation. Some of the TOCmeasurements for the periodof 1978–1991 were missing and the TOC concentration values wereestimated based on measured chemical oxygen demand (CODMn)using the following regression Eq. (1):

cTOC = 1:1218 + 0:6435cCODMn R2 = 0:93 ð1Þ

where cTOC is the total organic carbon concentration (mgl−1) andcCODMn is the chemical oxygen demand (mgl−1). Missing monthlyTOC and CODMn concentration values were interpolated from thoseof the adjacent months. Use of Eq. (1) is supported by the fact thatKortelainen (1993) found a high correlation between CODMn and TOCconcentrations in Finnish lakes.

Mean annual and seasonal TOC concentration values werecalculated by averaging the weekly and biweekly values. Exportvalues were calculated by multiplying monthly concentrations bymonthly catchment discharge values. The spring season refers toMarch, April and May, summer to June, July and August, and autumnto September, October and November. Because of the low samplingfrequency during December, January and February, the winter seasonwas excluded from the study of seasonal trends in TOC concentra-tions. However, the data from the winter months were interpolatedand included to the assessment of annual trends.

2.3. Meteorological data and atmospheric deposition

Daily meteorological data were obtained from the Finnish Meteo-rological Institute weather stations at Valtimo (63°40′N, 28°49′E) andSotkamo (63°57′N, 28°26′E), located 10–15 km from the catchments. AtValtimo, precipitation (snowfall and rain), daily mean, maximum andminimum air temperature, relative humidity, and snow depth wererecorded. At Sotkamo only precipitation was recorded. The bulkdeposition of TOC, SO4, NO3, NH4 and total N to our catchments wasobtained from a depositionmonitoring station at Kuhmomaintained bythe Finnish Environment Institute. The effective temperature sum (sumof daily mean temperatures exceeding +5 °C) was calculated based onthe air temperature data from Valtimo. The range of variability of thehydrometeorological variables is presented in Table 2.

2.4. Catchment characteristics

The main characteristics of the catchments and the range of theirvalues are given in Table 2. The soils and forests in each catchmentwereinventoried once during 1990–1994. Based on the field inventory data,the MOTTI growth and yield simulator (Hynynen et al., 2002; Salminenet al., 2005)wasparameterized to reconstruct the annual volumes of thetree stands over the study period. These data were converted into totalstand biomass values using the biomass equations presented byLehtonen et al. (2004). The estimated average stand volume of thecatchments in 2006 varied from 43 to 195 m3ha−1, and the meanannual volume growth varied from 1.1 to 4.4 m3ha−1.

Spatial estimates of peat layer thickness were produced by kriginginterpolation using discrete field measurements and core samplingdata collected by the Geological Survey of Finland. The total peatvolume for the catchments was estimated based on the spatialdistribution of the peat layer thickness. The volume of the humuslayer covering the upland soils was derived from measured humuslayer thicknesses and mapped areas. We used a bulk density value of120 kgm−3 for both peat and humus layers to calculate catchmentsoil organic matter (SOM) contents. The total mass of terrestrialorganic material in a catchment was calculated by summing the drymasses of tree stand, peat and humus layer. Because the temporalchanges in the masses of peat and humus layer are likely to be smallover the monitoring period, they were assumed to be constantthroughout the monitoring period.

Site type classification data was used to estimate the percentagecover of nutrient-rich sites within each catchment. A site wasconsidered to be nutrient-rich if it belonged to the Vacciniummyrtillus-site type (MT) or one of higher fertility in the case of uplandareas and equivalent mire sites types (higher fertility than MT) in thecase of peatland areas.

2.5. Statistical analyses

Linear trends in annual and seasonal TOC concentrations (mgl−1),TOC exports (kgha−1yr−1 or kgha−1season−1), runoff, timing of thepeak runoff (Julian day), mean snow depth, and meteorologicalvariables over the monitoring period were calculated using the non-

Fig. 1.Mean annual air temperature (a) and annual precipitation (b) in eastern Finlandduring the monitoring period 1979–2006. Significant trend (P<0.05) is marked in theparenthesis.

95S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

parametric Seasonal Kendall test (Gilbert, 1996). The Seasonal Kendalltest is suitable because it is insensitive to outliers, missing data, andnon-normality of the time series. In the seasonal-Kendall test, the Senslope estimation method was used to discover trends in the timeseries. The trend was considered statistically significant using a risklevel of 5%.

Significant hydrometeorological variables and catchment charac-teristics affecting the mean annual and seasonal TOC concentrationsand export values were identified using mixed model regressionanalysis (Searle, 1997). The mixed model approach considers possibleautocorrelation, which may be involved in the time series of repeatedobservations. In the mixed model regression (Eq. (2)) two hierarchi-cal levels of variation were identified in the data: 1) betweencatchments, and 2) within the time series from a catchment

yij = αij + β1x1ij + β2x2ij + … + βnxnij + uj + eij ð2Þ

where yij is the response variable (TOCmgl−1, kgha−1yr−1 or kgha−1

season−1) for a time period i in the catchment j, x1ij−xnij are theexplanatory (fixed) variables (see Table 2), αij is the intercept, β1−βn

are the parameters of the explanatory variables, uj is the random effectof a catchment j, eij is the random error accounting for the within-catchment variation between successive seasonal or annual observa-tions, and n is the number of explanatory variables. The underlyingassumption is that the random variables uj and eij are uncorrelatedwhite noise and follow normal distributions with zero means and thevariancesσu

2 andσe2 foruj and eij, respectively. The explanatory variables

were included in the model as such or linearized with, e.g., the naturallogarithm, when their relationship against the dependent variable wasnonlinear.

The mixed model analysis was conducted by using the MLwiNsoftware, which estimates the fixed and random parameters simul-taneously (Rasbash et al., 2001). The standard error of the parameterestimates was used to determine the significance of the parameter. Aparameter was determined to be significant when its absolute valuewas larger than the estimate of the standard error. The value of −2(log-likelihood) was used to compare the overall goodness-of-fit ofthe models of increasing number of explanatory variables. Thebuilding of the mixed model was done in the following way: Firstly,an initial likelihood measure was derived from the model includingonly the intercept. Thereafter, explanatory variables were one afteranother added to the model based on the premise that the addition ofa variable must maximise the decrease of the likelihood measure.Based on the decrease of the likelihood measure for each explanatoryvariable, the importance of the variable in explaining the TOCconcentration or export is ranked. Variables are added to the modeluntil there is no significant improvement in the likelihoodmeasure, orone or more of the explanatory variables becomes non-significant.The models were estimated with the restricted iterative generalizedleast-square (RIGLS) method recommended for small samples.

To evaluate the reliability and accuracy of the model, thesystematic bias (B) and the relative systematic bias (Br) werecalculated as follows for each season:

B =1m

∑m

j=1

1nj

∑nj

i=1yij−yijÞ� i"

ð3Þ

Br =1m

∑m

j=1

1nj

∑nj

i=1

yij− yijyij

!" #ð4Þ

where m is the number of catchments, nj is the number of obser-vations in catchment j, and ŷij is the predicted value of the TOCconcentration or export at time period i in catchment j.

A correlation analysis was conducted to identify interdependenciesbetween explanatory variables used in the mixed model analysis.

Spearman correlations were calculated 1) between catchment specifictemporally changing variables (between the time series) and, 2)between temporally lumped static variables (between the catchments).

3. Results

3.1. Trends

During the period 1979 to 2006 the mean annual air temperatureincreased by 1.2 °C (Fig. 1a). The trend was the strongest for thesummer and autumn seasons (Fig. 2). An increase in the annual meantemperature of stream water was observed after mid 1990s (Fig. 3).The trend of increase in the stream water temperature was higher insummer than in autumn (Fig. 3). The annual precipitation (Fig. 1b) didnot change during the monitoring period, but the winter precipitationsignificantly increased by 38 mm (39%) from 1979 to 2006.

Of the 8 study catchments, two (Liuhapuro and Kangaslampi)showed a significant negative trend in annual runoff (Table 3) and threecatchments (Kivipuro, Liuhapuro, and Murtopuro) showed decreasingseasonal trends, mostly autumn. Summer and winter runoff did notchange in any of the study catchments. During the monitoring period,the average date of maximum daily runoff during spring snowmelt was2ndMay in 2006, 6 days earlier than in 1979. The peak runoff value alsodecreased; the decrease was the largest in Liuhapuro.

The deposition of SO4, NH4, NO3 and total N all decreasedsignificantly over the period 1980 to 2004 (Fig. 4). The steepestabsolute decrease was for SO4 deposition, which decreased from 7 kgSha−1a−1 in 1985 to about 2 kg Sha−1a−1 in 2004.

Mean annual TOC concentrations increased at 7 of the 8 catch-ments (Fig. 5). The examination of the seasonal TOC concentrationssuggested that an annual trend became detectable when there was atleast one season with a significant positive TOC concentration trend(Figs. 5 and 6). Most commonly a trend in TOC concentrations wasfound in spring, except at Suopuro where it occurred in summer(Figs. 5 and 6). Compared with the magnitudes of the trends, thevariation in the seasonal concentrations between years was large.

Fig. 2. Seasonal mean air temperatures during the monitoring period 1979–2006.Significant trends (P<0.05) are marked in the parentheses.

Table 3Mean annual and seasonal runoff during the study period.

Catchment Runoff, mm season−1

Winter Spring Summer Autumn Annual

Murtopuro 25.6 232.1 65.1 90.0 (−) 412.8Liuhapuro 20.1 210.0(−) 54.4 95.1 (−) 381.4(−)Suopuro 23.1 217.6 55.8 89.2 385.0Kivipuro 18.9 178.1 52.7 67.0(−) 316.6Välipuro 27.4 183.5 70.0 89.4 344.1Porkkavaara 33.0 180.1 84.7 74.7 369.0Kangaslampi 19.3 137.0 40.8 45.2 242.6(−)Korsukorpi 29.6 206.2 74.6 90.7 400.4

Bold values indicate significant trends (P<0.05) and the sign in the parenthesis denotesthe direction of the trend.

96 S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

When the annual TOC concentration (Fig. 5a–c) was combinedwith the annual runoff (Table 3) to produce the annual TOC export(Fig. 5d–f), no clear export trendswere found in 7 of the 8 catchments.Only Liuhapuro showed a significant (negative) trend in TOC export(Fig. 5d–f) and no seasonal trends were detected at any of thecatchments. The inter-annual variation in the export was very large.

Fig. 3. Mean annual temperature (a) and seasonal temperatures (b, c, d) of stream watertemperature data of the catchments Liuhapuro, Välipuro, Murtopuro, Suopuro and KivipKangaslampi, and Korsukorpi. Significant trends (P<0.05) are marked in the parentheses.

3.2. Factors related to the TOC concentration and export

According to the mixed model regression analysis, the percentageof peatland within the catchment was the most important variableexplaining both the annual and seasonal TOC concentrations(Table 4). It was more important than any temporally changingvariable. Stream water temperature was the second most importantvariable explaining annual, spring and autumn TOC concentrations(Table 4). In summer, the amount of precipitation was the mostimportant hydrometeorological variable; precipitation was a signif-icant variable explaining annual TOC concentrations as well. Catch-ment mean stand volume was also a significant variable explainingspring and annual TOC concentrations.

The percentage of peatlands within a catchment was the mostimportant variable explaining the annual, spring and autumn TOCexports (Table 4). The second most important explanatory variable

(±standard error) in the monitored catchments. The series 1979–2006 contains theuro; the series 1993–2006 contains the temperature data of catchments Porkkasalo,

Fig. 4. Annual SO4–S, NH4–N, NO2–N and total N deposition in eastern Finland in years1980–2004. Significant trends (P<0.05) are marked in the parentheses.

97S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

was precipitation for annual and autumn TOC exports, and snowdepth for spring exports. In summer, precipitation explained TOCexports the best.

Fig. 5. Time-series of annual stream water TOC concentrations (a–c) and exports (d–f)management operations were removed from the data. Significant trends (P<0.05) are mar

Compared with the ranges in the annual and seasonal TOCconcentrations, the bias between the measured and predicted TOCconcentrations was low (Table 4). The relative bias was systematicallynegative for the annual and seasonal models. The relative bias for theTOC export varied from −16.1% to −27.4% and the absolute value ofthe relative bias was higher for the TOC export than for the TOCconcentration model.

The models were able to explain 56–71% of the variance inseasonal and annual TOC concentrations across all catchments. ForTOC export values, the models were able to explain 46–69% of thevariance. Examination of the residuals revealed no systematic error inthe model predictions. For the annual and seasonal models, both therandom effect of catchment and random error were significant.

Precipitation and temperature sum correlated in the modeldescribing the springtime TOC export (Table 5). This implies thatthe TOC export model for spring cannot be applied withoutconsidering the dependency between the explanatory variables, i.e.not all combinations of precipitation and temperature sum are valid inthe model. Furthermore, the correlations between catchment char-acteristics (Table 6) revealed a dependency between the streamwatertemperature and the mean tree stand volume, both of which arepresent as explanatory variables in the models of annual and springTOC concentrations. This dependency should be considered inapplying the concentration models.

in the study catchments. In non-continuous series, the years affected by the forestked in parenthesis with signs (+ or −).

Fig. 6. Time-series of seasonal streamwater TOC concentrations in spring (a–c), summer (d–f) and autumn (g–i) in the study catchments. In non-continuous series, the years affectedby the forest management operations were removed from the data. Significant trends (P<0.05) are marked in parentheses with signs (+ or −).

Table 4Models for the mean annual and seasonal TOC concentration and TOC export in stream water.

Variable Concentration Export

ln(TOC, mg l−1) ln(TOC, kg ha−1yr−1) ln(TOC, kg ha−1season−1)

Annual Spring Summer Autumn Annual Spring Summer Autumn

α 0.009 0.859 1.051 0.9102 2.489 −0.147 −1.735 −1.933ln(Peatland %) 0.639 0.507 0.496 0.487 0.668 0.714 0.521 0.691Precipitation 0.0005 0.001 −0.001 0.004 0.011 0.012Snowfall 0.001 −0.008Snow depth 0.011Temperature sum 0.002ln(Stream water temperature) 0.169 0.081 0.367 0.517Mean stand volume 0.002 0.0026u 0.079 0.071 0.057 0.098 0.057 0.070 0.026 0.060e 0.014 0.022 0.042 0.030 0.115 0.055 0.195 0.233Bias 0.31 −0.56 0.43 −0.57 −0.51 −1.79 −0.24 −0.22Bias% −3.70 −8.97 −6.26 −11.19 −18.60 −16.13 −24.70 −27.36

The hydrometeorological variables are either annual or seasonal depending on the model. All presented values of the parameters (α, β, u, and e in Eq. (2)) are significant (P<0.05).See Table 2 for the details of the explanatory variables.

98 S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

Table 5The number of study catchments out 8 where correlations between differenthydrometeorological variables were significant.

Mean stand volume Runoff Precip. Tsum

Runoff 3 (−)Precipitation 7(+)Mean snow depth 2 (−) 2 (+) 3 (+)Snowfall 1 (+) 1(+)Tsum 5 (+) 7 (−) 6 (−)Water temp. 1 (+) 2 (+)

Only variables that are significant in the mixed models of TOC concentration and exportwere included (see Table 4). The sign of the correlations is presented in the parentheses.

99S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

4. Discussion

4.1. TOC concentrations

The increasing trends in the annual and seasonal streamwater TOCconcentrations we found are in accordance with previous observa-tions from boreal and temperate streams, rivers and lakes (e.g. Arvolaet al., 2004; Worrall et al., 2006; Vuorenmaa et al., 2006; Monteith etal., 2007; de Wit et al., 2007; Ågren et al., 2008). During the last10 years of the study period (1997–2006), the mean annual TOCconcentration in our catchment streams increased between 6 and 24%compared to the levels one decade earlier. Because the data for theperiod affected by forest management had been excluded from ourstudy, the increase in the concentration must result from changes inenvironmental conditions.

We found clear increasing trends in both air and stream watertemperatures and decreasing trends in the amounts of atmosphericdeposition since the mid of 1980s. However, deposition levels werenot among the most important variables explaining the stream TOCconcentrations in the catchments of this study (see also Clair et al.,2008; Worrall et al., 2008). This may have been because our aciddeposition levels are considerably lower than in studies wheredeclining acid deposition loads have been found to be important inexplaining increasing TOC concentrations (e.g. Monteith et al., 2007).The decrease in acid deposition trends in our studymay thus have hadlittle impact on the dissociation of organic acids in the soil (Krug andFrink, 1983). Furthermore, the chloride deposition has been very lowin our catchments.

Stream water temperature was found to be a more importantvariable explaining the TOC levels than the air temperature (Table 4).Changes in stream water temperature correlate strongly with thechanges in air temperature, but the streamwater temperature slightlyvaries between the catchments. Since stream water is hydrologicallyconnected with catchment soils, the stream water temperaturebecomes a better explanatory variable for TOC concentration thanthe air temperature (see Arnell, 2002). TOC concentrations during thesummer did not depend on stream water temperature but insteaddepended on the amount of precipitation. The observed increasingtrends in stream water TOC concentrations may indicate an increasedtemperature effect (increased TOC production) during the spring andautumn period but the importance of an increased flushing (precip-itation) effect during the summer.

Table 6Correlations between catchment characteristics used in the mixed models of TOCconcentration and export (see Table 4).

Peatland% Mean stand vol. Organic material Runoff

Peatland% 1Mean stand volume −0.299 1Organic material 0.994 −0.238 1Runoff 0.383 −0.429 0.333 1Water temperature 0.359 −0.857 0.286 0.405

Significant correlations (P<0.05) are marked with bold face.

Differences in TOC concentrations between the catchments weremainly related to the percentage of peatland in the catchments. Thepool of terrestrial organic matter in the catchments is stronglycorrelated to the amount of peatland percentage, and since peatlandsare strongly associated with the riparian zone, the peatlands arepresumably the main source of TOC to the streams. The importance ofpeatlands in supplying TOC to boreal streams has been indicated in anumber of studies (e.g. Schiff et al., 1997; Dillon and Molot, 1997;Freeman et al., 2001a; Arvola et al., 2004; Laudon et al., 2004). Wefound that the peatland percentage was the main factor explainingthe TOC level in all seasons, including spring. Laudon et al. (2004),however, found that DOC concentrations in runoff from peatlandswere low in spring. When surface runoff from snowmelt flows overfrozen soil it is not effectively in contact with the peat. Thus, thepeatland percentage should have a minor effect on the springtimestream water DOC concentrations (Laudon et al., 2004; Ågren et al.,2007, 2008). The positive association between TOC concentrationsand peatland cover during spring in our study may be because ourpeatlands were mainly forested, unlike those in the study by Laudonet al. (2004). Snowmelt is likely to generate more surface runoffon treeless fens than on forested peatlands, which may explain thecontradiction.

4.2. TOC export

Although TOC concentrations increased, annual and seasonal TOCexports, with the exception of one of our catchments, did not showsignificant trends over the study period, because the variationbetween years and seasons in the TOC exports was higher thanthose in the TOC concentrations. Since export is derived as a productof discharge and concentration, changes in runoff strongly contributeto the export. Pastor et al. (2003) used small scale manipulationexperiments to study DOC input and export from isolatedmesocosms,and found out that the elevated TOC concentration in the dischargedoes not necessarily result in an increased TOC export. In our studythe amount of precipitation was the main hydrometeorological driverfor the TOC export. The amount of precipitation becomes an importantfactor, because it has a direct impact on export through runoff. Theimportance of the water flux in determining TOC exports has beenemphasized in other studies (e.g. Pastor et al., 2003; Holmberg et al.,2006; Clair et al., 2008).

However, difference in TOC export between catchments wasmostly explained by percentage coverage of peatland, which is inaccordance with earlier studies (e.g. Freeman et al., 2001a; Lepistöet al., 2008). The relationship between peatland percentage and theTOC export was more significant during spring and autumn thanduring summer. In Swedish boreal streams, the peatland and forestcoverage were found to be the primary factors controlling TOCexports throughout the snow-free season (Ågren et al., 2007).

The TOC export dynamics vary between seasons (Ågren et al.,2007, 2008). According to climate change scenarios, substantialchanges are predicted to occur in the hydrological conditions in theboreal zone (Bates et al., 2008). For example, winter precipitationmayincrease and the snow cover period may shorten. These changes arelikely to increase winter TOC exports to aquatic ecosystems. In ourstudy catchments, the winter period was excluded from the seasonalexamination of the TOC trends due to missing data in most years.Available measurements suggest that so far the role of the winter TOCexport in our catchments is small, because the winter export formedonly a small part of the total annual export. This situation may changein the future and therefore more attention should be paid in gatheringwater quality data during winter as well.

In our study, the variables affecting TOC concentrations and exportwere identified with a statistical model. According to Futter and deWit (2008), statistical modeling is useful in analyzing the factors thatcontrol element concentrations and export on a catchment scale. In

100 S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

the modeling process, careful interpretation of the explanatoryvariables is important for formulating reasonable and reliable models(Searle, 1997). In our models, all the identified explanatory variableswere logical from the point of view of process understanding.However, when using models in predicting future loads or applyingthe models to other areas, a critical evaluation is needed. Uncertaintyin the model predictions is generated by unexplained randomvariation in the model performance, and the differences in theinherent TOC dynamics between catchments.

5. Conclusions

We looked for trends in stream water TOC concentrations andcatchment export loads from boreal forested catchments located in anarea receiving low atmospheric deposition and unaffected by forestmanagement using long-term runoff and water quality data. Whileincreasing annual and seasonal trends in stream water TOC concentra-tions were revealed, no trends in TOC export loads were found. Therewas large variation in both TOC concentrations and export loadsbetween years, seasons and catchments, and the observed changes inTOC concentrations over time were considerably smaller than thevariability between years and catchments. The levels of stream waterTOC concentration and export are primarily controlled by thepercentage of peatland in the catchment. The trends in the TOCconcentration were found to be related to temperature and precipita-tion trends, but not to changes in acid deposition. We concluded thatstream water TOC concentrations and exports are mainly driven bycatchment characteristics and hydrometeorological factors rather thantrends in atmospheric acid deposition.

Acknowledgements

We warmly thank Kone Foundation in Finland for contributing tothe funding of this study andMarkus Hartman for revising the English.

References

Ågren A, Buffam I, Jansson M, Laudon H. Importance of seasonality and small streamsfor the landscape regulation of DOC export. J Geophys Res 2007;112. doi:10.1029/2006JG000381 Art. No. G03003.

Ågren A, Buffam I, BerggrenM, Bishop K, JanssonM, Laudon H. Dissolved organic carboncharacteristics in boreal streams in a forest-wetland gradient during the transitionbetween winter and summer. Geophys Res 2008;113. doi:10.1029/2007JG000674Art. No. G03031.

Ahtiainen M. Effects of clear-cutting and forestry drainage on water quality in theNurmes study. Symposium on the hydrology of wetlands in temperate and coldregions, Vol. 1, Joensuu, Finland 6–8 June 1988, Publications of the Academy ofFinland; 1988. p. 206–19. 4/1988.

Ahtiainen M, Huttunen P. Long-term effects of forestry managements on water qualityand loading in brooks. Boreal Environ Res 1999;4:101–14.

Alcamo J, Moreno JM, Nováky B, Bindi M, Corobov R, Devoy RJN, et al. Europe. Climatechange 2007: impacts, adaptation and vulnerability. In: Parry ML, Canziani OF,Palutikof JP, van der Linden PJ, Hanson CE, editors. Contribution of Working GroupII to the fourth assessment report of the intergovernmental panel on climatechange. Cambridge, UK: Cambridge University Press; 2007. p. 541–80.

Arnell N. Hydrology and global environmental change. Pearson Education; 2002. 346 pp.Arvola L, Räike A, Kortelainen P, Järvinen M. The effect of climate and landuse on TOC

concentrations and loads in Finnish rivers. Boreal Environ Res 2004;9:381–7.Bates BC, Kundzewicz ZW, Wu S, Palutikof JP, editors. Climate Change and Water.

Technical Paper of the Intergovernmental Panel on Climate Change. Geneva: IPCCSecretariat; 2008. 210 pp.

Buffam I, Laudon H, Temnerud J, Mörth CM, Bishop K. Landscape-scale variability ofacidity and dissolved organic carbon during spring flood in a boreal streamnetwork. J Geophys Res 2007;112. doi:10.1029/2006JG000218 [Art. No. G01022].

Cajander AK. Forest types and their significance. Acta For Fenn 1949;56:71.Clair TA, Dennis IF, Vet R, Laudon H. Long-term trends in catchment organic carbon and

nitrogen exports from three acidified catchments in Nova Scotia, Canada.Biogeochemistry 2008;87(1):83–97.

Clark JM, Chapman PJ, Adamson JK, Lane SN. Influence of drought induced acidificationon the mobility of dissolved organic carbon in a peat soil. Glob Chang Biol2005;11:791–809.

deWit HA, Mulder J, Hindar A, Hole L. Long term increase in dissolved organic carbon instream waters in Norway is response to reduced acid deposition. Environ SciTechnol 2007;41:7706–13.

Dillon PJ,Molot LA. Effect of landscape formon export of dissolved organic carbon, iron, andphosphorus from forested stream catchments. Water Resour Res 1997;33:2591–600.

Eikebrokk B, Vogt RD, Liltved H. NOM increase in Northern European source waters:discussion of possible causes and impacts on coagulation/contact filtrationprocesses. Water Sci Technol: Water Supply 2004;4(4):47–54.

Ellis T, Freeman C. Export of carbon from peatlands under changing climate. In: ManderU, Köiv M, Vohla C, editors. 2nd international symposium on wetland pollutantdynamics and control, WETPOL 2007, vol. 1. Extended abstracts; 2007. p. 93–4.

ErlandssonM, Buffam I, Fölster J, Laudon H, Temnerud J,Weyhenmeyer GA, et al. Thirty-five years of synchrony in the organic matter concentrations of Swedish riversexplained by variation in flow and sulphate. Glob Chang Biol 2008;14(5):1191–8.

Evans CD, Freeman C, Monteith DT, Reynolds BR, Fenner N. Climate change— terrestrialexport of organic carbon — reply. Nature, 415. ; 2002. p. 862-862.

Finér L, Ahtiainen M, Mannerkoski H, Möttönen V, Piirainen S, Seuna P, et al. Effects ofharvesting and scarification on water and nutrient fluxes. A description ofcatchments and methods, and results from the pretreatment calibration period.Research PapersThe Finnish Forest Research Institute; 1997. p. 1-38.

Freeman C, Evans CD, Monteith DT, Reynolds DB, Fenner N. Export of organic carbonfrom peat soils. Nature 2001a;412:785.

Freeman C, Ostle O, Kang H. An enzymic latch on a global carbon store. Nature2001b;409:149.

Futter MN, de Wit HA. Testing seasonal and long-term controls of streamwater DOCusing empirical and process-based models. Sci Total Environ 2008;407:698–707.

Gilbert RO. Statistical methods for environmental pollution monitoring. New York: VanNostrand Reinhold; 1996. 320 pp.

Holmberg M, Forsius M, Starr M, Huttunen M. An application of artificial neuralnetworks to carbon, nitrogen and phosphorus concentrations in three borealstreams and impacts of climate change. Ecol Model 2006;195:51–60.

Hudson JJ, Dillon PJ, Somers KM. Long term patterns and dissolved organic carbon inboreal lakes: the role of incident radiation, precipitation, air temperature, southernoscillation and acid deposition. Hydrol Earth Syst Sci 2003;7:390–8.

Hynynen J, Ojansuu R, Hökkä H, Siipilehto J, Salminen H, Haapala P. Models forpredicting stand development in MELA system. Research Papers. The Finnish ForestResearch Institute; 2002. p. 1-116.

Jonsson A, Meili M, Bergström A-K, Jansson M. Whole-lake mineralization ofallochthonous organic carbon in a large humic lake (Örträsket, N. Sweden). LimnolOceanogr 2001;46(7):1691–700.

Kalbitz K, Solinger S, Park J-H, Michalzik B, Matzner E. Controls on the dynamics ofdissolved organic matter in soils: a review. Soil Sci 2000;165:277–304.

Kortelainen P. Content of total organic carbon in Finnish lakes and its relationships tocatchment characteristics. Can J Fish Aquat Sci 1993;50:1477–83.

Kortelainen P, Saukkonen S. Leaching of nutrients, organic carbon and iron from Finnishforestry land. Water Air Soil Pollut 1998;105:239–50.

Kortelainen P, Mattsson T, Finér L, Ahtiainen M, Saukkonen S, Sallantaus T. Controls onthe export of C, N, P and Fe from undisturbed boreal catchments, Finland. Aquat Sci2006;68:453–68.

Krug EC, Frink CR. Acid rain on acid soil: a new perspective. Science 1983;221:520–5.Laudon H, Köhler S, Buffam I. Seasonal TOC export from seven boreal catchments in

northern Sweden. Aquat Sci 2004;66:223–30.Laurén A, Heinonen J, Koivusalo H, Sarkkola S, Tattari S, Mattsson T, et al. Implications of

uncertainty in pre-treatment dataset on estimation of treatment effects frompaired catchment studies: loads of phosphorus from forest clear-cuts. Water AirSoil Pollut 2009;196. doi:10.1007/s11270-008-9773-1.

Lehtonen A, Mäkipää R, Heikkinen J, Sievänen R, Liski J. Biomass expansion factors(BEFs) for Scots pine, Norway spruce and birch according to stand age for borealforests. For Ecol Manag 2004;188(1–3):211–24.

Lepistö A, Kortelainen P, Mattsson T. Increased organic C and N leaching in a northernboreal river basin in Finland. Glob Biogeochem Cycles 2008;22. doi:10.1029/2007GB003175.

Mattsson T, Kortelainen P, Räike A. Export of DOM from boreal catchments: impacts ofland use cover and climate. Biogeochemistry 2005;76:373–94.

Mattsson T, Ahtiainen M, Kenttämies K, Haapanen M. Avohakkuun ja ojituksenpitkäaikaisvaikutukset valuma-alueen ravinne- ja kiintoainehuuhtoumiin. In:Kenttämies K, Mattsson T, editors. Metsätalouden vesistökuormitus. MESUVE-projektin loppuraportti. Suomen ympäristökeskus. Helsinki: Finnish EnvironmentCenter; 2006. p. 73–81. (in Finnish).

Monteith DT, Stoddard JL, Evans C, de Wit H, Forsius M, Hogasen T, et al. Dissolvedorganic carbon trends resulting from changes in atmospheric deposition chemistry.Nature 2007;450:537–41.

Nieminen M. Export of dissolved organic carbon, nitrogen and phosphorus followingclear-cutting of three Norway spruce forests growing on drained peatlands insouthern Finland. Silva Fenn 2004;38(2):123–32.

Pastor J, Solin J, Bridgham SD, Updegraff K, Harth C, Weishampel P, et al. Globalwarming and the export of dissolved organic carbon from boreal peatlands. Oikos2003;100:380–6.

Rasbash J, Browne W, Goldstein H, Yang M, Plewis I, Healy M, et al. A user's guide toMlwiN. Centre for Multilevel Modeling, Institute of Education, University ofLondon; 2001. 278 pp.

Salminen H, LehtonenM, Hynynen J. Reusing legacy FORTRAN in theMOTTI growth andyield simulator. Comput Electron Agric 2005;49(1):103–13.

Schiff SL, Aravena R, Trumbore SE, Hinton MJ, Elgood R, Dillon PJ. Export of DOC fromforested catchments on the Precambrian Shield of Central Ontario: clues from 13Cand 14C. Biogeochemistry 1997;36:43–65.

Searle SR. Linear models for unbalanced data. New York: Wiley; 1997. 560 pp.Seuna P. Hydrological effects of clear-cutting and drainage in the Nurmes-study.

Proceedings of the international symposium on the hydrology of wetlands in

101S. Sarkkola et al. / Science of the Total Environment 408 (2009) 92–101

temperate and cold regions, Joensuu, Finland, 6–8 June 1988, vol. 4/1988.Publications of the Academy of Finland; 1988. p. 122–34.

Uusivuori J, Kallio M, Salminen O, editors. Vaihtoehtolaskelmat Kansallisen metsäohjel-man 2015 valmistelua varten, vol. 75. Working Papers of the Finnish ForestResearch Institute; 2008. 104 pp. (in Finnish).

Vuorenmaa J, Forsius M, Mannio J. Increasing trends of total organic carbonconcentrations in small forest lakes in Finland from 1987 to 2003. Sci TotalEnviron 2006;365:47–65.

Worrall F, Burt TP. Flux of dissolved organic carbon from U.K. rivers. Glob BiogeochemCycles 2007;21(1):GB1013.

Worrall F, Harriman R, Evans CD, Watts C, Adamson JK, Neal C, et al. Trends in dissolvedorganic carbon in UK rivers and lakes. Biogeochemistry 2004a;70:369–402.

Worrall F, Burt TP, Adamson J. Can climate change explain increases in DOC flux fromupland peat catchments? Sci Total Environ 2004b;326:95-112.

Worrall F, Burt TP, Adamson JK. Trends in drought frequency — the fate of DOC exportfrom British peatlands. Clim Change 2006;76:339–59.

Worrall F, Burt TP, Adamson JK. Linking pulses of atmospheric deposition to DOCrelease in an upland peat-covered catchment. Glob Biogeochem Cycles 2008;22.doi:10.1029/2008GB003177.