Ecosystem responses to increased organic carbon concentration: comparing results based on long-term monitoring and whole-lake experimentation

ABSTRACT Recent increases in terrestrial dissolved organic carbon (DOC) concentrations in northern inland waters have many ecological consequences. We examined available data on carbon cycles and food webs of 2 boreal headwater lakes in southern Finland. Basic limnology and catchment characteristics of a pristine lake, Valkea-Kotinen (VK), were monitored over the past 25 years while the lake has undergone browning and DOC increased from ∼11 to 13 mg L−1. Pronounced changes in the early 2000s represent a regime shift in DOC concentration and color. Lake Alinen Mustajärvi (AM) was manipulated for 2 years by additions of labile DOC (cane sugar), raising the DOC concentration from ∼10 to 12 mg L−1, but not changing light conditions. The 2 different approaches both revealed increased concentrations and efflux of carbon dioxide (CO2) from the lakes and thus net heterotrophy and changes in the pelagic community structure following an increase in DOC concentration. Long-term monitoring of VK revealed a decline in phytoplankton primary production (PP) along with browning, which was reflected in retarded growth of young (1–2-year-old) perch. In the experimentally manipulated lake (AM), PP was not affected, and the growth of young perch was more variable. The results suggested the importance of a pathway from labile DOC via benthic invertebrates to perch. Although provided with this extra resource, the food chain based on DOC proved inefficient. Long-term monitoring and whole-lake experimentation are complementary approaches for revealing how freshwater ecosystems respond to climate and/or atmospheric deposition-induced changes, such as browning.


Introduction
During recent decades, increased concentrations of terrestrial dissolved organic carbon (DOC) and associated water browning have been reported in many lakes in the Northern Hemisphere (e.g., Jennings et al. 2010, Couture et al. 2012. This phenomenon has been related to various processes, including recovery from earlier acidification, increasing precipitation and runoff, climate warming connected with increase in vegetation cover and prolonged frost-free period of soils, and land-use changes (Vuorenmaa et al. 2006, Monteith et al. 2007, Lepistö et al. 2014, Finstad et al. 2016). In addition to colored terrestrial DOC, increased concentration of iron can also contribute to water browning (Krizberg and Ekström 2012, Weyhenmeyer et al. 2014). The future impact of browning on boreal lakes is difficult to predict because of several potential impacts on lake properties and processes (Solomon et al. 2015, Creed et al. 2018. First, the increased water color reduces light penetration, affecting both light quantity and quality, which in turn may reduce both pelagic and benthic primary production as well as the balance between these processes (e.g., Karlsson et al. 2009). Second, terrestrial organic matter entering lakes provides additional substrate for bacteria (Tranvik 1988, Tulonen et al. 1992, increasing production of CO 2 and shifting lake metabolism toward net heterotrophy (e.g., del Giorgio and Peters 1994, Hanson et al. 2003, Sobek et al. 2003). These changes may also have implications for food web structure and energy transfer efficiency (Jansson et al. 2007, Carpenter et al. 2015. Third, terrestrial organic matter loading will also bring nutrients such as phosphorus (P) and nitrogen (N), potentially available for primary producers (Jones 1992, Seekell et al. 2015. Disentangling the consequences of these parallel potential effects of terrestrial DOC increase has been extremely challenging and will require different but complementary approaches. Three basic approaches have been used to investigate the effects of lake browning and DOC increase. Spatial surveys comparing the properties of lakes that currently exhibit different concentrations of DOC assume that future temporal change will reflect present spatial variation, but the assumptions underpinning this "space-fortime" approach are likely at best weak and at worst invalid (e.g., Clark et al. 2010, Solomon et al. 2015, Kelly et al. 2016. Monitoring the properties of particular lakes over a long time span can reveal the impact of browning on lake properties, but the generality of the findings will be uncertain. Moreover, neither of these approaches can properly identify the mechanisms underpinning the patterns revealed, a process that will require experimentation. However, experiments using mesocosms to investigate the impact of DOC on lake properties (e.g., Arvola et al. 1996, Faithfull et al. 2012 suffer from lack of realism or relevance because mesocosms only represent a part (usually the openwater pelagic) of an entire lake ecosystem, are hydrologically isolated from the critical terrestrial loadings from the catchment, and only yield meaningful results for short (days to few weeks) time periods. Hence, experimental investigation of the effects of DOC increase are needed at the whole-lake scale, which is logistically challenging and has only rarely been attempted (e.g., Blomquist et al. 2001, Peura et al. 2014, Kelly et al. 2016).
Here we used data available from 2 small boreal headwater lakes in southern Finland to compare how an increase in DOC concentration affected their carbon cycles and food web structures. The data reflect 2 different approaches: long-term monitoring and whole-lake manipulation experiments. Since 1990, Lake Valkea-Kotinen has been monitored for several limnological variables, primary production (PP), and the food web up to fish. During this period, the DOC (measured as C) concentration of the lake increased from ∼11 to 13 mg L −1 and water color values increased from ∼100-150 to 150-200 mg Pt L −1 , whereas the mean daily PP (measured as C) in the photic layer (0-1 m) declined from ∼130 to 70 mg m −3 d −1 . Lake Alinen Mustajärvi was experimentally manipulated by increasing the epilimnetic DOC concentration from ∼10 to ∼12 mg L −1 for 2 ice-free periods by adding cane sugar to the lake. The sugar addition was intended to increase the availability of labile bacterial substrate without affecting either light penetration or nutrient concentrations, and hence to isolate and study one of the parallel potential effects of lake browning. In addition, the fate of the added DOC could be traced in organic and inorganic carbon fractions and onward through the pelagic and benthic food webs up to fish (Peura et al. 2014) based on the δ 13 C differences between organic carbon from a C4 plant (cane sugar) and local terrestrial organic carbon from C3 plant origin. Here we (1) evaluate whether the experimental whole-lake manipulation results can help us better understand the long-term ecosystem changes in response to changes in color and DOC concentrations and (2) discuss the relative merits of different but complementary approaches in the study of lake carbon cycles and food webs.

Methods
Lake Valkea-Kotinen (VK) in southern Finland (61°1 5 ′ N; 25°04 ′ E) is a small boreal headwater lake located on a pristine old-forest nature conservation area (Ruoho-Airola et al. 2014). The lake and its catchment are part of the Finnish Long-Term Socio-Ecological Research network (FinLTSER) and have been continuously monitored for over 25 years . This lake has an area of 0.04 km 2 , a catchment area of ∼0.3 km 2 , a maximum depth of 6.5 m, and an estimated volume of 1.03 × 10 5 m 3 . The whole-lake manipulation was conducted in Alinen-Mustajärvi (AM), another small boreal headwater lake just 4 km from VK (61°1 2 ′ N; 25°06 ′ E). During the study years no forestry operations were conducted in the catchment area. The lake has an area of 0.007 km 2 , a catchment area of ∼0.04 km 2 , a maximum depth of 6.5 m, and an estimated volume of 0.31 × 10 5 m 3 . Both lakes are typically ice covered from around late November to late April and during summer develop a steep and stable stratification with an epilimnion depth of ∼2 m and an anoxic hypolimnion. Littoral macrophyte vegetation is highly restricted in both lakes because light penetration into these humic lakes is poor; however, floating littoral vegetation mats with associated periphyton can be highly productive in such lakes (Vesterinen et al. 2016) and can harbor abundant macroinvertebrates. Both lakes contain reproducing populations of European perch (Perca fluviatilis) and northern pike (Esox lucius). Further information about the characteristics of the 2 lakes can be found elsewhere (e.g., Peltomaa 2013, Peura et al. 2014. In AM, water chemistry and carbon cycle variables were measured during one open-water period (2007) before cane sugar was added. During the open-water periods of 2008 and 2009, 66 kg of cane sugar was added monthly to the epilimnion (May-Oct), equivalent to a concentration of 2 mg L −1 of DOC, or a mean daily loading of 0.07 mg L −1 to the epilimnion (22 g m −2 yr −1 ).
The carbon addition was made assuming the proportion of labile DOC in the lake to be ∼15% of the total DOC (e.g., Tranvik 1998) and represents an increase in labile DOC to the level expected for the fourth quartile of boreal lakes according to the Nordic Lake Survey (Henriksen et al. 1998). Prior to the addition to the lake, the cane sugar was dissolved in lake water, and the sugar solution was then distributed evenly across the lake surface from a row boat.
Most basic data used in this paper have been reported previously, and the methods used to collect the data have been described in detail elsewhere and will not be repeated here. For VK monitoring, see Vuorenmaa et al. (2014) for chemistry, Arvola et al. (2014) for plankton metabolism, Peltomaa et al. (2013) for phytoplankton, Lehtovaara et al. (2014) for zooplankton, and Rask et al. (2014) for fish. Respective methods used in AM studies are described by Peura et al. (2014) and Jones et al. (2018). Only the results of variables measured during the summer stratification period (Jun-Sep) in both lakes using the same methods and the same instruments at Lammi Biological Station were compared.
To detect possible abrupt changes in DOC concentration and/or water color during 1990-2009 in VK, a regime shift index (RSI) was calculated with SRDS software (www.climatelogic.com) according to Rodionov (2004), based on monthly mean values (May-Sep) of the variables measured at 1 m depth. The SRDS algorithm is based on sequential Student t-tests for detecting time points of significant changes in the variable mean in a time series. For the RSI calculation we set the probability level at 0.05 and cut-off length as 50, corresponding to a 10-year period. A clear rise in DOC concentration occurred around 2000 and 2001 ( Fig. 1), which was also detected by RSI. Thus, we compared VK data separately for the period of "lower-DOC" years 1990-2000 and of "higher-DOC" years 2001-2009 (later referred to as low-DOC period and high-DOC period , respectively). We also compare specifically data pooled for the 3 years with the lowest (1990, 1998, 2000low-DOC yrs ) and the highest (2002,2004,2005; high-DOC yrs ) DOC concentration. Sampling for most variables took place weekly or at 2-week intervals, and thus n = 8-16 per ice-free season (i.e., Jun-Sep). Because normal distribution and data independence could not be ensured, a nonparametric Mann-Whitney U test (SigmaPlot 13 software) was applied to test for statistical differences in the measured variables within the lakes between low-DOC period and high-DOC period as well as between low-DOC yrs and high-DOC yrs . In AM, the pre-DOC addition data from June to September 2007 were compared with those of the DOC addition years (2008 and 2009) and tested with Mann-Whitney U test, similar to the approach with VK data. During 2007 and 2008, most samples in AM were taken twice a month but in 2009 only once a month. Thus, the results of both DOC addition years were pooled.
Although some differences in chemical properties (color, original DOC, total nitrogen [TN], total phosphorus [TP]) between VK and AM were noted (Tables 1, 2), we estimated possible changes caused by cane-sugar DOC addition on epilimnetic metabolism and phytoplankton community composition in AM by using VK as a "reference lake" before (2007) and during the DOC addition years (2008)(2009). Because the number of measurements per month varied (1-4) and the sampling did not always take place during the same day, monthly mean values of the variables were compared with a paired t-test (SigmaPlot 13). In a case of failed normal distribution (4 of 24), a nonparametric Wilcoxon signed-rank test was used. We compared basic results of water chemistry (samples from 1 m depth, except for dissolved inorganic carbon [DIC] from the 0-0.30 m layer), phytoplankton PP, (0-1 m layer), epilimnetic community respiration (CR; 0-1 m layer), and carbon dioxide (CO 2 ) efflux to the atmosphere collected during the summer stratification period (Jun-Sep). Phytoplankton biomass based on microscopical counts and crustactean zooplankton abundance in VK were from the epilimnion (0-1 m), but in AM these abundances were averaged for the epilimnion and metalimnion (0-3 m). Thus, for these variables only within-lake differences between low-DOC and high-DOC periods and years were tested. Non-metric multivariate scaling (NMS) was used to visualize differences in monthly phytoplankton community composition between different years (2007)(2008)(2009) in VK and AM. The variables were plotted in the NMS figure as vectors. Differences in within-year variation of phytoplankton community composition were tested with the homogeneity of multivariate dispersion (PERM-DISP) routine in PRIMER 6 and PERMANOVA+ addon (Anderson 2006). CO 2 efflux from each lake was estimated with the boundary layer (BL) method Caraco 1998, details in Kankaala et al. 2013) based on DIC concentration and pH measured from the 0-30 cm top layer of the lakes and wind speed measurements at VK (Huotari et al. 2011). Although shown to underestimate CO 2 fluxes in VK (Huotari et al. 2011, Gålfalk et al. 2013, Erkkilä et al. 2018, the BL method was used because high-frequency measurements of water temperature and meteorological drivers were lacking during the 1990s in VK as well as in AM studies and are required for more accurate model estimations. The BL method does not take into account possible changes in physical processes in the surface mixing layer related to DOC and color increase. Especially in the gas exchange of small lakes, the relative importance of convection may exceed that by wind mixing during summer stratification (Read et al. 2012). Although the bulk of the CO 2 effluxes from these small headwater lakes usually occurs during the autumn turnover period (Huotari et al. 2011), the data for October and November during the whole study period (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009) were too sparse for reliable estimates for the whole open-water period. However, the results should provide a reliable indication of the potential direction of changes related to DOC increase.
Perch growth was back-calculated from opercular bones in both lakes, and the results provided here are for 3 years (2005)(2006)(2007) prior to DOC addition to AM, during the DOC addition years (2008-2009), and for years 2010 and 2011 after the manipulation (for detailed methods see Rask et al. 2014). Between-lake differences in the growth of young (first-and second-year) perch were tested with a Mann-Whitney U test, as well as for those in AM before (2007) and during (2008 and 2009) the sugar manipulation.

Results and discussion
Long-term and short-term changes in VK Water chemistry and carbon flux The long-term increase in DOC concentration and water color ( Fig. 1) in VK from the 1990s to the next decade (from ∼11 to 13 mg L −1 and from 140 to 160 mg Pt  Rodionov (2004); see methods for details. Table 1. Median values (with 25-75% range) of water chemical variables at 1 m depth (except DIC at 0-30 cm), CO 2 efflux from lake surface, phytoplankton primary production (PP), community respiration (CR, uppermost 1 m layer), phytoplankton biomass (PB), and zooplankton density in June-September (epilimnion) during the "low-DOC period " (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000) and "high-DOC period " (2001-2009) as well as those of perch density and growth during the first and second year in Lake Valkea-Kotinen during the same periods. CO 2 efflux, PP, and CR measured as C. Statistical differences between the periods were tested with Mann-Whitney U test (p denotes significance, ns = not significant). Bold values indicate statistically significant results. Low-DOC period (1990-2000 High-DOC period (2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009) Table 2. Mean monthly values and standard deviation (SD) of June-September temperature and those of chemical variables (1 m depth, except DIC from 0 to 0.3 m depth) as well as carbon flux (CO 2 efflux from lake surface and primary production [PP] and community respiration [CR] in the 0-1 m layer, measured as C) in Valkea-Kotinen (VK) and Alinen-Mustajärvi (AM) before (2007) and during the cane sugar addition to AM (2008AM ( -2009). The differences were tested with paired t-test of monthly mean values based on 1-4 measurements per month (for t-test n = 4 in 2007, 8 in 2008-2009 in each lake). The results for the total phytoplankton biomass (BP) and cladoceran and copepod abundances in VK are only for epilimnion but in AM are averaged for epilimnion and metalimnion, and thus statistical differences were not tested. L −1 , respectively) was strongly related to recovery from acidification of the lake and its catchment, indicated by a significant increase in alkalinity (Lehtovaara et al. 2014; Table 1). Water pH varied more than alkalinity and showed only a minor increase, presumably because of the impact of organic acids transported from the catchment by inflow ).
The RSI for DOC indicates that a clear sudden increase took place around 2000-2001 ( Fig. 1) but not simultaneously with the increase in color values (most pronounced shift in 2004). In this long-term dataset, color explained only 56% of the DOC variation by a linear regression equation: DOC (mg L −1 ) = 0.04 × color (mg Pt L −1 ) -6.05 (R 2 = 0.56, n = 100). Thus, as shown in many previous studies, browning is also influenced by other factors such as iron leaching from soils and its complexation with DOC (Krizberg and Ekström 2012, Weyhenmeyer et al. 2014) as well as differences in molecular structure and the proportion of colored substances in DOC, shown by spectral absorbance characteristics (e.g., Köhler et al. 2013, Adams et al. 2018).
In a 3-year study (2007)(2008)(2009)) at the riparian zone and in the lake, Rasilo et al. (2015) observed that DOC concentration in VK was positively correlated with catchment runoff but negatively with weekly rainfall and air, soil, and lake-surface water temperatures. During the whole study period (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)) the annual variation in DOC concentration in VK was not clearly related to the annual variation in mean precipitation and air temperature (Supplemental Fig. S1). For example, of the 3 years with the highest annual precipitation (1998, 2004, and 2008), only 2004 was one of the high-DOC yrs in our dataset. In whole ecosystem-scale carbon flux measurements during 15 years in 2 small boreal forest catchments, Pumpanen et al. (2014) showed that precipitation largely determined DOC fluxes, but the forest (∼50-year-old Scots pine [Pinus sylvestris] stand) net ecosystem exchange and litter production during the previous year also had a small but significant effect on DOC runoff. Thus, presumably time lags in DOC leaching from the soil and reaching the lake also affected the DOC concentration variation in the headwater of lake VK.
The concentrations of DIC in the lake surface layer and CO 2 efflux to atmosphere as well as CR in the epilimnion were significantly higher (Mann-Whitney U test, p = 0.021, 0.029, 0.024, respectively; Table 1,  Supplemental Table S1) and oxygen concentration lower (p < 0.001) during high-DOC period than values during low-DOC period . However, when only the measurements from low-DOC yrs (1990,1998,2000) and high-DOC yrs (2002,2004,2005) were compared, no significant differences were detected except for oxygen concentration (p < 0.001; Supplemental Table S1), suggesting that interannual variation in CO 2 efflux may not be directly coupled with DOC and CR. The bulk of annual CO 2 release takes place during erosion of stratification and water column mixing in late summer and autumn (Huotari et al. 2011), which were not fully covered in our dataset of samples every 2 weeks from June to September. Moreover, CO 2 originating from catchment and riparian soil respiration also contributes to CO 2 concentrations and fluxes from headwater streams and lakes (Rasilo et al. 2012, Dinsmore et al. 2013, uncoupled from DOC decomposition in the lake itself. No significant changes in TN or TP concentration took place during the study period in VK (Mann-Whitney U test, p > 0.05; Table 1), although the concentration of dissolved inorganic nitrogen (DIN; i.e., NO 3 -N, NO 2 -N, and NH 4 -N) slightly increased . DIN increase was presumably more related to in-lake processes such as changes in phytoplankton nutrient uptake in the light-limited environment and to hydrology ) than to processes taking place in the catchment; N deposition declined (Ruoho-Airola et al. 2014) and net retention of inorganic N in the catchment was high (Forsius et al. 2005). However, the TN concentrations were significantly higher during high-DOC yrs than N concentrations during low-DOC yrs (p < 0.001; Supplemental Table S1). Thus, although not clearly seen in the VK long-term data, TN input seemed to be connected with DOC during high-DOC yrs . In VK, the relationship between N and DOC as well as between lake DIN and atmospheric N deposition differed from those observed on broad regional and/or time-scale surveys in the same climatic and vegetation zone. In data collected from unmanaged boreal forest catchments from southern to northern Finland (n = 21) during several years, annual median and/or mean export estimates for total organic carbon and N were strongly correlated, and TN consisted of, on average, 85% of organic N (Kortelainen et al. 2006). However, in the long-term parallel measurements in VK, DOC and organic N concentrations were not correlated . In a large dataset covering 78 headwater streams and 95 nutrient-poor lakes in Sweden sampled during 1998-2013, average epilimnetic (Jul-Sep) DIN concentrations decreased, and DIN:TP declined in a strong correlation with diminishing atmospheric N deposition (Isles et al. 2018). This phenomenon was not seen in VK during the study period 1990-2009, but no obvious explanation exists. Presumably, several counteracting, nonsynchronous, and catchment-specific factors such as variable organic matter loading, decreased N deposition, and changes in-lake processes influenced N processing in the old-forest catchment and the lake.
Along with browning and DOC increase, a long-term decrease in PP was evident in VK (Peltomaa 2013, as also seen in the data grouped for low-DOC and high-DOC periods and years (Table 1, Supplemental  Table S1). Seekell et al. (2015) detected nonlinearity in the response of PP to allochthonous DOC in small arctic and boreal lakes. Below a DOC concentration of ∼5 mg L −1 , primary producers benefit from additional N and P transported with allochthonous organic matter, but at higher DOC concentrations the effect is negative. In VK the DOC concentration was already >8 mg L −1 at the beginning of the monitoring period, thus supporting the conclusion drawn by Seekell et al. (2015). However, the long-term change in VK was not clearly detected in short-term measurements. In weekly measurements during 3 years (2002)(2003)(2004), Peltomaa and Ojala (2016) observed no clear immediate response in pelagic metabolism (PP, CR, and 14 C-leucine uptake) after sudden rain events. After a long rainy period with some extreme rain peaks (20-50 mm d −1 ) in 2004, water color in the epilimnion increased markedly, but no clear immediate increase was observed in DOC concentration. The photic layer of the lake became shallower, which caused a decrease in PP at 1 m depth but did not affect PP in the uppermost 0.5 m layer where most PP occurred.

Plankton community
Total phytoplankton biomass, based on microscopical counts, was significantly lower in VK during the high-DOC years and periods (Mann-Whitney U test, p = 0.002, p < 0.001, respectively; Table 1, Supplemental  Table S1), although the differences based on chlorophyll a were significant only when low-DOC yrs and high-DOC yrs were compared (p = 0.013; Supplemental Table  S1). Changes in phytoplankton community composition also took place: the biomass of chlorophytes declined, whereas that of cryptophytes and chrysophytes slightly increased during high-DOC period . Many taxa among cryptophytes and chrysophytes are potentially mixotrophic (e.g., Jones 2000) and thus presumably benefitted from high-DOC conditions and associated bacteria. Flagellate taxa, capable of diurnal vertical migration for nutrient uptake from the dark nutrient-rich hypolimnion and photosynthesis in the euphotic zone, are typical in humic lakes (Salonen et al. 1984, Lepistö andRosenström 1998). Drakare et al. (2002) observed that the proportion of autotrophic and mixotrophic flagellates in phytoplankton increased along a DOC gradient from 3 to 31 mg L −1 in small lakes in northern Sweden.
During the study period 1990-2009, changes in crustacean zooplankton species dominance were recorded both in cladocerans and copepods. Ceriodaphnia quadrangula was the dominant cladoceran in most years in the 1990s but was gradually replaced by Bosmina longirostris during the years of regime shift (2002)(2003)(2004) for DOC and water color. Among copepods, Mesocyclops leuckarti was most abundant in the early 1990s, but subsequently Thermocyclops oithonoides dominated (Lehtovaara et al. 2014). The changes in zooplankton community were associated with increasing carbon load and recovery from acidification, but changes in the microbial community and phytoplankton, including the decrease in primary production ), may also have had an influence. Within the food web, bottom-up regulation of zooplankton seemed to exceed the importance of top-down control (Lehtovaara et al. 2014). However, no significant differences in the total abundance of cladocerans and copepods were noted between high-DOC and low-DOC periods and/or years (p > 0.05; Table 1, Supplemental Table 1). This finding differs from the results obtained for 2 small lakes in northeastern Pennsylvania, USA, undergoing recovery from acidification and browning during a 27-year monitoring period. In these lakes, the abundance of Daphnia and calanoids decreased, whereas the abundance of cyclopoids increased or was unchanged (Williamson et al. 2015). The DOC concentration in these lakes was originally much lower than in VK and increased ∼1 mg L −1 (from ∼1 to 2 mg L −1 and from 5 to 6 mg L −1 , respectively) during the monitoring period.

Fish
In VK, long-term changes took place in the population dynamics of perch, evidenced by the occurrence of strong year-classes every 4 years in the 1990s, levelling off in the 2000s ; Table 1). Around 2000-2005 (i.e., the years of the regime shift discussed earlier), a significant decrease was also recorded in perch growth in the first 2 years of life. This decrease was related to the increase in water color and subsequent decrease of the general productivity of the lake, decrease in volume of oxygenated habitat, and deterioration of light conditions for perch, which is a visual feeder ). The slight increase in the growth of older and larger perch in the same years suggested their lower dependence on the pelagic food web and feeding on macroinvertebrates ).

Short-term effects in AM in comparison to VK
Water chemistry and carbon flux During the AM study (2007)(2008)(2009), the reference lake VK was already in the high-DOC phase. Rasilo et al. (2015) estimated that during these years the annual load of terrestrial DOC to VK varied between 1500 and 4100 kg, which translates to ∼37-100 g m −2 yr −1 . Assuming that terrestrial DOC input to AM was of the same order of magnitude as that to VK, the cane sugar addition in 2008 and 2009 was ∼18-37% of the annual DOC input to the lake. However, when the labile lowmolecular weight fraction is generally <20% terrestrial DOC (Münster 1993, Jonsson et al. 2007, Hulatt et al. 2014, the cane sugar addition most likely exceeded the input of labile DOC transported from the surrounding catchment; the DOC concentration increased by 2 mg L −1 , thus reaching the level prevailing in VK during those years (Table 2).
Because monthly mean temperatures, measured at 1 m depth in both lakes, did not differ significantly (paired t-test, p > 0.05; Table 2), we assumed that seasonal weather-induced variations in both lakes were similar, and thus the paired mean monthly data could be used for the comparisons. During the pre-DOC addition year 2007, CO 2 efflux, PP, CR, and water pH did not differ significantly (p > 0.05) between the lakes, even though DOC, TP, TN, and chlorophyll a concentrations as well as color values were significantly higher in VK, (all p < 0.05; Table 2). Cane sugar caused no change in water color, as expected, and no marked changes in PP took place during the study years (Peura et al. 2014; Table 2, Supplemental Table S2), as expected.
In AM, significantly greater surface DIC concentrations and CO 2 effluxes were measured during the DOC addition years (paired t-test, p < 0.001 for both), suggesting an immediate increase in DOC decomposition. However, this increase was not clearly seen in the measurements of epilimnetic CR, bacterial production, or oxygen (O 2 ) concentration (Supplemental Table S2; Peura et al. 2014). The measurements were taken 2 weeks after DOC amendments, and presumably some microbial activity peaks were missed. However, a slight but significant decrease in pH and alkalinity in AM during the DOC addition years was presumably related to increased concentration of CO 2 (Mann-Whitney U test, p < 0.001, p = 0.017; Supplemental Table S2,  Table 2), thus in the opposite direction from measurements taken in VK.

Plankton community
Phytoplankton community composition in VK and AM clearly differed. In June-September 2007-2009, 58-81% of the biomass in VK consisted of the raphidophyte G. semen (on average 67%), and in only one case was the proportion of chrysophytes higher (69%) than that of G. semen. Cryptophytes, dinophytes, chrysophytes, diatomophytes, and chlorophytes averaged 4%, 9%, 15%, 2%, and 1%, respectively, of the biomass. In AM, before cane sugar addition in 2007, chrysophytes and chlorophytes averaged 60% and 39%, respectively, of the phytoplankton biomass. The proportion of cryptophytes, dinophytes, and G. semen was then ∼1-2% of the biomass. During the DOC addition years the total phytoplankton biomass was at the same level as in 2007 (Supplemental Table S2). However, changes in the community composition occurred; the proportion of chrysophytes decreased and that of G. semen increased. In the latter half of the second DOC manipulation season (Aug-Sep 2009), 37-75% of the phytoplankton biomass consisted of this taxon. Thus, although no water browning took place, G. semen was favored by labile DOC increase and presumably also by the lower pH (see Lebret et al. 2018 and earlier discussion). The biomass of cryptophytes was slightly but significantly higher during the DOC addition years than that in 2007 (Mann-Whitney U test, p = 0.028; Supplemental Table S2). Also, the proportion of heterotrophic taxa (e.g., Bicosoeca spp., Petalomonas sp., Katablepharis ovalis) increased from ∼1% in 2007 to 4% during the DOC addition years. PERMDISP analysis indicated that the seasonal variation in phytoplankton community composition increased in AM during the DOC addition years (F 2,9 = 9.069, p = 0.02; Fig. 2) while the variation in VK did not change (F 2,9 = 2.981, p = 0.299). Hence, the phytoplankton community seemingly became more stochastic in AM following the DOC additions.
Zooplankton community composition also differed between VK and AM. In VK copepods were more abundant than cladocerans (Table 2). In AM the crustacean zooplankton community was dominated by the cladocerans Holopedium gibberum, C. quadrangula, and Daphnia sp. and by small cyclopoids. Enriched stable carbon isotope values (δ 13 C) of particulate organic matter and zooplankton clearly showed transfer of cane-sugar DOC to zooplankton during the DOC addition years (Peura et al. 2014), although no significant differences were observed in the abundance of cladocerans and copepods during the study years (Mann-Whiney U test, p > 0.05; Supplemental Table S2). This result differs from that obtained by Kelly et al. (2016) in a DOC manipulation experiment of an oligotrophic temperate lake in which the catchment-origin DOC concentration in the treatment basin increased from 8 to 11 mg L −1 , whereas in the reference basin DOC concentration decreased from ∼8 to 6 mg L −1 . The authors concluded that zooplankton in the treatment basin benefitted from improved resource quality because of an increase in PP associated with transported P (see also Seekell et al. 2015 for the positive impact on PP, but Williamson et al. 2015 for negative long-term changes in zooplankton in the lakes with DOC concentration <6 mg L −1 ). In our study lakes, DOC concentration was originally higher (∼10-11 mg L −1 ), and the long-term effect on PP was negative in VK and the short-term effect insignificant in both lakes, resulting in no clear effect on crustacean zooplankton density.
The first-year median growth of perch was higher in AM than in VK during 2005-2010 (U = 36, p = 0.004). In VK both the firstand second-year growth of perch showed less variation than in AM. The second-year growth of perch was similar in both lakes during 2005-2007, but during the years of DOC manipulation (2008 and 2009) the growth was significantly faster in AM (Mann-Whitney U = 408 and 352, p = 0.04 and < 0.001, respectively; Fig. 3). When only the "reference year" 2007 and DOC "manipulation years" 2008 and 2009 were considered, the first-year growth of perch in AM was significantly higher in the reference year (Mann-Whitney U = 561, p = 0.005), but the second-year growth during the manipulation years was greater (Mann-Whitney U = 721, p < 0.001). Enriched stable carbon isotope values (δ 13 C) revealed the importance of the route of benthic invertebrates to fish, leading to a cane-sugar car- bon proportion of 18.7% of perch carbon biomass at the end of the sugar addition ). The extra cane-sugar DOC input seemed to temporarily stimulate more the growth of benthivorous second-year perch than that of primarily planktivorous young-of-the-year perch. The diet of perch was not examined in this study, but the increasing proportion of benthic food during the early years of perch has been recorded in other studies and lakes in the study area Arvola 1985, Estlander et al. 2010). The slight increase in the growth of older and larger perch in VK  suggests that increasing DOC may positively affect benthivorous fish also in natural conditions. Terrestrial DOC vs. cane-sugar DOC and long-term monitoring vs. short-term manipulation -What can be learned? The extensive dataset collected during long-term monitoring shows clear ecosystem-scale changes and a regime shift due to DOC increase and browning in VK; poorer light penetration into the water column was followed by changes in community structure, a decrease in PP and in the growth of young perch, and an increase in CO 2 efflux to the atmosphere. In the short-term (2 years) manipulation with cane sugar in AM, the pathway of DOC up to fish could be traced by δ 13 C, and an increase in CO 2 efflux could also be observed. However, detecting significant changes in a strict statistical sense was limited by marked seasonal and interannual variation in the measured variables and rather low frequency of measurements. For example, epilimnetic bacterial cell number and production as well as CR were not clearly related to DOC addition to AM (Peura et al. 2014). However, in short (2-3 weeks) replicated mesocosm experiments, a significant increase in these variables indicating heterotrophy was generally seen after DOC addition (Kankaala et al. 2010, Faithfull et al. 2012.
Glucose or sucrose, here in the form of cane sugar, is commonly used to simulate labile DOC in experiments (e.g., Blomquist et al. 2001, Smith and Prairie 2004, Faithfull et al. 2012) and could represent fresh biodegradable detritus of both autochthonous and allochthonous origin (cf. Søndergaard et al. 2000, Kiikkilä et al. 2011. In addition to carbohydrates (including sugars), low-molecularweight terrestrial DOC contain other compounds such as amino acids and carboxylic acids, which also are easily degraded by microbes (Jonsson et al. 2007, Berggren et al. 2010, Kiikkilä et al. 2011). However, more than half of terrestrial DOC consists of high-molecular-weight humic compounds, which are recalcitrant to immediate degradation by aquatic microbes (e.g., Jonsson et al. 2007, Räsänen et al. 2018. Thus, the impacts of colored humic substances are negative rather than positive on lake food webs by reducing light availability for primary producers (Thrane et al. 2013, Kelly et al. 2018, thus impairing food resource availability for the higher trophic levels , Taipale et al. 2016. Moreover, colored substances promote steeper temperature and oxygen stratification in summer, narrowing potential habitats for fish survival and worsening foraging opportunities for visually feeding fish (Estlander et al. 2010, Williamson et al. 2015. Although DOC enhances bacterial production, growth efficiency of bacteria utilizing both low-and high-molecular-weight DOC is strongly related to availability of inorganic nutrients (N, P; Smith and Prairie 2004, Jansson et al. 2006, Räsänen et al. 2018, which also impacts the magnitude of DOC-origin carbon available for consumers and/or respired and finally released as CO 2 to the atmosphere. Although available for grazers, bacteria are poor-quality diets for zooplankton and zoobenthos because they lack essential polyunsaturated fatty acids and sterols (Martin-Creuzburg et al. 2011, Taipale et al. 2012. Even when the experimental addition of labile DOC to AM (22 g m −2 yr −1 ) was of the same order of magnitude as annual PP (AM 18-20 g m −2 yr −1 ; Peura et al. 2014), only <20% of top consumer (perch) carbon originated from labile DOC . This finding indicates the inefficiency of DOCbased food chains and consumer diet requirements to include high-quality food resources produced by algae (Taipale et al. 2016).
The δ 13 C label of DOC addition could be detected in AM consumers (crustacean zooplankton, benthic invertebrates, fish) but not in the components of seston <100 µm (phytoplankton, bacteria, detritus, protozoans, rotifers), which could not be separated and analyzed, so that the actual microbial food chain routes could not be followed. In addition to pelagic and benthic microbial food chains, flocculation of cane-sugar DOC to particulate organic carbon (von Wachenfeldt and Tranvik 2008) could have been another route for benthic consumers, contributing to their δ 13 C enrichment. Although the flocculation of cane sugar in AM is speculative, the results from a whole-lake scale experiment with maize leaf addition to 2 German lakes showed that benthic invertebrates could play a significant role in transferring terrestrial POC to fish (Scharnweber et al. 2014). However, the contribution of maize leaves, corresponding to the autumnal alder leave input, was only 1-3% of omnivorous fish carbon, and the benthic microbial food chain in these lakes proved to be inefficient (Lischke et al. 2017). More investigations are needed to clarify how DOC increase and browning influence benthic-pelagic coupling in lakes.
In conclusion, long-term monitoring of a boreal headwater lake confirmed the general trend of increasing DOC concentration and browning in northern lakes and associated changes in pelagic community structure as well as in productivity. The decrease in lake PP was also reflected in higher trophic levels, here represented by the growth of young perch. Both long-term monitoring and whole-lake DOC manipulation showed an increase in CO 2 efflux to the atmosphere and changes in pelagic community structure, followed by an increase in DOC concentration. Long-term integrated monitoring of aquatic ecosystems is without question important in revealing responses of lake parameters to environmental change. However, given normal high between-lake variability, data obtained from an individual lake and catchment will not necessarily support trends based on correlation or regression analyses of parameters measured on large regional and/or time scales. The wholelake manipulation approach has the advantage of detecting ecosystem-scale responses to one major driver, here to labile DOC, but not simultaneously affecting other factors, like browning, that also suppress light availability for primary producers. This experimental manipulation, applying δ 13 C differences in cane-sugar and catchmentorigin DOC, revealed the importance of benthic pathways from DOC to perch . Benthic DOC pathways were not detected in the long-term monitoring of VK and in the mesocosm experiments, although both these approaches detected increased pelagic heterotrophy after DOC increase (Table 1; Kankaala et al. 2010). Short-term whole-lake manipulation experiments can provide insights into the processes behind the patterns and trends demonstrated in longterm studies.