Accounting for littoral primary production by periphyton shifts a highly humic boreal lake towards net autotrophy. RH: Humic lake net autotrophy

Abstract


Introduction
Most studies concerning primary production (PP) in lakes have included only phytoplankton and have ignored benthic algae in littoral habitats (Vadeboncoeur, Vander Zanden & Lodge, 2002). Over the last decade, a consensus has emerged that littoral and pelagic food webs are linked and that a combination of the PP of phytoplankton, benthic 45 algae and macrophytes together with allochthonous organic carbon (OC) forms the energetic base of lake food webs Vadeboncoeur et al., 2008;Reynolds 2008;Karlsson et al., 2009;Cole et al., 2011). Although the view of discrete pelagic and littoral food webs has changed, quantitative integrations of littoral processes into whole-lake dynamics remain under-represented in the literature (Vadeboncoeur & Steinman, 50 2002). Of the few studies investigating littoral productivity and dynamics, most are from clear-water lakes, both oligotrophic (eg. Ask et al., 2009) and eutrophic (eg. Van de Bogert et al., 2007), and only very few from dystrophic humic lakes. There is strong evidence that benthic algae can make substantial contributions to whole-lake PP in clear-water lakes (Vadeboncoeur et al., 2003;Andersson & Brunberg, 2006;Ask et al., 2009) and can provide a 55 major food resource for consumers even in large lakes, despite a low benthic contribution to whole-lake PP (Hecky & Hesslein, 1995;Hobson & Welch, 1995;Devlin, Vander Zanden & Vadeboncoeur, 2013). In contrast, poor light penetration into humic lakes is considered to severely constrain primary production, and especially benthic primary production . 60 Allochthonous (terrestrially derived) organic matter can affect several fundamental functions in lake ecosystems (e.g. constraining primary production) and allochthony increases along a humic gradient (Jones 1992;Carpenter et al., 2005;Jansson et al., 2007).
Incorporation of allochthonous OC into lake food webs increases heterotrophy, leading to community respiration (CR) that exceeds gross primary production (GPP) and overall net 5 heterotrophy (del Giorgio & Peters, 1994;Duarte & Agusti, 1998). However, most studies investigating whole-lake carbon flux concern only pelagic habitats without measurements of benthic autotrophic production or metabolism in the littoral zone. Ask et al. (2009) measured biomass production of algae and bacteria in both pelagic and benthic habitats in subarctic clear-water lakes and found that pelagic habitats were net heterotrophic with respiration and 70 net CO 2 production dominating carbon fluxes, whereas primary production was dominant in benthic habitats. Those authors also concluded that allochthonous OC supported a significant part of the total consumer biomass production in both pelagic and benthic habitats. Humic or highly turbid lakes, where light effectively limits phytoplanktonic primary production to the uppermost water layers, are expected to have very low benthic contribution to whole-lake PP 75 (Vadeboncoeur et al., 2008). This is especially pronounced if a lake has a high depth ratio (DR = z̄/z max ) and lacks illuminated benthic habitats. Consequently, and possibly due to more demanding sampling and measurement techniques for benthic algae and benthic PP in complex littoral habitats compared to relatively simple pelagic PP measurements, inclusion of littoral habitats has been rare in studies of humic lakes. 80 We measured primary production and respiration in both pelagic and littoral habitats of a small, highly humic lake, Mekkojärvi, which has been well studied (e.g. Kairesalo, Lehtovaara & Saukkonen, 1992;Järvinen & Salonen, 1998;Salonen et al., 2005;Taipale et al., 2008). There are numerous similar lakes in Finland and globally, with the estimated total number of lakes with surface area from 0.1 to 1 ha being 2.77·x 10 8 (Downing et al., 2006). 85 Finnish lakes are predominantly shallow due to flat topography and consequently allochthonous carbon inputs are diluted in a small volume of water resulting in higher TOC concentrations and water colour than in deeper lakes (Kortelainen, 1999). Climate change is expected to increase terrestrial exports of DOC to lakes (Tranvik & Jansson, 2002) affecting the light climate in these lakes. Previous studies have revealed that carbon fluxes in the pelagic zone of Mekkojärvi are dominated by heterotrophic processes and the lake has been considered to be net heterotrophic. However, the littoral zone of Mekkojärvi consists of a surrounding floating moss bed with submerged, inert, sedge stems supporting profuse epiphytic algal mats (periphyton) under well-illuminated conditions in summer. The productivity of this periphyton has not been measured before. Species diversity and biomass 95 of animals in the littoral zone of Mekkojärvi is much higher than in the pelagic zone, and those animal communities have been assumed to have strong reliance on periphyton (Kairesalo et al., 1992). Our main hypothesis was that the periphyton productivity in the littoral zone should be high in order to support the abundant macroinvertebrate consumers, despite the argument that benthic PP should be low considering the trophic status, light 100 regime and basin shape of the lake (Vadeboncoeur et al., 2008). Our second hypothesis was that, although the pelagic zone in Mekkojärvi is strongly dominated by heterotrophic processes, the littoral zone is more dominated by autotrophic processes and would offset pelagic heterotrophy even to the extent of making the lake net autotrophic.

Study site
The study was conducted at Mekkojärvi (61º13'N 25º3'E) in the Evo forest area in southern Finland (Fig. 1A) during the open water season in 2012. Mekkojärvi is a small (0.35 ha) and highly humic (water colour 300-800 mg Pt l -1 ) headwater lake with maximum depth 110 of 4.3 m and mean depth of 2 m. The lake receives a high loading of terrestrial organic matter from its catchment area, and has high DOC concentrations and low pH (Table 1). Despite being so shallow, the very high water colour and shelter from wind by the surrounding coniferous forest cause the lake to develop a very steep temperature and oxygen stratification gradient in summer, when the thermocline lies between 0.5-1.0 m and anoxia generally occurs below 1.0 m. The lake cannot sustain fish populations due to complete under-ice anoxia during winter, and the absence of planktivorous fish has led to very dense populations of the large bodied cladoceran Daphnia longispina in summer. Our study was conducted as part of a whole-lake food web experiment in 2012, when Mekkojärvi was divided into two basins with a plastic curtain and adult perch (Perca fluviatilis) (n = 100) were introduced to 120 one basin in early July, while the other basin remained fishless. Both lake basins were always sampled simultaneously.
Mekkojärvi has a depth ratio of 0.47 and lacks illuminated benthic surfaces due to its dark brown water colour and extremely low light penetration (light-attenuation coefficient, K d , ranges from 4.5 to 7.5). Despite its moderate DR, Mekkojärvi does not have a clearly 125 defined littoral zone, as the lake is surrounded by a floating moss bed with fallen submerged terrestrial sedges, and a water column lies underneath the vegetation mat leading to a more cylindrical lake morphometry than implied by the DR. The littoral zone is thus composed mainly of Sphagnum and Warnstorfia moss species, with submerged and emergent macrophytes such as Menyanthes trifoliata, Phragmites australis, Carex spp. and Utricularia 130 sp. associated with the moss mat. The littoral vegetation mat does not extend further than ca. 1 m from the lake edge and not deeper than ca. 0.5 m, but it forms a complex physical structure, which is well-illuminated and has a large surface area for epiphytic algal development. Relatively thick (ca. 1 cm) periphyton communities, rich in pennate diatoms (Bacillariophyceae), develop early in June, and achieve their highest biomass from July to 135 mid-August. The area of littoral vegetation mat, estimated from a satellite image by using an area calculator tool (in Retkikartta.fi, © Metsähallitus, MML, CGI 2015), is ca. 0.086 ha, which covers ca. 25% of the whole lake area and gives a ratio of pelagic to littoral zone by area of 3.0.

Physical and chemical parameters
Light attenuation was measured at each sampling period using a single planar LI-COR ® (Li-192) Underwater Quantum light sensor. Light was measured at 0.5 m intervals from the surface to the bottom of the lake and light attenuation coefficients were calculated (Kirk, 2003).

Pelagic phytoplankton biomass, production and respiration
Composite water samples were taken with an acrylic tube sampler (Limnos) by combining three pulls covering each of three strata (0-0.5, 0.5-1.5 and 1.5-3 m) 155 corresponding to epi-meta-and hypolimnion in both basins of the lake. Samples were taken 10 times between May-October in 2012. Phytoplankton biomass was estimated as chlorophyll a (chl a) concentration in the composite water samples. Samples were filtered onto Whatman GF/C glass fibre filters and chl a was measured by UV-spectrophotometry (Shimadzu UV-1800) according to a standard spectrophotometric method (SFS 5772, Finnish 160 Standards Association).
Primary production in the pelagic habitat was measured eight times from composite water samples from four strata (0-0.2 m, 0.2-0.5 m, 0.5-1 m and 1-3 m) from both basins according to the 14 C incorporation method described by Schindler, Schmidt & Reid (1972).
Subsamples were pipetted into 20 mL glass vials (containing NaH 14 CO 3 ) with four replicates (one dark sample in each depth). Samples were then incubated in situ in each stratum for 24 h. The 14 C activity was measured with a Packard Tri-Carb ® liquid scintillation counter. These 24 h determinations were assumed to approximate most closely to net primary production (NPP). Primary productivity of the water column was converted from volumetric basis to areal basis by multiplying the values from each stratum by the total depth of the stratum and 170 then summing over the whole water column. Whole-lake daily PP estimates were calculated by multiplying daily areal values by the lake surface area.
Pelagic community respiration (CR) was measured eight times, together with PP, as the increase in DIC concentration during dark incubations. Subsamples (two initials and two samples for dark incubation) were taken from composite water samples from each stratum 175 into 50 mL glass BOD bottles. Half of the bottles were wrapped in aluminium foil and incubated in situ for 24 h in each stratum and the other half (t 0h ) were placed on ice and taken immediately to the laboratory, where DIC concentration was measured with a DIC-analyser (Salonen, 1981). This was repeated for t 24h samples the next day. Areal values and whole-lake estimates for CR were calculated as for pelagic PP. Pelagic NPP was converted to GPP as the 180 sum of NPP and CR values in the photic zone, assuming that most of the CR was attributable to primary producers.

Littoral primary production, respiration and periphyton biomass
Primary production in the littoral habitats was measured eight times, concurrently 185 with pelagic PP, from in situ incubations of periphyton samples in both dark and transparent 20 ml glass vials where the consumption or production of DIC over time indicates either respiration or production. Periphyton PP on moss and on sedges per unit dry mass of substratum was compared and there was no statistically significant difference in PP rates (ttest, t 39 = -0.2005, P = 0.8421, data not shown). This allowed us to sample routinely from submerged sedges instead of from the more complex moss substratum. Furthermore, sampling of sedges was appropriate because while lying under the water surface they were metabolically inert. To insure that the sedge substratum was indeed metabolically inert, epiphyton was carefully removed from sedges and the cleaned substratum was incubated in both light and dark. No change in DIC was measured (data not shown). The lake was sampled 195 at eight sites around the shoreline, four sites from each basin. Samples were collected by clipping random sedge pieces into 20 mL glass vials and filling with surrounding littoral water. Trial incubations of this littoral water without any epiphyton or substratum demonstrated no substantive changes in DIC concentration over time. Four replicate samples of surrounding water to obtain initial DIC concentrations were taken from each sampling site. 200 Samples were incubated in situ for 2 h at noon, half of the samples in dark conditions and the other half in the surface water in the same light conditions as those from which the periphyton originated (barely submerged under the water surface). After the incubations, samples were put into ice and dark and taken immediately to the laboratory, where DIC was measured with a DIC-analyser (Salonen, 1981). Periphyton was then entirely removed from 205 the collected substratum by scraping with a spatula and was filtered onto glass fibre filters (Whatman GF/F) and chl a was determined as described above for phytoplankton. Periphyton biomass was estimated as total chl a per dry weight of substratum in a sample. Substrata were dried at 60 ºC for 24 h and dry mass was recorded.
Gross primary production (GPP) was calculated from the difference in DIC between 210 dark and light samples (GPP = DIC dark -DIC light / incubation time); CR was calculated from the difference in DIC between dark and initial samples (CR = DIC dark -DIC initial / incubation time). Net primary production (NPP) was then calculated as NPP = GPP -CR, assuming that most of the CR was attributable to primary producers. Values for each incubated sample were normalized to mg C g dry weight (DW) substratum -1 . To estimate whole-lake littoral primary production, we quantified the availability of littoral substratum, which consisted of macrophytes and moss. Littoral substratum along 40 cm of lakeshore at 12 sites within each basin was entirely physically removed, carefully collected and dried at 60º C. We calculated the average dry weight of littoral substratum per meter of lake shore (20.1 and 22.5 g DW substratum m -1 , respectively) in both basins and, since the substratum consists mainly of flat 220 and horizontally aggregated macrophytes and moss leaves, we divided the weight by two in order to conservatively estimate the mass of the illuminated side of the substratum. It should be noted that more than half the substrate collected had substantial development of epiphyton and that this approach is likely to have yielded a conservative estimate of available substratum. We compared the use of this method of whole-lake extrapolation to a simple areal 225 extrapolation method and found that the results were within 20% at the whole-lake scale.
Given the vast oversimplification of defining the heterogeneous, extremely rugose and complex physical nature of the littoral substratum as a plane, we considered that extrapolation by substratum weight was more realistic. Daily estimates for periphyton PP were calculated from a slightly modified version (Eq. 1) of the calculation described by Vadeboncoeur et al. 230 (2008). For whole-lake estimates, we calculated PP per m of lake shore and multiplied that by the total shoreline lengths of each basin (135 and 185 m, respectively). Daily estimates for periphyton respiration were calculated by multiplying the mean values at noon by 24.
where ∆t is the time increment (h), PP max is the light-saturated primary production (mg C g dry-weighted substrate -1 h -1 ) derived from incubating samples right under the surface at noon , tanh is the hyperbolic tangent function, I zt is the light at depth z (µmol m -2 s -1 ) and time t (h) and I k is the light intensity for onset of saturation, which was set to 300 µmol m -2 s -1 240 according to Hill (1996). We used the irradiance data and day lengths obtained from the Finnish Meteorological Institute in Jokioinen. To define I z , the photosynthetically active radiation (PAR) on the surface was multiplied by 0.75, which was the average fraction of light at the incubation depth over the whole sampling period (Fig. 1B).

Statistical analyses
Since the lake was divided into two basins with a plastic curtain, and one basin was fishless and the other had fish, we used multivariate repeated measures of ANOVA (RMA) to test if there were statistically significant differences in PP, CR and chl a between the sampling occasions and the basins. The data were log-transformed, when necessary, to 250 achieve the requirements of parametric analyses. A non-significant basin effect (betweensubject factor) revealed that the fish experiment and the lake division did not affect the production and respiration values, which allowed us to pool the data. The daily estimates for periphyton PP (Eq. 1) were performed using R project for Statistical Computing version 3.0 (R Core Team, 2013). Other tests were performed using IMB SPSS Statistics 20.0.0.2. If not 255 expressly noted, all the descriptive statistics are means + S.E.

Pelagic primary productivity, respiration and phytoplankton biomass
Physical and chemical parameters are reported in Table 1. Light attenuation 260 coefficient, K d , ranged from 4.5 to 7.5. Pelagic NPP ( Fig. 2A) was highest in late May (331 ± 274 mg C m -2 d -1 ), indicating a clear spring maximum of phytoplankton. PP then decreased towards autumn, being lowest in late September (9.0 ± 1.7 mg C m -2 d -1 ). PP values for 19 unreliable. RMA was not executed for the pelagic PP data, since variation was substantial and 265 assumptions for repeated measures were not met; however, a t-test showed no differences in mean PP for the whole sampling period between the two basins (t-test, t 40 = 0.254, P = 0.801), or in pairwise comparisons between the basins on each date, allowing us to pool the data from the two basins.
Community respiration also varied substantially ( Fig. 2A), and exhibited rates over 270 one order of magnitude higher than those for PP. The highest mean CR on 18 June was 3210 ± 1655 mg C m -2 d -1 , of which the hypolimnetic fraction was 78%. The mean hypolimnetic fraction of the CR in the water column over the sampling period was 77% ± 5. CR values for 1 and 15 August include values from only one basin. RMA for CR from May to August showed a significant main effect of dates (F 3, 48 = 11.88, P < 0.01), indicating change over 275 time, but there was no significant interaction between the basins and the dates (F 3 = 2.03, P = 0.12), allowing us to pool the data from the two basins.
Chlorophyll a concentration in the epilimnion was rather constant from spring to autumn (8.0 ± 1.1 mg m -3 ) and substantially lower than in the metalimnion (27.2 ± 6.2 mg m -3 ) and hypolimnion (55.4 ± 4.4 mg m -3 ) (Fig. 3A). However, the high chl a concentrations in 280 both meta-and hypolimnion presumably contain appreciable quantities of bacteriochlorophyll (BChl), which is abundant in Mekkojärvi and which overlaps with chl a in spectrophotometric determination (Karhunen et al., 2013). Indeed BChl may also have contributed to epilimnetic chl a, as chl a did not follow the same pattern as pelagic PP. When the basins were compared for the mean chl a over the sampling period, chl a was higher in 285 the fishless basin (37.8 ± 5.4 mg m -3 ) than in the basin which had fish (22.6 ± 4.2 mg m -3 ) (ttest, t 58 = -2.200, P < 0.05).

Littoral primary productivity, respiration and periphyton biomass
Littoral periphyton NPP was highest in spring and throughout the summer except for a 290 rapid decrease in early July (Fig. 2C), and showed a pattern of decline towards autumn, decreasing substantially from late-August to late-September (Fig. 2B, C). The mean NPP at noon was 35.8 ± 3.2 mg C g DW substratum -1 h -1 and the highest value was determined on 20 July (47.2 ± 11.0 mg C g DW substratum -1 h -1 ). RMA revealed a significant main effect of time (F 7, 98 = 6.940, P < 0.01), which was not affected by basin treatments (F 7, 98 = 1.098, P = 295 0.366). The only significant difference in NPP was observed between 30 August and 27 September (F 1, 14 = 3.462, P < 0.05), although the main effect of basin was not significant (F 1, 14 = 3.982, P = 0.066), allowing us to pool the data from both basins. Mean daily littoral NPP was 364 ± 66 mg C g DW substratum -1 d -1 and the highest mean rates were observed on 20 July (663 ± 2 mg C g DW substratum -1 d -1 ), which was later than the highest noon rates 300 observed on 4 July (Fig. 2B, C). This, as well as the decrease in daily NPP values on 30 August, was caused by rapid changes in weather and light conditions, which gave lower daily estimates. NPP rates remained constant until August, when variation increased between the basins.
Littoral CR was low relative to the highest NPP rates (Fig. 2B, C). The mean CR at 305 noon was 16.7 ± 2.4 mg C g DW substrate -1 h -1 ; the highest rates were observed on 4 July (37.0 ± 11.3 mg C (g DW substratum) -1 h -1 ) and the lowest on 27 September (4.6 ± 1.3 mg C (g DW substratum) -1 h -1 ). RMA revealed no significant main effect of time (F 2.88, 40.36 = 1.530, P = 0.222) and no interaction between time and the basins (F 2.88, 40.36 = 0.252, P = 0.852), or any main effect of basin (F 1, 14 = 0.013, P = 0.858). When the daily CR values were 310 calculated from the hourly rates, the mean CR over the sampling periods was 395 ± 65 mg C (g DW substratum) -1 d -1 .

Whole-lake estimates for net primary production and respiration
Whole-lake estimates showed that littoral contributions dominated whole-lake NPP 320 from spring to early autumn (Fig. 4A, B). The mean littoral contribution to whole-lake NPP over the sampling period was 81 ± 8%. Pelagic NPP contribution most early in the spring (24%) and in late autumn (54%), while in the summer the pelagic contribution varied from 2 to 20%. The highest estimate for the whole-lake NPP in the littoral was 4.1 kg C d -1 on 21 June and the lowest was 17.5 g C d -1 on 27 September, while the mean was 2.4 ± 0.7 kg C d -1 . 325 The highest whole-lake estimate for pelagic NPP was 1.1 kg C d -1 on 31 May and the lowest was 32.4 g C d -1 on 27 September, while the mean was 338 ± 165 g C d -1 . Whole-lake estimates for July could not be calculated due to missing pelagic data from the fishless basin, but the littoral contributions in the basin with fish remained high in July (77 and 91%).
Whole-lake estimates of pelagic CR showed higher values in spring when the highest 330 peak, observed on 21 June, was 10.4 kg C d -1 (Fig. 4C, D). The lowest estimate was 385 g C d -1 on 30 August while the mean was 3.2 ± 1.9 kg C d -1 over five sampling periods. Littoral CR was relatively stable, ranging from 766 g C d -1 in late September to 4.4 kg C d -1 in late August, while the mean was 2.5 ± 0.5 kg C d -1 (Fig. 4C, D).

Production to respiration ratio (GPP:CR)
The mean GPP:CR ratio in the littoral zone was 2.2 ± 0.4, n = 8), but was significantly lower in the pelagic zone (0.4 ± 0.1, n = 5) (t-test, t 11 = 3.185, P < 0.01). The highest GPP:CR ratio in the littoral was 4.8 on 20 July, and the lowest was 1.0 on 27 September. For the pelagic zone, the highest GPP:CR ratio was 1.0 on 20 July, and the lowest 340 was 0.2 on 16 August. The whole-lake average of five sampling periods was 1.6 ± 0.5. The results indicated net heterotrophy (P:R<1) in the pelagic zone but net autotrophy (P:R>1) in the littoral zone (Fig. 5). When whole-lake pelagic and littoral results were combined, the outcome indicated net autotrophy for the complete lake ecosystem (Fig. 5).

Discussion
Our results demonstrate an overwhelming dominance of littoral PP in whole-lake autotrophic production in Mekkojärvi. The lake is very humic, which limits light penetration, while its depth ratio and complex littoral structure leave the majority of the lake bottom in darkness. However, littoral vegetation lining the lake perimeter is well-illuminated and 350 supports highly productive periphyton assemblages, with mean contribution to whole-lake NPP over 80% throughout the sampling period. Intense respiration of allochthonous OC in the pelagic habitat substantially exceeded phytoplankton PP, indicating strong net heterotrophy with low GPP:CR ratios in the open water of the lake. In contrast, welldeveloped and highly productive periphyton mats led to net autotrophy in the littoral habitat, 355 and when incorporated into whole-lake PP the system GPP:CR ratio was net autotrophic.
Estimating PP and CR for structurally complex littoral zones is very challenging, and our estimates of whole-lake littoral PP and CR undoubtedly have higher uncertainty than the pelagic values. Nevertheless, the littoral values are so high that the associated uncertainty cannot alter their essential message. Moreover, the values for littoral NPP must be 360 underestimates, as we obtained these by subtracting CR from GPP, and the true respiration attributable to primary producers must be less than the CR. Hence our results demonstrate that, despite high terrestrial inputs of DOC and dominance of heterotrophic processes in the pelagic zone, some humic lakes can be metabolically balanced or even net autotrophic when highly productive periphyton communities are included in measurements of whole-lake 365 metabolism. Our findings challenge the current theoretical framework regarding whole-lake metabolism in humic systems and provide some answers to recently posed questions about the carbon budgets in lakes (Hanson et al., 2014).
Lakes with high DOC, high colour and low pelagic primary production are often assumed to be heterotrophic due to the allochthonous OC (Cole et al., 2000). However, 370 littoral primary production is often overlooked in lakes like Mekkojärvi, as many factors that regulate possible littoral contributions to whole lake autochthony would appear to seriously constrain the potential significance of periphyton. Indeed, based on Mekkojärvi's depth ratio, light attenuation and P content, equations presented by Vadeboncoeur et al. (2008) predict that the overall contribution of littoral PP to whole-lake PP should be a maximum of 10%. 375 However, these models do not account for sub-surface littoral vegetation lining the lake perimeter that provides substantial substratum for colonization by epiphytic algae.
Previous work demonstrating that benthic algal contribution to whole-lake PP declines rapidly with increasing DOC concentration in northern boreal lakes, offers a perspective for autotrophy in humic lakes when epiphytic benthic algae is not measured and 380 included as a source of PP (e.g. Ask et al., 2009). When metabolism on sediment and by periphyton colonizing only the lake bottom are measured, logically the reduced light penetration caused by high DOC concentrations explains why benthic algae could not contribute substantially to humic lake autotrophy. However, to fully account for and evaluate the contribution of benthic algae in humic lakes, the extremely productive, albeit easily 385 overlooked, PP from shallow periphyton needs to be included.
It is still likely that large north temperate lakes are generally net heterotrophic, as bacterial respiration of allochthonous OC can proceed in the whole water column during the entire year, while photosynthesis in most of the lakes is restricted to the uppermost water layers (Jansson et al., 2000). That is evident if a lake does not have extensive littoral 390 vegetation and the ratio of pelagic area to the area of littoral vegetation is very high.
However, northern areas have millions of lakes with mean depth less than 3 m and a littoral zone that dominates the pelagic zone (Wetzel, 1990). In Finland, where ca. 10% of the surface area is covered by lakes, over 95% of the 190 000 lakes are smaller than 1 km 2 (Raatikainen & Kuusisto, 1990) and a majority of these are humic with a high concentration 395 of organic matter (Kämäri et al., 1991). Although our results are from only one lake, we can assume that in small humic lakes, where littoral vegetation comprising floating moss and macrophytes is a widespread characteristic, estimates of PP and lake metabolism based only on measurements from the pelagic zone and ignoring epiphytic PP are likely to be seriously erroneous. 400 Pelagic PP in Mekkojärvi was similar to that in previous studies of the lake (Salonen et al., 2005;Taipale, Kankaala & Jones, 2007): a clear spring peak was followed by a decrease towards autumn with mean rates under 100 mg C m -2 d -1 and relatively high variation. Salonen et al. (2005) measured the CR in Mekkojärvi in situ from a 1 m water column in the epilimnion and found respiration rates 5.2 times higher than PP. Even so, they 405 concluded that total water column CR was likely to be considerably underestimated, as CR in the anoxic hypolimnion has been shown to contribute significantly to whole-lake CR (Cole & Pace, 1995). Our finding of a 77% contribution from the hypolimnetic fraction to total pelagic CR, with a mean rate of almost 1 g C m -2 d -1 is consistent with this view. One of the greatest uncertainties in measuring PP by 14 C incorporation is whether it measures net or 410 gross primary production, or something in-between (Marra, 2009). We used a 14 C incorporation method with 24 h incubation, as has been widely used at other nearby humic lakes (Arvola et al., 1999), as a longer (24 h) incubation time has been considered appropriate in unproductive lakes, and such an incubation time should give values approximating NPP.
Other uncertainty is involved in deriving GPP from NPP by adding CR, as this includes the 415 respiration of organisms other than autotrophs and gives an overestimate of GPP. The true autotrophic respiration must be less than CR and therefore the true PP:R ratios of the lake will be higher than our estimates; hence the lake must actually be more autotrophic and our interpretations remain conservative.
Oxygen data from the pelagic (Table 1) (Table 1). DO in the littoral was measured just above the moss from several sampling sites and from the epilimnion (0-0.5 m) in the pelagic. In the littoral zone oxygen saturation varied between 105 and 120%, whereas in the pelagic it was under 70%. These results give 425 support to our high PP rates and demonstrate marked spatial heterogeneity in oxygen in the lake. Hence the undersaturation of oxygen in the open water is not inconsistent with an overall metabolic balance for the lake.
The extremely high littoral production we measured in Mekkojärvi is likely to support a productive littoral food web. Van Duinen et al. (2013) found that periphyton constituted a 430 significant food resource for aquatic macroinvertebrates in small and highly humic bog pools in southwest Estonia. Lau et al. (2014) also showed that autochthonous resources are the main driver of secondary production in five dystrophic lakes in Sweden. While they did not directly measure primary production, our results of high epiphytic production support their findings. The most diverse and abundant macroinvertebrate community in Mekkojärvi lives 435 within the littoral moss, and it is likely that periphyton is the basal food resource in this habitat. Moreover, the most widespread fish species in small humic lakes, European perch (Perca fluviatilis), relies mainly on littoral macroinvertebrates for food in such lakes (e.g. Nyberg et al., 2010). The contribution of periphyton to macroinvertebrate diets and the lake food web remains unclear in Mekkojärvi and deserves further study. 440 Our findings demonstrate that littoral zones in humic lakes offer a highly productive environment for lake biota, and are also as critical to whole-lake metabolism as the pelagic processes that have been much better documented in the past. Littoral primary production can dominate pelagic PP, and in doing so can offset the high heterotrophic pelagic respiration even to the extent of producing a balanced or net autotrophic whole-lake metabolism. This 445 contradicts the most widely held current view (Del Giorgio, Cole & Cimbleris, 1997) and demonstrates that there is much more to be learned regarding the role of littoral processes in humic lake metabolism. However, our estimates cover only the ice-free period. Respiratory processes, mainly due to decomposition of OC, continue in boreal Finnish lakes at a low levels throughout the winter ice cover , which may still make these 450 lakes net heterotrophic on an annual basis.
We estimated that the ratio of pelagic to littoral zone by area in Mekkojärvi is 3.0, which was derived from areal calculations of both habitats from satellite imagery. Littoral area was determined as originating at the land-water interface and ending at the edge of the vegetation. Although crude, we believe this approach provides a reasonable estimate. The 455 number of lakes with the similar pelagic-littoral ratios approaches 10 5 worldwide (Peters & Lodge, 2010) with the majority other lakes having lower values. Yet, it is hard to give quantitative estimates of the prevalence of humic lakes ringed by similar floating vegetation with highly productive epiphyton as occurs in Mekkojärvi. Bryophyte mats and sedges are characteristic of oligotrophic humic lakes (Rintanen, 1996;Heino & Toivonen, 2008) and 460 therefore there is a high probability that many lakes that are similar to Mekkojärvi in size and region contain similar vegetation features. Aquatic moss stands in small, humic Finnish lakes have expanded in recent decades showing a clear trend of filling the entire waterbody (Rintanen, 1996). A recent overview of the biodiversity and ecosystem functioning in natural bog pools (Beadle et al., 2015) highlighted the need for future studies, as knowledge of these 465 ecosystems is limited. While lakes like Mekkojärvi differ in some regards from bog pools, there are relevant similarities in humic contents, vegetation and structure of the basins. Those bog pools are assumed to be net heterotrophic, even though there is evidence that invertebrates significantly rely on periphyton (Van Duinen et al., 2013). We suggest that littoral primary production and respiration should also be taken into account in the future 470 studies of humic systems, as our data demonstrate that they are likely to be more productive and less heterotrophic then previously assumed.  August. for pelagic CR) have values only from one basin. Littoral NPP and CR represent mean (± S.E.) net primary production and community respiration of periphyton per gram dryweighted substratum per hour (B) and per day (C). Differences in daily light regime account for inconsistencies in proportional scaling between hourly and daily rates on individual dates. 630