In 2013, the daily mean air temperature remained above 0 °C after the 11th of April, with no cold spells after that. In the warm winter 2013–2014, the mean air temperatures rose above 0 °C for the first time already in late December 2013, then varied around 0 °C resulting in a slowly warming spring (Fig. 2). Air temperature during winter and spring months (December to May) was higher in 2014 (p < 0.004, n = 360), but daily mean precipitation during the same period was the same in both years (p = 0.976, n = 360; Fig. 2). In 2013, the ice cover period was longer and additionally both the ice thickness as well as the snowpack thickness was greater in 2013 (Table 1). In 2013, the snow cover above ice cover reached the maximum, 18 cm, in mid-March, while in 2014 the maximum was already in early-February, 28 cm, and there was less than 1 cm of snow on the ice after 20 February. Also, in the forest, the snow cover lasted longer and was thicker in 2013 than in 2014 (Table 1).
Differences in snow accumulation and melt were also reflected in the discharge pattern of the streams (Fig. 3). In 2013, the discharge decreased gradually during the period of ice-cover and increased rapidly after the mean air temperature reached 0 °C in April (Fig. 2). The start of the freshet was sudden due to the rapid increase in air temperature followed by the quick thawing of the accumulated snowpack (Table 1). The event reached its peak discharge (1.57 m3 s−1 and 1.82 m3 s−1 in the inlets and the outlet, respectively) five days after the onset of the freshet and the discharge then decreased gradually towards the end of May. In the following winter, the peak of melting event and the period of highest discharge (peak 0.46 m3 s−1 and 1.01 m3 s−1 in the inlets and the outlet, respectively) started already in mid-December and lasted until late January. Even though the snow cover was thickest in February, and it persisted until the end of March (Table 1), there was no clear spring freshet. Due to periods of high precipitation, the discharge increased slightly in three events in March–April, but the events were small in comparison to the thaw in January (Figs. 2, 3). In 2013, the snowmelt event completely masked the influence of precipitation. The water output from the lake was higher than the input through the streams, and the difference between the inputs and output was most prominent during the high flow events during the ice cover periods (Fig. 3). In 2013, the main inlet discharge was 51% and in 2014 58% of that in the outlet.
In all inflowing streams, the total discharge was higher in 2013 than in 2014 (p < 0.000; Fig. 4a, b). The mean discharge in the outlet was three times higher in 2013 than in 2014, i.e., 0.32 m3 s−1 and 0.10 m3 s−1, respectively. In 2013, water was flowing in 15 inlet streams whereas in 2014 five out of the 15 inlets (S1, S4, S13, S14, S15) were dry. The secondary inlets were of greater importance in 2013; they covered 21 and 11% of the discharge in the outlet in 2013 and 2014, respectively. Taking together, the inlets covered 72% and 69% in 2013 and 2014 of that in the outlet. The lake thus receives water inputs also as overland flow from the riparian zone, or through direct groundwater inputs, which were not directly measured.
CO2 and DOC concentrations
In general, the CO2 concentrations showed differences within and between the streams, but despite these substantial hydrological differences, there were only small concentration differences between the years (Fig. 4c, d). In the medium-sized streams the concentrations of CO2 were the same in both years (p = 0.29, n = 43), whereas in the large-sized inlet streams, the concentrations were higher in 2013 than in 2014 (p < 0.028, n = 75), indicating stronger connection with shallow groundwater in 2013. However, in the large outlet stream, the concentrations were the same in both years (p = 0.303, n = 34). The small streams were completely dry in 2014. However, when present, the small streams had the same CO2 concentration as the other streams (p = 0.448, n = 125).
The DOC concentrations had high instream variability and did not show apparent differences between the years in the medium (p = 0.773, n = 43) or large-sized streams (p = 0.428, n = 77; Fig. 4e, f). We measured very high DOC concentrations in 2013 in the stream number 4, whereas in 2014 that stream remained dry. DOC concentrations were lowest in small-sized streams, and of similar size in medium and large-sized streams (p = 0.032, n = 136). In the outlet stream, DOC concentrations were higher in 2013 than in 2014 (p < 0.000, n = 34).
Between the study years, the relationship between the concentrations of CO2 and DOC and discharge was stronger in 2013 than in 2014 (Table 2). In 2013, significant correlation between the variables were found in various stream size groups. In 2014, the relation between concentration and discharge was indistinguishable, and significant correlation with discharge was only found with DOC concentrations in medium sized streams and with CO2 concentrations in large sized streams. In all stream size groups, the significant relationship between DOC concentrations and discharge was always positive, whereas the response of CO2 concentrations to discharge differed between the stream size groups. In medium sized streams, the increasing discharge in 2013 resulted as decreased CO2 concentrations, while in large sized streams and in the outlet this relationship was positive and the CO2 concentrations increased with increasing discharge. No clear relationship between DOC concentrations and discharge was found in small sized streams.
Lateral transport of CO2 and DOC in streams
The total amount of CO2–C and DOC–C transported into the lake was almost 1.5-fold higher in 2013 than in 2014 (Fig. 5). However, the difference in daily lateral transport between the years was statistically significant neither with CO2–C (p = 0.173, n = 364) nor DOC–C (p = 0.174, n = 364). The DOC–C transport dominated the C transport in streams in both years. In 2013 and 2014, during the six months study period, the CO2–C input was 5760 and 3818 kg, respectively, and the DOC–C input was 43,794 and 29,308 kg, respectively. Thus, the transported CO2–C and DOC–C in 2014 were 66% and 67% of the amounts transported in 2013, respectively following the much higher discharge in all streams in 2013 than in 2014. The output from the lake exceeded the total input of CO2–C as well as DOC–C in both years and similarly to input, the total output from the lake was higher in 2013 than in 2014, i.e., 7217 and 4744 kg CO2–C, in 2013 and 2014, respectively, and 65,659 and 37,442 kg DOC–C, in 2013 and 2014, respectively. Thus, the output of CO2–C was 25% and 24% higher than the input in 2013 and 2014, respectively. Corresponding values for the DOC–C was 50% and 28% in 2013 and 2014, respectively. The differences in daily output between the years were statistically significant with both CO2–C (p = 0.028, n = 364) and DOC–C (p = 0.000, n = 364). In both years, the small and medium-sized streams only contributed 2 and 3% of the total transport of CO2 and DOC–C, respectively.
The timing of the lateral transport differed between the years and was connected to events with higher discharge. In 2014, the lateral transport mainly took place during the ice cover period, while in 2013 the transport was more evenly distributed, even though the highest inputs into the lake took place during the freshet (Fig. 5). In 2013, most of the transport in the inlets occurred during the freshet, when 51% of CO2–C and 54% of DOC–C was transported. The corresponding values for the ice cover period were 33% of CO2–C and 24% of DOC–C. In 2014, the input was highest during the ice cover period, due to two different hydrological events: the small snowmelt, which induced small flooding in January, and the precipitation lasting few consecutive days at March. During these periods, 68% and 64% of CO2–C and DOC–C, respectively, was transported. During the freshet in 2014, the input was small and covered only 11% of CO2–C and 6% of DOC–C transport. The relation between CO2–C and DOC–C indicate higher input of DOC–C during melting events and lower during cold periods, although the relation during the whole study remained constant; 12% of the laterally transported input C was CO2–C and 88% of DOC–C during both years.
The timing of the output of CO2–C seemed to have a similar pattern with the highest transport in a freshet in 2013 (48%) and ice cover period in 2014 (82%). The outputs of DOC–C, however, were equal in size during the ice cover period and freshet in 2013, 41% in both periods indicating that the discharge was not the only factor defining the output. In 2014, 82% of the DOC–C output took place in the ice cover period. Similar to the input, the total output was dominated by the DOC–C and with no differences between the years: 10% of the C was in CO2–C and 90% in DOC–C.
In 2013, when the lake was ice-covered for 155 days, the lake showed inverse stratification with temperatures close to 0 °C below the ice (Fig. 6a). At the onset of the freshet, temperatures from the depths of 0.5 m to 4.5 m decreased further. From 19 April onwards, the surface temperature increased gradually and showed a diurnal variation. The lake was stratified at the onset of the freshet, with the thermocline at a depth of 4.0 m. However, the mixing started in the middle of the freshet, but the turnover was not complete until 2 May, i.e., one day after the ice-out on 1 May.
The warm period in January 2014 did not affect the under-ice thermal stratification, although the surface water (< 2.5 m) temperatures fluctuated slightly after the hydrological peak entered the lake (Fig. 6b). In general, the warmer year resulted in higher lake water temperatures. In March, the snowless ice-cover was already thin (Table 1), and as a result, the below-ice temperatures were higher than in 2013. During the last ten days before ice-out, the surface water under the ice was 2.0 °C warmer than in 2013. Similar to 2013, there was a diurnal variation in water temperature under the ice. The surface water then warmed up gradually, and the water column was homothermal at 3 °C from mid-March until the ice-out on 12 April. After that, the water column warmed and finally stratified at the end of May.
CO2 and DOC concentrations in the lake
During the ice cover period in 2013, CO2 concentrations at 7.0 m depth increased, while concentrations at 1.5 and 2.5 m remained constant (Fig. 7a). On the first week of the ice cover period, the mean concentrations of CO2 at all depths were similar, i.e., 3.0, 2.6 and 3.4 mg L−1 at the depths of 1.5, 2.5 and 7.0 m, respectively. Concentrations at 7.0 m then increased clearly, so that just before the start of the freshet, the concentrations in the two upper depths were only half of that at 7.0 m depth. When the freshet started, the concentrations at 1.5 m depth increased within four days close to concentrations at 7.0 m. At the same period, the lake stayed stratified. The concentrations at 2.5 m depth increased smoothly after ten days from the onset of the freshet, simultaneously with the lake surface water mixing. During the freshet, the concentrations at 7.0 m did not increase anymore. The mean concentrations during the freshet were 8.3, 4.3, and 10.0 mg L−1 at the depths of 1.5, 2.5, and 7.0 m, respectively. At the ice-out, the CO2 concentration was lowest at 2.5 m, 5.8 mg L−1, and highest at 7.0 m, 8.4 mg L−1. Since CO2 accumulated in the water column was rapidly released to the atmosphere after the ice-out (results not shown), CO2 concentrations at all depths dropped clearly at the beginning of May. In mid-May, the concentrations at 7.0 m started to increase again, while concentrations closer to surface decreased. The mean concentrations during the open water period were 5.4, 5.7, and 7.3 mg L−1 at 1.5, 2.5, and 7.0 m depths, respectively.
In 2014, the CO2 dynamics was more unstable than in 2013, and thus in agreement with water column stratification (Fig. 7b; 6b). At the beginning of the ice cover period, the concentrations were low; 2.0, 2.5, and 2.6 mg L−1 at 1.5 m, 2.5 m, and 7.0 m, respectively. The warm period in December-January resulted in smooth increases in CO2 concentrations at 1.5 m depth, but it did not have clear influences on the concentrations at 2.5 m and 7.0 m, which increased steadily until the freshet. The mean concentrations during the freshet were 5.9, 7.0, and 9.7 mg L−1 at depths of 1.5, 2.5, and 7.0 m, respectively. The water mixing started one week after the onset of the freshet but did not affect deeper (> 6 m) layers before the four last days of the freshet, and the turnover completed on 11 April, one day before the ice-out on 12 April. The concentrations at all depths increased to their maxima just before the ice-out simultaneously with complete turnover. After the ice-out, concentrations rapidly declined until May. The mean concentrations during the open water period were relatively low, 2.6, 3.6, and 4.3 mg L−1 at 1.5, 2.5, and 7.0 m, respectively.
The mean concentrations during the ice cover period were significantly lower in 2014 than in 2013 at 1.5 m (p = 0.00, n = 14) and 7.0 m (p = 0.00, n = 14), but the concentrations at 2.5 m did not differ (p = 0.08, n = 14). Concentrations during the freshet in 2014 were lower at 1.5 m (p = 0.00, n = 36) and at 2.5 m higher (p = 0.00, n = 36) than in 2013 and at 7.0 m depth the concentrations were the same in both years (p = 0.48, n = 36). In 2014, the concentrations during the open water period were lower than in 2013 at all depths (1.5 m, p = 0.00, n = 82; 2.5 m, p = 0.00, n = 82; 7.0 m, p = 0.00, n = 82).
The mean DOC concentrations in the lake surface during the ice cover period were 16.2 and 13.2 mg L−1 in 2013 and 2014, respectively (Table 3). After the ice-out, the concentrations dropped to 14.0 and 10.7 mg L−1 in 2013 and 2014, respectively. The concentrations close to the bottom (12.0 m) were lower in comparison to the surface concentrations, but the decline from the values of the ice cover period to open water period was small.