Growing Season Water and Carbon Dynamics of a Subalpine Wetland with Complex Terrain
Across the entire 2018 study period (96 days), cumulative evapotranspiration (ET) reached a total of 157.7 mm. During Snow Melt (17 days in duration), mean ET was 1.0 mm/day (±0.7) and increased throughout the seasonal phase (Figure 3), because of evaporation from snowmelt runoff. During this season ET was most likely dominated by evaporation, since the majority of the wetland vegetation was still under a blanket of snow and without leaves. Cummulative ET during snowmelt was 16.6mm. ET continued to increase into Green Up , reaching the seasonal maximum for the entire study period (3.5mm/day) from 12 – 17 July (DoY 193-198). Overall, Green Up had a daily average ET of 2.3 mm/day (±1.1), contributing 57.5 mm to the overal cumulative ET. Green Up represented only 28% of the study period (27 days), but made up 37% of total study period ET.Throughout Snowmelt and Green Up the site experiencedSteady shade (around 2 hours of shade daily). After Green Up, ET was influenced by the increasing horizon shadow effect and followed a decreasing trend, similar to Rg through the remainder ofPeak Growing Season and into Late Growing Season (Figure 2a and Figure 3). Therefore, daily Peak Growing Season ET was lower than Green Up with a daily average of 2.0mm/day (±0.5). Since Peak Growing season was longer (35% of study period, 34 days), it resulted in a greater cumulative total ET (66.6 mm) compared toGreen Up despite the lower daily mean ET. Peak Growing Season contributed 42% of total study period ET. During Late Growing Season, ET remained low (mean of 0.94 mm/day (±0.3)) and contributed only 11% (17.0 mm) to total study period ET. Therefore,Green Up and Peak Growing Season accounted for the largest (124.1 mm, 79%) chunk of cumulative ET during the course of this study.
In this study we use the term “carbon flux” to represent only carbon dioxide (CO2) exchange between the wetland and the atmosphere. The carbon (C) flux was extremely variable during Snow Melt and represented a strong source, defined as negative Net Ecosystem Exchange (NEE= GPP- Reco), releasing an average of 1.4 gCm-2 day (±2.2) into the atmosphere for a total release of 23.8 gC over this period (Figure 4). During Snow Melt , there was no photosynthetically active vegetation present so almost all NEE was defined by Ecosystem Respiration (Reco). The wetland then fluctuated between a source and sink during Green Up, when variability in CO2 uptake/release continued until the 6th of July. During this time the ground surface became increasingly snow-free and vegetation began leafing-out. Following the 6th of July, the wetland was a net C-sink (i.e. CO2 uptake with positive NEE) until theLate Growing Season . During the Green Up season, the wetland took up an average of 0.58 gCm-2day (±0.7) for a total carbon sink of 15.6 gC. Maximum CO2 uptake from photosynthesis (defined as Gross Primary Production: GPP) occurred from 29 July to 2 August, once the entire site became snow-free and vegetation began to grow, with an average daily GPP of 6.4 gCm-2day and Reco of 4.5 gCm-2day, respectively. This equated to an average NEE of 1.9 gCm-2day (±0.5), representing a net carbon sink of 60.5 gC over the Peak Growing Season . Net C sequestration largely took place between the hours of 08:00 and 19:00 duringPeak Growing Season from the middle of July to the end of August (Figure 7). At this time, cumulative net CO2 uptake was high enough to offset CO2 losses during the Snow Melt period and shift the wetland into a cumulative net sink (Figure 4b). Following the seasonal maximum, there was a decreasing trend in the net C flux for the remainder of the study. During the Late Growing Season period, the site remained a consistent C sink with a net C uptake of 0.6 gCm-2day (±1.0) for a cumulative sink of 10.6 gC. NEE spiked to 3.3 gC on 7 September because of abnormally warm air temperature (9.4°C) compared to the Late Growing Season’s average of 5.5 °C. Over the entire study period, Bonsai was a net sink of 63gC; however, it is possible that the wetland is an overall annual source of CO2 if emissions during the snow-covered period are similar to what we found during Snow Melt .
Environmental Factors driving Temporal Variability in Observed Carbon and Water Fluxes
Bonsai carbon flux was influenced by shade and had a statistically significant negative relationship with GPP (R2=0.75; p<0.01) and Reco (R2=0.39, p<0.05), and a statistically significant positive relationship with NEE (R2=0.73, p<0.01) duringDynamic Shade (Figure 5b). This indicates that each hourly increase of shade during Dynamic Shade decreased GPP by 0.8 gCm-2day (y=-0.8x+6.7), overall decreasing the C sink strength (NEE) by an average of 0.5 gCm-2day (y=-0.5x+3.9). Therefore, horizon shade negatively impacted C uptake at the site. The period ofStable Shade yielded non-significant results for Reco (Figure 5a). Thus, other environmental variables influenced the temporal variability in water and carbon fluxes inStable Shade .
ET also had a statistically significant negative relationship (R2=0.66, p<0.01) with hours of shade per day (hrs/day) during Dynamic Shade (Figure 5b), but not duringStable Shade (R2=0.04, p>0.05) (Figure 5a). During Dynamic Shade , each hour of shade decreased ET losses by 0.42 mm/day (y=-0.4x+3.3). Therefore, there was a statistically significant relationship between ET and available energy over the study period (R2=0.79, p<0.01), indicating that increased shade (and lower available energy) decreased evaporative losses and established a greater potential for water storage.
Figures 6 and 7 displayed similar results to the statistical analysis above, in which ET and GPP were strongly influenced by the horizon shadow. In a uniform and non-complex environment, fluxes typically display a bell-curve pattern, increasing in the morning, peaking in the afternoon, and gradually decreasing into the evening. However, at Bonsai there was a noticeable shift from this pattern in the middle of July, when half hourly ET had a sudden reduction at 16:00 hours (0.13 to 0.05mm, Figure 6). This trend became more intense into August and the period of Dynamic Shade , when shade decreased energy and water fluxes in the morning from 09:00 to 11:00 hours and at 15:00 hours in the afternoon (Figure 7). The response of ET to horizon shade followed a nearly identical pattern as that of Rg and Q* (Figure 6). ET (and Rg) should have declined from the afternoon until sunset; however, sharp reductions were observed at 16:00 hours in the middle of July, indicating that horizon shade shaped the energy and water fluxes during the Peak Growing Season . Similar to the results found in the statistical analysis, GPP did not follow the same pattern as ET, Rg, or Q* during constant shade, but did have a narrowing bell curve shape in the portion of increasing shade in Peak Growing Season . High Reco during the Snow Melt period resulted in a C source (averaging 1.4 gCm-2day-1) with the largest C-source contributions occurring from advection overnight, in the middle of June between 22:00 and 06:00 hours. C uptake through GPP was greatest midday from the middle of July to the middle of August between the hours of 08:00 and 18:00. In the middle of August there was a rapid decline in C uptake, as the wetland transitioned from 0.3 gC/hh to 0.06 - 0.13 gC/hh and closer to net neutral C uptake during daytime hours.
Using general additive modelling and the list of key environmental variables that were available to us from measurements at the site, we were able to explain up to 84% of temporal variability in ET (Table 4), but only 71% of GPP (Table 5). For both water and carbon fluxes, we found that the most significant variable in explaining the temporal variability was incoming solar radiation (Rg). For GPP, Rg was the most important variable, reducing model performance the most, if excluded. Interestingly, the model for ET had more variables that were significant in explaining its temporal variability (Table 4), including surface soil temperature (at 2 cm depth), vapour pressure deficit, shade – as a factor variable, and time of day (Hour). In contrast, the model for GPP had day of the year (DoY) as the only other variable that turned out to have statistically significant estimated coefficients in the best-fitting model (Table 5). All other variables tested were not significant. This finding supported the spectrographs of the data (Figure 6), where the changes in ET seemed to be more visually correlated to changes in Rg than GPP. Our analysis suggests that ET was related to the temporal variability in environmental variables, while GPP was dependent on plant physiology (which was not measured in this study) such as leaf area and greenness factor. This is further evidenced by the fact that GPP continued to increase throughout the Peak Growing Season , despite decreased Rg, having a higher mean GPP value inDynamic Shade versus Stable Shade (Table 1). Although, the difference was not found to be statistically significant. Another interesting finding was that ET was nonlinearly related to Rg and that the two curves differed between Dynamic Shade versusSteady Shade (Table 4 and Appendix 2). The estimated parametric coefficient for shade-factor in the ET model was also negative for the Steady Shade group (Appendix 2: fShadeSteady=-0.0096116), which meant that the curve for Steady Shade was offset from Dynamic Shade (lower intercept), such that for a given Rg value, ET will be higher in Dynamic Shade thanSteady Shade . This was evident if we looked at our observed data using binned-Rg (Appendix 3). Dynamic Shade had a lower mean ET because shade reduced overall Rg, resulting in more values at lower Rg during that period (Appendix 3). Interestingly, this was not the case for GPP, where shade factor was not found to be significant in the best-specified gam model (Table 5 and Appendix 1). Here, GPP was higher with lower Rg-bins (Dynamic Shade ), but not much different from higher Rg-bins (Steady Shade ) (Appendix 3). These finding have implications for the local water balance and support the notion that shaded ecosystems have reduced overall water loss from shade, because of greater reductions in evaporation than transpiration.