3. RESULTS
3.1. Seasonal and inter-annual variations of meteorological
Environmental factors varied greatly in different seasons from January 2007 to December 2016 (Fig. 1). The peaks of monthly environmental factors (Ta, Ts, GDD, PPFD, PPT, RH and VPD) were not consistent (Fig. 1). The maximum value of monthly GDD occurs in July (Fig. 1a). The average annual air temperature (Ta) and soil temperature (Ts) were -1.0 ± 0.5 and 3.0 ± 0.8°C, respectively. Only in April, May and June, the average monthly Ta is higher than monthly Ts (Fig. 1a). This may be caused by melting of snow and ice, which absorbed heat from the soil. Based on a threshold deviation from average annual Ta of 20%, the years 2014 and 2015 were classified as warmer years (>0.8 Ta), the years 2007, 2010, 2013, 2016 were considered normal years, and the years 2008, 2009, 2011, 2012 were defined as cooler years (<1.2 Ta) (Fig. 1). The average annual precipitation (PPT) was 456.5 ± 86.2 mm. The minimum annual rainfall was 323.1 mm occurred in 2012 year, and the maximum annual rainfall was 578.1 mm occurred in 2013. The years 2007, 2013, 2014 were classified as wetter years (>1.1 PPT) and 2012, 2015, 2016 were classified as drier years (<0.9 PPT) (Fig. 1).
3.2. Seasonal and inter-annual variations of GPP, RES and NEE
Both monthly GPP (162.7 ± 8.4 g C m−2month−1, mean ± S.E., with the same notation hereon in) and monthly NEE (−61.1 ± 6.3 g C m−2month−1) peaked in July, while the maximum monthly RES (109.4 ± 8.8 g C m−2month−1) appeared in August (Fig. 2a). The wetland ecosystem is a carbon sink from June to August with a negative NEE (Fig. 2a). During the growing season, linear regression analysis indicated that compared to monthly RES (r2= 0.44, p < 0.001), monthly GPP (r2= 0.88, p < 0.001) played a more important role in controlling the change of monthly NEE.
Annual GPP was 500.3 ± 59.4g C m−2year−1 and annual RES was 620.7 ± 74.2g C m−2year−1, respectively (Fig. 2b). Both monthly GPP (r2= 0.11, p =0.33) and monthly RES (r2= 0.28, p = 0.12) in the peak growing season (July and August) were not predominantly determined for variations of annual GPP and annual RES. The10-year mean NEE was 120.4±34.8 g C m−2year−1 and a range of 76.3 g C m−2year−1 in 2010 to 184.8 g C m−2year−1 in 2014 (Fig. 2b). The change of annual NEE was neither directly controlled by annual RES (p = 0.06) nor directly controlled by annual GPP (p = 0.61), which was controlled more strong by annual RES to a certain extent compared to annual GPP. The sum of monthly NEE of July and August was also not predominantly determined for variations of annual NEE (r2= 0.12, p =0.32). The non-growing season RES accounts for a small proportion of annual RES, but it is significantly positively correlated with annual NEE (r2= 0.50, p < 0.05) and annual RES (r2= 0.45, p < 0.05). This finding suggested that CO2fluxes in alpine wetland during the non-growing season were crucial for annual carbon balance of alpine wetland.
3.3. Influence of environmental factors on seasonal variations of GPP, RES and NEE
The value of proportional reduction in error is more than 0.92, indicating that the analysis results of CART on monthly CO2 fluxes (GPP, RES and NEE) are acceptable. The results of CART indicated that GDD was the main control factor that affecting the change of monthly GPP and NEE. GDD accounted for 78% and 79% change of monthly GPP (Fig. 3a) and NEE (Fig. 3c), respectively. Furthermore, the splitting values of GDD were all around 230℃d, and appeared at the beginning of July or the end of August, indicating that thermal condition was the main control factor that determined the change of monthly GPP and NEE during the peak of plant growth stage. The factor of the second right root node of GPP is Ta, suggesting that Ta was crucial to the photosynthetic and carbon fixation capacity of plants in wetland ecosystem during the peak growth period. Variability in monthly RES was predominantly determined by Ts in the whole year (Fig. 3b). The factor of the second right root node of NEE was Ts, suggesting that the carbon dynamics of wetland ecosystem to some extend was controlled by RES more than GPP in non-peak growth period. In addition, during the growing season, the influence of GDD on variations of monthly GPP was stronger than the change of monthly RES (p < 0.001) (Fig. 4). Furthermore, the growing season NEE had no significant correlation with growing season Ta (r2=0.06, p=0.48) and PPT (r2= 0.15, p=0.27). However, the non-growing season NEE significantly correlated with non-growing season Ta (Fig. 5a) and PPT (Fig. 5b).
3.4. Influence of environmental factors on annual GPP, RES and NEE
Annual GPP and RES had no significant correlation with annual environmental factors (p > 0.05, Table.1). The correlation analysis showed annual GPP had significant positive and negative correlation with the growing season Ta (p<0.05) and PPT (p<0.05), respectively, suggesting that the increase of temperature and decrease of precipitation in growing season was beneficial to the improvement of GPP. The result was noteworthy that the Ts during the non-growing season had significant impact on annual GPP of the following year (p<0.05), indicating that the warmer Ts during the non-growing season promoted the activities of microorganisms and accelerates the decomposition of litters and soil organic matter. Therefore, the soil of wetland contained a lot of nutrients and promoted the photosynthesis and carbon sequestration of vegetation in the following year. The further partial correlation analysis showed that annual RES had significant positive correlation with the Ta and Ts (p<0.05) of the non-growing season. Annual PPT exhibited significant impact on variation of annual NEE (Table 1), and linear regression analysis indicated that the non-growing season Ta (Fig. 5c) and PPT (Fig. 5d) had a major impact on the change of annual NEE. However, the further partial correlation analysis showed that non-growing season air temperature (Ta, p = 0.05), rather than precipitation (PPT, p = 0.25) was a predominant determinant on the change of annual NEE. Overall, hydrothermal condition especially in non-growing season was the most important variable for describing variation on annual carbon fluxes.