Haitao Li, Institute of Geographic Sciences and Natural Resources Research, the Chinese Academy of Sciences, Jing Xie, Beijing Forestry University, Michael Deal, University of Toledo, Jianye Xu, University of Toledo, and Jiquan Chen, The University of Toledo.
Background/Question/Methods and Results/Conclusions: The response of carbon flux to biophysical control factors at multi-time scales is of both scientific and practical interest. To understand the temporal dynamics of ecosystem carbon cycling, CO2 fluxes were measured over an oak forest ecosystem in Ohio. The measurements were made from 2004 to 2009 using the eddy covariance technique. Annual net carbon exchange (NEE) ranged from −93 to 190 gC m−2 year−1 above the canopy. Inter-annual variability in NEE was significantly related to length of growing season for the tree canopy. We used wavelet analysis to examine the multi scale structure of NEE and local climatic factors (e.g., light, vapor pressure deficit (VPD), temperature and net radiation), based on six year data series with the time interval of 30 minutes. We further used wavelet-coefficient regression, in which the dependent and independent variables were wavelet transformed prior to analysis, as a means to formalize scale-specific relationships between NEE and environmental factors. We also applied Granger’s causality test to analyze the causality between NEE and environmental variables at different time scales. Increases in VPD were associated with reduced NEE at all scales(P<0.05), with the greatest apparent constraints occurring at the scale of 60 minutes. Correlations between NEE and all other environmental variables subtly differed among scales, with soil temperature showing a negative correlation and net radiation showing a positive correlation(P<0.05). Although time scale was shown to influence the relationships of NEE and biophysical control factors, radiation was verified to “Granger cause” NEE at all scales by Granger’s causality test. We found that the wavelet transform and wavelet-coefficient regression efficiently characterized scale-specific temporal pattern in these data. We also found that different environmental factors showed up as good predictors of NEE at different scales and that these differences in scale greatly facilitated interpretation of the mechanisms relating biophysical control factors to ecosystem functionality.