Multi-scale dynamics and environmental controls on net ecosystem CO2 exchange over a temperate semiarid shrubland

Abstract

Our understanding of the variability in net ecosystem CO2 exchange (NEE) across different timescales is still limited, as terrestrial carbon cycle models often mismatch data at multiple timescales. Especially, multi-scale environmental controls on NEE are less well understood in semiarid shrublands than in mesic ecosystems. We collected eddy-covariance measurements of NEE over five years (2012–2016) from a semiarid shrubland in northern China, and then used continuous wavelet transform (CWT), wavelet coherence (WTC), and partial wavelet coherence (PWC) analysis to investigate how photosynthetically active radiation (PAR), air temperature (Ta), vapor pressure deficit (VPD), and soil water content (at 30-cm depth, SWC30) modulate the variability of NEE in the time-frequency domain. CWT revealed that NEE not only had clear daily and annual periodicities, but also oscillated strongly at intermediate scales (days, weeks to months). At the 1-day period, NEE showed significant WTC with PAR, Ta, and VPD during growing seasons, with NEE leading PAR by about 1.0 h and leading Ta and VPD by over 3.5 h. At the 1-year period, NEE also showed strong WTC with PAR, Ta, and VPD throughout time, with NEE lagging Ta by 19 days and lagging PAR and VPD by about 40 days. At intermediate periods, non-continuous areas of significant WTC were observed between NEE and environmental factors, notably between NEE and PAR at 4–32-day periods during growing seasons. PWC revealed a greater modulating effect of PAR than that of Ta on NEE at intermediate periods. However, the intermediate-scale PAR effects were largely weakened during spring or summer drought periods (i.e., low SWC30). In addition, drought events were identified as hotspots of WTC between NEE and SWC30 at monthly or longer timescales. This study highlights the need for a multi-scale approach to understanding the temporal dynamics of NEE. Modeling efforts should take into account these multi-temporal correlations between NEE and environmental factors in order to improve model-data agreement across timescales.