Multiscale spectral analysis of temporal variability in evapotranspiration over irrigated cropland in an arid region

The temporal patterns of evapotranspiration (ET) and its biophysical and physiological controls (e.g., soil water content, solar radiation, and canopy conductance) occur over a wide range of time-scales ranging from seconds to decades. Thus, there is increasing interest in understanding at which scales the main temporal correlations between ET and its controlling factors occur across different ecosystems. For this study, we used eddy covariance measurements of ET over 2 years at an irrigated maize field in an arid inland region of northwest China. We applied the wavelet transform as a novel technique to examine spectral characteristics of ET and its controlling factors. The ET power spectra displayed a −1 power law in turbulent inertial subrange at <1-h time-scale, and showed substantial power at daily, seasonal and annual time-scales. The cospectra of ET and soil water content (SWC) showed significant temporal correlation at 5-days, which has implications for calculation of ET using the soil water balance method in this region. We found that ET synchronized with the change of net radiation, and led vapor pressure deficit and air temperature for ∼2h at the 1-day time-scale (i.e., positive lags), but the phase relationship between ET and SWC was influenced by irrigation patterns. Canopy conductance influenced ET variability at the 1-day time-scale, but the effect was not consistent across the growing seasons. Our results show the importance of irrigation practices and its influence on the multi-temporal correlations of ET and its controlling factors. Irrigation can sharply change the phase angle relationship of ET and SWC from −10h to 10h. These results are important for understanding water cycle processes, improving water management, and address food security issues across irrigated croplands in arid regions.