Time-frequency insights: Uncovering the drivers of reference evapotranspiration across China

Abstract

Reference evapotranspiration (ET_o) is an important variable required in many disciplines and is influenced by many factors. However, the bivariate and multivariate relationships between ET_o and affecting factors across multiple time-frequency domains remain unknown. Here, we identified the primary factors affecting ET_o across time-frequency domain in 653 meteorological stations of mainland China based on the combination of wavelet transform coherence (WTC) and multiwavelet coherence (MWC) methods. The results indicated that ET_o and all affecting factors (solar radiation, R_s; vapor pressure deficit, VPD; air temperature, T_a; wind speed, u₂) during 1967–2016 exhibited a frequency ranging from 2 days to 211 months, and had a continuous annual (374 d) periodicity (except u₂) for almost all sites. Results of percentage area of significant coherence (PASC) of WTC indicated that VPD or R_s is the dominant single factor driving variations of ET_o across time-frequency space in majority sites (66.3 % and 32.0 %, respectively), while u₂ is only dominant in limited (11) sites. This quite differs from the daily scale, where daily ET_o was primarily influenced by daily R_s at 361 sites, daily VPD at 286 sites, and daily T_a at 6 sites. Results of MWC showed that the explanation for the time-frequency variations of ET_o can be further improved using two-factors in 40.7 % of all sites as indicated by absolute increased PASC of MWC by 5 %. Overall, we found that the variation of ET_o across time-frequency domain can be well explained by using only one variable (VPD or R_s) in 59.3 % of all sites, while by combinations of VPD-R_s and VPD-u₂ in remaining sites. This study provides novel insights into understanding the variations of ET_o across multiple time-frequency spaces.