Diagnosing environmental controls on actual evapotranspiration and evaporative fraction in a water-limited region from northwest China

•A boosted regression tree method is used to quantify controls on ETa and EF.•Rn plays an important role in affecting ETa in the study region with arid climates.•The impact of RH on ETa is also substantial in the study region.•The impact of LAI on ETa and EF varies with soil water conditions.•RH plays a pivotal role in controlling EF at rain-fed sites. The knowledge of the impacts of environmental factors on actual evapotranspiration (ETa) and evaporative fraction (EF) is important for understanding a range of land surface processes and for estimating ETa. However, rare attempts have been made to quantitatively assess various controls on ETa and EF. To this end, a comprehensive dataset was obtained from four study sites with different land cover conditions within an arid region from northwest China. The boosted regression tree (BRT) method was applied to assess the relative controls of different factors (e.g., net radiation-Rn, soil water content-SWC, leaf area index-LAI, relative humidity-RH, air temperature-Ta, and wind speed-WS) on ETa and EF. In general, the importance of SWC in controlling ETa increased with decreasing SWC across the sites; however, the BRT analysis revealed that although an arid climate was prevalent in the study region, Rn might outweigh SWC in affecting ETa at both irrigated and rain-fed sites. In addition, RH also played a nonnegligible role in affecting ETa at rain-fed sites, while the relative control of LAI varied with SWC. Moreover, EF at the irrigated site was mainly controlled by LAI and SWC when SWC was low, while by Ta when SWC was higher; RH was the most important in explaining the variability in EF at the rain-fed sites, which was followed by LAI and SWC. This study underscores the importance of available energy and other environmental factors (e.g., RH and LAI) in modulating ETa and EF in water-limited regions, which has important implications for parameterizing hydrological and land surface models for simulating ETa and for identifying EF regimes more accurately.