Assessing spatiotemporal variation in actual evapotranspiration for semi-arid watersheds in northwest China: Evaluation of two complementary-based methods

SummaryWater vapor generated locally by actual evapotranspiration (AET) is important both to the recycling of water regionally and to the long term sustainability of desert–oases in the semi-arid-to-arid region of northwest (NW) China. An accurate assessment of AET is central to describing the hydrologic status of watersheds. Conventional methods of estimating AET from meteorological point data are generally not appropriate for regions with high spatial variability, particularly with respect to landcover and topography. Insufficient monitoring stations make it particularly difficult to estimate AET that is spatially representative of large areas. The objective of this study was to estimate spatially-distributed monthly AET for a complex landscape, consisting of deserts, oases, and mountains, with climate and landcover data generated primarily from remote sensing (RS) data. In this study, we used two complementary relationship (CR)-based methods to estimate monthly reference evapotranspiration (ETo) and AET over a 10-year period (2000–2009) for two large watersheds in NW China. In evaluating the performance of CR-based methods, we compared point-estimates of ETo and AET generated with the two methods (generated either by using climate-station data or by extracting point-estimates from end products produced from RS-data) against (i) climate-station-based estimates of ETo calculated with the FAO Penman–Monteith (P–M) equation and from pan-evaporation data, and (ii) geographically-corresponding point-estimates of AET extracted from the MODIS global product of AET (MOD16) recently developed by Mu et al. (2011, Remote Sensing of Environment, 115, 1781–1800). Point-extractions of AET from MOD16-products were the least representative, when compared to ETo and AET calculated with the other methods. Between CR-based methods, the Venturini et al. (2008, Remote Sensing of Environment, 112, 132–141) method provided the best comparison with ETo calculated with the P–M equation and from pan-evaporation data. Due to its independence from wind velocity, the Venturini method is rated the most suitable for regional application, especially for the complex landscapes of NW China.

► We calculate actual evapotranspiration (AET) for ungauged watersheds.
► We compare widely used complementary relationship (CR) methods of AET.
► We evaluate MODIS global AET product in a regional application for NW China.
► CR-based estimates of AET give better overall results than global MODIS products.