Khan, MS; Baik, J; Choi, M (2020). Inter -comparison of evapotranspiration datasets over heterogeneous landscapes across Australia. ADVANCES IN SPACE RESEARCH, 66(3), 533-545.

The selection of optimal actual evapotranspiration (ET) dataset for hydro -meteorological applications requires a detailed accuracy assessment and uncertainty characterization. This evaluation study was conducted at the site scale to compare the accuracies of ET mea- surements extracted from four widely used satellite- and model -based ET datasets (MOD16, GLEAM, GLDAS, and AWRA-L). Site - scale ET values were compared with five eddy -covariance -based in -situ flux tower measurements representing different land cover types: forest, grassland, and savanna, over heterogeneous landscapes across Australia. The results show moderate to good index of agreement 0.71, >0.88, and >0.46 for forest, grassland, and savanna, respectively, producing a root mean square errors in the range of 8.16-31. 52 mm/month. AWRA-L demonstrated its best performance in forest ecosystems, GLDAS in grasslands, and GLEAM performed well in both forest and grassland biomes. Compared with the other datasets tested in this study, MOD16 performed relatively poorly across Australia. AWRA-L consistently showed the best agreement with the flux tower measurements, with a normalized standard deviation (o - n ) of -0.98, followed by GLEAM (o - n -0.92). Based on our accuracy assessment, agreement between the satellite- and model -based ET products and the flux tower measurements followed the order: AWRA-L > GLEAM > GLDAS > MOD16. The results of this study further reveal that existing ET datasets each have a particular advantages and limitations for a specific land cover types. Therefore, an integrated and merged ET product is needed to take advantage of the complementary strengths of each ET dataset. Overall, our quality evaluation of four widely used satellite- and model -based ET measurements can serve as a benchmark study of ET dynamics over dif- ferent land cover types and be used to select an appropriate dataset for improved quantification of water consumption across Australia. (c) 2020 COSPAR. Published by Elsevier Ltd. All rights reserved.