Abstract:

Evapotranspiration (ET) is an indispensable part of water resources consumption in arid region, and it is also one of the key indicators used for assessing water resources. As the vegetation here is extremely sparse and has relatively high spatial heterogeneity, it is a challenge to accurately estimate ET in arid region. Most of the common ET products based on fusion of multi-source remote sensing data and land surface modes showed the different ET results in arid region. The results of different ET products with eddy covariance (EC) measurements were generally compared in previous studies, but the issues of spatial scale mismatch between EC data and ET products are unresolved yet. It is difficult to obtain the true values of ET in large scale. The sources of uncertainty may be from the different model structures, calculation methods, forcing datasets, and so on. In this paper, three Cornered Hat (TCH) methods without requiring a priori knowledge of the observed ET values were used to estimate the uncertainties of ET output in Xinjiang based on the Common Land Model (CoLM), Global Land Evapotranspiration Amsterdam Model (GLEAM) and Moderate Resolution Imaging Spectroradiometer ET products (MOD16). The results showed that the spatial patterns of GLEAM and MOD16 ET products were similar, and the ET values in the mountains were significantly higher than that in the plains. The ET values of CoLM in the Altay and Tianshan mountains were significantly higher than that in the plains. In addition, the mean annual ET values in Xinjiang during the period from 2000 to 2014 differed markedly, and they were 142.36±11.11 mm·a^-1, 126.66±19.12 mm•a^-1 and 78.64±4.61 mm·a^-1 obtained by the GLEAM, CoLM and MOD16 respectively. The CoLM and MOD16 products revealed that there was no significant decrease trend in ET during the period from 2000 to 2014 (-1.03 mm•a-1 and -0.24 mm•a^-1 respectively, P> 0.05). The GLEAM product showed that there was no significant increase trend in ET (0.02 mm·a^-1, P> 0.05). The uncertainty based on the TCH method of CoLM was the highest (24.18 mm•a^-1) in Xinjiang, that based on the GLEAM was moderate (21.58 mm•a^-1), and that based on the MOD16 was the lowest (17.66 mm•a^-1). In the shrub lands, Tianshan Mountains and semi-humid region, the uncertainties of CoLM (33.47, 30.46 and 32.11 mm/a) were significantly higher than those of GLEAM (28.31, 24.73 mm•a-1 and 24.19 mm·a-1) and of MOD16 (9.50, 14.80 mm•a^-1 and 14.34 mm•a^-1). It was crucial to quantify the uncertainty of regional ET based on the CoLM, GLEAM and MOD16 ET products for accurately estimating water consumption and providing appropriate ET product for long-term water cycle change in Xinjiang. The method used in this paper might provide a reference for analyzing the uncertainty.

Key words: uncertainty, evapotranspiration, Cornered Hat method, land surface mode, remote sensing data