Publications

Liu, M; Tang, RL; Li, ZL; Yan, GJ (2019). Integration of two semi-physical models of terrestrial evapotranspiration using the China Meteorological Forcing Dataset. INTERNATIONAL JOURNAL OF REMOTE SENSING, 40(6-May), 1966-1980.

Abstract
Combining surface evaporation and plant transpiration, evapotranspiration (ET) is critical to surface water and heat balances as it links water, carbon cycles and energy exchanges. Many models have been developed and are presently used to estimate terrestrial ET. However, there are large model uncertainties among the different models, which present a problem. By combining meteorological reanalysis data from the China Meteorological Forcing Dataset (CMFD) with remote sensing data and observational data during 2002-2009, two semi-physical models, the modified satellite-based Priestley-Taylor (MS-PT) model and a semi-empirical Penman equation-based (SE-PM) model, are used to estimate ET and are validated using in situ measurements collected at 22 flux tower sites in China. Then support vector machine (SVM) method is used to integrate these two semi-physical models to improve the accuracy of ET estimates for eight different vegetation types separately, as well as all of these types together. The integrated model likely explains 56-94% of the land surface ET changes indicated by the observations collected at the flux tower sites. The ET predictions obtained by driving the models with the reanalysis data (for which the relative bias and the relative root mean square error (RMSE) for all types of the SVM were 6 and 51W m(-2), respectively, and the maximum decrease of the relative RMSE for different types is nearly 20W m(-2)) are less accurate than those obtained by driving the models with the observational data (for which the relative bias and the relative RMSE for all types of the SVM are 4 and 43W m(-2), respectively). Compared to the individual semi-physical models, the results produced by the integrated model display significantly decreased bias (less than 5W m(-2) for all types) and RMSE (for which the maximum decrease is nearly 71W m(-2)) in the validation.

DOI:
10.1080/01431161.2018.1482026

ISSN:
0143-1161