Evaluation of the mechanisms and performances of major satellite-based evapotranspiration models in Northwest China

Highlights

• SWH, SEBAL and MOD16 models are used to estimate ET in Northwest China.

• SWH model provides superior applicability in Northwest China.

• Model simulation effects are discussed with respect to the simulation mechanism.

• SWH model has higher sensitivity to driving data.

• Correlations between the models are high during the vegetation growth period.

Abstract

Evapotranspiration (ET) is an important component of the water cycle and surface energy balance system. Accurate measurements and estimations of ET can be used to manage and allocate regional water resources and in agricultural water management under climate change. Satellite-based ET models extrapolate site ET values to regional scales, but there are uncertainties in their simulations. In this study, the use of satellite-based ET models including the Shuttleworth–Wallace–Hu (SWH), Surface Energy Balance Algorithm for Land (SEBAL), and the moderate-resolution (MOD) imaging spectroradiometer ET product (MOD16) was validated, and their applicability in Northwest China were compared. In addition, the sensitivity of these models to driving data and the correlation between the models were analyzed. Results showed that the SWH model provided superior applicability in Northwest China, followed by SEBAL and MOD16. The correlation coefficient, mean relative error, root mean square error, and index of agreement of the ET simulated by SWH compared with the eddy flux of the Haibei station were 0.90, 0.39 mm•d⁻¹, 0.55 mm•d⁻¹, and 0.87, respectively. Furthermore, the sensitivity to driving data in SWH was the highest among all models. The correlation between the models was high during the vegetation growth period but low in winter. The model simulation effects were also compared with respect to their mechanisms, and it was evident that further improvements and validation could be achieved by investigating the model structures and examining the ET estimations and critical parameters of each model.

Introduction

Evapotranspiration (ET) is an important component of the water cycle and surface energy balance system. Land surface ET can affect precipitation, and the accompanying latent heat has a cooling effect. ET is thus regarded as a core climatological process and a crucial linking component within the hydrothermal system (Jung et al., 2010). In recent years, global warming has become increasingly significant and has caused changes in ecological water demands and irrigation requirements (He and Shao, 2014). Therefore, conducting accurate measurements and estimations of ET is of great importance when investigating environmental issues and managing agricultural water allocation.

The application of ET methods on a large scale is difficult owing to spatial variations in the geometrical and physical properties of the underlying surface. For a long time, land surface ET has been the part of the hydrological cycle that was most difficult to measure directly. Although many ET methods continue to be developed, currently they are only applicable for use in certain areas with small scale. An increasing number of studies have recently focused on ET because of the intensification of global problems, such as global warming and water shortage. Remote sensing technology provides temporally and spatially continuous information over vegetated surfaces and is useful for regional measurements and monitoring of the surface biophysical variables that affect ET (Los et al., 2000). Recent developments of remote sensing have enabled ET to be estimated over large-scale. An increasing number of studies are now being conducted to estimate the actual ET (ET_a) from the land surface using remote sensing inversion (Feng and Wang, 2012). Three types of methods have been developed to estimate ET from remote sensing data (Mu et al., 2011): (1) empirical/statistical methods; (2) physical models that calculate ET as the residual of surface energy balance (SEBPM); (3) and other physical models (OPM).

Based on the energy balance principle, Bastiaanssen et al. (1998a) proposed the widely used Surface Energy Balance Algorithm for Land (SEBAL) model, which eliminated the errors caused by the spatial interpolation of meteorological data and performed surface temperature corrections; however, issues with spatial ambiguity were evident. The Penman–Monteith (PM) model is the standard method for calculating reference crop ET, as recommended by the Food and Agriculture Organization of the United Nations (FAO) in 1998. However, the FAO–PM model is a single-source large-leaf model which considers the underlying surface as an entirety (Hu, 2009), and when the vegetation coverage is low, the FAO–PM model error is large. Based on the FAO–PM model, Shuttleworth and Wallace (1985) proposed a dual-source coupled model (SW) that divides ET into soil evaporation and vegetation transpiration which is suitable for considering the underlying surface of sparse vegetation (Ji et al., 2004). Hu et al. (2009, 2013) improved the SW model and proposed the Shuttleworth–Wallace–Hu (SWH) model, which introduced the Ball–Berry stomatal conductance model to estimate canopy stomatal conductance, replaced photosynthetic rate with gross primary productivity (GPP) and addressed key issues relating to regional-scale application (Zhang et al., 2015). Mu et al. (2011) further improved the ET algorithm (MOD16 model) presented in Mu et al. (2007), by separating the dry canopy surface from the wet and dividing the soil surface into a saturated wet surface and a moist surface. The MOD16 model performs well on the global scale but poorly in certain regions. Satellite-based ET models have different ranges of application due to different assumptions and mechanisms, which need to be verified and corrected in specific regions; therefore, there is currently no satellite-based ET model that can adapt to all environments.

This study was conducted to evaluate the performances of major satellite-based ET models in Northwest China and compare their mechanisms. The following models were selected to simulate ET: the SEBAL model (a SEBPM), the SWH model (a simple OPM) and the MOD16 model (a complex OPM). The ET simulated by the FAO–PM model was used to represent the maximum possible ET. The sensitivities and correlations of these satellite-based ET models were analyzed, and comparisons were made between their simulation performances, distributions, and mechanisms. This paper provided a methodological basis for further studies on the temporal and spatial evolution of regional ET, which would enable more accurate remote sensing inversions on regional scales. Finally, this paper provided theoretical support to policymakers as an important reference for water resources management, and as a methodological basis for research on climate change.