Spatial-scale effect on the SEBAL model for evapotranspiration estimation using remote sensing data
Highlights
► Spatial effect of the size of both AOI and satellite pixel on SEBAL was examined. ► Analytical equations were proposed to quantify the spatial effects on SEBAL. ► Effective variables were introduced to bridge the high- and low-scale difference. ► Data used covered a wide range of vegetation cover and soil moisture condition.
Introduction
Land surface evapotranspiration (ET) consumes 60% of the annual land precipitation in the global hydrological cycle (Oki and Kanae, 2006). Quantifying the distribution of the land surface ET at different temporal and spatial scales is critical to meteorology, hydrology, global climate change, and almost all other water-related scientific fields. There are more than 140 sites on five continents in the Fluxnet network that currently measure the exchanges of energy, water, and carbon dioxide between the land surface and the atmosphere (Baldocchi et al., 2001). However, these sparsely distributed tower sites are insufficient for local, regional, and global ET studies due to the heterogeneity of the earth's surface. Remote sensing provides a potential method to estimate ET at various scales.
Over the past two decades, various remote sensing-based ET models, roughly classified as the end-member-based single-source models and the soil and vegetation energy/temperature separation two-source models, have been developed. These models include the Surface Energy Balance Algorithm for Land (Bastiaanssen, 1995, Bastiaanssen et al., 1998), surface temperature versus vegetation index triangle/trapezoid space (Price, 1990, Moran et al., 1994, Jiang and Islam, 1999, Tang et al., 2010), Simplified Surface Energy Balance Index (Roerink et al., 2000), Surface Energy Balance System (Su, 2002), Mapping Evapotranspiration with Internalized Calibration (Allen et al., 2007), the two-source N95 model (Norman et al., 1995), and the Atmosphere-Land Exchange Inverse model (Anderson et al., 1997). Li et al. (2009), among others, provides an overview of the limitations, advantages, and uncertainties of these commonly used ET models. The Surface Energy Balance Algorithm for Land (SEBAL), proposed by Bastiaanssen (1995) and Bastiaanssen et al. (1998), assumes a linear relationship between near-surface air temperature difference and surface temperature. It has been used to estimate the surface sensible heat flux and latent heat flux at different spatial and temporal resolutions in more than 30 countries (Bastiaanssen et al., 2005, Timmermans et al., 2007, Singh et al., 2008, Teixeira et al., 2009). Typical accuracies of the estimated ET in the SEBAL model are 85%, 95%, and 96% at daily, seasonal, and annual scales, respectively. The SEBAL model has several advantages over other approaches in estimating surface-energy fluxes from remote sensing data, including: (1) it uses minimal ground-based data, (2) the near-surface air temperature is not mandatory, unlike in many other bulk transfer models, and (3) a self-calibration process is automated in each region of interest through the identification of dry and wet pixels and the determination of the near-surface air temperature difference. The most current version of the SEBAL model (SEBAL2008) has incorporated three major refinements: (1) the effect of the land-surface slope and aspect in mountainous areas, (2) a correction of the advection effect, and (3) improved estimates of surface albedo and soil heat flux (Bastiaanssen et al., 2010).
A basic premise associated with the SEBAL model, in general, is that under the given atmospheric conditions, the area of interest exhibits an extreme contrast in surface hydrological condition so that the dry and wet pixels where sensible heat flux (Hdry and Hwet) and latent heat flux (LEdry and LEwet) are known a priori could coexist. Bastiaanssen et al. (2005) have suggested that the wet pixel can be identified in open water bodies or in well-irrigated agricultural fields whereas the dry bare soil surface may be a good candidate for the dry pixel when no other knowledge can be used for calibration. After determining the locations of the dry and wet pixels, a linear relationship between the near-surface air temperature difference (dTs) and surface radiometric temperature (Ts) constructed by fitting data to the dry and wet pixels is further applied to estimate the pixel-by-pixel dTs under intermediate hydrologic/surface temperature conditions (Fig. 1).
When SEBAL and other end-member-based models are used to estimate surface energy components, a concern is how large the area of interest (AOI) should be. A lack of general agreement on this issue brings about some uncertainties when the size of the AOI is varied, as the slope and intercept of the linear line between dTs and Ts may have changed. To the best of our knowledge, there have been few studies dedicated to the effect of the size of the AOI on the sensible heat flux and latent heat flux derived using the SEBAL. Long et al. (2011) noted this heat flux difference by comparing the pixel-by-pixel sensible heat fluxes calculated for three different-sized sub-watersheds within the Baiyangdian watersheds in North China, and they concluded that the discrepancies in the sensible heat fluxes using the SEBAL are systematic for given atmospheric conditions when the AOI is expanded. However, it is unclear under which circumstances and how the sensible heat flux will change. Determining the mechanism of this change in the SEBAL-derived surface energy fluxes is important for understanding the uncertainties that result from the selection of dry and wet pixels for given atmospheric conditions within a given AOI.
For a given AOI, the SEBAL model is also affected by the pixel size of the satellite data. In most cases, extremely dry or wet pixels may not exist in the low-resolution satellite view of the area of interest due to areas that contain both soil and vegetation. Compared with the low-resolution data, high-resolution satellite data are better able to capture the spatial variability of the surface, and therefore, the extremely dry and wet pixels are more readily identified. The SEBAL-derived surface energy components (surface net radiation, soil heat flux, sensible heat flux, and latent heat flux) may differ over an area when the satellite surface temperature at one scale is changed to a different scale. In previous studies (Long et al., 2011, Gebremichael et al., 2010, Hong et al., 2009, Compaoré et al., 2008), the uncertainty that results from the variation of pixel size in the SEBAL has generally been evaluated by comparing the ET estimated at a targeted lower resolution (e.g., MODIS) and the spatially averaged value derived by aggregating either output (output up-scaling) or input (input up-scaling) of the SEBAL with high-resolution data (e.g., TM/ETM+ or ASTER). Long et al. (2011), among others, have demonstrated the significant differences of the SEBAL-estimated sensible heat flux and latent heat flux using satellite data with different pixel sizes. They attributed the heat flux discrepancy to the difference in the range and magnitude of surface temperatures acquired at the different spatial resolutions. When the satellite pixel size is changed, estimates of the sensible heat flux and latent heat flux are influenced by a number of other factors, including variations in surface albedo, emissivity, and aerodynamic resistance (surface roughness length), in addition to the variation in surface temperature. Those inconclusive evaluations are far from an all-around apprehension of the mechanism implicit in the scaling transfer of the sensible and latent heat fluxes when the satellite pixel is coarsened.
The objective of this paper is to comprehensively quantify the spatial-scale effect of the SEBAL model caused by variation in the sizes of the AOI and satellite pixel on estimating the regional sensible heat flux and latent heat flux by proposing generalized analytical equations and through model application. We examine under which circumstances the sensible heat flux is overestimated/underestimated/unaltered in the SEBAL model when the size of the AOI is increased. We also investigate how the sensible and latent heat fluxes vary when high-spatial-resolution satellite data are aggregated with or replaced by low-resolution data. This study differs from previous studies primarily in two aspects: (1) the data cover a broad range of vegetation growth stages under a full range of vegetation cover and soil wetness conditions from January 2010 to late October 2011, allowing the inclusion of a variety of meteorological, surface vegetation, and soil moisture conditions; and (2) analytical equations are proposed to reveal the mechanism of the scaling transfer of the SEBAL model, which is different from previous studies that evaluate the spatial-scale effect of the SEBAL model by simple comparison and make inconsistent conclusions that depend on the meteorological, soil, and vegetation conditions. Moreover, these equations can help understand how uncertainties in the intermediate variables in the SEBAL model propagate to the resulting surface energy components due to the alteration of dry and wet pixels when the size of the AOI is changed or when using satellite data with different spatial resolutions. Section 2 presents the study area used as the AOI, the MODIS dataset, the SEBAL model theory, the spatial scaling theory, and how the simulations are carried out. In Section 3, the spatial-scale effect of the AOI size and satellite pixel size on the SEBAL model are presented and the proposed analytical equations are verified by comparing the pixel-by-pixel sensible (latent) heat flux with that estimated by directly implementing the SEBAL model. Section 4 summarizes how the spatial-scale effect affects the SEBAL model.
Section snippets
Study area
The study area (Fig. 1, hereafter referred to the large AOI) is part of the North China Plain. It is located from 34.5°N to 38.5°N and 114.8°E to 118.8°E and covers approximately 155,650 km2 (349 samples × 446 lines at 1 km spatial resolution). The climate is humid and sub-humid in the warm temperate zone, and the soil is brunisolic. The surface elevation (Fig. 1a) varies primarily from 0 to 400 m above sea level. The land cover types are mainly arable (wheat, corn, cotton, sorghum, peanut), forest,
Effect of the size of the AOI
When the size of the AOI is increased, as mentioned in Section 2.4, the surface temperatures at the wet and dry pixels are lowered and raised, respectively. For example, the surface temperatures of the dry pixels in the small AOI are less than or equal to those in the large AOI. Table 2 shows the variations in the locations (also as indicated in Fig. 1b) and temperatures of the dry and wet pixels for the 23 clear-sky MODIS/Terra and MODIS/Aqua overpass times. Of the 23 clear-sky MODIS
Summary and conclusions
The spatial effect of the sizes of the area of interest (AOI) and satellite pixels was comprehensively evaluated by applying the SEBAL to MODIS/Terra and MODIS/Aqua data at 23 clear-sky overpasses covering different soil water contents and crop growth stages from January 2010 to late October 2011. Considering the uncertainty in the estimated surface net radiation and the limited validation of the surface energy components (due to a lack of ground-based measurements of net radiation, soil heat
Acknowledgments
We thank the hard-working staff at the Yucheng Comprehensive Experimental Station who cooperatively provided the ground- and near-surface measurements used in this study. This work was partly supported by the National Natural Science Foundation of China under Grant 41201366, 41101332 and by the China Postdoctoral Science Foundation funded project under Grant 07Z7602MZ1.
References (31)
- et al.
A two-source time-integrated model for estimating surface fluxes using thermal infrared remote sensing
Remote Sens. Environ.
(1997) - et al.
A remote sensing surface energy balance algorithm for land (SEBAL): 1. Formulation
J. Hydrol.
(1998) - et al.
Evaporation mapping at two scales using optical imagery in the White Volta Basin, Upper East Ghana
Phys. Chem. Earth
(2008) - et al.
Dependence of remote sensing evapotranspiration algorithm on spatial resolution
Atmos. Res.
(2010) - et al.
Up-scaling of SEBAL derived evapotranspiration maps from Landsat (30 m) to MODIS (250 m) scale
J. Hydrol.
(2009) - et al.
Satellite-derived land surface temperature: current status and perspectives
Remote Sens. Environ.
(2013) - et al.
A two-source approach for estimating soil and vegetation energy fluxes from observations of directional radiometric surface temperature
Agric. For. Meteorol.
(1995) - et al.
S-SEBI: a simple remote sensing algorithm to estimate the surface energy balance
Phys. Chem. Earth B
(2000) - et al.
An application of the Ts–VI triangle method with enhanced edges determination for evapotranspiration estimation from MODIS data in arid and semi-arid regions: Implementation and validation
Remote Sens. Environ.
(2010) - et al.
An intercomparison of three remote sensing-based energy balance models using large aperture scintillometer measurements over a wheat–corn production region
Remote Sens. Environ.
(2011)
Reviewing SEBAL input parameters for assessing evapotranspiration and water productivity for the low-middle São Francisco River basin, Brazil. Part A: calibration and validation
Agric. For. Meteorol.
An intercomparison of the Surface Energy Balance Algorithm for Land (SEBAL) and the Two-Source Energy Balance (TSEB) modeling schemes
Remote Sens. Environ.
Satellite-based energy balance for mapping evapotranspiration with internalized calibration (METRIC)-model
J. Irrig. Drain. Eng.
FLUXNET: a new tool to study the temporal and spatial variability of ecosystem-scale carbon dioxide, water vapor, and energy flux densities
Bull. Am. Meteorol. Soc.
Cited by (88)
geeSEBAL-MODIS: Continental-scale evapotranspiration based on the surface energy balance for South America
2024, ISPRS Journal of Photogrammetry and Remote SensingEstimating evapotranspiration and yield of wheat and maize croplands through a remote sensing-based model
2023, Agricultural Water ManagementA physical method for downscaling land surface temperatures using surface energy balance theory
2023, Remote Sensing of EnvironmentRemote sensing-based evapotranspiration modeling using geeSEBAL for sugarcane irrigation management in Brazil
2022, Agricultural Water ManagementCitation Excerpt :Yet, these models use orbital imagery and ground-based meteorological data, and depending on the model, can also use field data to estimate the components of the energy balance (latent and sensible heat, and soil heat flux) and ETa. Among several surface temperature (Ts)-based models, SEBAL computes ETa from satellite images and weather data using the surface energy balance and has been successfully applied worldwide across different climates and land cover conditions for water resources management presenting accurate results (Bastiaanssen et al., 2005; Bhattarai et al., 2012; Ruhoff et al., 2012; Tang et al., 2013; Wagle et al., 2017; Yang et al., 2012). The main component of the algorithm is the internal process to estimate the near-surface temperature gradient (dT), which is based on the selection of endmembers that represent the extremes of the wet (cold) and dry (hot) ETa spectrum (Allen et al., 2007; Bastiaanssen et al., 1998a).