Evaluation of SEBS, SEBAL, and METRIC models in estimation of the evaporation from the freshwater lakes (Case study: Amirkabir dam, Iran)

Highlights

• Implementing SEBAL, METRIC, and SEBS evapotranspiration estimation methods.

• Evaluating selected methods in a freshwater lake using evaporation pan measurements.

• Estimating daily evaporation from Amirkabir dam reservoir by remote sensing methods.

Abstract

Evapotranspiration (ET) estimation is of great importance due to its key role in water resource management. Surface energy modeling tools such as Surface Energy Balance Algorithm for Land (SEBAL), Mapping Evapotranspiration with Internalized Calibration (METRIC), and the Surface Energy Balance System (SEBS) can estimate the amount of evapotranspiration for every pixel of the satellite images. The main objective of this research is evaporation investigation from the freshwater bodies using SEBAL, METRIC, and SEBS. For this purpose, the Amirkabir dam reservoir and its nearby agricultural lands in a semi-arid climate were selected and studied from 2011 to 2017 as the study area. The implementations of this study were accomplished on 16 satellite images of Landsat TM5 and OLI. Then, SEBAL, METRIC, and SEBS were implemented on the selected images. Moreover, the corresponding pan evaporate measurements on the reservoir bank were considered as the ground truth data. Regarding to the results, SEBAL is not a reliable method to evaluate freshwater evaporation with the coefficient of determination (R²) of 0.36 and the Root Mean Square Error (RMSE) of 5.1 mm. On the other hand, METRIC with RMSE and R² of 0.57 and 2.02 mm and SEBS with RMSE and R² of 0.93 and 0.62 demonstrated a relatively good performance.

Introduction

Various methods have been developed to estimate evapotranspiration by meteorological data in different environmental conditions and land covers. These methods mostly use local stationary data to evaluate evapotranspiration, which makes them suitable only at local scale (Bastiaanssen et al., 1998a). Furthermore, regarding to the dynamic nature and regional changes in evapotranspiration, these methods cannot be generalized to the larger areas (Seemann et al., 2003).

In recent decades, with developments in satellite sensors and remote sensing methods, using their images attracts a great attention for estimation of evapotranspiration (Bastiaanssen et al., 1998b, Jana et al., 2016). One can estimate evapotranspiration using satellite images, with no regards to soil condition, crop, and farm management (Bastiaanssen et al., 2005). Using these images, the spatial distribution of required factors and their changes are provided in two or more sequential images (Allen et al., 2002). Among the remote sensing algorithms of evapotranspiration estimation, Surface Energy Balance Algorithm for Land (SEBAL) (Bastiaanssen et al., 1998a), Mapping Evapotranspiration at high Resolution and with Internalized Calibration (METRIC) (Allen et al., 2005), and Surface Energy Balance System (SEBS) (Su, 2002) are the most frequently used methods.

SEBAL was initially introduced by Bastiaanssen et al. (1998a) to estimate evapotranspiration from the vegetated surfaces. Then, it was modified by Allen et al. (2002) to make it usable for other land conditions. SEBAL was implemented in various environmental regions using different satellite images with acceptable performance.

National Water Commission of Australia represented the annual evaporation of Wetherell and Pamamaroo Lakes in their annual report using SEBAL, pan evaporate, and water balance. Results showed that remote sensing methods are reliable especially in remote places where terrestrial data is not available (NationalWaterCommission, 2009). Sima et al. (2013) developed a distributed model to estimate the evaporation of Urmia saline lake using SEBAL method and remote sensing data. It was accomplished by considering salinity effects and spatial distribution of variables. Yang et al. (2015) estimated actual evapotranspiration of corn farms (summer) and wheat farms (winter) in Hwang Hwai Hai Plain in China using SEBAL. Local analysis showed a linear relationship between actual evaporation, Normalized Difference Vegetation Index (NDVI), and Land Surface Temperature (LST). Jana et al. (2016) used Landsat 5 images to calculate evapotranspiration of Doon Valley in India using SEBAL. The results showed the variability of ET in at the regional scale and unreliability of point in situ measurements of ET in the relatively vast areas. Genanu et al. (2017) calculated actual evapotranspiration for sugarcane farms in Wonji using SEBAL, the Simplified Surface Energy Balance (SSEB), and the operational Simplified Surface Energy Balance (SSEBop). Regarding this research, estimated evapotranspiration of well-watered sugarcane farms in plant growth season, was higher than terrestrial measurements.

METRIC algorithm is an image processing tool to calculate evapotranspiration based on the energy budget balance on the land surface. This algorithm is an important derivative from SEBAL. METRIC was initially developed by Allen et al. (2005), and extended by Allen et al. (2007).

Allen et al. (2005) allpied METRIC algorithm in Idaho and compared it with Lysimeter measurements. They obtained 4% and 1% error in evapotranspiration for lawn and sugar beet in plant growth season, respectively. Trezza (2006) investigated and compared SEBAL and METRIC at a field in Venezuela. He found out that both models were able to estimate evapotranspiration with an acceptable precision. However, METRIC overestimated SEBAL as 7% in this study, and SEBAL had better results. Folhes et al. (2009) calculated actual evapotranspiration of farm lands at the north-east of Brazil using METRIC and Landsat images. They compared the acquired ET from METRIC with observations from Eddy Correlation (EC) method. This study demonstrated that METRIC is a good tool for water resource management in semi-arid regions of north-east of Brazil. Trezza et al. (2013) implemented METRIC in New Mexico using images from Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat. The main challenge for this study was the calibration of METRIC to utilize MODIS images. Well-calibrated METRIC model with MODIS as well as Landsat images showed reliable monthly and annually ET estimation. Morton et al. (2013) used METRIC in the water budget estimation of Nevada. According to the results, uncertainty for regions with less evapotranspiration was higher and vice versa. French et al. (2015) implemented Two-Source Energy Balance (T-SEB) and METRIC in Arizona, using Landsat to estimate ET. Both models showed similar results with a daily difference of 1.9 mm for plant growth season. Estimated evapotranspiration amounts for two models were between 5.77 mm and 10.75 mm per day.

SEBS was introduced by Su (2002) to estimate daily, monthly, and annual evapotranspiration in a semi-arid environment, and for drought monitoring.

Chinyepe (2010) modified SEBS for evaporation of Mutirikwi Lake in Zimbabwe. He used MODIS/Terra images, and compared the results with pan evaporate measurements. He concluded that SEBS results were slightly higher than pan evaporate measurements, but it was negligible. Moreover, SEBS was proved as a reliable method to estimate evaporation in large scales. Abdelrady (2013) modified SEBS to estimate evaporation from fresh and saline waters using AATSR/Envisat images. The coefficient of determination (R²) for freshwater and saline water evaporation was acquired as 0.98 and 0.88, respectively (Abdelrady, 2013). Zhuo et al. (2014) calculated distribution of actual and potential evapotranspiration on the land surface in Tibet using SEBS and Penman-Monteith equation. Regarding to this investigation, both actual and potential evaporations were changed with the same distribution from the south-east to the north-west of the study area. Comparing the results of SEBS with Penman-Monteith equation confirmed the efficiency of SEBS. However, there were variations in regions with highly-dense vegetation such as forests. Elhag (2016) employed SEBS to estimate evapotranspiration for Nile Delta using weather stations data and satellite remote sensing data. Analysis of the result along with evapotranspiration maps confirmed the efficiency of SEBS for farm lands. On the other hand, some short-fallings were reported for SEBS such as using fractional vegetation cover to estimate the ground heat flux and saturating this parameter rapidly with increasing LAI and consequently underestimating soil heat flux (Timmermans et al., 2013). Moreover, poor performance of the SEBS for complex surfaces is another disadvantage (Ma et al., 2015).

Regarding the previous studies, different algorithms of evapotranspiration estimation using satellite images have different performance in various environmental condition and land covers. Moreover, enough studies have not been accomplished about the efficiency of remote sensing based evapotranspiration algorithms in freshwater lakes. Then, the main aim of this research is investigating the efficiency of SEBAL, METRIC, and SEBS to estimate evaporation from freshwater lakes. In this regard, Amirkabir (Karaj) dam reservoir in Iran was chosen as the study area, and the results of these three algorithms were compared with Pan evaporate measurements.