Evaluation of Actual Evapotranspiration Estimations of GLEAM Model and GRACE Satellite in the Gharehsu-Gorganrud basin

Introduction

Improving water efficiency in agriculture especially in the face of global warming, requires an accurate of evapotranspiration. The Gharehsu-Gorganrud Watershed with a complex topography is located in Golestan province,north of Iran. Remote sensing methods can provide acceptable estimation of ET in larfe areas with inadequate ground observations. However, these methods have lower accuracy compared to ground-based techniques and require regional validation using water balance or lysimeter approaches. Selecting suitable satellite datasets for water management planning in a specific study area is a fundamental challenge that needs validation through physical methods and ground data. Previous studies show that the GLEAM model wwhich is based on satellite data provides reliable outputs for the Karkheh basin, west of Iran and can be used as an alternative to empirical and conventional methods for estimating crop water requirements. Gharekhani et al. (2020) investigated the uncertainty of actual evapotranspiration in the Gharehsu-Gorganrud basin using two climate databases and a remote sensing-based model. The study demonstrated that the ERA-Interim, GLEAM, and ETPT-JPL databases performed well in reducing uncertainty. Another study by Hafezparast et al. (2022) utilized GRACE satellite data to monitor changes in groundwater levels in the Mianrahān aquifer, revealing critical conditions in some aquifers.Overall, the research aimed to investigate the uncertainty in actual evapotranspiration estimates using GRACE satellite data and climate databases in the Gharehsu-Gorganrud Basin.

Materials and Methods

The study area is Gharehsu-Gorganrud basin which is a sub- basin of the main Caspian Sea Basin..The meteorological data used in this study were collected from synoptic stations of Iran meteorological organization. These data included average, minimum and maximum temperature, relative humidity, precipitation, wind speed, and total sunshine hours. The satellite-derived data provides estimates of actual evapotranspiration, as key variable for this study. However, to compare these estimates with the Penman-Monteith equation (FAO 56 PM) and determine the potential evapotranspiration, we need to consider the vegetation cover factor specific to the study crop of wheat. Hence, remote sensing techniques was employed to retrieve and acquire satellite images exclusively for the wheat fields, allowing us to accurately calculate the potential evapotranspiration by multiplying the actual evapotranspiration by the corresponding vegetation cover factor.The Global Land Evaporation Amsterdam Model (GLEAM) is an algorithm that estimates various components of evaporation and transpiration using satellite observations. The model outputs include potential evaporation, root zone soil moisture, surface soil moisture, and evaporative stress.This model utilizes solar radiation and temperature data to calculate potential evapotranspiration and multiplies it by the evaporative stress to obtain actual evaporation. The data is available on a daily, monthly, and yearly basis,and the grids are divided into 0.25-degree geographical resolution.GRACE satellite data is obtained from the GRACE spacecraft,measuring changes in Earth's gravity field due to water variations. These data, along with ground-based information like precipitation and runoff, enable the calculation of actual evapotranspiration. By assuming a water balance for a specific watershed and utilizing variables such as precipitation, runoff, and ΔS from GRACE, actual evapotranspiration can be determined.

Results and Discussion

Based on the comparisons, the best performance GLEAM model was obtained in Rezvan station, with an elevation of 1447 meters, dominant agricultural land use pattern, with statistical metrics of RMSE=0.32, MAE=0.30, R2 of 0.78, and MAPE =13.67.The lowest agreement was related to the "Kalaleh" station, with an elevation of 127 meters, non-agricultural land use pattern, and statistical indices of RMSE =0.77, MAE =0.60, R2=0.49, and MAPE =18.05. Overall, the results indicate that estimating evapotranspiration using the Penman-Monteith FAO equation performs better in high-elevation areas with agricultural land use patterns, while it yields less reliable results in low-elevation areas with non-agricultural land use patterns. The study by Gharekhani et al. (2020) also showed that the GLEAM model exhibits less uncertainty at elevations between 1400 and 1800 meters above sea level and in areas with agricultural land use patterns. For more precise explanations, further examination of the environment and comparison with field data is required. Based on the conducted comparisons, the best GRACE performance is associated with the "Rezvan" station, which has a drier climate compared to other stations, and statistical indices of 0.41 RMSE,0.38 MAE, 0.66 R2, and 17.46 MAPE. The worst performance is related to the "Kalaleh" station, with an elevation of 127 meters, non-agricultural land use pattern, and statistical indices of 0.91 RMSE, 0.77 MAE,0.45 R2, and 23.17 MAPE.



Conclusions

The use of satellite imagery can provide broader insights into various topics. In this study, the estimation of actual evapotranspiration was conducted using GRACE satellite data and the GLEAM model in the Gharehsu-Gorganrud region of Golestan province. The FAO Penman-Monteith equation was employed for evapotranspiration calculation. The results indicated that the best estimations belongs to Rezvan station, while the worste case performance was observed in Kalaleh station in estimating evapotranspiration based on the FAO Penman-Monteith equation measure and GLEAM data.More precise informatin on land cover maps in the region for ET estimation using vegetation cover dependent coefficents is necessary.