Remote sensing based estimates of reference evapotranspiration for the southwest region of Bangladesh

Abstract

Water resources, especially irrigation water in Bangladesh may have under threat due to many factors like climate change impact, upstream river flow control, and unsustainable groundwater consumption. These problems are more prominent in the Southwest (SW) part of Bangladesh, where irrigation-based agriculture dominated during the dry season. In order to assess the actual irrigation requirement for agriculture, it is a prerequisite task to understand the evapotranspiration (ET), especially reference evapotranspiration (ET0). Direct measurement of ET0 is a difficult task and the equation-based assessment requires a complex set of meteorological data. In many places in the world, the existing networks of meteorological gauge stations are insufficient to capture the spatial heterogeneity of meteorological data and the actual assessment of ET0 is quite difficult in these places. The purpose of this research is to investigate whether it is possible to attain reliable estimation of ET0 from remote sensing technology for the SW region of Bangladesh. Three air temperature based empirical ET0 equations i.e. the Hargreaves-Samani, the Thornthwaite, and the Blaney-Criddle have been chosen for this purpose. Land surface temperature (LST) data retrieved from the MODIS/Terra sensor are evaluated as an alternative input for the equations. Four hundred and sixty LST images, considering the years from 2007 to 2016, are used to extract point LST data for five selective locations to fit the equations. As the equations require local calibration before use, five meteorological stations with complete climate datasets are used to calibrate the selected equations using FAO-56 Penman-Monteith (FAO-56 PM) method as a standard. The results suggest that the LST derived from satellite images are strongly correlated (R2= 0.81) with air temperature and the ET0 derived from the satellite data using empirical equations show moderate to high correlation with FAO-56 PM method. It is found that the coefficient of determination (R2) of the empirical ET0 equations are 0.85 (high correlation), 0.81 (good correlation), and 0.79 (moderate correlation) for the Hargreaves-Samani, Thornthwaite, and Blaney-Criddle equation, respectively. The results of the statistical analyses show that the mean RMSE, MAE, and MRE (%) of the selected empirical equation are 0.53 mm/day, 0.48 mm/day, and 13.93%; 1.17 mm/day, 1.10 mm/day, and 32.25%; 1.94 mm/day, 1.92 mm/day, and 61.99% for the Hargreaves-Samani, Thornthwaite, and Blaney-Criddle equation, respectively. The mean RMSE, MAE, and MRE (%) of the selected empirical equations after calibration are 0.44 mm/day, 0.39 mm/day, and 12.40%; 0.46 mm/day, 0.40 mm/day, and 11.62%; and 0.47 mm/day, 0.42 mm/day, and 11.83% for the Hargreaves-Samani, Thornthwaite, and Blaney-Criddle equation, respectively. It is found that the Hargreaves-Samani derived ET0 from the MODIS LST data is the most appropriate for the SW region of Bangladesh. It is also found that the remote sensing technologies can be used to estimate ET0 with a good accuracy for the SW region of Bangladesh. Using the best fitted calibrated equation, spatiotemporal mapping of the ET0 has also been carried out for the SW region. The derived remote sensing based estimates of ET0 can be used for assessing irrigation demand with low cost, which will help to improve the irrigation management system of the study area.