Daily scale evapotranspiration prediction over the coastal region of southwest Bangladesh: new development of artificial intelligence model

Among several complex hydrological process elements, Evapotranspiration (ET) is the most complex one. Estimation of ET is very challenging compared to other hydrological variables as it depends on complex interactions of several hydrometeorological variables. In the current research, the estimation of daily ET from maximum and minimum temperature was established. For this purpose, Dynamic Evolving Neural-Fuzzy Inference System (DENFIS) and Multivariate Adaptive Regression Spline (MARS) were hybridized with two advanced metaheuristic optimization algorithms [i.e., Whale Optimization Algorithm (WOA) and Bat Algorithm (BA)]. Daily ET and temperature data estimated at 3 locations in the coastal region of southwest Bangladesh for the period 2005–2016 were used to develop and validate the models. The results showed a good performance of DENFIS-WOA model with minimum values of normalized root mean square error (NRMSE = 0.35–0.54) in estimating ET using only temperature in the complex climatic setup of southwest Bangladesh. DENFIS-BA also showed reasonable performance (NRMSE = 0.43–0.62), while the performance of MARS–WOA (NRMSE = 0.54–0.97) and MARS-BA (0.60–1.13) was found satisfactory in terms of most of the statistical indices. Obtained results were also evaluated using innovative visual presentations of model outputs, which revealed the better capability of only DENFIS-WOA in estimating mean, variability and distribution of ET for all the months and locations. The results indicate the potential of DENFIS-WOA to be used for reliable estimation of daily ET from the temperature in a tropical humid coastal region.