Quantum-driven satellite imagery for enhancing evapotranspiration estimation in coastal areas

Accurate assessment of water requirements for crops and large-scale, real-time water usage measurement is essential in water and soil management. Utilizing remote sensing data, which provides extensive spatial and temporal coverage, has emerged as a viable approach for evaluating evapotranspiration. The FAO-Penman-Monteith method is widely regarded as the most precise technique for computing reference evapotranspiration (ET۰). Extreme learning machines (ELM) models were employed alongside quantum computing optimization strategies (QIS-ELM), Bayesian optimization (BO-ELM), and particle swarm optimization (PSO-ELM). These models excel in processing complex datasets and recognizing patterns, thereby enhancing estimation accuracy. Daily meteorological data comprising temperature, relative humidity, wind speed, and sunlight hours, along with ET satellite imagery data from MODIS, covering the Ramsar and Babolsar stations from ۲۰۰۱ to ۲۰۲۳, were used as inputs for the models. To transform this data into daily figures, the Kalman filter and cubic spline interpolation techniques were applied. The performance evaluation of the models at both stations revealed that the PSO-ELM-۸ model for Ramsar and the QIS-ELM-۸ model for Babolsar achieved the highest accuracy, with error values of RMSE ۰.۱۹ and ۰.۲۸ mm/day, respectively, using satellite image data. Thus, the QIS-ELM and PSO-ELM models improve ET₀ estimation for coastal water management.

Keywords:

evapotranspiration , Extreme learning machine , Hybrid Model , satellite image data