Modelling soil moisture and daily actual evapotranspiration: Integrating remote sensing surface energy balance and 1D Richards equation

Evapotranspiration (ET) is a crucial component of the soil - plant-atmosphere system. In semi -arid Mediterranean regions, most land water loss occurs through ET, encompassing both evaporation from the earth ' s surface and plant transpiration. A comprehensive understanding of the actual ET spatiotemporal dynamics is critically important for hydrological modelling and effective water resource management. This significance is further pronounced considering the growing stress on water resources and the potential influence of climate change on water fluxes. Remote sensing (RS) provides long-term, high-resolution data that can contribute to the monitoring and management of natural ecosystems. Surface energy balance (SEB) methods relying on satellite remote sensing have proven effective in measuring actual evapotranspiration (ET a eb ) across different scales. However, their applicability may be constrained by interruptions in image acquisition caused by cloud cover and/or the spatio-temporal resolution limitations of satellites. In this research, a model-based methodology is suggested for simulating the dynamics of the soil - plant-atmosphere system and for estimating the daily actual evapotranspiration (ET p act ) of a Mediterranean Maquis ecosystem in northwest Sardinia. The model integrates ET a eb estimates obtained from the SEBAL model utilizing Landsat-8 data, satellite-derived vegetation indices, on-site measurements of potential evapotranspiration, and the mono-dimensional transient flow Richards equation for simulating soil moisture within the root zone. By combining these elements, the proposed model provides a more comprehensive and accurate estimate of the ET p act between Landsat acquisitions. The SEBAL model showed satisfactory performance in estimating actual evapotranspiration (ET a eb ) on satellite acquisition days, with an average error of 17 % compared to Eddy Covariance measurements. In addition, the integrated modelling approach yielded an ET p act average estimation error of +/- 37 % in the whole studied period. The soil moisture simulation by the model had a notable accuracy with an average error of 7.1 %. Temporal analysis showed that the model effectively simulated ET p act and soil moisture under both dry and wet conditions, exhibiting similar monthly and daily variations as observed data. Furthermore, a sensitivity analysis revealed that the stress index significantly improved the model ' s accuracy, while vegetation dynamics had a lower impact. Overall, the proposed model is a valuable tool for estimating ET p act in semi -arid Mediterranean regions, providing important information for water resource management and conservation efforts. Further application and validation of the model are recommended as new data becomes accessible, especially in areas characterized by cropped and irrigated agriculture. In future work, we aim to spatialize the Richards equation and integrate a multidimensional water balance hydrologic model.