Improving soil moisture simulation by optimizing soil texture profiles with a micro-genetic algorithm

Soil texture, defined by percentages of clay, silt, and sand (PCSS), significantly affects soil moisture estimates in land surface models. However, PCSS profiles are often uncertain in soil databases and are sparsely observed, which can propagate into errors in soil moisture estimation. In this study, we propose an optimization framework to optimize PCSS profiles by evaluating soil moisture against in situ observations, using a micro-genetic algorithm (μGA) coupled with the University of Torino land surface Process model for Interaction in the Atmosphere (UTOPIA). We applied a multi-constraint approach, including prescribed parameter ranges based on a reliable database, to prevent unrealistic PCSS and ensure physically consistent soil texture profiles. Optimization was conducted at three Italian sites and, on average across sites and layers, led to a 27% reduction in RMSE and a 1.3% increase in R for soil moisture estimation compared with experiments using typical soil databases and in situ observations.