Modeling Soil Thermal Regimes During a Solarization Treatment in Closed Greenhouse by Means of Symbolic Regression via Genetic Programming

Modeling soil thermal regimes during a solarization treatment in closed greenhouse is useful to estimate the required duration of the treatment in relation to the climatic conditions, as well as the efficacy of the technique. Several studies have been carried out, based on two main strategies: modeling the physical processes of the soil-mulch-greenhouse system or applying numerical procedures based on neural networks (NNs). However, the application and reliability of physical models require accurate knowledge of the thermo-physical properties of each component of the system, while NNs do not give any symbolic function which can be easily used. Symbolic regression via genetic programming represents an alternative method for finding a function that best fit a given set of data. In this paper, a such model is proposed, which use air temperature and global solar radiation flux outside the greenhouse, depth into the soil, existence of mulch and time of day as input variables and provides soil temperatures at different depths as output. The results allowed to obtain an easy to use symbolic function that is able to estimate soil temperature with an accuracy comparable to that one attained with other simulation models.