A numerical approach for soil microbiota growth prediction through physics-informed neural network

Abstract

In recent years, there has been a growing interest in leveraging Scientific Machine Learning (SciML) techniques to address challenges in solving Partial Differential Equations (PDEs). This study focuses on forecasting the growth of microbial populations in soil using a novel numerical methodology, the Physics-Informed Neural Network (PINN). This approach is crucial in overcoming the inherent challenges associated with the general unculturability of soil bacteria. PINNs can be used to model the growth of bacterial and fungal populations, considering environmental factors like temperature, solar radiation, air humidity, soil hydration status, and external weather conditions. In this paper, some stability issues related to the mathematical model have been analyzed. Moreover, by utilizing field data and applying equations that describe the biological mechanisms of microbial growth, a PINN was trained to predict the development of the microbiota over time. The results demonstrate that the use of PINNs for studying microbial growth and evolution is a promising tool for enhancing agriculture, optimizing cultivation processes, and facilitating efficient resource management.