Modelling of Aerial Biomass and Its Physicochemical Properties Using Artificial Intelligence and Response Surface Methodology

Aerial biomass and its associated physicochemical properties are pivotal components in understanding ecosystem dynamics, with far-reaching implications for environmental and agricultural management. This study delves into the dynamics of aerial biomass and its correlation with physicochemical properties, utilizing a dataset derived from research on Spartina alterniflora in the Cape Fear Estuary of North Carolina. Machine learning (ML) models, including Response Surface Methodology (RSM), the Quadratic Model, Artificial Neural Networks (ANN), and the Adaptive Neuro-Fuzzy Inference System (ANFIS), are employed to evaluate the relationships between these properties and aerial biomass. The RSM model outperformed other models with a remarkable Mean Squared Error (MSE) of 0.0579 and a high R-squared value of 0.9518, emphasizing its efficiency in estimating biomass. The quadratic model follows closely, with an MSE of 0.1778 and an R-squared value of 74.61%, providing valuable insights into biomass variation. Furthermore, Particle Swarm Optimization (PSO) is applied to optimize the models. The results highlight RSM-PSO as the most efficient technique, with a PSO value of 0.9872 and an ML R-squared of 0.9518, underscoring the robustness of the RSM model when combined with PSO for predicting aerial biomass. The findings emphasize the significance of pH and potassium content in biomass estimation and recommend the RSM model, particularly when coupled with PSO, for efficient biomass prediction. These insights have critical implications for environmental and agricultural management and may serve as a valuable tool for ecosystem optimization.