Neuro-Fuzzy Modeling as Support for Decision-Making in the Production of Irrigated Coriander Under Mulch in the Semi-Arid Region

Abstract

Reducing water consumption by crops in semi-arid regions is an important factor for the sustainability of agriculture in these locations. In this sense, this study aims to evaluate the neuro-fuzzy inference method as a support for decision-making in irrigated coriander cultivation. The experiment was performed in two cultivation cycles in Pentecoste-CE, Brazil. The experiment was conducted in randomized blocks arranged in a split-plot design with five primary treatments, consisting of irrigation depths (50, 75, 100, 125, and 150% of the localized evapotranspiration, ETcloc), and five secondary treatments, consisting of different levels of bagana mulch (0, 25, 50, 75, and 100%, equivalent to 16 t ha −1 ). Neuro-fuzzy models with two input variables and eight output biometric variables were developed to evaluate growth (plant height, number of roots, and root length) and yield variables (productivity and shoot and root fresh and dry mass). In the first cycle, the best results occurred close to 55% ETcloc and between 40 and 50% of mulch; in the second cycle, water consumption returned results between 50 and 80% ETcloc. The fuzzy and multiple regression models showed MAE, MSE, and RMSE errors of 9, 22, and 10% lower, respectively. The neuro-fuzzy model might be a viable option for decision-making in irrigated crops, being able to optimize the use of natural resources and available water in semi-arid regions. The use of 55% of irrigation depth and a range of 40 to 50% of mulch can be a strategy for a higher water use efficiency.