Unsaturated Hydraulic Conductivity Prediction Using Artificial Intelligence and Multiple Linear Regression Models in Biochar Amended Sandy Clay Loam Soil

Improved productivity of crops grown in biochar amended soils largely depend on the unsaturated hydraulic conductivity (K (q)) and moisture content of the soil. However, their relationships in biochar amended soil have not been well elucidated in both field and laboratory studies. Moreso, it is important to propose a model, which can accurately predict the K (q), since its determination in field is laborious. The goals of this study were to determine the relationship between moisture content and K (q) in biochar amended soil; (ii) predict K(q) using multiple linear regression (MLR), artificial neural network (ANN) and adaptive neuro-fuzzy inference system (ANFIS), as innovative tools used in soil science; and (iii) evaluate their performances. Field experiments with five levels of biochar applications (0, 3, 6, 10 and 20 t/ha) were used during maize growing seasons. Soil moisture contents in relation to days after planting (DAP) of maize and biochar (B) application rates were recorded and used as model inputs. Results showed that measured K (q) decreased as moisture content increased in biochar amended soil. Also, ANN outperformed ANFIS and MLR in predicting K (q). The coefficients of determination, R2 were 0.98, 0.92 and 0.95 for the ANN, MLR and ANFIS during validation, respectively. Also, the root mean square error (RMSE) values were 1.80, 8.83 and 6.84 mm h−1 for the ANN, MLR and ANFIS during validation, respectively. Artificial neural network is most suitable for modelling water flow in biochar amended soil, and moisture content is important for its determination.