Method for estimating parameter γ in a two-dimensional approximate furrow infiltration model

• Published in: Computers and electronics in agriculture Vol. 235; p. 110403
• Main Authors: Li, Ge, Nie, Weibo, Li, Yuchen, Zhang, Wei
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.08.2025
• Subjects: Adaptive boosting (AdaBoost); agriculture; algorithms; Backpropagation neural network (BPNN); Edge effect; edge effects; electronics; Furrow irrigation; furrows; Gray wolf optimization (GWO); hydraulic conductivity; prediction; regression analysis; soil texture; Furrow irrigation; Gray wolf optimization (GWO); Adaptive boosting (AdaBoost); Backpropagation neural network (BPNN); Edge effect
• ISSN: 0168-1699
• DOI: 10.1016/j.compag.2025.110403

•The range of parameter γ is 0.60–0.94.•The edge effect’s relative contribution to the total furrow infiltration is 21.0%–41.7%.•A GWO–BPNN–AdaBoost model is proposed for estimating γ. This study expands the analysis of the parameter γ in the approximate furrow infiltration model (FIM) proposed by Bautista et al. On the basis of the obtained results, a gray wolf optimization (GWO)–backpropagation neural network (BPNN)–adaptive boosting (AdaBoost) regression model was developed, and its γ prediction performance was compared with that of a BPNN model and BPNN–AdaBoost model. The results indicated that furrow cross section, soil texture, and water depth (h0) considerably influence the edge effect (ΔI) and γ, with γ ranging from 0.60 to 0.94. Moreover, the edge effect’s relative contribution to the total furrow infiltration is 21.0 %–41.7 %. Three performance measures, namely Bias, root mean square error (RMSE), and coefficient of determination (R2), were employed to evaluate the performance of the proposed models. The results revealed that γ values of Bias, RMSE, and R2 were –0.0009, 0.062, and 0.851 for the GWO–BPNN–AdaBoost model has the high accuracy, respectively, with furrow depth (D), bottom width (B), top width (T), n, α, saturated hydraulic conductivity (Ks), h0, effective saturation (Se) as input factors. The two-dimensional cumulative infiltration was calculated using the γ values predicted by the models. Among these predictions, those produced by the GWO–BPNN–AdaBoost model most closely aligned with the simulated values acquired using Hydrus-2D, with the Bias, RMSE, and R2 values being –0.53, 1.41 cm, and 0.993, respectively. Based on analysis of the obtained results, it is evident that GWO–BPNN–AdaBoost can estimate γ more accurately.