Abstract Suitable design and management of drip irrigation systems depends on an accurate knowledge of the moisture distribution patterns around the emitters. An important parameter in the design of surface and subsurface drip irrigation systems is the wetted pattern of soil area in the up and down wetted areas of drippers. Accurate estimation of the up and down wetted areas in surface/subsurface irrigation systems is essential for enhancing the efficiency of irrigation systems and optimal management of water resources in the field. In this study, artificial neural network (ANN) and nonlinear regression (NLR) methods were used to develop equations for the estimation of the up and down wetted areas around the dripper installation position. Experiments were carried out in a transparent physical Plexiglas with dimensions of 3 m × 1.22 m × 0.5 m . In this study, three different soil textures (i.e., light, medium, and heavy) were used. The drippers at four installation depths, including 0, 15, 30, and 45 cm, were evaluated. By considering an irrigation time equal to 6 h, the emitter discharge rates (i.e., 2.4, 4, and 6 L/s) were applied. Nine different variables, including soil hydraulic conductivity, emitter discharge, soil apparent density, irrigation application time, dripper installation depth, and the soil texture (i.e. amount of clay, silt and sand)were applied as inputs for the developed NLR and ANN models for estimating the up and down wetted areas of drippers. The comparison results between the measured and simulated values showed that the ANN and NLR models have appropriate performances and that the statistical error indices are within an acceptable range. The use of these models for design goals can be helpful in choosing the accurate distance between laterals and emitters as well as the suitable depth of emitters to minimize water losses via deep percolation in surface/subsurface drip irrigation. Furthermore, the results of the ANN and NLR models were compared with an experimental model (developed based on dimensional analysis (DA)) and confirmed the superiority of the ANN and NLR to DA models.

中文翻译：

---

滴灌系统湿球湿面积建模：增强的经验模型和人工神经网络

摘要 滴灌系统的适当设计和管理取决于对发射器周围水分分布模式的准确了解。地表和地下滴灌系统设计中的一个重要参数是滴头上下浸湿区土壤面积的润湿模式。准确估计地表/地下灌溉系统的上下湿润面积对于提高灌溉系统的效率和田间水资源的优化管理至关重要。在这项研究中，人工神经网络 (ANN) 和非线性回归 (NLR) 方法被用于开发用于估计滴头安装位置周围上下湿润区域的方程。实验在尺寸为 3 m × 1.22 m × 0.5 m 的透明物理有机玻璃中进行。在这项研究中，使用了三种不同的土壤质地（即轻、中和重）。评估了四种安装深度的滴头，包括 0、15、30 和 45 厘米。通过考虑等于 6 小时的灌溉时间，应用了发射器排放速率（即 2.4、4 和 6 L/s）。九个不同的变量，包括土壤导水率、发射器排放量、土壤表观密度、灌溉应用时间、滴头安装深度和土壤质地（即粘土、淤泥和沙子的数量）被用作开发的 NLR 和 ANN 模型的输入。估计滴头的上下浸湿面积。实测值和模拟值的比较结果表明，ANN 和 NLR 模型具有合适的性能，统计误差指数在可接受的范围内。将这些模型用于设计目标有助于选择支管和发射器之间的准确距离以及合适的发​​射器深度，以通过地表/地下滴灌中的深层渗透将水损失降至最低。此外，将 ANN 和 NLR 模型的结果与实验模型（基于维度分析 (DA) 开发）进行了比较，并证实了 ANN 和 NLR 对 DA 模型的优越性。