A multivariate polynomial regression modeling (MPR) framework is developed to estimate total dissolved solids (TDS) in stormwater runoffs from rangelands in the Colorado River Basin in the Southwestern United States. An accurate TDS estimation model is needed to simulate terrestrial and aquatic salt transport processes on rangelands, identify critical source areas, and manage these sources effectively. However, modeling stormwater TDS runoff on rangeland sodic soils is challenging due to its complex correlation with variables in many aspects, such as topography, climate, soil, and vegetation. We propose a two-stage MPR framework based on field data collected from multiple rainfall simulator experiments: (1) variable selection with factor analysis and (2) TDS modeling via MPR, considering the nonlinear relationships between variables. Tabu search (TS) is used to optimize the TDS model in MPR. The proposed framework achieved a high prediction accuracy of 74.7% in estimating the TDS runoff transport.

开发了一个多元多项式回归模型 (MPR) 框架来估计来自美国西南部科罗拉多河流域牧场的雨水径流中的总溶解固体 (TDS)。需要一个准确的 TDS 估计模型来模拟牧场上的陆地和水生盐分迁移过程，识别关键源区并有效管理这些源。然而，对牧场钠质土壤上的雨水 TDS 径流建模具有挑战性，因为它与地形、气候、土壤和植被等许多方面的变量具有复杂的相关性。我们提出了一个基于从多个降雨模拟器实验中收集的现场数据的两阶段 MPR 框架：（1）通过因子分析选择变量和（2）通过 MPR 进行 TDS 建模，考虑到变量之间的非线性关系。禁忌搜索 (TS) 用于优化 MPR 中的 TDS 模型。所提出的框架在估计 TDS 径流传输时实现了 74.7% 的高预测精度。