Dynamic monitoring of soil pH changes across different seasons is crucial to understanding soil alkalization and preventing land degradation in arid regions. Here, we explored the potential use of ground-measured hyperspectral data and Landsat 8 OLI remote-sensing images in the inversion of soil pH during the dry and wet seasons. The measured and image data of spectral reflectance and corresponding soil pH data obtained in the laboratory were used to analyze the spatiotemporal patterns of soil pH in the Yinbei region, Ningxia, China. First, the hyperspectral data were resampled to match the range of the image bands; then 11 spectral indices were calculated based on the two sets of spectral data. Principal component analysis (PCA), stepwise regression (SR), and gray relational analysis (GRA) were used to select feature variables (sensitive bands and spectral indices) from the spectral data. The back propagation neural network (BPNN), support vector machine (SVM), ridge regression (RR), and geographically weighted regression (GWR) were used to construct soil pH inversion models. A total of 24 models established using the different methods were compared in terms of the determination coefficients of the calibration set (R_c²) and prediction set (R_p²), root mean square error (RMSE), and relative percent deviation (RPD). The results revealed that the mean soil pH in the study area was higher during the dry season (9.28) than the wet season (9.11). The resampled hyperspectral data correlated well with the image data under different levels of soil alkalization (R² > 0.9652). Of the different models examined, the global regression models (BPNN, SVM, and RR) were superior to the local regression model (GWR), with better inversion results obtained in the wet than the dry season. The BPNN and SVM models performed better than the RR model, and the PCA-SVM model based on measured data in the dry season achieved the best overall performance (R_p² = 0.9724 and RPD = 5.76). These findings provide valuable information on the distribution of Solonetzs land in the study area, facilitating management of soil alkalization and degradation in the Yinbei region.

动态监测不同季节的土壤pH值变化对于了解土壤碱化和防止干旱地区的土地退化至关重要。在这里，我们探索了地面测量的高光谱数据和Landsat 8 OLI遥感影像在干旱和潮湿季节反演土壤pH值的潜在用途。利用实验室获得的光谱反射率的实测数据和图像数据以及相应的土壤pH数据，分析了宁夏银北地区土壤pH的时空分布。首先，对高光谱数据进行重新采样以匹配图像带的范围；然后根据两组光谱数据计算出11个光谱指数。主成分分析（PCA），逐步回归（SR），使用灰色关联分析（GRA）和灰色关联分析（GRA）从光谱数据中选择特征变量（敏感谱带和光谱指数）。使用反向传播神经网络（BPNN），支持向量机（SVM），岭回归（RR）和地理加权回归（GWR）来构建土壤pH反演模型。根据校准集的确定系数，比较了使用不同方法建立的总共24个模型（R _c ²）和预测集（R _p ²），均方根误差（RMSE）和相对百分比偏差（RPD）。结果表明，研究区的平均土壤pH在旱季（9.28）高于雨季（9.11）。在不同土壤碱化水平下，重采样的高光谱数据与图像数据很好地相关（R ²> 0.9652）。在检查的不同模型中，全局回归模型（BPNN，SVM和RR）优于局部回归模型（GWR），在潮湿季节获得的反演结果要好于干旱季节。BPNN和SVM模型的性能优于RR模型，基于干旱季节实测数据的PCA-SVM模型获得了最佳的总体性能（R _p ² = 0.9724和RPD = 5.76）。这些发现提供了有关研究区域所罗门兹土地分布的宝贵信息，有助于管理银杯地区的土壤碱化和退化。