基于3种地理加权回归方法的安徽省土壤pH空间预测制图对比研究

基于安徽省140个采样点的土壤pH数据,综合考虑土壤、地形、气候、生物等因子对土壤pH的影响,采用地理加权回归(Geographically Weighted Regression, GWR)、主成分地理加权回归(Principal Component Geographically Weighted Regression, PCA-GWR)和混合地理加权回归(Mixed Geographically Weighted Regression, M-GWR)3种模型对安徽省土壤pH空间分布进行建模预测,揭示环境因子对土壤pH的影响在空间上的差异,最后以多元线性回归模型(Multiple Linear Regression, MLR)为基准比较3种GWR模型的精度。研究表明：(1)安徽省土壤pH具有空间异质性,且集聚特征明显。(2) 3种GWR模型中M-GWR模型略优,GWR、PCA-GWR和M-GWR的建模集调整后决定系数(R_adj²)分别为0.59、0.62和0.63;对比MLR模型,3种GWR模型的R_adj^2<...

Comparison and analysis of spatial prediction and variability of soil pH in Anhui Province based on three kinds of geographically weighted regression

Geographically weighted regression (GWR), principal component geographically weighted regression (PCA-GWR) and mixed geographically weighted regression (M-GWR) were used to model and map the spatial distribution of soil pH in Anhui Province. Based on the soil pH data from 140 sampling sites in Anhui Province, and the effects of macroscopic factors such as soil, topography, climate and biology on soil pH were also taken into consideration. And then, the spatial distribution characteristics of the effects of environmental factors on soil pH were explored. Finally the accuracy of three GWR models was compared based on multiple linear regression model (MLR). The results showed that: Soil pH in Anhui Province has spatial heterogeneity and obvious agglomeration characteristics. Among the three GWR models, the M-GWR model is slightly better, and the modeling sets R_adj² of GWR, PCA-GWR and M-GWR are 0.59, 0.62 and 0.63, respectively. Compared with the MLR model, the R_adj² of the three GWR models increases by 23%, 31% and 35%, respectively. The prediction map generated by M-GWR is smooth in space, and the modeling effect is stable. The prediction results show that the area north of the Huaihe River in Anhui Province is mostly alkaline soil, and the south of the Yangtze River is mostly neutral or acidic soil, which accords with the characteristics of southern acid and northern alkali. The results show that GWR and its improved model can effectively predict soil pH attributes and reflect the influence of environmental factors on soil pH in different spatial locations, while M-GWR has both global and local effects of variables, which improves the interpretation ability of the model and provides an important reference method for digital soil mapping in large areas.