Comparative Research On Different Regional Digital Soil Mapping Methods

Conventional map of soil investigation cannot reflect the the continuity and graded properties of soil in space. In this study, the new theory and technology of Pedometrics and DSM was used to build soil spatial inference models and accomplish the digital mapping on regional scale, which would promotes the development of pedogeography and has demonstrative value to other related subjects.The study area locates in the Yellow River alluvial plan, As this area is influenced by the Yellow River, burst flood and human activities, the geomorphological types are complicated. In this thesis, firstly, Fuzzy C-means clustering method was applied in continuous soil classification. Secondly, spatial regression model and quantitative soil-landscape model are used to fit the classification results and environmental covariate. Lastly, building the spatial prediction models and realizing the digital soil mapping. The main results are summarized as follows:(1) The OS AC A classification system, which is based on Fuzzy C-means clustering method was used to develop unsupervised classification on soil profiles, and soil types were divided into five categories, the taxonomic distance between the five center soil profiles and the40soil profiles were calculated.(2) The results of spatial regression analysis reflect that the use of spatial regression model on regional digital soil mapping is reasonable and necessary. The determinable taxonomic distance shows the spatial difference of natural factors between different soil types. The undeterminable taxonomic distance indicates the spatial soil variation that resulted from unnatural factors. It is clearly shown in the digital soil map that Typic Endorusti-Ustic Cambosols is soil type with the biggest distribution percentage, while Parasalic Siltigi-ustic Cambosols is the soil type with the second biggest coverage, Haplic Endorusti-Ustic Cambisol has the smallest area, spatial soil variation mainly locates in the Yellow River dike, Dagong crevasse splay and southeast of study area.(3) The results of quantitative soil-landscape model indicates that topographic factors cannot be used as combination of environmental factors alone to predict the distribution of soil types. It is clearly shown in the digital soil map that Pandian Series,Haplic UapUstic Cambisol, is soil type with the biggest distribution percentage, while Salinic Warpic Ustic Cambosols is the soil type with the second biggest coverage, then the third and fourth are Haplic Warpic Anthric Entisol and Yingju Series,Haplic UapUstic Cambisol, Haplic Endorusti-Ustic Entisol has the smallest area.(4) Comparing the results of spatial regression model with quantitative soil-landscape model, the digital soil map produced by spatial regression model displayed an evident advantage to reveal the spatial distribution of soil types, and it is found to be more sensitive to spatial soil variation than quantitative soil-landscape model. Moreover, the influence of natural factors to soil development can be better reflected by spatial regression model than quantitative soil-landscape model on regional scale. the soil types produced by quantitative soil-landscape model are single and concentrate on one area, this is inconsistent with the principle of soil genesis. The spatial regression model is more suitable for digital soil mapping in the study area than quantitative soil-landscape model.