Numerical soil classification using fuzzy K-means algorithm and predictive soil mapping at regional scale

摘要: In the studied region,choosing texture,organic matter content,soil color,pH,electronic conductivity and soil layer thickness as variables,numerical classification based on 40 profiles was conducted by use of fuzzy K-means algorithm,and resulting central 01,18,37,38 were allocated into a hierarchical according to Key Chinese Soil Taxonomy.They belonged Pandian Series,Haplic Uap Ustic Cambisol,Yingju Cambisol,Haplic Endorusti-Ustic Warpic Anthric Entisol Salinic Cambisol respectively.On basis known taxonomic distances(Here,Euclidean distance employed) amongst sampled soils above profiles,the between unknown at any sites figured out using Geostatistic techniques.In this way,continuous region through predicted relationship distances soils.For better visualization,the soft borders from different classes hardened means defuzzification defined thresholds.And output predictive mapping thus had same visual appearance with conventional map.From mapping,it could be clearly seen that cover in dominated Haplic Cambisols Cambisols,which accounts for large percentage region′s total area(46.3% 33.2% respectively).By contraries,Haplic Entisols discontinuously distributed totally occupied only around 10% land area region.It concluded integration numerical geostatistic interpolation would one most recommendable approaches mapping.