Classification Of Soil Family Based On Chinese Soil Taxonomy By Using Soil Taxonomic Distance

A large amount of manpower,material and financial resources and time are required when making traditional soil maps and the literal results of soil information cannot be applied directly in the research and production practice of soil,agriculture,environment,ecology and the like today,a digital and information age.As a result,digital soil information emerged based on measuring soil science,3S and fuzzy logic,etc.After developing more than 20 years,digital soil mapping has gradually improved,which is now widely used in soil survey and mapping,as well as in various fields of research and practice.However,it is difficult to apply the classification standard of literally descriptive soil classification system directly when digital soil mapping.And the soil types determined by the multivariate factors could not be showed directly on the soil map as the soil property maps.Numerical soil classification of the pedons sampled in the study area,in which the soil diagnostic horizons and diagnostic characteristics involved in the hierarchical classification at various levels of Chinese Soil Taxonomy were employed as the input variables,is a bridge between the soil taxonomy and digital soil mapping.Meanwhile it transformed the text variable of soil types to numerical variable,and multivariate soil attributes to a single variable--soil types.The pedons sampled in the study area were numerically classified by Fuzzy K-means algorithm,in which the soil diagnostic horizons and diagnostic characteristics involved in the hierarchical classification at various levels of Chinese Soil Taxonomy were employed as the input variables.The fuzzy clusters were generated from the sampled pedons by the FKM algorithm and their centroid soil families identified by checking the centroid values of the input variables.Besides,the taxonomic distances between the sampled pedons and centroid soil families were calculated,thus laying the foundation for evaluating reasonability and validity of the hierarchical classification completed according to Chinese Soil Taxonomy.It was found that the most criteria for classifying the soil group seems reasonable,while some are open to question,which was caused by unreasonable soil classification criteria or the input variables of fuzzy K-means algorithm.In this case study,numerical soil classification was demonstrated by fuzzy K-means algorithm,in which diagnostic horizons and characteristics were assigned and employed as input variables combining with the weights and membership,thus guaranteeing genetic connections of the pedons gathered in a cluster.Meanwhile numerical soil classification provides a quantitative data basis,on which measurable relationships between sampled pedons and environmental covariates can be established and spatial soil prediction functions fitted.In that respect,numerical soil classification is becoming one of the most critical driving forces to promote spatial soil prediction and digital soil mapping approaches.