Disaggregation of conventional soil maps: a review

Abstract

The disaggregation of conventional soil maps is an active research line inside the Digital Soil Mapping framework that aims to generate new cartographies by disclosing the non-explicit soil distribution pattern within the polytaxic or multi-component cartographic units. This article shows a comprehensive review of methodologies found after a bibliographic search in the Web of Science and Scopus databases. They are analysed regarding common factors such as the conventional soil map, environmental data sources and covariates, classification methods, and evaluation; likewise, those specific to the leveraging of conventional maps as the main source of soil information such as sampling scheme and assignment of soil categories for the classification. The applications were frequently carried out in small and medium areas with intensive and extensive conventional soil maps and featuring supervised classification methods. The definition of the training sets is a critical task that has a strong influence on their performance. The comparative analysis noted the potential of the reviewed disaggregation methodologies that adopted two-stage strategies: first, areas potentially associated with soil categories are delimited; and second, supervised models are built on them. Ultimately, the development of new disaggregation methodologies is possible by combining those strategies within each factor that yielded the best results in terms of accuracy.