UC Davis

Soil Taxonomy has proven to be a valuable tool for communicating soils information. Since its inception in 1975, soils have been classified by manually working through criteria established in the Keys to Soil Taxonomy. Navigating this system is often time consuming, prone to error, and requires significant pedologic knowledge and training. Despite recent advancements in soil data storage and analysis, no method currently exists to automatically calculate soil taxa. Here, a new suite of computer algorithms is presented to aid in the determination of soil classification. This tool is designed to take in soil lab and field observations, calculate diagnostic features, and output a final classification to the suborder level for all orders and great group level for Gelisols, Histosols, and Alfisols. Demonstrations prove that this tool can reach classifications that match human-verified results. However, the tool’s accuracy is highly dependent on the quality and completeness of input data. The automated process requires a massive collection of data and observations for every possible classification outcome. Thus, the tool inherently lacks the human/pedologist expertise that would allow a user to simplify the classification process and reduce data demand by only considering regionally relevant taxa. However, the tool can be constrained by users to better mimic expert behavior. This tool shows incredible potential for reducing the burden of updating outdated pedon taxonomy in NRCS databases by processing large amounts of data quickly and accurately. This tool also provides a computer resource that will help make Soil Taxonomy more accessible.