Gaussian Processes for Statistical Soil Modeling of the Tropics

Abstract

Soil maps are essential resources to soil scientists and researchers in any fields related to soils, land use, species conservation, hunger reduction, social development, etc. However, creating detailed soil maps is an expensive and time consuming task that most developing nations cannot afford. In recent years, there has been a significant shift towards digital representation of soil maps and environmental variables that has created the field of predictive soil mapping (PSM), where statistical analysis is used to create predictive models of soil properties. PSM requires less human intervention than traditional soil mapping techniques, and relies more on computers to create models and predict properties. However, because most of the research funds for soil research come from developed nations, the research in this field has mostly focused in temperate zones (where most developed nations are located). The areas of the world with more needs in terms of hunger and poverty are mostly located in the tropics, and require different statistical models because of the unique characteristics of their weather and environment. Through to the v-unit/TechBridgeWorld initiative at Carnegie Mellon we were able to work with a group of soil scientists from the International Center for Tropical Agriculture (CIAT) and develop statistical soil models for Honduras. Thanks to this joint work, we were able to leverage the knowledge of the soil science and computer science communities, and create a model that matches or advances the state of the art for PSM