Detailed mapping of key indicators to assess soil quality from 0 to 100 cm deep throughout the Brazilian territory

Abstract

The quality of soils refers to the soil's capacity to provide essential services to humans through the environment, and this capacity integrates multiple chemical, physical, and biological soil indicators. Often, it is not easy to determine these indicators over large land extents, which complicates the creation of soil quality maps in vast areas. However, one way to obtain these indicators is through the use of pedotransfer functions, utilizing easily acquired soil attributes such as clay content, organic carbon content, and total oxide content. New methodologies based on remote sensing and machine learning algorithms are aiding in the construction of soil attribute maps with minimized uncertainties and high levels of detail. Therefore, the goal of this project is to conduct a detailed digital mapping of key indicators to assess soil quality between 0 and 100 cm deep across the entire Brazil, using data acquired through remote sensing and fieldwork, modeled by machine learning algorithms. As such, the proposition of this postdoctoral fellowship consists of four stages aimed at achieving this objective. These stages are as follows: in Stage 1, the datasets will be organized and filtered for subsequent processing; in Stage 2, soil attributes will be modeled from multispectral images and terrain attributes using the Random Forest machine learning algorithm; in Stages 3 and 4, the indicators will be spatialized via pedotransfer functions and geostatistical techniques. At the conclusion of this project, it is anticipated that maps of key indicators for evaluating soil quality will be generated on a national scale, with minimized levels of uncertainty and a spatial resolution of 30 meters at different depths (0-100 cm). (AU)