Dados do Trabalho

Título

MAPPING SOIL ATTRIBUTES FOR LATIN AMERICA AND THE CARIBBEAN: A COMPREHENSIVE GEOSPATIAL ANALYSIS

Resumo

Latin America and the Caribbean (LAC) encompass a vast expanse of 20,139,378 km2, playing a crucial role in sustaining the livelihoods of 670 million people and housing some of the planet's most diverse ecosystems. In this study, we present detailed maps of various soil attributes across the entire LAC region, utilizing a 90 m resolution. To achieve this, we collected georeferenced soil data from nearly 100,000 locations at different depths and employed spline interpolation to estimate values at a 0-20cm depth range. Incorporating 13 covariates, such as surface reflectance of bare soils, potential natural vegetation, precipitation, temperature, elevation, slope, multi-scale topographic position index, and Euclidean distance to water, we developed parsimonious prediction models using the Google Earth Engine Python API and Random Forest algorithm. Stratifying our models by biome, we generated spatial predictions with 100 repetitions (bootstrap) to derive lower (5%), mean, upper (95%), and standard deviation values for each soil attribute. The resulting 90m predicted maps include clay, silt, sand, bulk density, plant available water, carbon, carbon stock, pH (water), nitrogen, phosphorus, potassium, sum of bases, and base saturation. These maps exhibit good prediction performance, with R2 values ranging from 0.28 to 0.75, and can be accessed and visualized at https://raulrpoppiel.users.earthengine.app/view/lac-soil-attributes. Our methodology provides reliable soil maps that are invaluable for conducting diverse environmental evaluations, including comprehensive assessments of soil health in the LAC region.

Palavras-chave

pedometrics; remote sensing; data science; earth engine; python; soil mapping, environmental assessment; public policies.

Instituição financiadora

This study was supported by the Brazilian National Council for Scientific and Technological Development (CNPq) and the São Paulo Research Foundation (FAPESP, grant number 2021/05129-8).

Agradecimentos

We express our gratitude to the researchers and institutions from Latin American and Caribbean (LAC) countries for their valuable contributions in providing soil data.