Potential productive of Brazilian agricultural soils

• Ano de defesa: 2022
• Autor(a) principal: Greschuk, Lucas Tadeu
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Aprendizado de máquinas; Atributos do solo; Digital soil mapping; Machine learning; Mapeamento digital de solos; Monitoramento do solo; Soil attributes; Soil monitoring
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/11/11140/tde-12072022-164028/

The main issue for the world\'s good is food security. Brazil plays an important role in global agriculture and can help meet global food demand in the future. The chemical, physical and biological properties of the soil directly influence the production of cultivated plants. Understanding the productive potential of Brazilian agricultural land is of great importance. Therefore, the objective of this study is to develop a strategy to map and quantify the productive potential of Brazilian agricultural soils through digital soil mapping techniques. Approximately 70,000 soil samples (0 - 1.0m) georeferenced with information on chemical, physical and biological properties of Brazilian agricultural soils were used. Each soil attribute was evaluated using literary information. Then, a weighted artifice equation was developed through the analysis of principal components of the soil samples. Therefore, each sample received a score ranging from 0 to 100 referring to its soil potential to produce plant biomass, that is, the higher the score, the higher its potential. Sample scores were spatialized via machine learning. 80% (205 million hectares) of Brazilian agricultural areas were mapped, considering areas of cultivated plants and pastures. Through the territorial quantification of the productive potential of soils, the possibility of seeing the best and worst categories of agricultural soils in each Brazilian biome. The best agricultural soils in Brazil that belong to the very high and high SoilPP categories were found in the Atlantic Forest (11.8 Mha), Ama zon (7.6 Mha) and Cerrado (4.4 Mha) biomes. On the other hand, soils that varied from medium/high to very low potential were found in the Cerrado (77.9 Mha), Caatinga (28.9 Mha) and Atlantic Forest (34.5 Mha) biomes. Through the evaluation of agricultural land and average productivity values, we observed that of 2304 soybean producing municipalities, 896 have the capacity to increase average soybean productivity to a certain level. For sugarcane, out of 2468 municipalities evaluated, 1056 can increase their average productivity. Therefore, this technique can assist in the development of global food and soil security policies and be replicated in other countries, regions, municipalities or farms. The SoilPPmap for Brazilian agricultural areas had as main limitations the spatial representativeness for all types of agricultural soils in Brazil and also the low accuracy of the prediction model, possibly related to the evaluation of soil chemical attributes. However, the strategy used was effective in mapping and quantifying the productive potential of Brazilian agricultural soils.