Coupling geostatistics and MLA for modelling and predicting the spatio-temporal dynamics of SOC stocks, drivers, and trade-offs in Brazilian integrated agricultural systems

• Grant number: 23/05122-9
• Support Opportunities: Scholarships abroad - Research Internship - Post-doctor
• Start date: November 01, 2023
• End date: October 31, 2024
• Field of knowledge: Agronomical Sciences - Agronomy - Soil Science
• Principal Investigator: Maurício Roberto Cherubin
• Grantee: Chukwudi Nwaogu
• Supervisor: Budiman Minasny
• Host Institution: Escola Superior de Agricultura Luiz de Queiroz (ESALQ). Universidade de São Paulo (USP). Piracicaba , SP, Brazil
• Institution abroad: University of Sydney, Australia
• Associated to the scholarship: 21/11757-1 - Assessment of soil carbon stocks in integrated agricultural systems in Brazil: a geoinformatics approach, BP.PD

Abstract

Soil organic carbon (SOC) is vital for soil biological, chemical and physical processes, provides essential information on changes in soil quality and land use, as well as controls food production and climate change. The assessment of SOC dynamics and its drivers is essential for improving soil C sequestration in integrated agricultural systems (IAS). This study applies advanced digital soil mapping (DSM) techniques such as geostatistical and machine learning algorithms (MLA) to model and predict the spatio-temporal dynamics of SOC stocks and drivers in Brazilian IAS in the past 3 decades. The study identifies and maps the distribution of IAS across Brazil. By combining different relevant data sets on soil and environmental covariates (e.g., SCORPAN) from Brazilian soil spectral library (BSSL), legacy data, remote sensed data, and using machine learning methods the study models and predicts the SOC stock in Brazilian adopted IAS. It also determines the influence of the environmental variables on the distribution of SOC stocks especially in the top soil layer (0-30cm). This study further determines the best SOC prediction model(s) by testing different MLA (artificial neural networks, support vector machines, random forest, multiple linear regression, boosted regression trees and Cubist). The findings from the study will increase the knowledge of the stakeholders in soil-climate-food production systems about SOC stock in IAS. It will further support the Brazilian government to formulate, define and re-structure relevant land use and carbon sequestration policies to meet the Nationally Determined Contributions (NDCs) contained in the Paris Agreement. The output of this project might contribute to the SOC stocks baseline for the country. (AU)