Machine Learning Approaches for Gap-Filling in Irregular Greenhouse Gas Time Series

• Grant number: 25/13458-2
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: August 01, 2025
• End date: July 31, 2027
• Field of knowledge: Agronomical Sciences - Animal Husbandry - Pastures and Forage Crops
• Principal Investigator: Ana Cláudia Ruggieri
• Grantee: Gabriela Oliveira de Aquino Monteiro
• Host Institution: Faculdade de Ciências Agrárias e Veterinárias (FCAV). Universidade Estadual Paulista (UNESP). Campus de Jaboticabal. Jaboticabal , SP, Brazil
• Associated research grant: 21/11922-2 - Science Center for the Development of Climate Neutrality for Beef in Tropical Regions, AP.CCD

Abstract

Brazilian agriculture and livestock play a significant role in global food security, but it is also one of the sectors most responsible for national greenhouse gas (GHG) emissions, especially methane (CH¿), nitrous oxide (N¿O) and carbon dioxide (CO¿). Brazil has one of the largest pasture areas in the world, but much of it is in low or medium vigor, which compromises productivity, favors GHG emissions and reduces the soil's carbon sequestration capacity. Faced with this challenge, integrated production systems, such as Crop-Livestock Integration (CLI), Crop-Livestock-Forest Integration (CLI) and consortium between grasses and legumes, have been proposed as promising strategies to restore degraded areas, increase soil biodiversity, improve nutrient use efficiency and mitigate emissions associated with livestock production. However, there are still important gaps in understanding the mechanisms involved in the dynamics of GHG emissions in these systems, especially with regard to soil carbon quality and the interaction of soil chemical, physical, and biological factors with pasture management practices. In addition, the detailed characterization of GHG emissions from agriculture still faces methodological challenges, such as spatial and temporal data variability and the scarcity of standardized data to feed inventories and projections. In this scenario, analytical tools such as machine learning can contribute significantly, offering greater accuracy in predicting emissions, identifying explanatory variables, and reconstructing historical series. In view of this, this project aims to integrate experimental and computational approaches to quantify, understand, and predict GHG emissions in tropical pasture systems. To this end, experiments will be conducted involving pastures in monoculture, integrated systems, intercropped systems, and native vegetation, with evaluation of soil physical-chemical and biological attributes and direct measurements of CH¿, N¿O, and CO¿. At the same time, predictive algorithms capable of integrating empirical data and projecting future scenarios will be developed, supporting public policies to mitigate climate change and sustainably intensify livestock farming in Brazil. (AU)