Soybean Production and Agricultural Resource Demand under The Changing Climate

Abstract

Soybean is a globally important agricultural commodity, essential for human consumption, animal feed, and bioenergy production. Latin America, particularly Argentina and Brazil, significantly contributes to global soybean supply, influencing international food security and agricultural trade. However, increasing climatic variability and extreme weather events, such as droughts, floods, and heatwaves, pose substantial threats to soybean productivity and agricultural stability. Effectively addressing these challenges requires improved yield prediction techniques, a deeper understanding of yield sensitivities, and reliable projections of future production under diverse climate and socioeconomic conditions. This dissertation addresses these needs through three complementary studies, employing econometric modeling, satellite-based remote sensing, and integrated scenario analysis.The first study evaluates the performance of traditional panel regression methods versus deep learning models (Long Short-Term Memory networks with attention mechanisms) for predicting soybean yields in Argentina using satellite-derived NDVI data. Results indicate that deep learning approaches provide more accurate and timely yield forecasts compared to conventional regression methods, particularly when utilizing satellite imagery corresponding to key vegetative growth stages. The second study assesses how climate extremes affected soybean yields in Argentina and Brazil from 2001 to 2022. By applying detailed climate indices within a panel regression framework, the analysis identifies drought as the predominant driver of yield reductions in both countries. The results also highlight notable regional differences in climate vulnerability, emphasizing the need for region-specific adaptation strategies. The third study utilizes an annually configured version of the Global Change Analysis Model (GCAM) to project future soybean production, resource demands (land, water, and fertilizer), and international trade dynamics under various climate scenarios (SSP-RCP pathways) up to 2050. Findings demonstrate that annual yield variability significantly influences soybean land-use decisions, irrigation water demand, and market stability. Additionally, the analysis explicitly differentiates between soybean used for food and biofuel production, revealing greater volatility in biofuel-related soybean demand in response to climate-driven productivity fluctuations, underscoring the complex interactions between agricultural and energy sectors. Together, these three studies enhance understanding of how soybean production systems respond to climatic variability, providing improved predictive methodologies, detailed assessments of historical yield impacts, and robust scenario analyses. The findings offer a scientific foundation for informed adaptation strategies, effective resource management, and resilience-building practices in the context of ongoing climate change.