Dados do Trabalho

Título

IMPACT OF CLIMATE CHANGE ON SUGARCANE PRODUCTIVITY IN A SOUTHEASTERN BRAZIL: A DIGITAL MAPPING AND AGRO-HYDROLOGICAL MODELING APPROACH.

Resumo

Soil is a natural capital that provides vital goods and services to terrestrial systems and human well-being. This capital is measured in terms of variability of soil properties (inherent or manageable). Biomass production is a key function linked to sustainable development goals (UN, 2015), including zero hunger, climate action, and life on Earth. Climate changes are demanding the adaptation of food systems. This research aims to map agricultural areas susceptible to loss of sugarcane productivity caused by climate changes (extreme events). Our hypothesis is that, based on soil hydraulic properties, it is possible to identify areas with higher climatic risk for agricultural production. The study was conducted in an area of 2,574 km2 in southwestern Brazil, São Paulo State. Soil samples were collected in 1752 sampling points located at rainfed sugarcane fields. We applied a data-driven approach using laboratory-determined soil attributes (clay, silt, sand, and organic matter), weather station data, pedotransfer functions, remote sensing derived products, machine learning, and a mechanistic model (SWAP) to estimate annual sugarcane production (tons of sugarcane per hectare; TCH) during the period 1988-2022. A mean TCH map was constructed in raster format using a Random Forest algorithm. This map had a 30 m spatial resolution and presented goodness-of-fit metrics of R2 0.54, RMSE 3.84 TCH, and CCC 0.60. The results showed that 60% and 36% of the area are susceptible to loss between 11-15 and 16-20 TCH, respectively. The risk of productivity loss occurs in the following soil classification order: Arenosol/Leptsol > Cambisol > Lixisol/Acrisol/Alisol > Ferralsol. Our approach provides detailed spatial information for site-specific management of soil and water resources, allowing stakeholders to implement integrated environmental, social, and governmental (ESG) based solutions, which could minimize the negative impact of climate change on food production.

Palavras-chave

Digital soil mapping; Machine learning; Soil functions; Climate change

Agradecimentos

We would like to thank the National Council for Scientific and Technological Development - Brazil (CNPq) for granting the scholarship to the first author (140288/2023-6), the São Paulo Research Foundation (FAPESP, Brazil) for granting Project number 2021/05129-8 and to the Geotechnologies on Soil Science group (GeoCIS, esalqgeocis.wixsite.com/english) for supporting this research.