Changing Sugarcane Yield Estimation In São Paulo

Overview: Brazil’s Sugarcane Market

Brazil’s sugarcane industry is a global powerhouse, producing vast quantities of sugar and ethanol that dominate international markets. In the 2023/24 marketing year, Brazil achieved record-breaking sugarcane production of 705 million metric tons (MMT), fueled by favorable weather conditions, significant investments in field renovation, and yield-improving technologies ^{Brazil: Sugar Annual. (2024). USDA, Foreign Agricultural Service.}.

However, projections for 2024/25 signal a downturn, with production expected to drop by 8.5 percent to 645 MMT due to unusually dry weather early in the planting season. Despite this, sugar and ethanol production are forecast to remain steady, buoyed by uncrushed stocks from the previous cycle. São Paulo leads the nation in production, contributing roughly half of the total output, followed by Minas Gerais and Goiás.

Beyond its sheer scale, sugarcane plays a vital role in Brazil’s economy, representing over 10 % of the national agricultural production value and contributing significantly to global supply chains ^{Ozbun, T. (Apr 29, 2024). Production of sugar cane in Brazil from crop year 2010/11 to 2024/25. Statista.}. In 2022, Brazil reclaimed its title as the world’s largest sugar producer, surpassing India and exporting an impressive 28 million metric tons ^{Davis, M., Dizon F. F., Rocha, L. B., & Sgarbi, F. (2023). Pathways for Brazil to sustainably transform its sugar industry. World Bank.}.

Ethanol production, though slightly declining in recent years, remains critical as a renewable energy source and an alternative to fossil fuels, reinforcing Brazil’s leadership in biofuel innovation. The sector’s impact extends beyond exports and energy, employing 8% of the national agribusiness labor force and driving economic development in rural regions. As global demand for sugar and bioethanol evolves, Brazil’s sugarcane industry remains a cornerstone of sustainable agricultural and energy transitions.

Challenge: Accurate Prediction Of Sugarcane Yields

Brazil has long been synonymous with agricultural excellence, and sugarcane is one of its crown jewels. As a critical crop for sugar and ethanol production, sugarcane plays a significant role in domestic and global markets. However, growers in São Paulo, one of Brazil’s leading sugarcane-producing regions, face ongoing challenges in yield prediction.

Traditional yield estimation methods have struggled to keep pace with the dynamic factors influencing sugarcane growth. São Paulo’s climate is as varied as it is unpredictable, with weather patterns that can dramatically affect crop cycles. The region’s diverse soils — from highly fertile patches to areas with poor water retention — add another layer of complexity. Farmers and agricultural stakeholders also grapple with inconsistent data sources, as historical yield records are often aggregated at the county level. These limitations make achieving the field-level precision needed to optimize resource allocation and improve productivity is difficult.

For EOSDA, the challenge was clear. Could the advanced yield estimation model accurately predict sugarcane yields in São Paulo’s dynamic environment? And, just as importantly, could this model offer actionable insights to make life easier for farmers while being scalable for broader application?

These questions framed the 2023 project, with the company focusing on a 110×110 km area of interest (AOI) in eastern São Paulo. The team’s goal was to validate their model’s accuracy and identify the potential for improvement and set the stage for more extensive deployment.

We chose this specific area of interest since São Paulo is a significant state from an agricultural point of view, and it grows big volumes of the target crop.

Oleksii Neskorovnyi

Business Analyst at EOS Data Analytics

Solution: Open-Data Sources

To address these challenges, EOSDA adopted a comprehensive and data-driven approach. At the core of their solution was the WOFOST model, a process-based crop simulation tool known for its adaptability and precision. WOFOST simulates crop growth by focusing on key physiological processes, including light interception, CO2 assimilation, and water balance.

However, sugarcane’s unique growth cycle required the EOSDA team to modify the model. They focused solely on the vegetative growth stage, as the reproductive stage — associated with flowering — reduces sucrose levels in the stalk, making it less relevant for sugarcane farming.

Calibrating the model was a meticulous process. The team used six years of historical yield data from the Global Yield Gap Atlas, focusing on data from São Simão, Sorocaba, and São Sebastião do Paraíso, three stations near the AOI. Calibration involved tweaking crop parameters like temperature thresholds, leaf growth rates, and moisture sensitivity until the model’s predictions aligned with actual observations. This iterative process, while time-consuming, was essential for building a model tailored to São Paulo’s specific conditions.

Data integration was another cornerstone of the solution. Daily weather data from NASA POWER provided the necessary insights into São Paulo’s climate variability. Soil properties were mapped using IRSIC SoilGrids, which offered detailed information on water retention and physical characteristics. Publicly available sugarcane crop calendars helped define optimal planting and harvesting periods, ensuring the model reflected real-world agricultural practices.

NASA POWER, in general, is a fairly common provider of free weather data that many people use in modeling for large-scale projects.

Oleksii Neskorovnyi

Business Analyst at EOS Data Analytics

The project also incorporated advanced geospatial analysis. EOSDA used raster-based crop classification data and vectorized field boundaries to map sugarcane fields rapidly. This enabled the team to calculate yields and production volumes for each field by multiplying the estimated yield per hectare by the field’s total area.

A significant aspect of the solution was the focus on visualization and reporting. The EOSDA’s science team delivered results as interactive vector layers and maps, making it easy for stakeholders to understand and act on the data. These visualizations highlighted sugarcane yields in tons per hectare and total production in tons, offering a clear picture of agricultural performance across the AOI.

Outcome: Visualization And Predicted Yields

The project results were a testament to the power of advanced crop analytics and EOSDA’s innovative approach. Despite the limitations of aggregated input data, the WOFOST model delivered highly accurate yield predictions. These estimates were validated against historical data, confirming that the model could account for the complexities of São Paulo’s environment.

For farmers and agricultural stakeholders, the benefits were clear. The field-level precision of the model meant they could identify high-performing areas and focus resources on fields that needed intervention. This granular approach to yield estimation promises to improve decision-making, reduce waste, and enhance overall productivity.

Beyond immediate results, the project demonstrated the scalability of EOSDA’s methodology. The model’s reliance on open-source data and its efficiency and adaptability made it a cost-effective solution for large-scale agricultural planning. São Paulo’s AOI was just the beginning. The same approach could be applied to other Brazil or global regions.

However, the project also highlighted opportunities for improvement. More granular historical data, such as field-specific yield records, could enhance the model’s accuracy. Incorporating higher-resolution weather data and satellite-derived vegetation indices like NDVI or LAI could add predictive power. Refining soil data inputs could improve water balance calculations, making the model even more robust.

Ultimately, EOSDA’s work in São Paulo was more than a technical success. It was a proof of concept that showed how advanced technology could bridge the gap between theoretical models and practical agricultural needs. Combining cutting-edge science with a deep understanding of local challenges, EOS Data Analytics laid the groundwork for a new era of precision farming in Brazil.

Through this project, EOSDA validated its yield estimation model and set a new standard for what’s possible in agricultural analytics. The insights gained in São Paulo pave the way for more intelligent, more sustainable farming practices that promise to benefit farmers, stakeholders, and the planet alike.