Novel Approach to Identify Coffee Cultivation Using Machine Learning

Strategic Insights -

Coffee demand continues to rise globally, and producing countries face challenges due to climate change, resulting in yield losses. Researchers from the State University of Campinas (UNICAMP) have developed a novel approach to identify coffee cultivation using machine learning. This innovative method proposes to identify coffee cultivation considering four phenological stages and employs a hierarchical classification framework to isolate coffee pixels and identify their respective stages in one of Brazil's most important coffee-producing regions. The study utilized a dense time series of multispectral bands, spectral indices, and texture metrics derived from Harmonized Landsat Sentinel-2 (HLS) imagery, combined with an ensemble learning approach based on decision-tree algorithms.

Key Takeaways:

  • The proposed approach achieved unprecedented sensitivity and specificity for coffee plantation detection, consistently exceeding 95% with Random Forest (RF).
  • The classification of coffee phenological stages showed balanced accuracies of 77% (ST) and from 93% to 95% for the other classes.
  • The novel method provides a scalable framework to monitor climate-resilient coffee management practices.
  • The study utilized a dense time series of multispectral bands, spectral indices, and texture metrics derived from Harmonized Landsat Sentinel-2 (HLS) imagery with an average revisit time of ~3 days.
  • The research employed an ensemble learning approach based on decision-tree algorithms, specifically Random Forest (RF) and Extreme Gradient Boosting (XGBoost).
  • Taya Cristo Parreiras, Claudinei de Oliveira Santos, Edson Luis Bolfe, Edson Eyji Sano, and other researchers contributed to the study.

Statistics:

  • The accuracy of coffee plantation detection exceeded 95% with Random Forest (RF).
  • The classification of coffee phenological stages showed balanced accuracies of 77% (ST) and from 93% to 95% for the other classes.
  • The study utilized a dense time series of multispectral bands, spectral indices, and texture metrics derived from Harmonized Landsat Sentinel-2 (HLS) imagery with an average revisit time of ~3 days.

Sources:

  • Dense Time Series of Harmonized Landsat Sentinel-2 and Ensemble Machine Learning to Map Coffee Production Stages. Remote Sensing, 2025,17(18):3168. (Remote Sensing - http://www.mdpi.com/journal/remotesensing/).
  • MDPI AG, publisher of Remote Sensing.
  • State University of Campinas (UNICAMP), Graduate Programme in Geography, Taya Cristo Parreiras, and other researchers.
  • National Council For Scientific And Technological Development (Cnpq)/research Productivity Fellowship.