Remote sensing and machine learning techniques used to predict sugarcane (Saccharum spp.) yield

Resumen

Objective: To gather information generated using remote sensing and machine learning models. These tools are already used in decision-making processes to support sugarcane yield prediction.

Design/Methodology/Approach: This study includes a systematic review of scientific literature. Different indicators used in studies about the prediction sugarcane yield were extracted and synthesized.

Results: This review retrieved 386 relevant studies from five electronic databases. Subsequently, using exclusion and selection criteria, 47 studies were selected for in-depth analysis. According to the analyses, the most frequently used variables were climatic variables (temperature, precipitation, and evapotranspiration) and crop variables (number of harvests). The most commonly used algorithms were random forests and multiple linear regression.

Study Limitations/Implications: The limitations are included in the in-depth analysis of the studies. These studies have great potential for further research. In this case, the analysis was limited to remote sensing and machine learning.

Findings/Conclusions: Satellites, such as Sentinel 2 and Landsat 8, are commonly used in remote sensing methods. The most frequently used vegetation indices include: Normalized Difference Vegetation Index (NDVI), Green Normalized Difference Vegetation Index (GNDVI), Soil-Adjusted Vegetation Index (SAVI), and Enhanced Vegetation Index (EVI). The use of spectral bands, such as near-infrared (NIR) and shortwave infrared (SWIR), was recorded.