Monitoring of Sugarcane Harvest in Brazil Based on Optical and SAR Data

• Published in: Remote sensing (Basel, Switzerland) Vol. 12; no. 24; p. 4080
• Main Authors: Kavats, Olena, Khramov, Dmitriy, Sergieieva, Kateryna, Vasyliev, Volodymyr
• Format: Journal Article
• Language: English
• Published: MDPI AG 13.12.2020
• Subjects: algorithms; area; Brazil; fields; harvest; harvest date; monitoring; NDVI; normalized difference vegetation index; remote sensing; Sentinel-1; Sentinel-2; SLC; sugarcane; synthetic aperture radar; time series analysis; Brazil
• ISSN: 2072-4292; 2072-4292
• DOI: 10.3390/rs12244080

The algorithms for determining sugarcane harvest dates are proposed; the algorithms allow the ability to monitor large areas and are based on the publicly available Synthetic Aperture Radar (SAR) and optical satellite data. Algorithm 1 uses the NDVI (Normalized Difference Vegetation Index) time series derived from Sentinel-2 data. Sharp and continuous decrease in the NDVI values is the main sign of sugarcane harvest. The NDVI time series allows the ability to determine most harvest dates. The best estimates of the sugarcane areas harvested per month have been obtained from March to August 2018 when cloudy pixel percentage is less than 45% of the image area. Algorithm 2 of the harvest monitoring uses the coherence time series derived from Sentinel-1 Single Look Complex (SLC) images and optical satellite data. Low coherence, demonstrating sharp growth upon the harvest completion, corresponds to the harvest period. The NDVI time series trends were used to refine the algorithm. It is supposed that the descending NDVI trend corresponds to harvest. The algorithms were used to identify the harvest dates and calculate the harvested areas of the reference sample of 574 sugarcane parcels with a total area of 3745 ha in the state of São Paulo, Brazil. The harvested areas identified by visual interpretation coincide with the optical-data algorithm (algorithm 1) by 97%; the coincidence with the algorithm based on SAR and optical data (algorithm 2) is 90%. The main practical applications of the algorithms are harvest monitoring and identification of the harvested fields to estimate the harvested area.