Evaluating Multiple Sensors for Mapping Cropped Area of Smallholder Farms in the Eastern Indo-Gangetic Plains

remote sensing;crop areas;agriculture;India;School for Environment and Sustainability

Smith, HarrisonBergen, Kathleen ;

University of Michigan

关键词: remote sensing;    crop areas;    agriculture;    India;    School for Environment and Sustainability;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/149011/Smith_Harrison_Thesis.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Accurate, fine-scale agricultural statistics are critical for understanding trends in crop production throughout the world. In many areas of the world, however, on-the-ground crop area estimates may be difficult to acquire or are only present at state or national scales. In these areas, remote sensing can offer a cost-effective alternative for gathering fine-scale agricultural statistics. Many methods exist for mapping cropped area using remote sensing, but the majority of these are done using moderate-to-coarse spatial resolution sensors such as MODIS or Landsat. Though often finer in scale than state-level data, these sensors may not accurately estimate cropped area in smallholder systems, where a typical agricultural plot can be smaller than a single image pixel. The purpose of this study was to examine the tradeoffs of using four different sensors—MODIS, Landsat 8, Sentinel-2, and PlanetScope—for mapping cropped area in the eastern Indo-Gangetic Plains (IGP) region of India. We used NDVI time series imagery from each sensor to map cropped area for the 2017-2018 winter growing season, and assessed accuracy using classified maps created using random forest classification. We compared each sensor in terms of accuracy, data availability, and ease of use. We find that Sentinel-2 and PlanetScope both show increased accuracies compared to more commonly used sensors such as MODIS and Landsat 8. This indicates that coarse and even moderate resolution sensors, such as MODIS and Landsat 8, may not be sufficient for mapping fine-scale cropped area in smallholder systems. Our results highlight the importance of appropriate sensor selection when mapping cropped area in smallholder systems.