An Automated Method for Annual Cropland Mapping along the Season for Various Globally-Distributed Agrosystems Using High Spatial and Temporal Resolution Time Series

摘要: Cropland mapping relies heavily on field data for algorithm calibration, making it, in many cases, applicable only at the campaign scale. While recently launched Sentinel-2 satellite will be able to deliver time series over large regions, it not really compatible with current approach or available situ data. This research introduces a generic methodology annual cropland along season high spatial resolution use of globally baseline land cover and no need The is based cropland-specific temporal features, which are cope diversity agricultural systems, prior information from mislabeled pixels have been removed cost-effective classifier. Thanks JECAM network, eight sites across world were selected global benchmarking. Accurate maps produced end season, showing an overall accuracy more than 85%. Early also obtained three-month intervals after beginning growing these showed reasonable stage (>70% accuracy) progressive improvement season. trimming-based method was found key using spatially coarse and, thus, avoiding costly campaigns retrieval. timeliness proposed shows that has substantial potential operational agriculture monitoring programs.