Early mapping of canola in Jianghan Plain by integrating time series Sentinel-1/2 data
[Purpose]Jianghan Plain is the main crop production base in Hubei Province,and the accurate positioning of canola in the early stage is very important for crop growth monitoring and crop yield prediction.[Method]In this paper,with the help of Google Earth Engine(GEE)cloud platform,through the Random Forest(RF)model combined with phenological information,based on the remote sensing images from 1 October 2022 to 30 April 2023,the 10-day interval was used to form different fertility interval combinations,and the accuracy changes of each fertility interval combination with 10-day interval were observed and compared to find the Earliest Identifiable Timing(EIT)of canola.The early identification map of canola in Jianghan Plain was made.[Result](1)The addition of VV and VH polarization characteristics of radar images was more conducive to the identification of canola;compared with the research using only optical remote sensing data,the polarization characteristics of radar data increased the diversity and richness of data.(2)Based on Sentinel-1/2 data,the best period for early identification of canola obtained by RF classifier was from 1 to 11 December 2022,with an overall accuracy of 0.88 and an Fl score of 0.88,which can be identified and extracted five months before canola harvest.(3)The method proposed in this paper could realize large-scale early extraction and rapid mapping of canola.[Conclusion]The earliest identification time of canola based on the time series data integrated by Sentinel-1 and Sentinel-2 is five months before harvest.The comprehensive utilization of Sentinel-1 and Sentinel-2 data has a good effect in obtaining the earliest identification time and early crop mapping.This study can provide data support and scientific services for canola production management,agricultural planting structure adjustment and grain and oil security in the region.
NETL
NSTL
万方数据
