Extraction Of Early Paddy Rice Information In Nanchang County Of Jiangxi Province Based On Sentinel-1 Data

Optical remote sensing plays an important role in the field of agriculture monitoring,such as crop identification,crop yield estimation and drought monitoring.Owing to most rice growing in the rainy,hot and humid regions,it's difficult to retrieve rice growing areas accurately with only optical remote sensing data.In China,most rice distributes in the south where is warm and cloudy,so it is difficult to obtain clear and useful optical remote sensing images in the rice growing periods.For optical images often used in crop monitoring,such as MODIS and SPOT,there exist some drawbacks including low spatial resolution,mixed pixels and some other issues,which would limit their application on crop information extraction,especially for rice.However,for high spatial resolution images,e.g.,GF time resolution would be another problem;and the purchase of high spatial resolution images would increase the research costs.In view of the drawbacks of optical remote sensing in rice information extraction,in this work,we try to extract rice growing areas and farming process using SAR images with Sentinel-1 data.The Sentinel-1 satellites are composed of Sentinel-1A and Sentinel-1B.So far,Sentinel-1 data is the only high spatial resolution synthetic aperture radar data which is distributed regularly free.In this study,sentinel 1A data were used to extract early paddy rice information from Nanchang County,Jiangxi Province,including early rice area extraction and cultivation behavior detection.The method of extracting rice growing area combines Sentinel-1 different polarizations with rice phenology.Rice area extraction accuracy was verified using ZY-3 and field survey data.And early rice cultivation was detected using images which can highlight the occurrence of cultivated rice fields such as coherence coefficient and the average intensity map as well as their false color composite images.Main conclusions are as follows:(1)Rice has a unique underlying surface changing process with other ground features,which was the principle of rice field area extraction in this work.Filled with water,in the early growth stage,rice scattering intensity was relatively low.With rice growing,the backscattering coefficient increased steadily,mainly due to body scattering and canopy surface scattering.These are the characteristics that other crops do not have,and they are key to extract the distribution of rice fields.(2)The choice of images acquiring time and polarizations has an important influence in the rice field extraction results.With repetition cycle 12 days,there are more than one Sentinel-1A images in a certain rice growth period,and image polarizations are the factors determining images choice.VV polarization of Sentinel-1 reached its maximum in early May;while VH polarization,reached to the maximum in late May and remained stable until rice mature harvest,with a of downward after that.So we selected Sentinel-1 images from April 24,May 6 and June 11 with VV polarization,and selected images from April 24,May 18 and June 11 with VH polarization to make false color composite.Although the dates of the middle images of the two polarizations were different,they were both in the early rice tillering and jointing periods.(3)VV polarization of sentinel-1A was superior to VH polarization in the accuracy of area extraction.The response of VV polarization to rice was better than VH,and it was also closer to the real statistical data in the extraction area.(4)We used VV polarization data for coherence analysis.From the results,we can quickly know the occurrence of sowing behavior and the progress of each region in the county.The highlight of this work is extract rice field information combining rice key growing stages and different polarizations of Sentinel-1A SAR data.The results showed that VV polarization was better than VH in recognition of paddy field.And it is feasible to reflect farming process using false color composite method with the inverse coherence map and average intensity map.This study examined the practicality and feasibility of Sentinel-1A data in obtaining rice planting information.