Extraction Of Cropping Patterns Based On Multi-source Remote Sensing Imagery

Crop planting structure is a combination of crop-related agricultural information in a region or an agricultural production unit,mainly including the spatial distribution of cropland,types of croplands,types of crops,crop rotation patterns and crop maturity systems.Accurate and timely information on crop cultivation structure can provide an important theoretical basis for crop quality and yield assessment,food security,agricultural risk assessment and rational planning of land resources.The rapid development of remote sensing technology has provided powerful data support for agricultural mapping,and remote sensing technology is gradually being widely used in the field of agricultural information extraction by virtue of its large scope,timeliness and rapid data collection.However,the fragmented plots of cropland,the variety of crop types and complex planting patterns make accurate remote sensing extraction of crop planting structure information still challenging.The thesis takes Shandong Province as the research area,based on time series Sentinel-2 satellite images,field sampling data,topographic data,Google Earth images,agricultural statistics and other multi-source data,with the support of Google Earth Engine(GEE)cloud platform to extract the crop planting structure information,the main research methods and results are as follows:(1)This study proposed a method for identifying the type of cropland under rules constraints.Using filtered and pre-processed image data,a feature space is constructed,and based on rule constraints,the object-oriented image segmentation results are optimized against the pixel-based classification results of random forests to finely extract the secondary classes of cropland and obtain a high-precision cropland type map.The Detailed Cropland Map of Shandong Province shows that Shandong has a large proportion of cropland and there are clear regional differences.There are strong east-west differences in the distribution of different cropland types in Shandong.The western inland areas are generally irrigated croplands,rainfed croplands is concentrated in the hilly areas in the center and south-west of Shandong,and the paddy fields are mainly located in the southwest of Jining,the south of Linyi and the north of Dongying,all of them are next to rivers or lakes.(2)This study proposed a sub-seasonal method for crop information extraction.The method designed separate identification models based on the different growing seasons of crops and selected features separately according to the crop characteristics to ensure the accuracy of crop mapping.The crop rotation map was then mapped based on spatial analysis of the crop maps and research surveys.The results showed that the most areas in Shandong use crop rotation pattern to grown two crops a year,with wheat in the first growing season and maize in the second being the main crop rotation pattern.Large areas of wheat-maize are concentrated in the northwestern plain,the south-western plain,and the Jiaolai plain of Shandong.Rice is mostly planted after the wheat harvest,forming a wheat-rice crop rotation.Peanut is mostly in singleseason cropping areas,scattered in the hilly areas of the south-central region.Cotton is grown as the second season crop in rotation with other minor crops.In addition,features such as LSWI,NDVI,SAVI,red-edge bands and elevation of the peak crop growing season have important contributions in crop type identification.(3)On the basis of accurate extraction of cropping planting structure,this thesis combines the qualitative analysis method of Multiple Correspondence Analysis(MCA)with quantitative analysis based on spatial area statistics to investigate the relationship between the various elements of crop planting structure and the influencing factors.The analysis found that topography has a significant influence on the cropping planting structure,with low and gently varying terrain being more likely to be irrigated croplands,and more likely to produce two crops a year,such as wheat-corn rotation.Higher terrain with more undulating surfaces is more likely to be rainfed croplands,with a single crop,such as peanuts.There is also a correlation between the cropping planting structure and annual precipitation.Areas in Shandong with high precipitation tend to be rainfed croplands,which can meet the water needs of single-season crops.Areas with less precipitation are more likely to be irrigated croplands,where two crops a year are grown and additional irrigation activities are required to ensure crop growth.