| Crop planting pattern information should be grasped promptly and accurately because it is strategically significant to ensure sustainable agriculture,assess regional resource capacity and guarantee national food.Characterized by large-scale surfaces detection and short cycle,remote sensing technology is a key tool to monitor regional agricultural conditions.It is significant to study how the remote sensing technology aids in identifying and monitoring complex crop types and planting patterns.The southern agricultural areas of China are dispersed lands,diverse crop types,and complex planting patterns.Under the background,general methods should be explored to effectively and accurately identify crop planting patterns,so as to produce spatial datasets in large scope and with high precision.These datasets are particularly important for understanding agricultural conditions.However,precise identification of crop planting patterns is prone to many factors such as failure in taking into account space and time resolution of image data synchronously,difficulty to obtain lots of field sample data in real time,complex and changeable agricultural conditions,and subjective influence of farmers.In order to overcome the limitations,this study mainly explores Jianghan Plain where crop types and planting patterns are complex.It integrates Landsat-8 and Sentinel-2 time series images from 2017 to 2021 via Google Earth Engine platform.Based on field sampling data,sample expansion and temporal-spatial migration strategies are proposed.Moreover,three classification models are classified in view of field samples and expanded sample sets.Optimal model-sample combination is selected to successfully extract crop planting patterns in Jianghan Plain during 2017-2021,obtaining a spatial data set with 10-m resolution.Considering identification results,this study investigates proportion of area,temporal-spatial distribution,changing features and driving factors of main crop planting patterns.In the end,conclusions are drawn as follows.(1)Random Forest model combined with sample expansion strategy is the best model-sample combination suitable for revealing realities of research area.At the same time,it shows that sample expansion and temporal-spatial migration strategy are feasible in raising classification accuracy.The classification accuracy results show that when small number of field sample points are classified,gcForest model has the highest classification accuracy and the highest Kapp coefficient(81.00% and 0.77).Therefore,it can help achieve more precise classification effects and best accuracy.When expanded sample set is classified,classification accuracy of gcForest,RF and DNN is improved respectively.RF performs the best in classification accuracy and drawing effect and it even exceeds gcForest with the highest classification accuracy of field samples,with overall accuracy of 90.54% and Kappa coefficient of 0.91.It is proved that based on Sentinel-2 and Landsat-8 data,random forest model and sample expansion migration strategy are more applicable to identify complex crop planting patterns in Southern agricultural region.(2)Through classification identification and extraction by suitable method,this study acquires space distribution information and area change of main crop planting patterns of Jianghan Plain.There are 9 planting patterns in the region,including wheatrice,wheat-cotton,wheat-soybean,wheat-corn,rape-rice,rape-cotton,rape-corn,rapesoybeans,and single rice.Analysis is conducted on five planting patterns with large planting area and stable variation: During 2017-2021,single rice is the superior planting pattern covering the largest area in study area.Average planting area is about 492,050hectares;it increases year by year and is widely distributed in all areas of Jianghan Plain.Rape-rice ranks second in terms of planting area,with average area of 173,630 hectares.Planting area is in a trend of rise first and then continuous decline,while such planting pattern mainly gathers in the Yangtze River and Han River coasts in the middle of Jianghan Plain.Regarding wheat-cotton pattern,planting area declines stably and gradually,and average planting area is 68,690 hectares,which is planted mainly in wheat area in the north of study area.Average planting area of rape-corn is 61,130 hectares,and it experiences stable fluctuation.Such planting pattern is decentralized.The wheat-soybean planting pattern covers an area of 78,420 hectares and is mainly distributed near Tianmen City in the north,with planting area rising year by year.(3)The main planting patterns in Jianghan Plain demonstrate obvious spatial distribution;space correlation and agglomeration features are significant.According to standard deviational ellipse and trend analysis,spatial distribution of main crops is reflected as follows: within a few years,five planting patterns are distributed in southwest-northeast trend in space,and planting center locates near the Qianjiang in middle of study area.Except for a few planting patterns,most of planting patterns are distributed high in middle and low in both sides along the north-south direction and east-west direction(rape-corn is planted high in west and low in east along east-west direction,while single rice is cultivated high in the south and low in the north along north-south direction).In addition,spatial correlation and agglomeration effect are significant among various planting patterns.During 2017-2021,Moran’s I index of various planting patterns is above 0.30,indicating significant spatial positive correlation and spatial agglomeration effect.(4)Planting patterns have different areas,directional features and distribution laws in time and space,which attribute to influence of nature,policy,social economy,and other driving effects.Among natural factors,crops planting and their spatial distribution are greatly impacted by water network density,average annual rainfall,and average annual temperature.When factors interact in pairs,they will produce greater effect than an independent factor.Provided that all conditions permit,farmers will give priority to planting crops with higher economic income. |
PDF Full Text Request