Distribution Extraction And Spatiotemporal Changes Analysis Of Winter Wheat In Shandong Province Based On GEE Platform

Wheat is the main cerel crop in the world,and also one of the three major food crops in China.It plays an important role in food supply and its planting area is second only to rice in China.Timely and accurate monitoring of winter wheat planting area and spatial distribution is of great significance for food security evaluation and agricultural policy-making.However,due to the impact of cloud and rain weather and the revisit cycle of satellite platform,using finer spatial resolution data to extract the planting area of crops in large area is chanllenging.Here Landsat and sentinel-2 data were used to carry out winter wheat distribution extraction and spatial-temporal change analysed based on GEE(Google Earth engine)cloud computing platform in Shandong Province.Firstly,six phenological median composites of Landsat-8 OLI and Sentinel-2 MSI reflectance measures were generated by a temporal aggregation technique according to the winter wheat phenological calendar,which covered seedling,tillering,over-wintering,reviving,jointing-heading and maturing phases,respectively,then,Random Forest(RF)classifier was used to classify multi-temporal composites,and we compared classification accuracies when mono-sensor and multi-sensors’data,RF,Support Vector Machine(SVM)and Classification And Regression Tree(CART)were classified.Secondly,the six phenological median composite constructed by intergrating Landsat-8 and Sentinel-2 data are randomly combined,the random forest classifier was used to implement classification,and the classification accuracies of different image combinations were compared to determine the optimal temporal window and earliest identible timing to identify winter wheat using remote sensing data.Lastly,based on the Landsat and Sentinel-2 data from nearly 30 years,the distribution of winter wheat planting areas in Shandong Province from 1990 to 2018 was extracted,and its temporal and spatial variation characteristcs were analyzed.The main conclusions of the study were as follows:(1)The method of combining the RF classifier with the combination of temporally aggregated Landsat-8 and Sentinel-2 data based on the GEE platform could accurately and effectively extract winter wheat in a large area.Temporally aggregated Landsat-8and Sentinel-2 data provide abundant observations to derive robust phenological based indicators for identifying winter wheat compared to used Landsat-8 and Sentinel-2 data alone.Compared with RF、SVM and CART classifiers,RF classifier had the highest classification accuracy,CART is the lowest,but it took the shortest time.The classification accuracy of SVM is between RF and CART and SVM took the longest time.(2)The classification accuracies of image combinations from different growth phases of winter wheat were quite different,the overall accuracies(OA)were between76.4%and 93.4%,and the kappa coefficients were between70.2%and 91.7%.The classification accuracy of mono temporal image from tillering phase was the lowest,and all six temporal images participated in the classification had the best accuracy.As the number of temporal images increased,the classification accuracy was higher,and the best classification accuracy gradually converges after three temporal images.In all three temporal image combinations,the images combination from overwintering,reviving and maturing phases of winter wheat had the best classification accuracy,with an OA of 92.7%and kappa coefficient of 90.8%.(3)Using images before the reviving phase of winter wheat could generate a good classification results.The best image combination before reviving phase was the images from the seedling,tillering and reviving phases,with an OA of 90.7%and a kappa coefficient of 88.3%.Compared with all six-temporal images combination classification,the OA was only reduced by 2.7%,and the kappa coefficient was reduced by 3.4%,but the planting distribution information of winter wheat could be obtained two months before harvest.(4)The distribution maps of winter wheat plating area in Shandong Province from1990 to 2018 generated by Landsat and Sentinel-2 data with OA ranging from 90.7%to 93.4%,and the kappa coefficient between 88.3%and 91.7%,which correlated well with official agricultural statistics records for areas(P<0.05).Winter wheat in Shandong Province declined across the province from 1990 to 2005 with an average annual decrease of 738.89 km~2.After 2005,the area of winter wheat gradually increased,and the main growth area was in the western and Yellow Triangle areas of the province,with an average annual increase of 488.53 km~2 from2005 to 2018.In the past 30 years,winter wheat has been steadily distributed in the western plains and the southern part of the central mountainous area in Shandong Province.The low-frequency cropping areas are mainly distributed around cities and hilly areas.