Extraction And Analysis Of Rice Planting System Based On Phenological Characteristics And Random Forest

Rice is one of the most important food crops in the world,and the rice yield in China ranks first in the world.However,with the continuous development of urbanization and industrialization,a large number of cultivated land planting systems have been changed from double to single in the main rice producing areas in southern China,and the total area of rice planting has decreased year after year.Therefore,monitoring the temporal and spatial changes of rice planting system in real time is of great significance for ensuring national food security and understanding regional economy.Remote sensing has the characteristics of wide coverage and strong timeliness.Using remote sensing technology,the temporal and spatial changes of rice planting system can be analyzed quickly and widely.In this paper,Hunan province was selected as the experimental area,and Sentinel-2 remote sensing image data from 2018 to 2020 were collected.based on Google Earth Engine(GEE)platform,the phenological statistical characteristic parameters(mean,standard deviation,skewness,kurtosis and coefficient of variation)based on rice NDVI sequence curve were extracted,and the remote sensing classification model of rice planting system was established by using random forest algorithm.the research work and main conclusions of this paper are as follows:(1)With the phenological statistical characteristic parameters of rice NDVI sequence curve as input parameters,the remote sensing classification model of rice planting system established by using random forest algorithm has high accuracy,with the overall accuracy ranging from 91% to93% and kappa coefficient ranging from 0.87 to 0.89.(2)In the three years from 2018 to 2020,the sown area of rice in Hunan Province decreased year by year.The percentage of double-cropping rice planting area decreased from 49.14% to38.44%.The planting area of single-season rice increased from 46.44% to 57.03%.The abandoned land increased from 4.42% to 7.46%.Double-cropping rice in Hunan Province is mainly distributed in Dongting Lake plain in the north of Hunan Province,plain area in Xiangjiang River basin in the middle and middle area of southern mountains(such as Yuanjiang County,Xiangtan County and Guiyang County).Single-season rice is mainly distributed in the periphery of Dongting Lake double-season rice area and the hilly area in the middle and east(such as Changde City,Yiyang City and Ningxiang County),and the rest areas are mostly mixed planting of singleseason rice;The abandoned land is mostly distributed in the hilly areas in the north-central and eastern parts of Hunan Province(such as Liuyang City,Lixian County and Youxian County).(3)The distribution of rice planting system is influenced by many factors such as topography,climate,soil,population,social economy and policy.In Hunan Province,60% of rice planting is distributed below 100 meters in elevation,and more than 45% is distributed in areas with slope of0-2.The proportion of single-season rice and abandoned land increased with elevation,while the proportion of double-season rice decreased with elevation.Double-cropping rice accounts for the largest proportion in paddy soil area,single-cropping rice accounts for the largest proportion in red soil area,while yellow soil area has the highest abandonment rate.In areas with high accumulated temperature and rainfall,the proportion of double-cropping rice increased,while the area of single-cropping rice and abandoned land decreased.In terms of socio-economic factors,the increase of floating population will result in the decrease of rice planting area,and the increase of output value of primary industry has the greatest influence on the change from single-season rice to double-season rice.To sum up,the remote sensing classification model of rice planting system based on phenological statistical characteristic parameters of rice NDVI sequence curve is established by using random forest algorithm,which can effectively monitor the change of rice planting system and provide a new idea and method for large-scale extraction of rice planting system.