Study On Prediction Of Cultivated Land Quality Grade Of Paddy Field Based On GEE And Multi-source Remote Sensing

In order to meet the demand for rapid updating of cropland quality evaluation results in the new era,this study extracted features reflecting crop growth and development and cropland production potential based on Google Earth Engine cloud platform,ENVI and ArcGIS software platform,and es Tab.lished a random forest prediction model for rice cropland quality directly with the traditional comprehensive index of cropland quality.This study explored the spatial variability of the prediction results and the relationship between the distribution of paddy field quality classes and the measured rice yield under the traditional method,and provided technical guidance and data support for improving the quality of paddy fields,ensuring food security,and implementing precision agriculture and sustainable soil management.(1)The classification accuracy of rice field area extracted by random forest classifier in Feidong County in 2021 was high,with overall accuracy of 0.99,producer accuracy of0.99,user accuracy of 0.99 and Kappa coefficient of 0.98.Paddy fields were distributed in the north and middle of Feidong County,and the main land property of paddy fields was paddy fields,with an area of 72632.94 hm~2,accounting for 82.81%of the total land area of paddy fields;The main types of paddy soil were paddy soil and yellow cinnamon soil,with an area of 54469.07 hm~2and 29764.87 hm~2respectively,accounting for 62.10%and33.93%of the total area of paddy field respectively.(2)The prediction model of paddy land quality grade es Tab.lished by random forest has good fitting ability and high accuracy,with R~2of 0.818,RMSE of 0.016 and polynomial Y=1.4264X~2-1.6726X+1.2263.Among the predicted characteristic variables,the reflectance of NPP,REP,LSWI,NDVI and GF-6 was the most important.ST1-4 and SRC influenced the planting date of rice.Rice mature stage was affected by TPS,and the growth trend of LAILV corresponded with that of rice mature stage.(3)The prediction results were similar to the average quality grade of paddy land based on traditional methods,with a difference of only 0.15,and there were differences in the spatial distribution characteristics of paddy quality grade in Feidong County under the two evaluation methods.Compared with the results of traditional methods,the predicted area of high-yield fields and low-yield fields decreased significantly,with a decrease of2056.07hm~2in the first and second places,and a decrease of 2847.38hm~2in the ninth and tenth places,and the high-yield fields of the traditional rice fields were mainly distributed in the south and center,the middle-yield fields were mainly distributed in the north and west,and the low-yield fields were mainly distributed in the northeast,southeast and northwest.The predicted high-yield paddy fields were mainly distributed in the southern part of Feidong County,the middle-yield fields were mainly distributed in the north and east,and the low-yield fields were mainly distributed in the northeast and northwest.(4)The grades of cultivated land quality of paddy fields based on the traditional method and random forest prediction gradually increased with the decrease of the measured rice yield,indicating that the results of the two evaluation methods can reflect the rice production status to a certain extent.