County-scale Production Capacity Evaluation Of Farmland Using Long Time Series Of Remote Sensing Images

Farmland is an indispensable and important resource in human survival and social development,as well as a basic element in determining food production capacity.Objective and accurate evaluation of farmland production capacity is important for optimizing planting structure,improving low yield farmland and sustainable use of farmland.The purpose of this study is to investigate the quantitative assessment method of farmland production capacity at county scale based on long time series remote sensing images,and to explore the response law of farmland production capacity with soil texture and soil organic matter.Winter wheat was selected as the benchmark crop for farmland production capacity assessment,and the spatial distribution of winter wheat from 2009-2019(excluding2010)was mapped using medium-resolution satellite images.A multiple linear regression algorithm and multi-temporal remote sensing NDVI data were used to establish a winter wheat yield estimation model,and a stay-one cross-validation method was applied to evaluate the accuracy of winter wheat yield estimation.The mean and standard deviation(SD)of multi-year winter wheat yields were used to characterize the average farmland production capacity and its stability,using arable farmland plots as the basic unit.A double-threshold classification strategy conforming to the normal distribution in statistical principles was used to classify the farmland production capacity and instability at the county level.Then the combination of soil organic matter solvent sensing features was screened by correlating the soil organic matter content with the reflectivity of each waveband of remote sensing images,and construction and selection of a remote sensing inversion model for soil organic matter in farmland using random forest and linear regression methods for comparison;the response pattern of soil organic matter and farmland production capacity was analyzed plot by plot.And the spatial pattern pattern analysis of the farmland production capacity level was conducted from the perspective of soil texture and DEM.The results of the study are as follows:(1)The achievements of winter wheat withdrawn by SVM Category Methodology showed that the overall accuracy of the classified winter wheat from 2009-2019(excluding2010)was between 87.49% and 94.85%,and the kappa coefficient was in the range of0.80-0.90,which had high classification accuracy.The extracted winter wheat can be used as the base data for subsequent processing.(2)The multiple regression model was used to construct the remote sensing yields evaluation patterns for winter wheat year by year,and the accuracy was verified by the leave-one-out cross-validation method.The results showed that the yield estimation model had good stability and predictive ability.(3)The farmland production capacity was classified using a two-threshold classification strategy conforming to a normal distribution,and it was obtained that the farmland production capacity in the south and north of the study area was relatively high and stable,while the farmland production capacity in the central part was low and stable.(4)Using random forest method to build multi-band organic matter inversion model works best,and its accuracy and stability are better than linear regression model.The findings of remote sensing reversed soil organic matter content in the research region show that the farmland plot with high soil organic matter content is mainly concentrated in the north and south,which is generally higher than 25g/kg,while the soil organic matter content in the central and southeastern areas of the study area is generally lower,generally lower than25g/kg.Analysis of soil organic matter and cropland capacity revealed a significant positive correlation,with a Pearson correlation coefficient of 0.67(P < 0.01).The results showed that the spatial distribution of farmland production capacity was highly consistent with soil organic matter.(5)Combined with the DEM and soil texture analysis,the areas with high capacity accounted for the largest DEM range between 53-109 m,and their soil texture sand content was mainly between 40%-50% and clay content was mainly between 16%-21%,farmland production capacity has a good response relationship with soil texture.