| Cultivated land has the function of food production,which is an important resource and material basis for human survival and development.At present,the decline of quantity and the degradation of quality has become a global problem.Huang-Huai-Hai agricultural area is the main producing area of winter wheat in China.Maintaining and improving the cultivated land quality has become an important basis for sustainable agricultural development and food security.It is an objective demand for winter wheat production and efficient utilization and management of cultivated land resources to construct the estimation models by using remote sensing data at different scales and grasp the cultivated land quality and its dynamic change rule quickly and accurately.In this paper,we took the Huang-Huai-Hai agricultural region as the research area,and divided into two scales(county and large area)and two region types(plain and hilly).Typical hilly counties(Tengzhou city)and plain counties(Shanghe county)were selected as the core areas.Then,we extracted wheat field information from MODIS and Landsat images,evaluated the quality based on the conventional GIS method,analyzed the scale characteristics of evaluation results and constructed the conversion models.On this basis,spectral indicators such as remote sensing band,vegetation index,and moisture index were selected to explore the spectral response characteristics of cultivated land quality by simulating time series curves and significance and correlation analysis.At the county scale,we established the variable groups by the construction and screening of spectral parameters,constructed and selected the inversion models by MLR,BPNN,SVM,and RF algorithms,and explored the inversion model and its accuracy based on the evaluation system conversion.At the large area scale,we constructed the cultivated land quality inversion models based on single-temporal MODIS image,multi-temporal MODIS image,mosaic Landsat image,and multi-source data fusion,respectively,and compared and analyzed the inversion accuracy.Furthermore,the research put forward the upscaling inversion method based on the ratio mean and the numerical regression methods,and explored the best inversion method of cultivated land quality at large-scale.According to the inversion results,the wheat field cultivated land quality and its spatio-temporal dynamic change were analyzed,and the suggestion of regionalization and its protection and management were put forward.The main conclusions are as follows:(1)The evaluation results by conventional GIS method in Tengzhou city,Shanghe county,and Huang-Huai-Hai agricultural region are mainly high and medium grades,and the spatial variability of cultivated land quality at different scales is significant.The conversion models of evaluation system from county-scale hilly area to county-scale plain area,from county-scale hilly area to large-scale hilly area,and from large-scale hilly area to large-scale plain area are y=1.021x-4.989,y=0.801x-16.925,and y=0.959x-3.458.(2)The time series curve of spectral indicators of winter wheat has significant peak and valley changes,and the same type has similar curve changes.Three kinds of spectral indicators have significant spectral responses to the cultivated land quality from the regreening to the filling stage(P<0.01),and reach the strongest correlation at the jointing-booting stage(April 15 to April 22).(3)At the county-scale,the correlation between spectral indicators and cultivated land quality is as follows:combination index>vegetation index>moisture index>remote sensing band,the model accuracy is as follows:RF>BPNN>SVM>MLR,and the correlation is stronger and the model accuracy is higher in hilly areas.The RF model based on the combined spectral index at county-scale hilly area has good universality,with the modeling set:r~2=0.788,rmse=4.278 and the validation set:r~2=0.780,rmse=4.471.Meanwhile,it is suitable for the prediction of cultivated land quality at county-scale plain area after the evaluation system conversion,so it is the best inversion model.The prediction accuracy of inversion models based on single-temporal MODIS image and mosaic Landsat image at large-scale is relatively low.The correlation between spectral indicators constructed based on the multi-temporal MODIS image and cultivated land quality is increased to 0.461～0.637,and the model has r~2>0.523,rmse<3.099.Furthermore,the inversion model of multi-source data fusion has r~2>0.535 and rmse<3.062,which can effectively improve the model inversion accuracy.(4)In the upscaling inversion,the reflectance conversion coefficients of green,red,near-infrared,shortwave infrared 1,and 2 bands based on the ratio mean method are 0.543,0.561,1.097,0.781,and 0.788,respectively.And the reflectance conversion model of shortwave infrared 2 is a cubic model,and the rest are quadratic models based on the numerical regression method.The upscaling inversion results based on the numerical regression method are consistent with the spatial distribution of the evaluation results,and the consistency of inversion area is as high as 96.03%,which is the best inversion method for cultivated land quality at large-scale.(5)From 2001 to 2018,the cultivated area of wheat fields in the Huang-Huai-Hai region was basically stable,with a dynamic degree of-5%～5%.The cultivated land quality showed a steady upward trend.The average quality grades in 2001,2010,and 2018 are 4.011,3.902,and 3.761,respectively,and the area of high and medium grade land is increasing.Based on the spatial polarization theory,five aggregation areas of H-H,L-L,H-L,L-H,and N-N were divided,and corresponding cultivated land quality management and protection measures were proposed to provide a reference for cultivated land protection and utilization management.This study used multi-source remote sensing data to construct inversion models at different scales in the Huang-Huai-Hai agricultural region,and proposed an upscaling inversion method for the wheat field cultivated land quality.It provides a new technical means for quickly and accurately obtaining the wheat field cultivated land quality and has positive significance for the sustainable development of regional agriculture. |
PDF Full Text Request