Remote Sensing Inversion Of Soil Organic Matter In Cultivated Land With Different Terrains In Shandong Province

As an important component of soil,soil organic matter（SOM）is an important index to measure soil fertility and evaluate the quality of cultivated land.Analyzing the spatial distribution of SOM on the provincial scale is helpful for the national Department of natural resources to carry out the macro-layout and planning of regional agricultural development.It has important guiding significance for sustainable utilization of soil and national food security and stability.Traditional SOM survey acquisition methods consume time and energy,and it is difficult to extract the spatial distribution information in a large range.Multi-spectral remote sensing has become an effective way to quantitatively predict SOM in large areas.However,from the perspective of large-scale,topographic conditions are complex,soil heterogeneity is strong,and topographic factors affect the distribution of surface SOM content.The characteristic topographic factors of different topographic units affecting the spatial heterogeneity of SOM were investigated,and SOM content was inverted by combining spectral data and topographic factors,in order to improve the prediction accuracy of zoning modeling,and open up a new idea for the high-precision prediction of SOM in large areas.This paper takes SOM in cultivated land as the research object,and with the support of Google Earth Engine（GEE）platform,based on the Sentinel 2A Multi-Spectral Instrument（MSI）and Landsat 8 Operational Land Imager（OLI）,through the optimization and comparison of response bands,spectral parameters and spectral model construction and optimization,the optimal model is determined to achieve image optimization.Then,using the selected images as data sources,the three topographic regions of Shandong Province,namely plain,mountain and hill,were selected according to the topographic conditions,and SOM spectral characteristics and spectral indices of cultivated land with different topographic conditions were compared and analyzed,and spectral models were constructed.Finally,topographic factors were introduced to explore the characteristic topographic factors of SOM in different topographic areas,and a quantitative inversion model combining spectral data and topographic factors was constructed.Compared with the spectral model,the feasibility of introducing topographic factors for regional inversion of different topographic areas was explored,and the accuracy of SOM inversion in large and complex topographic areas was improved to provide data support for accurately mastering the distribution of regional SOM.The specific research content and results are as follows:（1）Optimization of multi-spectral image data sources for remote sensing inversion of SOMTaking Southern Shandong Province as the research area,SOM sensitive bands were selected based on MSI and OLI images,spectral index and model were constructed and optimized.The results show that the accuracy and stability of the model are comparable based on the common band of OLI and MSI and the spectral index modeling,which indicates that the common band of OLI and MSI images has the same spectral prediction ability for SOM.The inversion accuracy of SOM in the research area can be significantly improved by introducing Reg band in modeling analysis.The model with MSI containing Reg₁ was determined to be the best model,with the modeling set R²was 0.689 and RMSE was 5.960g/kg,and the verification set R² was 0.713 and RMSE was 5.855 g/kg,RPD was 2.176.Compared with OLI,the modeling set R²was increased by 0.083 and RMSE was decreased by 0.546 g/kg;the validation set R²was increased by 0.106,RMSE was decreased by 0.565g/kg and RPD was increased by 0.163.The results show that MSI has better prediction ability than OLI,Sentinel-2A MSI can be used as the optimal data source for remote sensing inversion of SOM.（2）SOM spectral characteristics and spectral models of cultivated land in different topographic areas of Shandong ProvinceThe three major topographic regions in Shandong Province were selected according to the topographic features.The sensitive bands of different topographic areas were screened based on the preferred MSI image,and the spectral index and model were constructed and the differences were compared.The results show that the spectral responses of SOM in different topographic areas are different.R and SWIR are the common sensitive bands of the three topographic areas.In addition,SOM is more responsive to Reg₁ and G in the western plain of Shandong Province,more responsive to Reg₁,G and B in mountainous region of central Shandong Province,more responsive to NIR and B in hilly region of eastern Shandong Province.The spectral index and model were constructed using the respective sensitive bands.The results showed that the accuracy and stability of the spectral models in different topographic regions were comparable.The variables involved in modeling in western plain of Shandong Province were SWIR₁、√R²+SWIR₂²+Reg₁²和SWIR₁-Reg₁/SWIR₁+Reg₁.The modeling set R² and RMSE were 0.661 and 2.740 g/kg,and the validation set R² was0.655,RMSE was 3.265 g/kg,and RPD was 1.854.The variables involved in modeling in mountainous region of central Shandong Province were SWIR₂and√B²+R²+Reg₁².The modeling set R² was 0.639,RMSE was 2.312 g/kg,verification set R² was 0.627,RMSE was3.446 g/kg,and RPD was 1.782.The variables involved in the modeling of eastern hilly region were NIR-SWIR₁和√SWIR₂²+R²+NIR².The modeling set R² and RMSE were0.650 and 2.409 g/kg,and the verification set R² was 0.646,RMSE was 3.557 g/kg and RPD was 1.683.SOM has different spectral response in different topographic areas.The response bands are different,and the variables involved in modeling are also different.Therefore,the same model cannot meet the needs of accurate SOM analysis of different terrains,so it is necessary to introduce terrain factor zoning modeling inversion to improve the prediction accuracy.（3）SOM inversion with topographic factors introduced into different topographic unitsFour topographic factors,Elevation（E）,Slope（S）,Aspect（A）and Relief Amplitude（RA）,were introduced to analyze the correlation between them and SOM and build a model based on spectral data to compare the differences.The results show that the correlation between topographic factors and SOM is not obvious in western plain of Shandong Province,so it is unnecessary to participate in the model construction.E and S are the characteristic terrain factors in mountainous region of central Shandong Province.In the model constructed,the modeling set R² is increased by 0.102 and RMSE is decreased by 0.162 g/kg;the verification set R²is increased by 0.095,RMSE is decreased by 0.238 g/kg,and RPD is increased by 0.129.E and RA are the characteristic terrain factors in the hilly region of eastern Shandong Province.In the model built with their participation,the modeling set R² is increased by 0.075 and RMSE is decreased by 0.171 g/kg;the validation set R²is increased by 0.067,RMSE is decreased by 0.236 g/kg and RPD is increased by 0.169.Therefore,the introduction of topographic factors can effectively improve the prediction accuracy of SOM in different topographic areas,and the SOM prediction effect is the highest in mountainous region of central Shandong Province,followed by the eastern hilly area of Shandong Province,and there is no need to introduce topographic factors in the western plain area of Shandong Province.（4）SOM partition inversion for different topographic regions in Shandong ProvinceThe optimal spectral data source and model were used to predict SOM in the research area.The results showed that the SOM average value was 22.31 g/kg in the western plain of Shandong Province.It was an inland plain,belonging to the Huang-Huai-hai Plain,and the cultivated land had high nutrient content.The SOM average value in mountainous region of central Shandong Province is 16.17 g/kg,and 18.25 g/kg in the hilly region of eastern Shandong Province.The two areas are affected by soil erosion and coastal salinization respectively,and the content of SOM is relatively low,this is consistent with the actual situation.This research can enrich the theoretical knowledge and technical methods of soil quantitative remote sensing,provide reference for provincial scale regional SOM spatial prediction mapping,and provide a more accurate data basis for soil fertility and quality management and improvement of regional soil carbon sequestration ability in Shandong Province.