| With the rapid development of China’s social economy,the contradiction between man and land has become increasingly prominent.In the face of the basic national condition of more people and less land,it is very important to realize the sustainable utilization of cultivated land resources,improve cultivated land production capacity and ensure food security for the improvement of people’s happiness.Protecting cultivated land resources is also the basic national policy that China has always strictly implemented.Based on the dynamic balance of cultivated land quantity in China,the protection of cultivated land quality is becoming increasingly prominent.At present,the monitoring and inversion of cultivated land quality has become the research focus of many scholars at home and abroad.The quality of cultivated land is mainly reflected in the content of soil organic matter.The degraded area of cultivated land in China is large,and the content of soil organic matter in cultivated land is low.Especially in the black soil area of Northeast China,the content of soil organic matter decreases rapidly and the soil nutrient is unbalanced.Analyzing the content of soil organic matter and mastering the quality of cultivated land will help to provide a scientific basis for the grading of cultivated land quality,the adjustment of crop planting structure and the formulation of land related policies.The content of soil organic matter plays an important role in protecting land resources and maintaining agricultural ecological environment.Facing the decline of soil organic matter content and severe degradation of cultivated land quality in Helen County,the change of soil organic matter content of cultivated land is analyzed by using remote sensing technology,so as to provide data and technical support for the protection of land resources.Taking Helen County as the study area,this paper analyzes the relationship between soil organic matter content and soil reflection spectrum value by using indoor hyperspectral reflectance data and Landsat series image data,quantitatively obtains the distribution map of soil organic matter content in Helen County in 1988,2009 and 2020,analyzes the changes of planting structure in the same period,and tries to analyze the changes of planting structure under human disturbance factors,The change of organic matter content is expected to put forward beneficial measures for further land protection in Helen County.The main conclusions of this paper are as follows:（1）There is a large amount of data in soil hyperspectral bands,and there is spectral information redundancy and overlap.Selecting characteristic variables through competitive adaptive reweighted sampling（CARS）algorithm can reduce the problem of high collinearity between spectral bands,so as to improve the accuracy and efficiency of prediction model.In this paper,through CARS algorithm,the selected characteristic bands not only compress the input bands to less than 16%of the number of full bands.After the spectral variables are screened by CARS algorithm,the determination coefficient of the model is increased by 0.167,and the estimation effect is better.It shows that CARS algorithm plays a key role in extracting feature key band variables and optimizing model structure.（2）In the research of predicting SOM from hyperspectral data,spectral index is often used as an important index to obtain depth signal from SOM content information.This paper discusses the relationship between normalized difference index（NDI）,normalized difference acquisition index（RDVI）,ratio index（RI）and SOM content.The correlation between the original reflectance data of the three soil types and the NDI,RDVI and RI indices of SOM was high,and all passed the significance test at P=0.01.The RI index of black soil has the highest correlation with SOM,with a correlation coefficient of 0.757,the RDVI index of meadow soil has the highest correlation with SOM,with a correlation coefficient of-0.784,and the RDVI index of swamp soil has the highest correlation with SOM,with a correlation coefficient of 0.922.The SOM sensitive band areas of the three soil types are mainly concentrated in the short wave infrared part,mainly around 1000,1900and 2200 nm.（3）Taking the indoor spectral reflectance of three soil types（black soil,meadow soil and swamp soil）in Helen County as the research object,combined with digital elevation model and spectral index as input,the random forest algorithm is used for SOM prediction.The research shows that SOM prediction through soil classification shows that there are differences in SOM adjusted determination coefficients of different soil types.The adjusted determination coefficient of swamp soil is the highest,followed by black soil,and the prediction accuracy of meadow soil is the lowest,only 0.674.The accuracy of the validation set of the soil classification local regression model is the best,the adjusted determination coefficient is 0.777,and the RPIQ reaches 2.689.Compared with the global regression（soil unclassified）model,the validation accuracy of the model is improved by 0.035.This provides an idea for the retrieval of soil organic matter by multispectral sensor.（4）In this paper,various mathematical transformations such as logarithm,reciprocal,first-order differential and second-order differential are adopted for the soil reflectance spectral rate extracted from Landsat Image.At the same time,three spectral indexes are constructed:Salinity index SI,soil regulating vegetation index OSAVI and soil brightness index BI.Seven micro slope factors and four macro slope factors are extracted and analyzed by using SRTM-DEM data as the basic data,The results show that the correlation between organic matter and spectral data is improved after mathematical transformation,in which the reciprocal correlation coefficient in blue band is the highest,reaching 0.565**,and passed the significance test with a confidence of 99%.The correlation between the extracted spectral index and organic matter is greater than 0.5,which is significantly correlated at the level of 0.01.The inversion accuracy of RF model based on mathematical transformation data combined with spectral index is higher than that based on original reflectivity.（5）Considering that topographic factors are effective for soil organic matter content inversion,this paper uses SRTM-DEM data as basic data to extract and analyze 7 micro slope factors and 4 macro slope factors,and finds that elevation（E）,slope（S）The correlation between relief degree of land surface（RDLS）and slope of slope and organic matter is at the level of 0.01,and the correlation between surface cutting depth（SCD）,slope of aspect（SOA）and profile curvature（PFC）and organic matter is at the level of 0.05.In this study,the introduction of spectral index and terrain factor can improve the accuracy of SOM inversion.Among them,the RF model with"mathematical transformation+spectral index+terrain factor"as the input is the best,R2:0.757,RMSE:0.542%,RPIQ:2.421.Considering the mathematical transformation of multi spectral satellite spectral information,spectral index and auxiliary variables,the accuracy of SOM inversion is the highest,and the model can accurately realize the inversion of SOM content.（6）By analyzing the inversion results of soil organic matter in 1988,2009 and 2020,the research shows that the content of soil organic matter in Helen County is gradually decreasing.The temporal and spatial variation trend and distribution characteristics of soil organic matter content were analyzed dynamically from the time dimension.From 1988 to 2009,the average content of soil organic matter in Helen County decreased from 5.23%to 4.73%,a decrease of 9.5%;From2009 to 2020,the average content of soil organic matter decreased from 4.73%to 4.62%,a decrease of 2.3%.Although the decline decreased,the overall level decreased,and the situation of cultivated land soil fertility is not optimistic.（7）According to the temporal and spatial variation characteristics of soil organic matter content in Hailun City,this paper analyzes its causes,and puts forward cultivated land protection measures.Due to the terrain over rivers and hills,the soil erosion is severe,the water and soil loss is intensified,and the soil fertility is reduced.In terms of human factors,due to the pressure of population,the lack of awareness of cultivated land protection and unreasonable land use have exacerbated the decline of land productivity.Therefore,in the face of the decline of soil organic matter content in the study area,it is necessary to adjust the planting structure,conserve water and soil,and carry out fallow rotation to restore the productivity of cultivated land. |
PDF Full Text Request