Hyperspectral Estimation And Spatial Inversion Of Soil Organic Carbon Content In Coal Mine Area

Soil organic carbon plays an important role in soil water holding capacity,soil biological activities,soil nutrient transformation and soil texture.It is the core of soil quality and the main index of soil fertility.It is also an important reference for land reclamation.Therefore,the rapid and accurate acquisition of soil organic carbon content information is not only the inevitable requirement of current precision agriculture,but also an important prerequisite for sustainable land use in mining areas and sustainable development of coal mining industry.In this study,107 soil samples were collected from Baodian Mining Area,Shandong Province.After laboratory analysis of soil organic carbon content,two sets of soil hyperspectral data were measured indoors and in the field respectively.At the same time,GF-1 WFV data images were used to explore the soil organic carbon content.The prediction model and spatial inversion model of soil organic carbon content in the mining area were studied,and the spatial distribution characteristics of soil organic carbon were revealed by combining land use types.The prediction model of soil organic carbon content was based on dry soil spectrum which was pretreated by Savitzky-Golay convolution smoothing(SG),multivariate scattering correction(MSC)and mathematical transformation.Partial least squares regression analysis was used to establish the prediction model of soil organic carbon content based on full band and significant band of soil spectrum,respectively.The spatial inversion model of soil organic carbon was based on wet soil spectrum.The narrow band of hyperspectral was fitted to the wide band of GF-1 WFV by various mathematical methods.The hyperspectral model of soil organic carbon content was established by multivariate linear stepwise regression analysis.Then the ratio correction method was used to correct the optimal hyperspectral model to a multi-spectral model,and then applied to the GF-1 image to obtain the soil organic carbon content in the study area.The results show that:(1)The spectral reflectance curves of different organic carbon contents were consistent in morphological characteristics and different in intensity.When organic carbon content decreased to a certain extent,its spectral response ability decreases,especially in the near infrared band.(2)In the prediction of soil organic carbon content,the saliency band range of SG and untreated mode was wider than MSC and MSC+SG pretreatment.Compared with full-band modeling,the salient band modeling method improved the model accuracy without processing mode,reduced the model accuracy under MSC and MSC+SG processing mode,and maintaind the model accuracy under SG processing mode.(3)In the study of spatial inversion of soil organic carbon content,band operation can significantly improve the correlation with soil organic carbon content.The reciprocal transformation was relatively good in single-band operation,but the correlation was still in the middle level;the ratio and difference of double-band operation were better,and the ratio can reach the strong correlation level;the correlation of spectral index produced by multiband operation could reach about 0.74.Under the full-band average fitting method,the optimal model was generated by three band operation methods.The inversion validation coefficient of the corrected multispectral model was 0.676.(4)Soil organic carbon content ranged from 1.08 to 31.70 g/kg,with an average of 11.76 g/kg.In the overall distribution,the soil organic carbon content in the mining area is mainly in grade III,followed by grade IV,and a little in grade II,V and VI,with grade I being the least,less than 0.1%.Soil organic carbon content is relatively high and concentrated in the central region,and low in the Northwest region.(5)The land use types in mining areas are still mainly cultivated land,which is the most widely distributed;forest land and grassland are less,mainly distributed in the northwest;mining land is mainly concentrated in the central region,which is equivalent to the road area;water area is relatively large,accounting for nearly one eighth of the whole region.The average quality level of cultivated land is above the medium level,but the high-quality arable land is very few;the overall organic carbon content of forest land and grassland is relatively close;the soil organic carbon content of mining land is the highest,which is significantly higher than that of agricultural land.In general,the soil organic carbon content in each land type was mining land > cultivated land > grassland > forest land.