| Soil organic carbon(SOC)is an important content to reflect soil fertility,and is a very active part of the atmospheric carbon source.Songnen Plain is an important production base of high quality and high yield of soybean and maize,and the improper use of cultivated land will lead to the decline of cultivated land environmental quality.Therefore,it is very important to understand the spatial distribution and change of soil organic carbon content.Because of the close relationship between the soil spectral reflectance and soil moisture content and organic carbon content,estimating soil organic carbon content using hyperspectral remote sensing technology in a short time has very important significance for realizing the quantitative quick acquisition of soil organic carbon,environmental protection and the development of precision agriculture.Most of the previous test for soil samples under different soil moisture content,this paper explored the impact of soil moisture content on the estimation precision of soil organic carbon content.In order to make the results accord with the actual conditions,choosing the thirty-two counties and cities of northern Songnen Plain as the study area analyzed physicochemical properties and collected spectral data of 112 soil samples,measured soil moisture content,organic carbon content,revealed the spatial distribution of organic carbon content,water content and the relationship between of them.Based on the correlation analysis of soil organic carbon content and five spectral mathematical transformation indicators including original spectral reflectance,first difference derivative,second difference derivative,continuum removal and log transformed spectra which had been pretreated(on average,the band removed,go to the overlapping and resampling,smoothing),the sensitive spectral parameters of soil organic carbon content inversion were comparative analyzed and determined.The estimation models of soil organic carbon content and reflectance spectrum was established by using partial least squares regression(PLSR)method.And using the coefficient of determination(R~2)and the root mean square error(RMSE)verified the accuracy of the model.After comparison of estimation accuracy,hyperspectral technology of estimating ability and practicability was discussed.Trying to overcome the shortcomings of traditional inversion methods and excavate the difference of models under different soil moisture content,which provided theoretical support for the rapid measuring of soil organic carbon content in large areas.The results show that:(1)According to the practically measured data of moisture content,point?s soil moisture content spatial distribution was analyzed and indicated that distribution of soil moisture content was uneven.The soil moisture content was decreasing,from the northeast to the southwest,which was divided into three water gradient zones:0~200g/kg(T1),200~300 g/kg(T2),300~500 g/kg(T3).Based on the One way ANOVA,it was found that there was a significant difference between the three kinds of moisture content gradients at the 0.01 level,and the coefficient of variation was in the range of0.1~0.5,which was moderate variability.The content of soil organic carbon decreased from northeast to southwest,indicating that the distribution of soil organic carbon was not balanced,and organic carbon content in the 3 kinds of moisture content in the study area showed moderate variability.It resulted that decomposed speed of animal and plant residues was slow gradually,and low temperature was bad for evaporation of water making soil organic carbon accumulated easily.(2)There was a extremely significant positive correlation between soil organic carbon content and soil water content at 0.01 level of all samples,and there was a significant positive correlation between of them under the 3 kinds of moisture content gradient(P<0.05).The results showed that the content of soil organic carbon increased with the increase of moisture content.And the effect of soil moisture content on soil organic carbon was very obvious.The soil organic carbon content showed a sensitive change under different moisture content gradient.Because of the excessive soil moisture content,the mineralization of organic carbon can be inhibited to a great extent.(3)In the full range of soil spectral curve showed a convex parabolic shape.Soil spectral curves of different organic carbon content decreased with soil water content increased.But reflectance spectroscopy showed the similar differences under different moisture content condition and with soil moisture entered the threshold(240 g/kg),and the rate of spectral reflectance values decreased significantly.When the soil moisture had identical fluctuation range,the spectral reflectance changes in the visible spectrum was less than near-infrared bands,and soil moisture content had the stronger effect on near-infrared spectral reflectance.With the increase of the wavelength,the spectral reflectance increased rapidly in the visible and near infrared forward region(400~1100nm),and the spectral reflectance increased slowly in others wavelength range(1100~2400 nm).The higher content of the soil water and the organic carbon had,the less obvious the wave crests and troughs were.(4)Through the correlation analysis of soil organic carbon content and the original spectrum and the four kinds of spectral transformations,the original spectrum reflected that the sensitive spectrum of soil organic carbon mainly concentrated in the visible bands.Before the differential transformation,the detection ability of soil organic carbon content in the near-infrared bands was weak than the visible bands.The spectral reflectance can clearly reflect the subtle changes of the spectrum after the first difference derivative,which can effectively extract the sensitive bands.The sensitive bands in the every moisture gradient appeared in different positions(The sensitive bands set of T1 were 580 nm,1080 nm,1955 nm,2020 nm and 2180 nm,the sensitive bands set of T2 were 685 nm,880 nm,1000 nm,1455 nm and 1955 nm,and the sensitive bands set of T3 were 450 nm,860 nm 960 nm and 1955 nm),reflecting the influence of moisture content on soil organic carbon characteristics of spectral bands.It was proved that the soil organic carbon estimation model under different moisture content should be based on moisture content.The correlation between soil organic carbon content and soil moisture content of T3 was significantly higher than that correlation of T2 and T1.(5)The prediction models of this study using partial least squares regression method were established between soil organic carbon content and different moisture content gradients of the reflection spectrum.The three models of the T1,T2 and T3both had well stability and precision(T1,T2,T3 estimation models of R~2 were 0.79,0.86,0.89,RMSE were 1.60,1.03,2.27;the verification models of R~2 were 0.83,0.82,0.82,RMSE were 1.80,2.59,3.76).It can be used for the field measurement of soil organic carbon content in a short period of time,which was feasibility and practical application value.By comparison,the estimation accuracy of spectral reflectance models with less soil moisture content was slightly better than that of high. |
PDF Full Text Request