Estimation Of Soil Total Salinity In The North Of Wei Fang Area Based On Satellite-ground Remote Sensing Data

Soil is the most basic renewable natural resource for human survival,and its quality will directly affect the development of social economy and human survival.Unreasonable development and utilization in the process of agricultural development has led to serious damage to the soil environment.Soil salinization has become one of the major factors affecting soil degradation in coastal areas.Salinized soil will affect the growth of crops,resulting in reduced grain production and thus affecting regional economic development.Therefore,the ability to obtain soil salinity information quickly and efficiently,and to dynamically and accurately monitor the characteristics of soil salinity changes over time,has important practical significance for the treatment of soil salinization and the realization of sustainable agricultural development.Under the support of the National Natural Science Fund Project(41371395),Shandong Province Geological Environmental Monitoring Station"the Yellow River delta Efficient Ecological Economic Zone sea salt water intrusion investigation and monitoring and early warning system construction",Shandong Province Science Fund"Construction project of remote sensing spectral database of land and resources in Shandong Province"and"Research on remote sensing monitoring of Longkou mining area and the surrounding coastal zone".In this paper,the sea salt-water intrusion area in northern Weifang is taken as the research area,and the near-surface measured hyperspectral data is used to estimate the total salinity of the soil.The model that has passed the test and has good stability is selected as the collaborative inversion estimation of satellite-to-land remote sensing data.Based on the use of collaborative remote sensing data to calculate the spectral index and combined with the total salt content of soil,a collaborative estimation model of soil salinity-terrestrial remote sensing data was established and a series of studies were conducted:the hyperspectral data of different soil salinity contents were analyzed,and the Spectral characteristics of soil under different salinization levels;Spectral transformation of hyperspectral measured data and correlation analysis to obtain sensitive feature bands to improve model accuracy,and constructed by multiple linear regression analysis and partial least-squares regression analysis.A hyperspectral model for estimating the total salt content of soils;using a satellite-to-terrestrial remote sensing data collaboration method,using ground-measured hyperspectral reflectivity data to simulate the multispectral data of remote sensing images and constructing a spectral index to establish a model for estimating soil salt content.The established model was used to remotely inspect the soil salinity in the study areaThe main conclusions are as follows:(1)In the range of 325nm~1075nm,the soil spectral curve of different salt content has the same change trend,the soil spectrum of different salt content has different regularity,and in the range of 325~1075nm,the soil spectrum curve of different salt content is the same as a whole.The hyperspectral data of soil samples with different salt contents showed that with the increase of total salt content(i.e.salinization degree),the reflectivity increased gradually with the increase of wavelength.The soil spectral reflectance increases gradually in each band,and the variation trend is relatively gentle.The more the band range is,the more obvious the increasing trend of soil reflectivity is.(2)The accuracy of soil total salt inversion and estimation is good by using soil hyperspectral data.The absorption reflectance characteristics of soil with different total salt content showed obvious differences.The hyperspectral original reflectance data were transformed by various types of spectral transformation to reflect the hidden information in the hyperspectral data.Using these subtle differences in hyperspectral data to select the feature bands,The theoretical basis for retrieving soil total salt from hyperspectral data is obtained.The sensitive band of the original spectral data is concentrated at 450~550 nm,and the correlation coefficient with soil total salt content can reach the highest of 0.4.The spectral transformation of the original spectral data(including the first order)is carried out.Second order differential transformation,continuum removal and reciprocal logarithmic transformation.After spectral transformation,the correlation between the spectral data and the total salt content of the soil is obviously improved.The correlation coefficient of some bands is more than 0.6,and the spectral transform can improve the accuracy of the model and reduce the computation of the model inversion,in which the hyperspectral reflectivity is the first order.The model R~2 is 0.885 and0.798,and the root mean square error(RMSE)is small.It is shown from the test results that the fitting effect between the measured data and the estimated data is good,and the model R~2 is0.885 and 0.798,respectively,and the root mean square error(RMSE)is small.The fitting curve R~2 is more than 0.8.The modeling results can provide theoretical and technical reference for soil salinity hyperspectral monitoring in northern Weifang.(3)On the basis of the better effect of near-surface hyperspectral inversion,this paper attempts to invert the soil salt content in the study area by combining the near-surface measured hyperspectral data with the satellite remote sensing data,and uses the ground measured hyperspectral reflectance data.Using ENVI to extract the band response function of Landsat-8image,the hyperspectral reflectance curve of pretreated soil was resampled to the range of visible light and near infrared.In order to retrieve soil total salt more accurately,the multispectral data of remote sensing image were simulated.The method of constructing spectral index by combination of different bands is used to improve the accuracy of the model.The stepwise linear regression model and the partial least square regression model are also used to model and estimate the model.The results of the model are compared and tested.The partial least square regression model R~2=0.779 and the model RMSE=0.882 are more accurate and more stable.After the inversion model is determined,the remote sensing image is inversed.The mixed pixel decomposition of the remote sensing image is used to remove some of the affected vegetation spectral information in the water body and the image.In order to reduce the difference between remote sensing reflectivity and hyperspectral reflectivity,the determined partial least square regression model is applied to remote sensing image.The inversion result of remote sensing image is obtained and the precision test is carried out by using the actual sampling data.The result shows that the result of remote sensing inversion is consistent with the actual sampling point data,which proves the feasibility of estimating soil salt content by co-inversion of satellite and ground remote sensing data.It provides an effective way for remote sensing investigation and dynamic monitoring of soil salt content in the region.