Spectral Inversion And Spatial Distribution Of Soil Ingredients

Soil not only supplies the necessary nutrients for plant growth,but also provides a living environment for some animals and microorganisms.Therefore,soil is not only a constituent element of a unified geographical environment,but also an important part of an area’s ecological environment.Based on the "food chain" theory,the root of the "people-oriented"political principle and the idiom " Hunger breeds discontentment" is soil.The studies of soil as an object include all aspects.The most human concern is no more than various elements in the soil,including those that are beneficial to humans or harmful.In recent years,with the rapid development of economicalization and urbanization,soil heavy metal pollution caused by human activities poses a serious threat to human health.In the Pearl River Delta,it is extremely urgent to control the heavy metal cadmium in the soil.Soil iron as a representative soil element in the subtropical region can not only stop the synthesis of amino acids,proteins and various corresponding enzymes by the heavy metal cadmium,but also can be used as a soil cadmium pollution remediation material.It’s content more or less can cause plants to produce undesirable traits.Therefore,this paper collates and analyzes the research of soil heavy metal inversion in the recent 20 years,and explores the quick inversion method to obtain real-time soil iron content.Because the point information is difficult to express the study area polygon information.So,the interpolated validation analysis of soil iron in the Pearl River Delta was carried out using different interpolation methods,and the spatial distribution map of soil iron in the study area was plotted at the end.The traditional method of soil elements determination is a combination of field sampling and laboratory measurement that requires a lot of manpower,material and financial resources.It has a long cycle,strong destructiveness,and low ratio between input and output,and it cannot meet the real-time,fast and precise goals asked by humans at the moment.Hyperspectral remote sensing has a wide range of wavebands,strong continuity of information,no destructiveness,and fast analysis speed.The emergence of this technology has made it possible to obtain experimental results quickly and accurately.This article mainly divided into three parts.The first part mainly analyzed the collection data.According to the research and analysis on the inversion of soil heavy metals during the past two decades,indirect inversion methods are generally used for research,soil elements have good correlation with soil heavy metals are mainly organic matter and soil iron,the processing of the spectrum has a significant influence on the accuracy of the inversion model.Most studies believe that the first-order differential effect is the best,but there are some studies that do not support this conclusion.So,what methods should be apply depends on specific situation.The second and third parts are exploratory analysis.The second part explores the rapid spectral inversion model of soil iron content,using the paddy soil sampling point of the Pearl River Delta.The data processing includes smoothing and denoising the original reflection spectrum.The original spectrum is divided into a prediction set and a verification set according to the Rank-KS partitioning method.The spectrum is preprocessed to extract the characteristic bands as the input parameters of stepwise multiple linear regression,partial least square regression,artificial neural networks and support vector machine,the results of the artificial neural network model is the best,the spectrum after differential treatment and soil iron significantly increased correlation,the characteristic spectral bands are 404,574,784,854 and 1204 nm,854 nm is the most significant.In the third part,the entire Pearl River Delta study area was interpolated and analyzed based on the soil iron measured values in the Pearl River Delta.By verifying the interpolation effects of various interpolation methods,the verification method was selected for cross validation and independent sample verification.The results are better for the interpolation of inverse distance weighting,ordinary Kriging,radial basis interpolation and simple Kriging.By calculating the mean relative error(RMSE)of different local coefficient of variation(CV),it was found that when the CV is greater than 0.07,any interpolation method is ineffective.However,in the range of 0.04 to 0.07,radial basis interpolation can highlight local variations,feature.Interpolation results showed that the soil iron content in the study area gradually decreased from the entrance to the surrounding area of the Pearl River Delta.