Study On Estimating Contents Of Soil Components Based On Hyperspectral Reflectance Data In Laboratory

Land resource is a key factor of environments system and a centralized reflection ofmany environments issues. Land resource utilization is an important part of sustainabledevelopment. Soil erosion, soil fertility degradation, soil degradation and desertification,environmental deterioration and other issues caused by irrational use of land resourcesbecome more serious. How to apply advanced technique for the soil monitoring is animportant research topic. Besides conventional remote sensing functions, adopted onanalyzing soil physical and chemical characteristic and soil fine spectral information, thehyperspectral remote sensing with its high spectral resolution have access to quantitativegain soil properties parameters. It is possible for estimating soil components andinterpreting hyperspectral image.Through investigating soil physical and chemical characteristic, this dissertationanalyses characteristic of space distribution of soil components and sums up spectrumdifference of paddy soil and red soil. The dissertation uncovers the genetic relationshipbetween soil components and their spectrum. The best spectral band is selected based onmathematic analysis of soil reflectance. Several multi-component linear regression modelsare established through cross-correlation analysis. Soil organic matter, its iron oxide andsoil mid-quantity elements could be estimated with a great success. Through overallanalysis and study, the following main conclusions are drawn:(1) Spectral curve of paddy soil and red soil has sameness and difference. Spectralreflectance of red soil at 550nm～1700nm is bigger 6.51%～17.71% than reflectance ofpaddy soil. We could distinguish red soil and paddy soil through compare spectralreflectance.(2) During study on estimating soil organic matter, the optimum band between SOMcontent and derivation of reciprocal of soil reflectance is 811 nm and the max correlationcoefficient is -0.8642. Through cross-correlation analysis, the derivation of soil reflectance at 477nm and 2152nm provided the best estimation model for soil organic matter.(3) During study on estimating iron oxide, the content of soil iron oxide has negativecorrelation with spectrum reflectance. The optimum band between iron oxide and spectrumreflectance is 497nm, the max correlation coefficient of this band is -0.88021. Thecorrelation between absorb peak depth and iron oxide content is poor through correlationanalysis between iron oxide and absorb parameters at 700nm, 900nm. The reflectance andderivation of square root provided the best estimation model for the content of soil ironoxide.(4) The optimum band is 480nm, 706nm for MgO and 818 nm, 1383 nm, 2162nmfor S.The content of soil MgO and S were estimated because estimation model is 90% ofprecision, bigger R square and adjusted R square, smaller RMSE and relative error. It is veryinteresting that the order of the estimation accuracy was the same as the order of their correlationcoefficients with SOM through correlation analysis with SOM. MgO may be estimatedindirectly because of bigger correlation coefficient between MgO and iron oxide.