Hyperspectral Inversion Of Soil Moisture Content Based On Hapke Model

Soil moisture content(SMC)is one of the important soil property parameters provided for thestudy fields of agriculture,hydrology,environment and meteorology.So the study of SMC quantitative inversion based on remote sensing technology has become the hot spot.Because of having a few image wavelength and SMC insensitivity to the small changes,the inversion accuracy by using images from microwave,thermal infrared and multi-spectrum are not very high.At present,the studies of SMC quantitative inversion based on hyperspectral remote sensing did not consider soil bidirectional reflection characteristics,and inversion models adopted are mostly statistical model.The inversion accuracy and model applicable scope are limited.So it has important significance for the study of SMC hyperspectral inversion based on the mechanism model to not only improve the inversion accuracy,but also expand the general applicability of the model.Taking the black soil and chernozem of farmland in Gongzhuling city of Jilin province as the study object,this paper conducted multi-angle spectrometry measurement to soil samples with different SMC from 5%to 45%(the interval of 5%)by using the ASD Fieldspec Pro spectrometer under the condition of indoor.The effect of different light source zenith angle,observing zenith angle,azimuth angle and SMC on soil BRDF were analyzed.SMC inversion by using Hapke model was conducted.The main research conclusions were as follows:1.Through dealing with the spectrum curves of different SMC by using envelope curve method,the sensitive wavelength were obtained as 490nm,650nm,1419nm,1919nm and 2210nm.2.Because of interaction between soil surface particles,soil BRDF increased with increasing of the light source zenith angle,and decreased with decline of the observing zenith angle.The change of soil BRDF with zenith angle was related to the wavelength.The BRDF in backscatter direction was the highest under different observing zenith angle,while it was the lowest in the forward scattering direction.In the range of 400~1400 nm wavelength,soil BRDF declined with SMC rising when SMC was less than the field water holding capacity,but soil BRDF raised with SMC rising when SMC was larger than the field water holding capacity.In the range of 1400~2400nm wavelength and under different observation position,there was no obvious rules for soil BRDF change caused by different SMC.3.The global and first-order sensitivities of six parameters(?,h,b,c,b',c')from Hapke model were examined by using EFAST method.The results showed that the parameter sensitivities in the order of high to low were as?>b'>b>c'>c>h,of which the average single scattering albedo(?)had the highest influence on the model simulation results,but the soil surface roughness(h)had the lowest sensitivity.The sensitivity of b',c'were higher than b,c.Those mean that the post-scattering sensitivity was greater than the forward scattering sensitivity4.The Hapke model parameters were optimized and obtained by using Particle Swarm Optimization(PSO)method under the different SMC gradients,and soil BRDF simulation results were verified.The results showed that SMC inversion accuracy based on Hapke model is higher for areas with the soil moisture content varies considerably,R~2 value was above 0.99 between the estimated values and the measured values of soil equivalent water thickness;While for the areas with SMC small changes,the inversion accuracy for 15%and 30%SMC were higher,but the inversion accuracy for other SMC base on Hapke was unsteadiness.The reason was need to further analysis.SMC inversion accuracy by using average parameters values was also higher with R~2 above 0.98,but lower than the non-average values.SMC model inversion accuracy decreased with increasing of observation Angle,while it was the highest under the vertical observation.