| Soil particle size,as one of the most important structural parameters of soil,is closely related to the hydraulic properties of soil and the ability to store plant nutrients.It is of great significance to study the distribution of soil particle size in ecological and agricultural applications in arid and simi-arid regions.The basis for characterizing the distribution of soil particle size by remote sensing is that there is a close relationship between reflectance and soil particle size,and soil surfaces with different particle sizes have high spatial heterogeneity in spectral reflectance.However,most of the current researches on the relationship between soil reflectance and soil particle size are empirical methods,and there are few physical models that can explain the essential relationship.Understanding the reflectance spectral characteristics of soil with different particle sizes through physical model,is helpful to quantify the relationship between soil particle size and soil reflectance.In previous study,the physics-based analytical radiative transfer model(SART model)simulated the reflectance factor of soil with different particle sizes in the wavelengths of the Landsat8 satellite at two different angles,such as the phase angle of 22° and 20°.The reflectance factors simulated by the model are in good agreement with the measured reflectance factors,because the same model parameters(absorptivity,k)are used at different angles,that is,the model is considered to be suitable for any incident angle and viewing angle.However,soil reflectance has strong anisotropy in the hemispherical space,and the applicability of a single model parameter may be weakened with the change of incident angle and viewing angle.Therefore,it is necessary to explore whether the SART model can be used to simulate the multiangular distribution of the reflectance factor of soil.In this study,46 soil samples with different particle sizes of four types of soil,including aeolian sandy soil,sandy soil,chestnut cinnamon soil and black soil,were used for multiangular spectral reflectance factor dataset measurement.The SART model is evaluated for the first time based on the multi-angular reflectance data set of soil with different particle sizes.The main conclusions are as follows: the accuracies(root mean square error,RMSE)of the model for simulating soil reflectance factors with different particle sizes in the visible blue wavelength of 480 nm,the near infrared wavelength of 830 nm and the short-wave infrared wavelength of 1650 nm of the Landsat8 satellite are greatly affected by the viewing angles.The model parameters inverted from the data of backward scattering 20° in the principal plane(the phase angle of 20°)were used to simulate the soil reflectance factor in nadir direction,and the RMSE value increased compared with that of the simulated soil reflectance factor of backward scattering 20° in the principal plane.This indicates that the same model parameters cannot be applied to the description of soil optical properties at different viewing angles.In order to improve the accuracy of the model for simulating the multi-angular reflectance information of soil with different particle sizes,this study optimized the parameter k of the model,and drew the optimal model parameter distribution diagram for different viewing angles to accurately simulate the multi-angular reflectance characteristics of soil with different particle sizes.The results showed that based on the optimal model parameters,the RMSE values of the four types of soil at 480 nm,830 nm and 1650 nm were all less than 0.04,indicating that the model could simulate the single-wavelength reflectance factors of soil with different particle sizes with high accuracy at different detection directions.In addition,previous study only simulated the reflectance factors of soil at single wavelength based on the SART model.This study found that the model can also simulate the continuous hyperspectral reflectance curve of soil,and then this model was used the model to simulate the multi-angular reflectance spectral curve(400-2500 nm)of soil with different particle sizes for the first time.Based on the optimal parameters,the SART model can simulate the multi-angular reflectance spectra of soil with different particle sizes with high accuracy.Under the multi-angular observation,the RMSE of the simulated soil reflectance spectral curve and the measured soil reflectance spectral curve is ≤0.03 for the four types of soil.And the simulated multi-angular reflectance spectra of soil with different particle sizes is consistent with the change of the measured reflectance spectra,the reflectance factor increases with the decrease of soil particle size and the decrease of the phase angle.These results show the SART model can accurately simulate the multi-angular hyperspectral reflectance characteristics of soil with different particle sizes based on the optimal parameters.In conclusion,this study found the angular dependence of the inverted parameters of the SART model based on soil multi-angular data.In order to improve the accuracy of the model to simulate the multi-angular reflectance data of soil,the parameters of the model were optimized and the multi-angular optimal model parameter distribution diagram was drawn.Based on the optimal parameters,the SART model can simulate single-wavelength and continuous hyperspectral soil reflectance factors with high accuracy under multi-angular detection.At any observation angle,as long as a finite number of soil particle size gradients and corresponding reflectance values are input,the reflectance information of any particle size of the soil can be simulated.The multi-angular optimal parameters are combined with this physical model to quantify the spectral characteristics of soil with different particle sizes.This provides a new method for simulating the reflectance information of soil with different particle sizes simply and effectively,and also provides a new way to develop new physical models. |
PDF Full Text Request