Analysis And Optimization Of BRDF Model Based On Canopy Reflectance

With the development of the remote sensing’s theory and technology, the quantificational and multiangular remote sensing become more and more important. Recently, several hundred models of quantificational remote sensing have been worked out, which are divided into three types:physical model, empirical model and semiempirical model. Particularly, the study of physical model is relatively complete. However, it should study more deeply about model’s more universal application, and explicit show of scaling problem in models, and the inversing of models with high accuracy. For further study, this article show research status of the Bidirectional Reflectance Distribution Function (BRDF) models, and do some introduction and simulation of several typical BRDF models, being corresponding to the four categories of physical model, such as, SAIL,4-SCALE, GEOSAIL and DART model. And we also do some comparison simulation analysis with measured data based on the study on the above, so we can optimizate or improve the4-SCALE model.This parameters in the article are the results of processing data, which are surveyed in the pangu forest station of pangu town, in greater khingan mountains of Heilongjiang province, in2010. For better analysis, we assume a custom sample plot according to the plot of larch pure forest. And parameters of optimizating the4-SCALE model are surveyed in jagdaqi forest station in2011and2012year respectively.The results of contrast analysis are shown as follows:(1) GEOSAIL model, which simulates the reflection with different cover fraction under the condition with the sole solar zenith angle, do not expain the hot spot phenomenon, and the vegetation’s three-dimensional structure information. Regarding the single band, with the cover fraction increases, the reflection decreases, and the absorbed photosynthetically active radiation increases.(2) DART model simulates the reflection of both the whole scene and the vegetation, ground, water etc. With the five times scattering of the ray of light, it is more truthfulness. The reflection curve provides a good illustration of hot spot phenomenon (PHI=0°, VZA=40°),and the reflection curve of near-infrared band shows the bowl side effect.(3)4-SCALE model, which simulates the reflection of the principal plane, sets about the repellency between crowns, the influence of the distribution of the entities below crown on BRDF. But not consider the influence of multiple scattering on the reflection strength of surface features which consist the shadow area, and the Non Lambert character of the vegetation-ground system, and the limit of the hypothesis of group structure. So when the vegetation distributes continuously and the smaller difference between the shadows and no shadows, it is the radiation transmission model with geometric corrections that is more stringent. When PHI=0°, VZA=40°, the result of simulation shows the hot spot phenomenon. When canopy density smaller, with the fraction of shadow crown increases, the BRDF in forward direction lessens. Which illustrates that, it can be ignored of the influence of multiple scattering on shadow crown inside canopy. And the bowl side effect occurs on considering multiple scattering niside canopy, which illustrates the multiple scattering mainly comes from the surface of the illumination crown.(4) The amount of vegetation canopy’s interception of radiation, the size and the direction of the scattered radiation are directly decided by leaf inclination angle distribution (LAD), which is the key parameter in quantitative remote sensing. Therefore, we simulated the leaf inclination angle distribution of the main tree species in Da Hinggan Ling forest region based on the Campbell ellipsoid distribution function and iterative method, and analyzed quantitatively the fitting result and the influence of age groups with/without stratify treament. Simulation results show that, all leaf inclination angle distribution of main tree species are planophile distribution in Da Hinggan Ling forest region, and the mean leaf inclination angle distribution of coniferous tree is smaller than the broad-leaved tree. With or without stratify treatment, the fitting result and the measured result are basically identical. For Be tula platyphylla and Larix gmelinii, the result of monadic linear regression analysis that, the correlation coefficients are0.8268,0.8192and the root mean square error are3.7%,4.3%respectively, proves that, it is reliable applied in forest canopy for Campbell model. Considering the effect of age groups, the trend of leaf inclination angle distribution has noting to do with age groups under the condition of stratify treament. But young Larix gmelinii’s mean leaf inclination angle is relatively smaller than mature trees. The result proves that, age groups have positive effect on the leaf inclination angle distribution and negative effect on the value of extinction coefficient.(5) A quantitative relationship is existence between leaf area index (LAI) and leaf inclination angle distribution (LAD), which are very important parameters in the study of vegetation canopy construction. The data of single line and multi-angle measuring by TRAC (Tracing Radiation and Architecture of Canopies), with the NC linear least-square inversion theory, can receive a very good inversion result of leaf inclination angle distribution, and its interpretation capacity of the measuring leaf angle distribution is up to more than90%. From the study, we can find out that, the leaf inclination probability density calculated by the NC theory is a little more than the real density, accordingly, the fitting result of Campbell theory is small, especially in the interval of [0°,10°]. Considering the complementary relationship between the results of two theories, we find out that, the correlation between the mean value of the results of two theories and the measuring result is up to more than96%. Besides, with using the NC theory, it is best that the number of the leaf angle interval is half of the number of the solar zenith angle. The quantitative inversion of LAI can provide strong authentication on a wide range of LAI measurements, and prove a new algorithm for LAD.(6) In addition, another major innovation of this research is the introduction of uniformity theory and the observation zenith angle distribution. According to the analysis between DART model, before and after optimization4-SCALE model, we can find that:with considering the blanked effect between crowns, the results of4-SCALE is less than DART model’s. And because the calculated extinction coefficient is greater than its experience value, the inversion results anr greater than before. So we can think that, it is improved that the ability of inversion by optimizating the parameters of4-SCALE model. So that, we can retrieval vegetation canopy reflectance more accurate, or other vegetation index, such as normalized difference vegetation index (NDVI), etc.