A New Hybrid Reflectance Model Suitable For Forest On Slopes

Vegetation structure significantly affects its exchanges of matter and energy with the atmosphere, and therefore vegetation structural parameters are important basic data for global change research. The use of forest reflectance models is an effective way to acquire canopy structural parameters at regional or global scale. Existing forest reflectance models generally presume that trees are growing on a horizontal background. However, trees are often growing on sloping terrains. This assumption may lead to large errors in the retrieved forest canopy parameters. Furthermore, the accuracy of simulated exchanges of matter and energy with the atmosphere over mountainous and hilly regions will also be strongly affected by the errors in retrieved canopy structural parameters with the flat terrain assumption. Many researches indicate that the topographic factors have strong influence on forest canopy structures and reflectance. However, topographic corrections are generally made to reduce topography effects on canopy bidirectional reflectance using empirical models rather than a theoretical forest reflectance model suitable for sloping terrains. For the purpose of quantitative remote sensing of forests over complex terrains, it would be useful to develop a physically based bidirectional reflectance model with consideration of topographical variations.For this purpose, the following objectives are addressed in this thesis:1) Developing a new geometric-optical (GO) model suitable for sloping terrains (GOST) based on the4-Scale GO model;2) Simulating the reflectance of the shaded components (shaded foliage and background) with consideration of radiation multiple scattering according to a radiative transfer theory (RT), and developing a new hybrid forest reflectance model suitable for sloping terrains by combining the advantages of GOST and the RT approach;3) Validating model results using measured reflectance datasets; and4) Quantitatively evaluating the topographic effects on sloping forest reflectance and vegetation indices based on the newly developed hybrid forest reflectance model. The major findings are summarized as follows:1) Based on the4-Scale model developed for flat terrains, GOST is developed to describe the canopy structure on slopes. The modeled area ratios of the4scene components (sunlit and shaded canopy fractions and sunlit and shaded background fractions) compare well with the simulated results of a3-D virtual canopy model using a computer graphics technique. It shows that GOST has the ability to simulate the gap fraction and the area ratios of the4scene components on sloping terrains;2) The new hybrid forest reflectance model has common advantages in both of GO models and RT models and the ability to simulate the reflectance of shaded forest components as well as the total BRF (Bidirectional reflectance factor) on slopes. Although it can be improved to include a within-canopy multiple scattering scheme, the new hybrid model already has a unique ability to model the bidirectional reflectance distribution of vegetation over sloping terrains;3) The topographic factors have pronounced influences on reflectance of shaded forest components. The reflectance of shaded components are negligible (less than3%of the total reflectance) in the red band, but the total forest reflectance is strongly influenced by multiple scattering (can reach up to40%of the total reflectance) in the NIR (near infra-red) band;4) The new hybrid forest reflectance model which considers the multiple scattering and topographic factors closely reproduced the MODIS surface reflectance in three selected sloping forest canopies (red band:R2=0.8614, RMSE=0.0055; NIR band R2=0.7573, RMSE=0.0214). If the topographic factors are considered and the multiple scattering is ignored (the reflectance of shaded components are negligible in the red band), the simulated reflectance is slightly underestimated (R2=0.7771and RMSE=0.0070). If the topographic factors are ignored and the multiple scattering is considered, the simulated reflectance is the worst among all simulations (R2=0.5176and RMSE=0.0302in the NIR band), suggesting the importance in considering topography in optical remote sensing;5) According to the quantitative evaluation of the topographic effects on reflectance (red and NIR bands) and its corresponding vegetation indices, such as SR (Simple Ratio vegetation index), and NDVI (Normalized Diffience Vegetation Index), the following conclusions are drawn. First, the topographic factors (slope and aspect) have great impacts on sloping canopy reflectance and vegetation indices. A case study shows that the modeled topographic index for reflectance|RER|, which is the relative error in modeled canopy reflectance by ignoring the topographic effects, can reach up to138%, while the maximum observed|RER|from a Landsat image is99%. Meanwhile, the modeled topographic index for vegetation indices|REV|, which is the relative error in simulated vegetation indices by ignoring the topographic effects, can reach up to64%, while the maximum obversed|REV|from the same Landsat image is178%. Therefore, the forest reflectance models considering the topographic factors are recommended for canopy reflectance and vegetation indices simulation and canopy parameters retrieval. Second, model simulations show that topography has different effects on the sloping canopy reflectance and vegetation indices in different view directions. These effects should be considered in analyzing multi-angle images. Third, topographic factors have different impacts on sloping canopy reflectance and vegetation indices of different canopy structures. Canopy structure should also be considered in topographic correction on reflectance images because topographic factors alone is insufficient for correcting all variations of reflectance with slope and view angle.