Impact Of Forest Background Relfectance On The Canopy Leaf Area Index Inversion

Forest is a significant terrestrial ecosystem. It accounts for30%of the land area and plays a key role in global change. As an important characterization of forest crown structure, Leaf area index (LAI) refers to a half the total developed area of leaves per unit ground surface area. It directly affects forest ecological processes such as photosynthesis, respiration, transpiration, precipitation interception, energy and material exchange between the surface and the atmosphere. Remote sensing is effective for mapping LAI at regional and global scales. In previous studies, background reflectance was always ignored or regarded as a fixed value just changing with land cover types. This will affect the accuracy of the LAI inversion since background reflectance changes spatially and temporally and impose noise on signals detected by sensors. Therefore, this paper researched on the retrieval of forest background reflectance in Tahe area using MODIS250m reflectance data. Then forest background reflectance was mapped and used in LAI inversion. The influence of temporal and spatial variations of background reflectance on LAI inversion was assessed. Therefore, the outputs from this study will be valuable for improving the accuracy of LAI inversion.Firstly, parameters in the kernel-based bidirectional reflectance factor (BRDF) model were fitted using MODIS reflectance and angle data. Then, based on the reflectance at specific angles simulated using fitted BRDF parameters, d, forest background reflectance at250m resolution was extracted throughout the2011using the4-scale geometrical optical model. Temporal and spatial variations of background reflectance analyzed. Finally, the series of8-day LAI were generated using the extracted background reflectance and an LAI inversion algorithm developed on the basis of4-scale geometrical optical model. Inversed LAI was validated using measured LAI scaled up with TM data. The changes in the effects of changing forest background reflectance on LAI with forest density, aspects, and season were also investigated. From this study following findings and conclusions can be drawn: (1) Retrieval of forest background reflectance using MODIS250m reflectance dataThe kernel based BRDF model parameters could be effectively retrieved using MODIS250m reflectance and corresponding angles. Then reflectance at specific angles can be simulated using the fitted BRDF parameters. With proportions of sunlit canopy, shaded canopy, sunlit background, shaded background simulated using the4-Scale model, forest background reflectance can be retrieved using the simulated reflectance. The indirect validation demonstrated the reliability of retrieved background reflectance.(2) The spatial and temporal variations of extracted forest background reflectanceThe background reflectance maintained high values in the winter because of understory snow and showed a downward trend in April due to snow melting. It reached the lowest value around day105. Then, reflectance of near infrared band increased gradually due to growth of understory vegetation and reached a maximum of40%around day185. While red reflectance weakly increased first due to snow melting and soil drying. Then it decreased with the growth of understory and approached a minimum of3%around day185. In the autumn, with the wilting of understory shrub and herb, reflectance of near infrared dropped slowly and reflectance of red changed inversely.(3) Changes of inversed LAI with the consideration temporal and spatial variations of background reflectanceLAI was inversed separately using spatially and temporally variant and invariant background reflectance. On average, LAI inversed using variant background reflectance was1.2smaller than that inversed using invariant background reflectance in the spring. This number decreased to0.5in the summer and fall. This could be explained by less background signals detected by sensors due to canopy in the summer and by weaker background signals owing to wilting of understory vegetation in the fall. The influence with variant background reflectance on LAI was related to tree density. The decreases of average LAI inversed with variant background reflectance were about1.3,0.8, and0.5when tree density was below1000trees/ha, ranged from1000to2000trees/ha, and from2000to3000trees/ha. If forest density was above3000trees/ha, the difference of LAI inversed with variant and invariant background reflectance was marginal. The influence of variant background reflectance on inversed LAI was also related to aspect. It was larger in three southbound aspects than in three northbound ones.