Study On Digital Pattern Of Multispectral Reflectance-LAI Based On PROSAIL Model

Leaf area index (LAI) is defined as half the total intercepting area per unit ground surface area. LAI as an important parameter of foliage, playing a significant role in land surface processes model, is widely applied to the relevant field of ecology, biogeochemistry and climate study. In the study of quantitative remote sensing of vegetation, exploring the relationship between vegetation canopy reflectance and LAI is one of the key issues. Deriving LAI from remotely sensed data acts an important driver to estimate Net primary productivity and assess ecosystem quality and other crutial problems.PROSAIL is a combination of the SAIL canopy bidirectional reflectance modle and the PROSPECT leaf optical properties modle. PROSAIL, which is mainly related to leaf biochemichal contents (chlorophyll a+b content, carotenoid content and equivalent water thickness and so on), canopy architecture(LAI, leaf inclination distributeion function and so on) and soil/vegetation contracts, has been used for deriving LAI for many years.The study used a hybrid approach that combined a look up table(LUT) and the PROSAIL radiative transfer modle to estimate LAI from Landsat7ETM+imagery. The main work is as follows:(1) Digital pattern of multispectral reflectance-LAI was built with PROSAIL. Distribution pattern of LAI in the spectrum space was divided into two parts:one was the LAI distribution in the plane of the red and near infrared band reflectance space, the other one was the LAI distribution in the space composed with brightness, greenness and wetness which transformed from tasseled cap. The study was based on simulation dataset. In order to get better understanding of LAI distribution patterns, the normalized difference vegetation index (NDVI) distribution in the corresponding spectral space was analyzed.(2)Modle sensitve analysis and field data provided basis for determination of model variables values range. Sensitivity analysis determined the distance between corresponding parameter selection, which reduced the number of similar simulation data. At the same time, the field data as prior knowledge did a lot of favour to choose soil line and determine the leaf inclination angels, which made simulation results more close to the real condition of study area. (3)In order to get closer value to the reflectance of satellite imagery, hyperspectral reflectance data was converted to the corresponding band reflectance of Landsat7ETM+. Hyperspectral reflectance data conversion consisted of two parts:one was the measured reflectance data, the other part was simulation reflectance with PROSAIL.(4)The simulated datasets were interpolated to obtain two lookup tables:one was established on the relationships among red, near infrared reflectance and LAI, the other one was established on the relationships among brightness, greenness, wetness and LAI. By using these two tables, LAI were retrieved from Landsat7ETM+data.(5)Using look up tables mapped LAI. The results showed that red-near infrared-LAI look up table got a better result.Then, the results were validated. Due to the measured data was not fit to validate the LAI results at region scale, so using MODIS LAI to validate the results.For comparison, Landsat7ETM+LAI inversion results were up-scaled to the same resolution of MODIS.LAI products. As results, derived LAI was reasonable.(6)The uncertainty of LAI inversion was discussed. The discussion found that during LAI inversion process, the inversion results affected by many uncertainty factors. Moreover, during the process of validation, some uncertainty influenced the results of the validation slightly.