Remote Sensing Modeling Of Leaf Area Index In Arid And Semi-arid Region Based On PROSAIL Model

Leaf area index (LAI), as an important parameter to characterize the biophysical structure and the amount of leaves in land surface systems, usually refers to a half of the total leaf area for a given surface area (Chen & Black,1992). LAI is also a key input variable of land surface process models, widely used in the study of ecology, hydrology, soil science and so on. As the current global LAI products, which are generally limited to kilometer level resolution and the homogeneous surface unit, can not reflect the differences and diversity of the local land surface and the complex ecosystem of vegetation and meet the needs of change detection on regional scale. Therefore, it is necessary to explore high quality LAI mapping of high resolution.Taking arid and semiarid areas in northwest of China as a case, the paper study remote sensing model for estimating LAI on regional scale. Base on the PROSAIL model, the paper establish look-up tables between the regional multispectral reflectance and LAI with field survey data. Then LAI mapping for Landsat5 TM image was performed. The main studying work are as follow.(1) The paper described the theory of the PROSAIL model thoroughly and analyzed he spectral feature space of vegetation cap tassels.(2) The sampling process was descripted in details through PROSAIL model. Based on the measured data and existing research data to determine the range of values of model parameters, which is suitable for arid and semi-arid regions. Combined with the range of model parameters and remote sensing image metadata to get 81804 hyper-spectral simulated data sets, then multispectral simulated data sets corresponding to TM sensor are calculated.(3) On the basis of multispectral simulated data sets, the paper had analysis of the LAI distribution in tasseled cap spectral feature space, established two kind of look-up tables between the multispectral reflectance (including Red, Nir, Brightness, Greeness, Wetness) and LAI, and also the inverse criterion for estimating LAI.(4) Based on the look-up tables and the inverse criterion, LAI inversion mapping were performed, then analyzed the differences between the two look-up tables, Red-Nir-LAI and Brightness-Greeness-Wetness-LAI look-up table respectively. Inversion results are similar by two kinds of look-up tables, the correlation coefficient was 0.8787. Using measured data to verify the theoretical accuracy of the model and the results show the former got a bit higher accuracy than the latter and the correlation coefficient were 0.86,0.67, RMSE were 0.54,0.9 respectively. Compared with the same period MODIS LAI products and the results show LAI inversion based on the brightness, greenness, wetness look-up table is more similar to MODIS LAI products closer, while LAI inversion based on the red, near infrared look-up table has less random disturbances.(5) Many factors had impact on the inversion results, including the uncertainty of the model in its nature, ill-posed nature of the inverse problem and the uncertainty during the data processing.