Forest biomass mapping using satellite, climate and field data.

摘要

The amount of carbon stored in forest biomass is one of the main sources of uncertainty in the terrestrial carbon budget. Remote sensing holds significant potential for estimating above-ground forest biomass over large areas, but robust and effective methods and data sources have proven elusive. This dissertation investigates the use of multisource remote sensing observations in combination with climate data and statistical methods to map forest carbon pools over large areas. The results show that coarse resolution optical remote sensing, when combined with climate and elevation data, can be effective in mapping forest biomass over large areas.; An empirical model based on 1-km Moderate Resolution Imaging Spectroradiometer (MODIS) data, topographic data, climate variables, and non-parametric statistical methods captured complex and non-linear relationships between forest biomass and these variables. Short wave infrared reflectance (MODIS Band 6) and precipitation data are the most significant variables in this regard.; Analysis of spatial patterns in biomass using remotely sensed data shows that the effectiveness of coarse resolution data for mapping biomass depends on the characteristics of the landscape. In areas characterized by relatively homogeneous vegetation, higher resolution remotely sensed data are required to capture spatial variability in forest biomass. To combine field observations with coarse resolution remotely sensed images, it is necessary to aggregate field information to a resolution consistent with the remotely sensed data. Stratification of the landscape into biomass strata based on high resolution Landsat ETM and climate data provides an effective method for this aggregation.; The results from this research show that reflectances observed from different view angles can be combined with nadir reflectance observations to explain spatial variance in forest biomass. Kernel weights used by semi-empirical BRDF models provide useful in formation regarding above-ground forest biomass. However, spectral information dominates the information content of multispectral multiangular data. More generally, remotely sensed data in combination with climate data can be used to map above ground biomass with good accuracy over large areas indicating the potential for improving estimates of continental and regional carbon budgets.