Abstract: Western savanna is one of the most complex ecosystems due to its horizontal and vertical heterogeneity. It is generally assumed that individual-tree based models are needed to capture canopy structure at utmost details so that the radiation and ecological processes can be modelled realistically. Accompanying that, there are two fundamental research questions for modeling such a heterogeneous landscape: (1) how to parameterize an individual-tree based model at the landscape level, especially with the aid of remote sensing, and (2) since it is unrealistic to apply individual-tree model at the regional and global scales, is it possible to develop simple models that can achieve comparable performance as individual-tree models? If yes, how? To address these two research questions, this dissertation is divided into two parts. This first part includes three chapters which introduce how I use an innovative remote sensing technology called LIDAR (Light Detection and Ranging) to extract the individual-tree structural information at the landscape level. In the second part, I propose an analytical approach to calculate clumping factors, which are used in a volume-integrated Markov model for estimation radiation and photosynthetic processes. It was found that the Markov model can achieve comparable performance as individual-tree based model, which indicates its potential in broad-scale biosphere-atmosphere modeling and global climate change studies.