Integration Ofwaveform Lidar and Hyperspectral Data to Estimate Structural Attributes of Tropical Forests

Estimation of forest structural attributes is important in biodiversity and global carbon cycle studies. These estimates can be used to predict important forest characteristics such as aboveground biomass (AGBM), which helps determine the amount of carbon in terrestrial vegetation pools. Traditional field sampling-based estimation methods are not only time consuming, expensive, and limited to local scale studies, but may also be biased by different crews and field conditions throughout the campaign period [1], [2]. To tackle these issues, both active and passive remote sensing technologies are being used in large scale studies. High resolution optical data have proven useful for determining canopy extent and horizontal structural information such as gaps and patches, but have limited information related to chemistry. Synthetic Aperture Radar (SAR) provides all-weather coverage over extended areas for determining biomass, but physical models are highly parameterized, and estimates saturate in dense canopies. Recently, interest has increased in discrete return and full waveform LIDAR (LIght Detection And Ranging) [3], [4], [5], [6] for characterizing vertical structure of forest and hyperspectral [7] image data for analysis of forest chemistry. Data acquired by different sensors provide information on different aspects of a target, potentially providing opportunities to exploit synergisms through multi-sensor analysis. Fusion techniques, which involve fusion of data or fusion of output of analysis derived from individual sensors, are currently of interest in many fields of remote sensing data analysis and have been applied to estimate forest structural attributes by several researchers. Some authors [8], [9] have investigated the integration of discrete return LIDAR and multispectral data. Others [10] studied fusion of SAR (Synthetic Aperture Radar) and multispectral data to estimate forest characteristics. While several combinations of multi-sensor data fusion have been investigated, to our knowledge no research has focused on the integration of full waveform LIDAR and hyperspectral data, which is the goal of this study. We propose a novel approach to integrate full waveform LIDAR and hyperspectral data which utilizes LIDAR waveform decomposition results to co-register waveform LIDAR with hyperspectral data, and perform forest structural attributes estimation using the integrated data. While synergisms clearly exist between hyperspectral and LIDAR data, integration strategies are complicated by differences in data acquisition architectures. Hyperspectral data are acquired in pixel based architecture; hence each pixel represents a spectral response of a single location, and pixels are regularly located over the Earth's surface. LIDAR data are acquired in an irregular point based architecture, …