Concept for Forest Parameter Estimation Based on Combined Imaging Spectrometer and Lidar Data

Both Imaging Spectrometry and LIDAR have been already investigated as independent data sources to describe and quantify forests properties. While Imaging Spectrometry provides information on the biochemical and biophysical properties of the canopy, LIDAR resolves the spatial and vertical distribution of the canopy structure (1, 2). The presented contribution outlines a concept how these two complementary information sources can be combined for an improved estimation of forest parameters based on radiative transfer modelling. INTRODUCTION Vegetation plays a major role in the terrestrial land surface processes such as the exchange and storage of water, energy and CO2. Driving parameters of the physiological processes, such as foliar biochemistry and the green leaf area, can be derived by remote sensing platforms such as Imaging Spectrometers and LIDAR sensors enabling an assessment of the terrestrial ecosystem functioning on a spatial scale. The characterization and quantification of biophysical and biochemical forest properties by remote sensing could thus provide spatial information on tree growth and canopy condition essential to foresters, fire and resource managers (3, 4). Remote sensing of vegetated surfaces, especially for forests, relies on the understanding of the interaction of the electromagnetic radiation within the canopy described by physically-based radiative transfer models (RTM). The radiative transfer within a forest canopy is dependant on the spectral properties, the spatial distribution of the canopy elements and on the subsequent complex radiative processes, such as multiple scattering, mutual shading of the crowns and shading of the understory (5, 6). Appropriate radiative transfer models can be used to explicitly exploit our knowledge of the physical processes governing the signal of a forest canopy recorded by an imaging spectrometer or LIDAR (7, 8). The information dimension observed by LIDAR provides information about the vertical canopy structure describing the canopy height and the vertical distribution of canopy elements. Whereas, the spectral information dimension provided by imaging spectrometers contains information about the biochemical composition of the canopy foliage such as chlorophyll and water content. Biophysical canopy parameters such as LAI (Leaf Area Index) and fractional cover are also inferable. Nevertheless, the leaf optical properties, which are directly related to the foliage biochemistry, scales to the canopy as function of canopy structure and spatially arrangement of canopy elements. Consequently the LIDAR signal, e.g. recorded as full waveform, can improve the accuracy and robustness of canopy parameter retrieval, especially the foliage biochemistry, by reducing uncertainties related to the canopy structure. On the other hand the accurate interpretation of the LIDAR signal depends on the spectral properties of canopy elements as well as the background. The two sensors and their different information dimension are thus mutually dependant and can complement each other. A simultaneous exploitation of the information dimensions observed by Imaging Spectrometry and LIDAR based on radiative transfer modelling will therefore provide a new approach to optimize the retrieval of forest foliage biochemical composition and the canopy structure. © EARSeL and Warsaw University, Warsaw 2005. Proceedings of 4th EARSeL Workshop on Imaging Spectroscopy. New quality in environmental studies. Zagajewski B., Sobczak M., Wrzesień M., (eds) In this contribution two separate case studies demonstrate the feasibility of Radiative Transfer Modelling for IS and LIDAR independently, a prerequisite for the proposed approach (9, 10). Further on a concept of combining these two sensors and their data content by joining the two employed RTM is presented. TEST SITE OF CASE STUDIES The test site for this study is located in the eastern Ofenpass valley, which is part of the Swiss National Park (SNP). The Ofenpass represents an inner-alpine valley on an average altitude of about 1900 m a.s.l with annual precipitation of 900-1100 mm. The south-facing Ofenpass forests, the location of the field measurement, are largely dominated by mountain pine (Pinus montana ssp. arborea) and some stone pine (Pinus cembra L.) (11, 12). These forest stands can be classified as woodland associations of Erico-Pinetum mugo (11). Unique ground based characterization of the canopy structure, biochemistry and optical properties of the canopy elements were conducted in summer 2002 using various instruments, ranging from non-destructive spectroradiometric measurements to dry biomass estimation of needles (10). RADIATIVE TRANSFER MODELLING: IMAGING SPECTROSCOPY The spectral reflectance of a vegetation canopy, provided by airor spaceborne imaging spectrometer, is known to be primarily a function of the foliage optical properties, canopy structure, background reflectance of understory and soil, illumination conditions and viewing geometry (5, 6). The complex radiative transfer within a canopy governing the signal recorded by imaging spectrometers can be described by physically based radiative transfer models (RTM), which take into account the above mentioned factors (7, 13). The use of such a RTM for a comprehensive retrieval of the biophysical and –chemical canopy properties from imaging spectrometer data was demonstrated in the following regional test case (1). During a large campaign in the Swiss National Park, in summer 2002, DAIS 7915 and ROSIS imaging spectrometer flights were carried out along with intensive ground measurements of the vegetation properties. Canopy structure was described by two canopy analyzers LAI2000 and hemispherical photographs following well known methods for the characterization of heterogeneous canopies such as coniferous forest (14, 15). Standard wet-laboratory procedures were used for determination of foliage water content and dry matter. Radiative Transfer Model The hybrid radiative transfer model (RTM) GeoSAIL (16) was employed to describe canopy reflectance at scene level. GeoSAIL was chosen due to its low computational costs and its comparable performance to the more sophisticated RTM FLIGHT (Kötz et al. 2004). The radiative transfer at foliage level was characterized by the model PROSPECT (17), which provided the foliage optical properties as a function of the biochemistry and was then subsequently coupled with the canopy RTM. GeoSAIL describes the canopy reflectance of a complete scene including discontinuities in the canopy and shadowed scene components. GeoSAIL is a combination of a geometric model with the SAIL model (18), that provides the reflectance and transmittance of the tree crowns. The geometric model determines the fraction of the illuminated and shadowed scene components as a function of canopy coverage, crown shape and illumination angle. All trees are assumed to be identical with no crown overlap nor does the model account for mutual shading.