Characterising forest ecosystem parameters with airborne hyperspectral data: the DARFEM (HySens IT2-01) experiment M. Meroni a c , C. Panigada b , M. Boschetti a , R. Colombo b , P.A. Brivio a , L. Busetto b , M. Rossini b , C.M. Marino b and J.R. Miller d a CNR-IREA, Via Bassini 15, 20131 Milano, Italy, email: michele.meroni@unimib.it b DISAT, University of Milano Bicocca, Milano, Italy c DISAFRI, University of Tuscia, Viterbo, Italy d Physics Dept., York University, Toronto, Canada ABSTRACT During the summer 2001 various experimental flights were carried out in the frame of EU-funded HySens project coordinated by DLR (Germany). This work presents the results achieved within the DARFEM (DAIS and ROSIS for Forest Ecosystem Monitoring) experiment designed to provide a better understanding of the capability of airborne hyperspectral observations in the retrieval of physiologically relevant vegetation parameters such as fractional cover, leaf area index and leaf chlorophyll content. The experimental site is located in the Ticino River regional park, northern Italy. Forest ecosystem is a poplar plantation that is a CARBOEUROFLUX site (Kyoto forest). Airborne surveys were planned in order to maximize the variance associated with the different viewing geometries of DAIS in the hyperspectral measurements over the forest. An intensive field campaign was conducted for collecting basic information on vegetation: biophysical parameters and optical properties were measured in the field and in laboratory. Methods developed to account for the understory contribution gave more accurate determination of relevant vegetation parameters. A comparison of the results obtained in the retrieval of those parameters from the application of semi-empirical models and from the inversion of radiation transfer models is presented, and the importance of the understory in sparse canopy forest ecosystem is discussed. Keywords: Hyperspectral data, leaf area index, fractional cover, chlorophyll content, forest understory, semi- empirical model, radiative transfer model. 1 INTRODUCTION Over the recent years the concern has grown within the scientific community about the need of accurate land surface parameterisation to improve our understanding of the state of the global climate system and to measure the responses of natural and managed terrestrial ecosystems to climate variability. In fact large uncertainties still remain in quantifying the interrelationships between the biosphere functioning and the climate system and climate change. The increase of greenhouse gases in the atmosphere, released by humans through their industrial activities and biomass burning, focused the attention on the role of the vegetation in the carbon bio-geo-chemical cycle. The importance of the vegetation and of the forest in particular within the carbon cycle found mention in a number of recent international agreements, such as the Kyoto Protocol of 1997 [1]. The presence and status of vegetation cover control in many ways the energy, gas and water exchanges between the land and the atmosphere [2]. Recent works show that forests play a key role also in the environmental distribution of Persistent Organic Pollutants POPs, such as the dioxins, since they act as filters of these chemicals, trapping them in the air compartment and transferring to forest soils consequently decreasing their atmospheric half-lives [3]. There is therefore an increased request for a systematic documentation on vegetation cover distribution and on vegetation characteristics and functioning both from the scientists and policy makers. Vegetation biophysical parameters such as fractional cover (Fc), leaf area index (LAI) and leaf chlorophyll content (Ch) are assimilated into soil vegetation atmosphere transfer (SVAT) or ecosystem biogeochemical cycle simulation models to calculate vegetation photosynthetic efficiency, gross and net primary production. These models benefit from the knowledge of spatial distribution of such vegetation parameters, derived from spatial interpolation of ground measurements or from remote sensing observations, are used either as input variables (forcing method) of spatially ecosystem functioning either as comparison of model output (recalibration method). Presented at the 3 rd EARSeL Workshop on Imaging Spectroscopy, Oberpfaffenhofen, May 13-16 2003 525