LEVERAGING DOMAIN ADAPTATION TECHNIQUES IN HYBRID APPROACHES FOR VEGETATION PROPERTY RETRIEVAL FROM HYPERSPECTRAL DATA

The hyperspectral satellite missions, which provide contiguous visible-to-shortwave infrared spectral information, are creating unprecedented opportunities for retrieving a wide range of vegetation traits with enhanced accuracy through novel retrieval methods. In this context, we utilized hyperspectral data cubes collected by the PRecursore IperSpettrale della Missione Applicativa (PRISMA ; https://prisma.asi.it) satellite operated by the Italian Space Agency. Our goal was to develop and test a hybrid retrieval workflow for mapping crop traits. We mapped crop traits over an agricultural area in Central Italy (Maccarese farm, RM) using PRISMA images acquired during February and March 2023. Canopy Chlorophyll Content (CCC) and Leaf Area Index (LAI) were estimated using a hybrid retrieval approach based on PROSAIL-PRO radiative transfer simulations combined with different machine learning algorithms. Domain Adaptation algorithms were employed to optimize the initial set of simulated data by selecting only the most informative samples. Results Crops and Hyperspectral data Reliable and accurate retrievals of vegetation traits are essential in numerous ecological applications such as precision agriculture. An effective monitoring of the crop status requires characterizing leaf biochemical traits as well as canopy structural traits across space and over time. Hybrid approaches that combine the adaptivity of non-parametric methods with the generic properties of the physically-based ones, make them optimal for developing processing chains for the retrieval of vegetation traits from satellite data.