PRISMA prototype algorithms for estimating environmental damage and vulnerability to land degradation: the SAPP4VU Project

The project "Development of PRISMA Prototype Algorithms for estimating environmental damage and VUlnerability to land degradation" (SAPP4VU), funded by ASI, aims at developing processing methods and algorithms for the full exploitation of PRISMA data to assess the vulnerability of ecosystems related to slow dynamic phenomena (e.g., Land Degradation) as well as to extreme events (e.g., fires, floods, extreme weather events). This is to foster sustainable and climate-smart farming strategies. Metapontum Plain, located in South Italy, shows both healthy and degraded conditions to test the PRISMA capability in evaluating ecological stability. For ecosystem damages and post-damage recovery, different study sites were selected across the Italian peninsula. In this context, the SAPP4VU main objective is to develop and optimize techniques and algorithms for new product development from PRISMA data to monitor ecosystems exposed to anthropogenic and meteo-climatic forcing. Results PRISMA imagery and Land Cover Land cover, or the biophysical cover of the earth's surface, plays an essential role in climate and environmental dynamics. The objective of the work is to explore the potential of the PRISMA multi-temporal hyperspectral imagery in generating new EO products to complement and improve the Copernicus' Land Monitoring Service for complex and fragile ecosystems such as the Metapontum Plain (Southern Italy) this by estimating land biological and economic productivity loss and land degradation vulnerability. CONCLUSIONS The preliminary geophysical products obtained using PRISMA imagery can be useful when integrated within mining procedures for identifying patterns and anomalies used for the modeling of ecosystem processes. Results showed that PRISMA is suitable for (a) retrieving vegetation traits map, (b) downscaling low resolution LST, and (c) mapping ecosystems at a higher level of detail with respect to satellite multispectral data. PRISMA 30m/pixel hyperspectral images can be used as valuable baseline information to study critical areas subdued to land degradation, even if sampled with a lower temporal resolution with respect to multispectral ones (e.g., S-2). Methods Study Area & data Going from the seaside to the internal plains we can observe a succession of a reforested Aleppo pine belt, which in the past years has been interested by local fires, interleaved humid areas that reach up to the coast, where the Mediterranean scrub dominates, followed by an intensive agricultural area characterized by orchards (i.e., olive groves, oranges, etc.), cereals, and horticultural crops. Preliminary activities deal with the optimization of the PRISMA data processing chain which includes a review of the quality of PRISMA data based on the state-of-the-art correction methods. Vegetation traits were retrieved by using a hybrid model approach. MLs were trained and validated using a large dataset of vegetation spectra obtained from the RTM (SCOPE and PROSECT-D/PROSAIL4) configured ad hoc to describe the variations in the biophysical parameters of the area. Land cover were obtained by: • clustering techniques • machine learning and deep learning classification techniques to identify the spatial patterns of the main agroforestry cover classes (i.e. ecosystems description) Downscale TIR data at the PRISMA spatial resolution to obtain a contemporary LST map at 30m Definition of a vegetation damage spectral base index to assess damage after a critical rapid event and the development of the unsupervised clustering Random Forest for identifying areas prone to land degradation