Abstract
Mountain grasslands in the European Alps play a crucial role in climate regulation, biodiversity safeguarding, landscape conservation, and soil quality preservation. The ownership of managed grassland is usually highly fragmented, and the management practices are very heterogeneous. Despite water being abundant in the past, water scarcity has started to raise concerns in the Alps because droughts are becoming more and more frequent, causing unstable income for mountain farmers. Risk management instruments, like insurances, are necessary to sustain grassland-based mountain agricultural systems and to allow them to overcome production shortcomings, and thus maintain their functioning over time. Traditional insurance schemes would require yield measurements and physical inspections by insurance appraisers to assess damages, but this approach is not economically sustainable due to their high cost compared to the low value of the production. Index-based insurance can overcome these issues because payoffs depend on an index related to grassland production that does not require physical checks. In this context, high-resolution satellite data from the Sentinel-2 constellation allow the development of accurate and low-cost tools to support risk management. In the project DRI2 (DRought Insurance - phase 2), we estimate yield losses due to drought in mountain grasslands by Sentinel-2 satellite data. DRI2 will be the basis for a digital insurance scheme which the agricultural consortia will test for the growing season 2022 over the Provinces of Bolzano and Trento in north-eastern Italy. The project builds on a close dialogue between researchers and stakeholders, including public administrations and agricultural consortia, which helps to match the needs of the forage production sector in the region of interest. We exploit a combination of Leaf Area Index (LAI) and a meteorological water stress coefficient as a proxy for grassland biomass production. We estimate LAI from Sentinel-2 satellite data by the SNAP biophysical processor and we investigate different gap-filling methods under cloudy sky conditions. We calculate the grassland production index as the growing season cumulate of the daily product between LAI and water stress. Finally, we estimate the drought index as the anomaly of the production index based on the preceding five years. We aggregate the index at the farm level based on the digital cadastral data available from the local authorities. To verify the ability of the model to reproduce the actual grassland biophysical parameters, we compare Sentinel-2 LAI with ground-based observations of LAI whose collection plan allows representing the variability of the vegetation within a Sentinel-2 pixel. We also compare Sentinel-2 LAI and measured above-ground biomass to verify whether they follow the same temporal dynamics. The preliminary results show an overall RMSE of 0.8 and r2 of 0.83 for LAI, with LAI ranging from 0.5 to 6, based on data collected from the year 2017 to the year 2020 at the Long-Term Ecological Research (LTER) experimental site of Matschertal/Val di Mazia. The comparison with measured wet biomass indicates that Sentinel-2 LAI has a high correlation with biomass, with r2 of 0.79, follows the same seasonality, and allows mowing detection. During the next project phase, we will extend the evaluation of the production index to a larger number of test sites, in which data collection is ongoing, to consider different management practices. Furthermore, we will assess the ability of the index to identify yield variations at a few experimental sites where yield data are available for several consecutive years.