INVESTIGATING THE DEPENDENCY BETWEEN SENTINEL-2 MULTISPECTRAL IMAGES AND GROUND-BASED FIELD MEASUREMENTS OF SOIL MOISTURE IN MENDATICA, LIGURIA, ITALY

This volume collects 34 selected articles that have been grouped into 4 chapters organized in thematic clusters. They refer to the impacts of human activity on the environment, highlighting the state-of-the-art potentialities of Earth Observation in the environmental analysis, monitoring, and protection, with particular concerns about climate change-related effects, natural resources, and human health. 1. Urban & cultural heritage focuses on contributions given by EO in the framework of environmental risks, urban planning, cultural heritage monitoring and documentation, and energetic efficiency. 2. Environment includes works representing studies at the territorial scale about different environmental components in relationship with threads that climate change has introduced. 3. Agriculture and Forestry, show the increasing role of EO in the agricultural and forest fields, related to their roles to the environmental sustainability of management and carbon stocks balances. 4. Algorithm & sensors presents and comparing new approaches and algorithms to support sensitive needs from civil society, like photovoltaic plants monitoring and location, oil spills, infrastructures monitoring. In the contribution, the potential of Sentinel-2 (S-2) multispectral satellite images for Surface Soil Moisture (SSM) estimate is investigated. For this purpose, dependency is looked for between S-2 images and an 18-months (from 1st of January 2020 to 30th of June 2021) dataset of hourly SSM measurements, acquired at four different depths (-10cm, -35cm, -55cm, -85cm) from each of the nodes of a monitoring network in Mendatica (Liguria, Italy). Data acquired by the sensors were previously calibrated, considering the soil-specific characteristics of the areas, and the reliability of the dataset was verified. After performing the required preprocessing on satellite images, the performance of three nonlinear regression methods, when applied to four different types of inputs (12 spectral channels, NDVI, NDWI and NDMI), was quantitatively assessed.