New sensors and methods for forest monitoring through remote sensing

Forests are sources of multifunctional services capable of meeting environmental, social, cultural, and economic demands. Therefore, climate research and international reporting activity need accurate data about forests. Remote sensing technologies nowadays provide data useful for such aim. In particular, satellite remote sensing technologies generate a constantly updated stream of data from different platforms with different characteristics, able to satisfy different purposes. New sensors and platforms sprout every year with improved capacity to meet research and operational goals and needs. In 2022, NASA alone launched four earth observation missions and planned over 10 to launch by 2030. ESA also planned to launch 9 satellite missions within the same year. Planned missions will provide new insights into the carbon and water cycle, vegetation, radiation budget, atmospheric and oceanic circulation, and much more. In parallel with the rapid advance in sensor technologies and available platforms, the capabilities for processing raw remotely sensed data into useful information also advanced thanks to the use of cloud computing technologies and Artificial Intelligence approaches. This thesis aims to explore the benefits and drawbacks of new remote sensing data and develop new tools and procedures for fully exploiting these emerging technologies. Four central studies are here presented. Study I was motivated by the very first release of data from the hyperspectral sensor carried on board the new PRISMA satellite the first mission for Earth Observation (EO) completely developed by the Italian Space Agency (ASI). So due to the very innovative type of data it was interesting to investigate their potential contribution for mapping forest areas in Italy. We analysed the band separability in two study areas, for two types of nomenclature systems and we compared the results against the well-known Sentinel-2's Multi-Spectral Instrument (MSI). We found that PRISMA sensor, allowed for a better discrimination in all forest types, increasing the performance when the complexity of the nomenclature system also increased. PRISMA achieved an average improvement of 40% for the discrimination between two forest categories (coniferous vs. broadleaves) and of 102% in the discrimination between five forest types based on main tree species groups. In the second part of my PhD I concentrated my effort on studying the use of remotely sensed data for monitoring forests on larger areas, and especially how to integrate this data with existing field-based systems such as the National Forest Inventory. In study II we investigated on how to optimise the wall-to-wall national growing stock volume estimation in Italy based on the lastly available national forest inventory (NFI) data (INFC2005). For such a purpose we compared several forest masks (FMs), and for each test we calculated model assisted estimations that were compared against the official national forest inventory estimates finding a negative correlation between the accuracies of the FMs and the differences between the model-assisted growning stock volume (GSV) estimates and the NFI estimate, demonstrating that the choice of the FM plays an important role in GSV estimation when using the model-assisted estimator. At the national and regional levels, the model-assisted GSV estimates based on the FM constructed as a mosaic of local forest maps were closest to the official NFI estimates with r2 = 0.986 and r2 = 0.972, for total and mean GSV, respectively. In Study III presents we were interested in integrating GEDI (Global Ecosystem Dynamics Investigation) data into such wall-to-wall spatial estimation of forest variables. GEDI is a cutting-edge spaceborne full waveform LiDAR specifically conceived to study vegetation dynamics and retrieve vertical vegetation structure. But since the elaboration of raw GEDI data is time consuming we presented the development of a new open-source R package (GEDI4R) that provides efficient methods for downloading, reading, clipping, visualizing, and exporting the new GEDI level 4A data. Finally, in the last study (IV) we presented a new methodology where the different elements developed in the previous studies were used to produce yearly high-resolution forest above-ground carbon pools and growing stock volume maps. The idea is to provide small-area estimations based on integrating several EO-based products with NFI data in a modeling environment. These new products allow the spatial analysis of the annual forest carbon stock changes for Italy to fit better the international reporting requirements, consistent with the IPCC guidelines. Additionally, during my Ph.D., I participated as a co-author inine otherne papers and in 21 conference contributions (5 of them as first author).