Remote sensing for the assessment of spatio-temporal variability of grassland cover and phenology

Global change is impacting grasslands through multiple processes and is driving severe consequences to their structure and functioning, and to the ecosystem services they provide. The recent advancements in remote sensing imagery availability offer new opportunities to tackle the challenge of grasslands monitoring, providing unprecedented revisiting frequency on wide areas at fine spatial resolution. Unfortunately, there are still few standardized indicators available to track grassland processes, and many processes still lack a thorough understanding. During my Ph.D., I focused on four main goals: i) assessing the grassland fractional vegetation cover prediction capability of newly available remote sensing products; ii) developing a easy to use, free, and cloud-based tool for grassland management intensity monitoring; iii) developing a workflow for grassland flowering phenology extraction using time-lapse cameras; iv) better understanding how plant phenology is shifting in climatically heterogenous mountain landscapes, and how this is affecting ecosystem productivity. Our findings demonstrated that the raw spectral signature of grasslands does not exhibit a linear variation across the fractional vegetation cover gradient, and that Sentinel-2 and PlanetScope have a higher fractional vegetation cover prediction capability compared to previously available imagery, especially in areas under patchy degradation and restoration processes. We introduced a model for estimating grassland mowing frequency, which can effectively be used under different management and environmental conditions. It was validated on small and fragmented parcels compared to previous studies, and it can be run using a provided ready to use code working on a cloud platform. We presented a new workflow for grassland flowering phenology extraction of single (or group of) species from time-lapse cameras. The workflow opened new possibilities for phenological studies, overcoming laborious and time-consuming ground-based vegetation observations. In the fourth study, we revealed substantial differences in the phenological response among vegetation types and across elevations in the European mountains over the last two decades. In grasslands, spring phenology was advanced at high altitudes and delayed at low altitudes, thus becoming more uniform along the elevational gradient, while in deciduous forests we observed the opposite trend. Remote sensing data indicated that growing season length has not been the primary factor limiting productivity over the last two decades. Therefore, it is crucial to incorporate the decoupling between phenology and productivity when simulating the potential carbon uptake of terrestrial ecosystems in future climate change scenarios. Overall, these four studies showed that remote sensing images and processing workflows can greatly contribute to a better understanding of human- and climate-induced processes impacting grassland and forest ecosystems.