BRIDGING PAST AND PRESENT: INTEGRATING RESURVEY DATA WITH LANDSAT TIME SERIES FOR ENHANCED VEGETATION CHANGE DETECTION IN ITALIAN COASTAL DUNES

Coastal dune ecosystems, among the most threatened habitats in Italy and Europe, host highly specialized plant communities shaped by strong environmental gradients (1). While historical vegetation resurvey data allow for the assessment of changes between discrete time points, they provide no insight into the dynamics occurring in the intervening period (2). The integration of remote sensing, particularly historical Landsat time series, could represent a powerful tool to capture continuous vegetation shifts and enhance our understanding of long-term ecological trajectories. This study explores the relationship between spectral changes and coastal dune habitats, improving our understanding and developing a faster, less labor-intensive monitoring tool. For resurveyed vegetation data we selected 244 plots from the ResurveyDunes database (3) based on size, time and habitat type, classified into three EUNIS habitats: N12 (annual drift vegetation), N14 (shifting dunes), and N16 (dune grasslands). We calculated metrics of change, such as Bray-Curtis dissimilarity, to quantify vegetation changes. For Remote Sensing, Landsat time series data (Landsat 5, 8, 9) were processed in Google Earth Engine, focusing on annual spring "greenest pixel" composites to maximize NDVI. Spectral bands, vegetation indices, and fractional cover estimates were extracted and analyzed with Generalized Linear Models (GLMs) to compare spectral and vegetation changes. Preliminary results from two time points show no significant differences in Bray-Curtis dissimilarity among habitats, but NDVI changes reveal significant productivity shifts in plots transitioning between habitat types, especially from N12 to N14. GLM analyses reveal a significant interaction between NDVI changes and vegetation dissimilarity across EUNIS classes (p < 0.01), with dune grasslands (N16) showing a strong negative relationship, indicating that productivity increases where communities remain stable (p < 0.001, estimate: -1.367). N12 and N14 show no clear trends, likely due to low vegetation cover. These findings suggest incorporating additional remote sensing variables, with future analyses planned to use the full Landsat time series to better understand long-term coastal dune vegetation changes.