ANALYSIS OF LAND COVER AND LAND COVER CHANGE THROUGH REMOTE SENSING IN THE DOLOMITI BELLUNESI NATIONAL PARK

Abstract

This work focuses on the spatial changes in semi-natural grasslands habitats over the past 60 years in the Dolomiti Bellunesi National Park, Italy. Semi-natural grasslands have been created and maintained over the centuries by human activities, resulting in biodiversity-rich landscape mosaics, but the progressive abandonment of traditional grassland management activities such as low-intensity grazing and haymaking after World War II resulted in a widespread phenomenon of forest expansion, declining biodiversity and shrinking or disappearance of grassland areas. Remote sensing technologies and Geographic Information Systems (GIS) represent invaluable tools for understanding the dynamics of this phenomenon; GIS software greatly improves image processing, data organization and quantitative analysis of the multi-temporal remote sensing da-tasets used for studying the dynamics of environmental patterns and processes at the spatial and temporal scale of interest. This in turn can provide support for decision-making and nature con-servation management, an aspect of fundamental importance for semi-natural ecosystems be-cause they require active management to sustain their ecosystem services. In the first part of this work, high-resolution maps of distribution of semi-natural grassland habitats within the Dolomiti Bellunesi National Park (Italy) was created based on two aerial image-ry datasets: the 1954 GAI flight (the oldest available photographic record of the study area, dating back to before the start of mountain abandonment) and the 2012 TELAER flight. The distribution maps were created through the visual interpretation of aerial photographs and manual digitization of grassland polygons with the aid of GIS software. The two distribution maps were subsequently combined and compared to assess the spatial distribution and magnitude of changes over the 58-years time period. The second part of the study explores the potential of studying the reduction of semi-natural grassland habitats by using Landsat satellite images, employing supervised pixel-based methods for classification and change detection and the NDVI vegetation index to increase the separability between different vegetated surfaces. First, a seasonality study was carried out to identify the most suitable period for ensuring maximum land cover separability during the vegetative season; for this purpose, a series of 5 Landsat TM images from 2005 (May-October) were classified inde-pendently through a semi-automated algorithm and their classification accuracies compared. Based on the best date identified by the seasonality study, two Landsat images (from the beginning and the end of Landsat TM data availability period) were chosen for NDVI calculation and NDVI differencing change detection. In this change detection procedure, the NDVI values of the two images were subtracted pixel-by-pixel, resulting in a ΔNDVI change image that can be classified by setting an appropriate threshold to distinguish significant changes (increase or decrease of vegetation cover, where an increase is associated to forest expansion or densification) from situations where the land cover remained essentially unchanged.