Fine spatial scale modelling of Trentino past forest landscape and future change scenarios to study ecosystem services through the years

Landscape in Europe has dramatically changed in the last decades. This has been especially true for Alpine regions, where the progressive urbanization of the valleys has been accom- panied by the abandonment of smaller villages and areas at higher elevation. This trend has been clearly observable in the Provincia Autonoma di Trento (PAT) region in the Italian Alps. The impact has been substantial for many rural areas, with the progressive shrinking of meadows and pastures due to the forest natural recolonization. These modifications of the landscape affect biodiversity, social and cultural dynamics, including landscape perception and some ecosystem services. Literature review showed that this topic has been addressed by several authors across the Alps, but their researches are limited in space coverage, spatial resolution and time span. This thesis aims to create a comprehensive dataset of historical maps and multitemporal orthophotos in the area of PAT to perform data analysis to identify the changes in forest and open areas, being an evaluation of how these changes affected land- scape structure and ecosystems, create a future change scenario for a test area and highlight some major changes in ecosystem services through time. In this study a high resolution dataset of maps covering the whole PAT area for over a century was developed. The earlier representation of the PAT territory which contained reliable data about forest coverage was considered is the Historic Cadastral maps of the 1859. These maps in fact systematically and accurately represented the land use of each parcel in the Habsburg Empire, included the PAT. Then, the Italian Kingdom Forest Maps, was the next important source of information about the forest coverage after World War I, before coming to the most recent datasets of the greyscale images of 1954, 1994 and the multiband images of 2006 and 2015. The purpose of the dataset development is twofold: to create a series of maps describing the forest and open areas coverage in the last 160 years for the whole PAT on one hand and to setup and test procedures to extract the relevant information from imagery and historical maps on the other. The datasets were archived, processed and analysed using the Free and Open Source Software (FOSS) GIS GRASS, QGIS and R. The goal set by this work was achieved by a remote sensed analysis of said maps and aerial imagery. A series of procedures were applied to extract a land use map, with the forest categories reaching a level of detail rarely achieved for a study area of such an extension (6200 km2 ). The resolution of the original maps is in fact at a meter level, whereas the coarser resampling adopted is 10mx10m pixels. The great variety and size of the input data required the development, along the main part of the research, of a series of new tools for automatizing the analysis of the aerial imagery, to reduce the user intervention. New tools for historic map classification were as well developed, for eliminating from the resulting maps of land use from symbols (e.g.: signs), thus enhancing the results. Once the multitemporal forest maps were obtained, the second phase of the current work was a qualitative and quantitative assessment of the forest coverage and how it changed. This was performed by the evaluation of a number of landscape metrics, indexes used to quantify the compaction or the rarefaction of the forest areas. A recurring issue in the current Literature on the topic of landscape metrics was identified along their analysis in the current work, that was extensively studied. This highlighted the importance of specifying some parameters in the most used landscape fragmentation analy- sis software to make the results of different studies properly comparable. Within this analysis a set of data coming from other maps were used to characterize the process of afforestation in PAT, such as the potential forest maps, which were used to quantify the area of potential forest which were actually afforested through the years, the Digital Ele- vation Model, which was used to quantify the changes in forest area at a different ranges of altitude, and finally the forest class map, which was used to estimate how afforestation has affected each single forest type. The output forest maps were used to analyse and estimate some ecosystem services, in par- ticular the protection from soil erosion, the changes in biodiversity and the landscape of the forests. Finally, a procedure for the analysis of future changes scenarios was set up to study how afforestation will proceed in absence of external factors in a protected area of PAT. The pro- cedure was developed using Agent Based Models, which considers trees as thinking agents, able to choose where to expand the forest area. The first part of the results achieved consists in a temporal series of maps representing the situation of the forest in each year of the considered dataset. The analysis of these maps suggests a trend of afforestation across the PAT territory. The forest maps were then reclassi- fied by altitude ranges and forest types to show how the afforestation proceeded at different altitudes and forest types. The results showed that forest expansion acted homogeneously through different altitude and forest types. The analysis of a selected set of landscape met- rics showed a progressive compaction of the forests at the expenses of the open areas, in each altitude range and for each forest type. This generated on one hand a benefit for all those ecosystem services linked to a high forest cover, while reduced ecotonal habitats and affected biodiversity distribution and quality. Finally the ABM procedure resulted in a set of maps representing a possible evolution of the forest in an area of PAT, which represented a similar situation respect to other simulations developed using different models in the same area. A second part of the result achieved in the current work consisted in new open source tools for image analysis developed for achieving the results showed, but with a potentially wider field of application, along with new procedure for the evaluation of the image classification. The current work fulfilled its aims, while providing in the meantime new tools and enhance- ment of existing tools for remote sensing and leaving as heritage a large dataset that will be used to deepen he knowledge of the territory of PAT, and, more widely to study emerging pattern in afforestation in an alpine environment.