Land abandonment and subsequent ecological transitions are key processes influencing the Italian landscape, driven by socio-economic and environmental factors. This research provides a national-scale assessment of these dynamics from 1985 to 2032, using two global land cover datasets: the ESA CCI Land Cover (300 m resolution, 1992–2022) and the GLC_FCS30D (30 m resolution, 1985–2022). The analysis was implemented within a GRASS GIS environment with automated scripting to ensure consistency and reproducibility. The methodology was organized around four components: (1) pixel-based statistical analysis quantifying the annual extent of each land cover class, with dynamic pixel-area calculation for cartometric accuracy; (2) computation of transition matrices to identify conversion pathways, particularly from agricultural to natural land covers; (3) comparative validation of the two datasets, resampling the 30 m data to 300 m to assess consistency through confusion matrices and accuracy metrics, showing moderate overall agreement (78–79%) with poor result for heterogeneous classes such as Shrubland (User's Accuracy ≈25%), and higher consistency for stable classes such as Water (>88%) and Cropland (>88%); and (4) development of a constrained Markov Chain model, integrating Italy-specific ecological and socio-economic constraints, to project land cover to 2032. The results indicate a landscape characterized by both persistence and change. In 2022, cropland accounted for approximately 48% of national territory and forest for 36%. Built-up areas expanded by 70–104% depending on dataset, representing the most significant transformation between 1985 and 2022. Forest cover showed a slight net decrease (–2.6%), while grassland declined more substantially (–15.2%). Projections to 2032 suggest continued urban expansion (+900 km², +5.3%), further forest loss (–376 km², –0.3%), and reductions in bareland (–27.5%) and permanent ice/snow (–19.1%). These findings suggest that although agricultural abandonment is occurring, its signal is less visible at the national scale than the expansion of urban areas. The comparison of datasets highlights the usefulness of high-resolution products for detecting fine-scale processes and the limitations of coarser-resolution data for capturing complex land cover transitions. The results provide information relevant to land management and policy, particularly in relation to balancing development pressures with the conservation of agricultural and forest ecosystems
L'abbandono dei terreni e le successive transizioni ecologiche sono processi chiave che influenzano il paesaggio italiano, guidati da fattori socio-economici e ambientali. Questa ricerca fornisce una valutazione a scala nazionale di tali dinamiche dal 1985 al 2032, utilizzando due dataset globali di copertura del suolo: ESA CCI Land Cover (risoluzione 300 m, 1992–2022) e GLC_FCS30D (risoluzione 30 m, 1985–2022). L'analisi è stata condotta in ambiente GRASS GIS, con l'ausilio di script automatizzati per garantire coerenza e riproducibilità. La metodologia è stata articolata in quattro componenti: (1) analisi statistica pixel-based per quantificare l'estensione annuale di ciascuna classe di copertura, con calcolo dinamico dell'area per garantire l'accuratezza cartometrica; (2) costruzione di matrici di transizione per identificare i percorsi di conversione, in particolare dalle superfici agricole a quelle naturali; (3) validazione comparativa dei due dataset, ricampionando i dati a 30 m su griglia a 300 m per valutarne la coerenza tramite matrici di confusione e metriche di accuratezza, con un accordo complessivo moderato (78–79%), risultato scarso per classi eterogenee come gli arbusteti (User's Accuracy ≈25 %), e maggiore coerenza per classi stabili come Acqua (>88 %) e Seminativi (>88 %); e (4) sviluppo di un modello di Catena di Markov vincolata, integrato con specifici fattori ecologici e socio-economici italiani, per proiettare la copertura del suolo al 2032. I risultati indicano un paesaggio caratterizzato da elementi di persistenza e trasformazione. Nel 2022 i seminativi rappresentavano circa il 48 % del territorio nazionale e le foreste il 36 %. Le aree urbanizzate sono aumentate del 70–104 % a seconda del dataset, costituendo la trasformazione più rilevante nel periodo 1985–2022. La copertura forestale ha mostrato una lieve riduzione netta (–2.6 %), mentre le praterie hanno registrato un calo più consistente (–15.2 %). Le proiezioni al 2032 suggeriscono un'ulteriore espansione urbana (+900 km², +5.3 %), una continua riduzione delle superfici forestali (–376 km², –0.3 %) e diminuzioni delle aree a suolo nudo (–27.5 %) e dei ghiacciai/nevi permanenti (–19.1 %). Questi risultati evidenziano che, sebbene l'abbandono agricolo sia presente, il suo segnale è meno rilevante a scala nazionale rispetto all'espansione urbana. Il confronto tra dataset mette in luce l'utilità dei prodotti ad alta risoluzione per individuare processi di piccola scala e le limitazioni dei dati a risoluzione più grossolana nel descrivere transizioni complesse di copertura del suolo. I risultati forniscono elementi utili per la gestione del territorio e per le politiche di pianificazione, in particolare rispetto alla necessità di bilanciare le pressioni dello sviluppo con la conservazione dei sistemi agricoli e forestali.
Land cover dynamics and transitions in Italy : a multi- temporal analysis using global land cover datasets
Zafari, Ghulam Abbas
2024/2025
Abstract
Land abandonment and subsequent ecological transitions are key processes influencing the Italian landscape, driven by socio-economic and environmental factors. This research provides a national-scale assessment of these dynamics from 1985 to 2032, using two global land cover datasets: the ESA CCI Land Cover (300 m resolution, 1992–2022) and the GLC_FCS30D (30 m resolution, 1985–2022). The analysis was implemented within a GRASS GIS environment with automated scripting to ensure consistency and reproducibility. The methodology was organized around four components: (1) pixel-based statistical analysis quantifying the annual extent of each land cover class, with dynamic pixel-area calculation for cartometric accuracy; (2) computation of transition matrices to identify conversion pathways, particularly from agricultural to natural land covers; (3) comparative validation of the two datasets, resampling the 30 m data to 300 m to assess consistency through confusion matrices and accuracy metrics, showing moderate overall agreement (78–79%) with poor result for heterogeneous classes such as Shrubland (User's Accuracy ≈25%), and higher consistency for stable classes such as Water (>88%) and Cropland (>88%); and (4) development of a constrained Markov Chain model, integrating Italy-specific ecological and socio-economic constraints, to project land cover to 2032. The results indicate a landscape characterized by both persistence and change. In 2022, cropland accounted for approximately 48% of national territory and forest for 36%. Built-up areas expanded by 70–104% depending on dataset, representing the most significant transformation between 1985 and 2022. Forest cover showed a slight net decrease (–2.6%), while grassland declined more substantially (–15.2%). Projections to 2032 suggest continued urban expansion (+900 km², +5.3%), further forest loss (–376 km², –0.3%), and reductions in bareland (–27.5%) and permanent ice/snow (–19.1%). These findings suggest that although agricultural abandonment is occurring, its signal is less visible at the national scale than the expansion of urban areas. The comparison of datasets highlights the usefulness of high-resolution products for detecting fine-scale processes and the limitations of coarser-resolution data for capturing complex land cover transitions. The results provide information relevant to land management and policy, particularly in relation to balancing development pressures with the conservation of agricultural and forest ecosystems| File | Dimensione | Formato | |
|---|---|---|---|
|
Ghulam_Abbas_Zafari_Thesis.pdf
accessibile in internet solo dagli utenti autorizzati
Dimensione
74.22 MB
Formato
Adobe PDF
|
74.22 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/242970