Glaciers worldwide are shrinking rapidly due to rising global temperatures. This retreat affects glaciers across all regions—regardless of size or elevation—including the Alps, Andes, Himalayas, Alaska, and the tropics. Even high-altitude glaciers, once considered stable, now show signs of accelerated thinning. In the European Alps, glacier retreat has intensified in recent decades. To assess climate impacts and predict future changes, this study focuses on estimating the Equilibrium Line Altitude (ELA)—the elevation where accumulation equals ablation, serving as a key indicator of glacier health. This study integrates remote sensing and numerical modeling to assess glacier response in the Monte Rosa region. Satellite imagery from Landsat (1985–2014) and Sentinel-1 (2015–2023) have been preprocessed through band normalization, cloud and shadow masking, and band mathematics (NIR–SWIR enhancement). The snow, terrain, and ice surfaces are classified using K-means clustering, with manual segmentation for validation to ensure accuracy and Otsu segmentation for comparison. We compute Snow Line Altitude (SLA) by zoning classified pixels over a digital elevation model (DEM). The computed time series is validated on two glaciers: the Careser glacier from eastern Italian Alps and Schwarzberg glacier from the swiss alps but it is part of Monte Rosa Massif. Afterwards, we employed a glacier climate model- Open Global Glacier Model (OGGM)- to compute ELA and volume of the region under study for the historical and future periods. The model has been calibrated using geodetic mass balance data and consensus thickness estimates to match with the SLA prepared by clustering. We have used climate datasets sourced from General Circulation Models (GCMs), EURO-CORDEX Regional Climate Models (RCMs), and high-resolution CPRCM, bias-corrected using the delta method. We find the image segmentation using K-means clustering significantly outperforms the Otsu thresholding in snow-ice-terrain classification, achieving an F1 score of 0.88 for the Careser Glacier, compared to 0.56 for Otsu. For the Monte Rosa Glaciers, K-means achieved an F1 score of 0.95. The SLA derived from segmented satellite imagery is validated against the World Glacier Monitoring Service (WGMS) ELA data and depicted a strong correlation of 0.7 for Careser and 0.57 for Schwarzberg. The modelled ELA achieved a correlation coefficient of 0.84 and coefficient of determination (R2) of 0.70 against observed SLA, with a mean absolute error of 67 m. Among climate datasets, CPRCM and RCM driven historical data yielded the bettter agreement with observations, showing a correlation of 0.58 (R2 = 0.34), compared to GCM (correlation = 0.20, R2 = 0.04). Future projections reveal scenario-dependent ELA increases, with ELA rising above 4200m ± 5.3% by the year 2100 under RCP8.5, threatening the persistence of mid-elevation glaciers. Modeled glacier volumes decline by 84% under RCP4.5 and 72% under RCP2.6, and up to 95% under RCP8.5 from 2005 to 2100 under RCM. The relationship between ELA and glacier volume reveals tipping-point behavior: when ELA crosses 3547m ± 43.6m, glacier volume loss accelerates non-linearly. This threshold marks a critical elevation beyond which the glaciers enter irreversible decline.
I ghiacciai in tutto il mondo si stanno riducendo rapidamente a causa dell’aumento delle temperature globali. Questo arretramento interessa i ghiacciai in tutte le regioni—indipendentemente dalla dimensione o dall’altitudine—incluse le Alpi, le Ande, l’Himalaya, l’Alaska e le regioni tropicali. Anche i ghiacciai ad alta quota, un tempo considerati stabili, mostrano ora segni di assottigliamento accelerato. Nelle Alpi europee, il ritiro dei ghiacciai si è intensificato negli ultimi decenni. Per valutare gli impatti climatici e prevedere i cambiamenti futuri, questo studio si concentra sulla stima della Quota della Linea di Equilibrio (ELA)—l’altitudine in cui l’accumulo equivale all’ablazione, servendo come indicatore chiave della salute dei ghiacciai. Questo studio integra il telerilevamento e la modellazione numerica per valutare la risposta glaciale nella regione del Monte Rosa. Le immagini satellitari di Landsat (1985–2014) e Sentinel-1 (2015–2023) sono state pre-processate tramite normalizzazione delle bande, mascheramento di nuvole e ombre, e matematica delle bande (potenziamento NIR–SWIR). Le superfici di neve, terreno e ghiaccio sono classificate utilizzando il clustering K-means, con segmentazione manuale per la validazione dell’accuratezza e segmentazione Otsu per confronto. Calcoliamo l’Altitudine della Linea della Neve (SLA) zonando i pixel classificati su un modello digitale di elevazione (DEM). La serie temporale calcolata è validata su due ghiacciai: il ghiacciaio del Careser nelle Alpi orientali italiane e il ghiacciaio Schwarzberg nelle Alpi svizzere ma appartenente al Massiccio del Monte Rosa. Successivamente, abbiamo impiegato un modello climatico-glaciale—Open Global Glacier Model (OGGM)—per calcolare ELA e volume della regione in studio per i periodi storici e futuri. Il modello è stato calibrato utilizzando dati di bilancio di massa geodetico e stime di spessore da consenso per essere coerente con la SLA ottenuta tramite clustering. Abbiamo utilizzato dataset climatici provenienti da Modelli di Circolazione Generale (GCM), Modelli Climatici Regionali (RCM) EURO-CORDEX, e CPRCM ad alta risoluzione, corretti per bias utilizzando il metodo delta. Abbiamo riscontrato che la segmentazione delle immagini mediante clustering K-means supera significativamente la soglia di Otsu nella classificazione neve-ghiaccio-terreno, ottenendo un punteggio F1 di 0.88 per il ghiacciaio del Careser, rispetto a 0.56 per Otsu. Per i ghiacciai del Monte Rosa, il K-means ha ottenuto un punteggio F1 di 0.95. La SLA derivata da immagini satellitari segmentate è validata rispetto ai dati ELA del World Glacier Monitoring Service (WGMS) e mostra una forte correlazione di 0.7 per Careser e 0.57 per Schwarzberg. L’ELA modellata ha ottenuto un coefficiente di correlazione di 0.84 e un coefficiente di determinazione (R2) di 0.70 rispetto alla SLA osservata, con un errore assoluto medio di 67 m. Tra i dataset climatici, i dati storici derivati da CPRCM e RCM hanno mostrato il miglior accordo con le osservazioni, con una correlazione di 0.58 (R2 = 0.34), rispetto ai GCM (correlazione = 0.20, R2 = 0.04). Le proiezioni future rivelano aumenti di ELA dipendenti dallo scenario, con ELA in aumento oltre 4200m ± 5.3% entro l’anno 2100 sotto RCP8.5, minacciando la persistenza dei ghiacciai a media altitudine. I volumi glaciali modellati diminuiscono dell’84% sotto RCP4.5 e del 72% sotto RCP2.6, e fino al 95% sotto RCP8.5 dal 2005 al 2100 secondo RCM. La relazione tra ELA e volume glaciale rivela un comportamento di tipping-point: quando l’ELA supera 3547m ± 43.6m, la perdita di volume glaciale accelera in modo non lineare. Questa soglia rappresenta un’altitudine critica oltre la quale i ghiacciai entrano in un declino irreversibile.
Characterization and Monitoring of Glacial Environments under a Changing Climate in the Alpine Region
AYUB, SOBIA
2025
Abstract
Glaciers worldwide are shrinking rapidly due to rising global temperatures. This retreat affects glaciers across all regions—regardless of size or elevation—including the Alps, Andes, Himalayas, Alaska, and the tropics. Even high-altitude glaciers, once considered stable, now show signs of accelerated thinning. In the European Alps, glacier retreat has intensified in recent decades. To assess climate impacts and predict future changes, this study focuses on estimating the Equilibrium Line Altitude (ELA)—the elevation where accumulation equals ablation, serving as a key indicator of glacier health. This study integrates remote sensing and numerical modeling to assess glacier response in the Monte Rosa region. Satellite imagery from Landsat (1985–2014) and Sentinel-1 (2015–2023) have been preprocessed through band normalization, cloud and shadow masking, and band mathematics (NIR–SWIR enhancement). The snow, terrain, and ice surfaces are classified using K-means clustering, with manual segmentation for validation to ensure accuracy and Otsu segmentation for comparison. We compute Snow Line Altitude (SLA) by zoning classified pixels over a digital elevation model (DEM). The computed time series is validated on two glaciers: the Careser glacier from eastern Italian Alps and Schwarzberg glacier from the swiss alps but it is part of Monte Rosa Massif. Afterwards, we employed a glacier climate model- Open Global Glacier Model (OGGM)- to compute ELA and volume of the region under study for the historical and future periods. The model has been calibrated using geodetic mass balance data and consensus thickness estimates to match with the SLA prepared by clustering. We have used climate datasets sourced from General Circulation Models (GCMs), EURO-CORDEX Regional Climate Models (RCMs), and high-resolution CPRCM, bias-corrected using the delta method. We find the image segmentation using K-means clustering significantly outperforms the Otsu thresholding in snow-ice-terrain classification, achieving an F1 score of 0.88 for the Careser Glacier, compared to 0.56 for Otsu. For the Monte Rosa Glaciers, K-means achieved an F1 score of 0.95. The SLA derived from segmented satellite imagery is validated against the World Glacier Monitoring Service (WGMS) ELA data and depicted a strong correlation of 0.7 for Careser and 0.57 for Schwarzberg. The modelled ELA achieved a correlation coefficient of 0.84 and coefficient of determination (R2) of 0.70 against observed SLA, with a mean absolute error of 67 m. Among climate datasets, CPRCM and RCM driven historical data yielded the bettter agreement with observations, showing a correlation of 0.58 (R2 = 0.34), compared to GCM (correlation = 0.20, R2 = 0.04). Future projections reveal scenario-dependent ELA increases, with ELA rising above 4200m ± 5.3% by the year 2100 under RCP8.5, threatening the persistence of mid-elevation glaciers. Modeled glacier volumes decline by 84% under RCP4.5 and 72% under RCP2.6, and up to 95% under RCP8.5 from 2005 to 2100 under RCM. The relationship between ELA and glacier volume reveals tipping-point behavior: when ELA crosses 3547m ± 43.6m, glacier volume loss accelerates non-linearly. This threshold marks a critical elevation beyond which the glaciers enter irreversible decline.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD SDC Thesis_Ayub Sobia.pdf
embargo fino al 13/11/2026
Licenza:
Tutti i diritti riservati
Dimensione
28.48 MB
Formato
Adobe PDF
|
28.48 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/310096
URN:NBN:IT:IUSSPAVIA-310096