Estimation of Snow Cover Changes on the Nevado del Ruiz Volcano Between 2019 and 2021 Using Sentinel-2 Imagery

This study presents a multitemporal evaluation of snow cover changes on the Nevado del Ruiz Volcano between 2019 and 2021, using Sentinel-2 images and the Normalized Difference Snow Index (NDSI). Annual mean NDSI values were calculated, and anomaly analysis was conducted by comparing each year to the overall median of the period. The results show fluctuations in the snow cover, with a possible recovery in 2020 and 2021, likely related to variations in temperature and precipitation. Annual NDSI values and anomaly histograms helped identify trends of snow reduction and recovery. Additionally, a visual check and confusion matrix analysis were performed, establishing that the optimal NDSI threshold for the study area is 0.4. These findings are crucial for understanding glacier dynamics in response to climatic variability and their impact on local water resources. Mountain glaciers in the tropical Andes are retreating rapidly, altering regional hydrology and highlighting the need for systematic snow-cover monitoring. This study quantifies interannual variations in the snow extent of Nevado del Ruiz (Colombia) from 2019 to 2021 using Sentinel-2 MultiSpectral Instrument imagery and the Normalized Difference Snow Index (NDSI). Annual median NDSI composites were generated after cloud masking and were compared with the multi-annual median to obtain anomaly rasters. Two detection thresholds (0.2 and 0.4) were evaluated to distinguish total snow cover from compact snow; classification accuracy was assessed against 39 visually interpreted PlanetScope scenes through confusion-matrix metrics. Results reveal a contraction of snow cover in 2019 followed by a relative expansion in 2020 and 2021. The proportion of positive anomalies increased from 32.9 % in 2019 to 65.4 % in 2021, consistent with lower mean air temperatures and higher annual precipitation recorded at the nearby Las Brisas meteorological station. The 0.4 threshold achieved higher overall accuracy (0.70) and specificity (0.90), indicating greater reliability for mapping dense snow, whereas the 0.2 threshold maximised sensitivity but overestimated marginal snow zones. These findings corroborate previous reports of threshold variability across mountain environments and underscore the importance of locally calibrated parameters.