Mapping glaciers using time-series of remote sensing data

Advances and retreats of mountain glaciers in Arctic and alpine areas are one of the most visible signs of climatic changes. Glaciers therefore play an important role as sensitive climate change indicators. Glacier products derived from remote sensing sources are important for change studies and as input to glaciological and climate models. The research in this thesis includes analyses that range from using the most common glacier mapping methods, to exploring and developing new glacier mapping methods for dense time-series data. A change study of glacier area and length changes of all glaciers in Norway was performed by comparing old analogue maps (from the period 1950-1980) with satellite images (from the period 1999-2006). The glacier retreat was found to be 11 % over a mean period of ~30 years, corresponding to a glacier loss of 326 km2 (ca. 11 km2 per year). The largest reduction of glacier area was found for the northernmost glaciers, with a 17 % decrease, and several glaciers had been downwasting and disintegrating. Currently, many of these glaciers in Finnmark only consist of ice patches, and will soon be completely melted away. The current multi-spectral band ratio method to retrieve glacier outlines was applied on new Sentinel-2A images, which resulted in more accurately derived glacier outlines compared to previous satellite sensors, especially due to improved spatial resolution. To meet the needs of newly launched optical satellite sensors that provide dense medium resolution time-series, new and robust glacier mapping methods were developed and investigated. By studying pixel values on glaciers in a chronological order, a seasonal pattern of snow and ice was found. This temporal variation on glaciers was found to be caused most likely by the sensitivity of the short wave infrared band to snow and ice grain size. This pixel pattern difference on glaciers compared to pixels outside the glaciers could be exploited to improve glacier mapping in the future. Optical satellite sensors cannot retrieve usable data under dark or cloudy conditions. This is especially challenging in maritime glacier regions with frequent cloud cover. By using radar satellites, which can receive transmitted signals though clouds and in the winter-time, glaciers can be observed when optical imagery is missing, although in a different way since the radar satellite sensors operate in different wavelenghts/frequencies than optical satellites. The existing and future extensive amount of multi-sensor timeseries data can be used for multiple purposes such as retrieving and tracking the transient snow line throughout the melt season, to study changes in glacier facies and firn evolution patterns, to detect winter weather events, and to derive glacier outlines. In a broad sense, this thesis has contributed to understanding how properties of glaciers are observed and manifested in dense satellite data time-series. These findings have implications for future glacier mapping methods, and provide a step towards implementing automatic mapping approaches.