From peak to planet: advancing multi-scale detection of snowmelt timing with satellite radar

Abstract

Snowmelt runoff onset represents a critical parameter in mountain hydrology, marking the beginning of increased water availability for over a billion people who depend on seasonal snowmelt, as well as being a key indicator of climate change. Despite its importance, systematic high-resolution observations of snowmelt timing across the Earth’s diverse mountain regions are limited with current monitoring approaches: sparse in-situ networks, cloud-obscured optical remote sensing, and coarse passive microwave observations.This dissertation advances the detection of snowmelt runoff onset from local to global scales using Synthetic Aperture Radar (SAR) observations from the European Space Agency Sentinel-1 mission. SAR overcomes key limitations of other snow monitoring approaches through its ability to observe through clouds and in darkness, high spatial resolution, and sensitivity to liquid water content changes in snowpack that coincide with snowmelt runoff onset. Chapter 2 establishes the methodological foundation of this work by demonstrating scalable detection of snowmelt runoff onset using Sentinel-1 backscatter time-series analysis over stratovolcanoes in the Cascade Range of North America. By integrating multiple orbital geometries, the approach achieves a median temporal resolution of 3.9 days at 10-meter spatial resolution. Validation with in-situ snow pillow measurements shows a median offset of 1 day and median absolute offset of 10 days for the SAR snowmelt runoff onset timing estimates. Analysis across elevation gradients reveals strong topographic control, with median delays of 4.9 days per 100 meter elevation gain, as well as dramatic interannual variability including 25-day early runoff onset during the 2015 snow drought. Chapter 3 scales this methodology, processing over 3.9 million Sentinel-1 images to create a global snowmelt runoff onset dataset spanning the 10-year period from 2015 through 2024. To enable robust detection of snowmelt runoff onset across diverse environments, we develop a custom MODIS-derived snow phenology dataset that provides spatial and temporal constraints for runoff onset identification. Systematic analysis establishes empirically-derived recommendations for snowmelt runoff onset dataset application based on forest cover, snow accumulation, and observation frequency. Validation against runoff onset estimates from over 900 in-situ snow pillows across the Western U.S. demonstrates robust dataset performance across different mountain environments. This dataset provides an unprecedented look at annual snowmelt runoff onset on a global scale, with 80-meter spatial resolution and 9.3-day average temporal resolution. Chapter 4 presents the first comprehensive global analysis of snowmelt timing patterns and controls across 150 major mountain ranges. Continental-scale analysis reveals systemic weakening of elevation gradients from mid-latitudes toward polar regions, as well as snowmelt runoff onset timing differences between sunny and shaded areas that varies seasonally but reaches maximum values of 20-60 days around early- to mid-spring. Mountain range-scale aggregation reveals a median runoff onset delay of 3.5 days per 100 meters of elevation gain, but with substantial variability, reflecting differences in climate, topography, and snowpack characteristics. Individual mountain ranges show variable aspect differences in runoff onset timing depending on local climate and topography, with some clear-sky continental ranges exhibiting differences exceeding 40 days while nearby cloudier ranges show minimal aspect differences. The tropical Andes and Tibetan Plateau mountain ranges display the highest interannual variability of snowmelt runoff onset timing, often exceeding 30 days. Temperature sensitivity analysis reveals that 72% of mountain ranges show correlations between spring (March-May) temperature anomaly and runoff onset timing, with most mid-latitude mountain ranges exhibiting runoff onset 8 to 13 days early for every 1°C warmer spring average temperature. Chapter 5 synthesizes these findings and examines implications for water-dependent populations through basin-scale analysis in High-Mountain Asia and western North America. The analysis reveals coherent regional patterns in snowmelt runoff onset during documented anomalous weather events, including the 2015 western North American “snow drought” and 2022 High-Mountain Asia “mega-heatwave”, demonstrating how synoptic-scale weather events can produce spatially coherent 20-40 day shifts in snowmelt runoff onset timing across entire regions. Preliminary vulnerability assessment identifies high-population basins with high interannual variability in snowmelt runoff onset timing, with implications for water security. This dissertation establishes the first systematic framework for observing snowmelt runoff onset at high resolution across global mountain regions. The decade-long record provides unprecedented detail of snowmelt timing patterns, quantifies fundamental physical controls operating across diverse environments, and documents substantial interannual variability linked to average spring air temperature. The methodology, open-source tools, and open datasets from this dissertation will enable improved understanding of snow hydrology and support more effective water resource management in an era of increasing environmental change.