Identification of Seasonal Snow Phase Changes From C-Band SAR Time Series With Dynamic Thresholds

Abstract

In cryospheric studies, the most critical variable is estimating the beginning and evolution of the seasonal snow accumulation and its melting process. We propose and evaluate a new method for identifying seasonal snow cover phase changes in the Argentinean Andes using time series of Sentinel-1 synthetic aperture radar data available in the Google Earth Engine platform. We used meteorological and optical Sentinel-2 data to validate snow presence. First, we investigated seasonal snow cover dynamics in different regions of interest (ROIs). We identified three land surface cover periods: bare soil, dry snow, and melting snow. This finding is significant because other studies show that bare soil and dry snow have similar backscattering responses in C-band. Our methodology uses time series derivatives and their positive and negative anomalies. Finally, we compare our results with those obtained with a fixed threshold change detection approach. Our method was able to detect the phase change between bare soil and dry snow period in 75% of the ROIs, while a fixed threshold of $-\text{2} \,\mathrm{dB}$ only detects it in 42% of the cases. Furthermore, the derivative method also detects in 92% of the ROIs time series the beginning of the melting period, showing that it is a promising methodology for operative systems.

Keywords

• Anomalies
• derivatives
• Google Earth Engine (GEE)
• synthetic aperture radar (SAR)
• seasonal snow
• time series