Master Thesis Defense by Anna Puggaard

Title: Greenland snow accumulation estimates from Sentinel-1

Abstract:
The mass gain of the Greenland ice sheet is almost solely governed by snowfall. However, the understanding of spatiotemporal variations of snowfall is limited and has yet to be quantified adequately. Due to the large size and harsh weather conditions of the ice sheet, in-situ measurements have proven to be difficult to obtain. Hence, remote sensing is an ideal tool for monitoring the Greenland ice sheet. Particularly, microwave remote sensing is well suited for measuring Greenland snow accumulation due to its ability to measure independently of cloud cover, weather conditions, and polar night.
Sentinel-1 offers high spatialresolution coverage of the Greenland ice sheet with a 6-day temporal resolution. This project aims to estimate snow accumulation from Sentinel-1 backscatter measurements using different linear and non-linear regression models. Due to the sparse distribution of the in-situ measurements, the regression models are calibrated using the CARRA regional climate model. Regression models based on the cross co-variance between backscatter images are able to explain 35-45% of the variance in the testing data sets. However, cross-validations show that regression models cannot estimate the same temporal variability seen in CARRA. Particularly, the regression models systematically underestimate periods with high snow accumulation, likely due to a total loss of cross co-variance between backscatter images. Cross-validation further shows that the models are able to extrapolate temporally but it appears to lack the same ability for spatial extrapolation. This study lays the foundation for further assessing the potential of using Sentinel-1 to estimate snow accumulation over the Greenland ice sheet.