| dc.description.abstract | The polar regions play a fundamental role in the ongoing climate warming and despite their remoteness and distance from human civilization, we have a great interest in understanding the changes taking place in the Arctic and Antarctic. In this thesis, the focus is on the sea-ice system: frozen ocean water floating on the surface. It is found in both the Arctic and Antarctic and covers up to 10% of the Earth’s oceans.
Between 2009 and 2019, NASA conducted the major, pioneering airborne Operation IceBridge program.The goal was to get high-resolution insights about all aspects of the polar regions. One of those aims was to measure snow depth on sea ice. Snow on sea ice until today remains as the largest uncertainty in this geophysical system. On the one hand the snow cover dictates physical aspects like the radiation balance, growth and melt patterns and thus the role of sea ice in the larger climate system. On the other hand lacking knowledge about snow loading (as a product of depth and density) is majorly hindering our observing capability of other sea ice
related variables such as ice thickness from satellite records. To get better, more accurate snow depth estimates, the so-called snow radar was implemented and further improved over the lifespan of Operation IceBridge (OIB).
The OIB observational snow depth on sea ice record is the single most dataset used in research domains as model initialization and evaluation as well as satellite validation.
In this study, I revisit the OIB snow radar data under the light of recent advancements in the field of snow remote sensing. This is done by performing a comparative analysis with a previously unused, extensive dataset of in-situ snow depths collected within the Canadian Archipelago in 2016.
I show that the widely-used, official National Snow and Ice Data Center (NSIDC) Quicklook (QL) data product does not resemble sub-kilometer snow depth distributions on Arctic sea ice. I further evaluate different methods to obtain higher quality snow depth estimates using the same in-situ data. Thus I am able to improve the fit from R = 0.25 and a negative, underestimating bias of 5 cm from the QL data product to R = 0.80 with no bias using an existing algorithm based on Continuous Wavelet Transform (CWT). By comparing with in-situ data, I can provide robust uncertainty estimates based on the snow radar parameters and the evaluation exercise.
Transferring and upscaling the results to all the pan-Arctic OIB flights in 2016, I find that snow depth differences between QL and the CWT approach are in the order of 25% (7 cm). These differences propagate into the sea-ice thickness estimates, which change accordingly by about -15% (30 cm).
In conclusion, I recommend caution when using the QL data product on small spatial scales. Further, I advocate for the recalculation of an Operation IceBridge sea ice dataset for the Arctic. While a start has been performed in this study for the snow depth variable, a potential recalculation will require additional research on the various other sea ice variables. | |
