Estimating sea ice volume inside the Arctic pole hole in the Envisat era

Sammanfattning

Arctic sea ice has been declining for several decades due to climate change and this in turn impacts both global and local climates. Satellite remote sensing has improved consistent long-term research of sea ice trends, and sea ice thickness and volume are important climate indicators for these changes. However, the majority of the satellite missions in orbit are inclination limited, resulting in gaps of missing data at the geographical poles, referred as the "pole hole". In this thesis, three methods with different level of complexity are presented to mitigate the uncertainty the pole hole introduces to pan-Arctic sea ice thickness measurements, and in turn estimation of sea ice volume. The halo method fills the pole hole with an average Envisat sea ice thickness, the ice age method does so by specified ice types, and the neural network method predicts the thickness from brightness temperature measurements using a trained multilayer perceptron. Validating these methods against CryoSat-2 data during 2010-2012, showed an underestimation of sea ice volume for all methods. Compared to the ice age method, the halo method underestimated higher volumes more, whereas the neural network method underestimated high volumes and overestimated low volumes relative to the other two methods. All methods showed a declining trend for estimated sea ice volume in 2002-2012. The average decline for halo was -199.5 km3/year, for ice age it was -208.4 km3/year, and for the neural network it was -137.9 km3/year. The largest variability between the three methods occurred during seasonal transitions, partly due to differences in how well they capture the ratio of different sea ice types and thickness. Since the ice age method includes more information, regarding both sea ice types and its location in the pole hole, and based on these results, we suggest that the ice age method is most appropriate for improving the long-term estimation of Arctic sea ice volume. Relative to CryoSat-2 in 2011, the ice age method estimated sea ice volume by 14% lower in October and 4% lower in April, while increasing the covering area with 2% more when the sea ice is near its minimum extent and 0.2% more at near maximum extent. Despite methodological differences and some required future work, all three approaches demonstrate the impact and importance of accounting for the full Arctic sea ice area when estimating sea ice volume trends.