Rethinking the relationship between the observed, simulated and real Arctic sea-ice evolution

Abstract

In this dissertation, I explore the large differences in Arctic sea-ice evolution between climate models and observations, and among individual climate models. First, I investigate the drivers of the long-term Arctic Ocean warming in a multi-model ensemble. I find that there is no consensus between the models about whether the excess energy is gained by the ocean through the net atmospheric surface flux or through the meridional oceanic heat flux. However, all models agree on the magnitude of the projected warming. The warming is small compared to the anomalies in the energy fluxes. This is because most of the energy gained through one energy flux is lost through the other energy flux due to a relationship between the magnitude of the increase in oceanic heat inflow and the increase in turbulent heat loss to the atmosphere. Second, I explore the feasibility of an observation operator for the Arctic Ocean. An observation operator translates the Arctic Ocean climate simulated by a climate model into a brightness temperature. The brightness temperature is the quantity directly measured by satellites from space. Hence, an observation operator enables us to circumvent the observational uncertainty currently inhibiting reliable climate model evaluation. Sea-ice brightness temperatures at 6.9 GHz are driven by the liquid water fraction profile inside the ice and snow, which is not resolved in most climate models. I show that in winter this profile can be described reasonably well by a linear temperature profile and a salinity profile prescribed as a self-similar function of depth. In summer, the melt-pond fraction is more important for the simulation of the brightness temperature than the internal structure of the ice. Third, I develop an Arctic Ocean Observation Operator for 6.9 GHz based on these findings. I compare brightness temperatures simulated from the output of an Earth System Model to brightness temperatures measured by satellites. The differences between simulated and measured brightness temperatures can mainly be explained by the uncertainty in the simulated state of the sea-ice concentration, the assimilation process, and the melt-pond parametrization. Differences attributable to biases in the observation operator itself are small. The operator is therefore a suitable method for climate model evaluation. In summary, I show different perspectives on the large differences in Arctic sea-ice evolution. On the one hand, I point out that the multi-model ensemble mean is not always representative for the simulated Arctic climate and should be interpreted with care. On the other hand, I introduce and develop an unconventional tool providing new opportunities for future climate model evaluation.