Dissertation Abstract
Initialization and predictability of Arctic sea ice in a global climate model
Tietsche, Steffen 2012 http://www.met.reading.ac.uk/~steffen
Department of Geosciences, University of Hamburg (Germany), 107 pp.
The state of sea ice, which plays an important role in shaping Earth’s climate, exhibits vast changes on multiple time scales. For the climate of the present day and the near future, this thesis investigates initialization and predictability of Arctic sea ice in a global climate model (GCM).
To predict the mean state, knowledge of nonlinear thresholds is required, where the mean state changes rapidly and irreversibly. I investigate the existence of such a threshold for the transition between perennial and seasonal sea-ice cover in the Arctic during the 21st century, which could in principle be caused by the destabilizing ice-albedo feedback. I find that Arctic sea ice at all times recovers even from extreme prescribed losses within typically two years, because the large-scale Arctic energy budget adapts to compensate the prescribed anomaly. This indicates that there is no such threshold.
To predict natural variability, knowledge of the initial conditions is required. I investigate the initialization of Arctic sea ice in a GCM by assimilating sea-ice concentration observations. I develop a Newtonian relaxation scheme that successfully constrains concentration and thickness of Arctic sea ice in the GCM to observations. The choice of the assimilation technique is supported by analysis of a conceptual model of local ice growth as well as analysis of assimilation error statistics.
I assess the relative importance of sea-ice initial conditions by performing ensemble predictions under the perfect-model assumption. Predictive skill achievable for Arctic sea ice is comparable between sea-ice-assimilated initial conditions and perturbed-perfect initial conditions for lead times of up to one year. I find that the large interannual negative anomalies that are expected to occur throughout the 21st century are partly predictable for up to three years ahead. However, sea-ice predictability decreases as Arctic sea-ice cover becomes thinner.
To predict the mean state, knowledge of nonlinear thresholds is required, where the mean state changes rapidly and irreversibly. I investigate the existence of such a threshold for the transition between perennial and seasonal sea-ice cover in the Arctic during the 21st century, which could in principle be caused by the destabilizing ice-albedo feedback. I find that Arctic sea ice at all times recovers even from extreme prescribed losses within typically two years, because the large-scale Arctic energy budget adapts to compensate the prescribed anomaly. This indicates that there is no such threshold.
To predict natural variability, knowledge of the initial conditions is required. I investigate the initialization of Arctic sea ice in a GCM by assimilating sea-ice concentration observations. I develop a Newtonian relaxation scheme that successfully constrains concentration and thickness of Arctic sea ice in the GCM to observations. The choice of the assimilation technique is supported by analysis of a conceptual model of local ice growth as well as analysis of assimilation error statistics.
I assess the relative importance of sea-ice initial conditions by performing ensemble predictions under the perfect-model assumption. Predictive skill achievable for Arctic sea ice is comparable between sea-ice-assimilated initial conditions and perturbed-perfect initial conditions for lead times of up to one year. I find that the large interannual negative anomalies that are expected to occur throughout the 21st century are partly predictable for up to three years ahead. However, sea-ice predictability decreases as Arctic sea-ice cover becomes thinner.