Constrainting Subglacial Heat Flux in Antarctica from Thermal Conductivity and Subglacial Lakes

Abstract

Developing accurate models for the dynamics of ice sheets requires detailed knowledge of the temperature field within. An important constraint on internal ice sheet temperature is provided by geothermal heat flux, the heat flow from the solid Earth to the base of the ice sheet (Fowler 2006). This flow of heat is not uniform, varying as a result of differences in thermal properties (i.e., thermal conductivity and heat production) and variations in heat transfer across the lithosphere asthenosphere boundary. Since temperature can affect a range of ice properties, from strain rate to hardness and melting rate (Paterson 1994), it is important that we have a detailed understanding of the heat flux both below and within the Antarctic Ice Sheet so that I can accurately map internal temperature. In this thesis, I examine the heat flux in Antarctic environments as well as the properties and factors that distort it. I also take an indirect approach to test geothermal heat flux models by using melting associated with subglacial lakes as a constraint. Heat can move both vertically and horizontally in order to find the path of least thermal resistance to the surface. The path is dictated by the thermal conductivity of the crustal material as heat will attempt to move through the most conductive material. In a subglacial valley, or buried bedrock high, most heat will move through the more conductive bedrock, resulting in heat being moved away from subglacial valleys and into bedrock in regions of geological contacts whereby heat will move into the more conductive of the two mediums. The result is the creation of localized regions where heat flux at the base of the ice sheet can be 80 to 120% of the regional heat flux creating localized regions of elevated/reduced temperature. Having demonstrated the underlying bedrock thermal geology is critical to mapping the flow of heat through the Antarctic ice sheet, I collected the thermal conductivity on 49 Antarctic rock samples and combined them with a larger global database to develop predictors for thermal conductivities in the inaccessible Antarctic lithosphere. From this dataset, I determine oxide and mineral contributions to the effective thermal conductivity for a range of igneous compositions. I exploit a correlation between high thermal conductivities and low seismic velocities to produce an empirical model, which is applied to a crustal tomography model to predict thermal conductivity of the Antarctic crust. The largest lateral conductivity variations are found in a region with high conductivity between 15 to 27 km, which also corresponds to an anomaly in proxy models for the geothermal heat flux beneath the Antarctic Ice Sheet. Several geothermal heat flux models for Antarctica have been made via proxy-based estimates due to limited sampling across the continent. Proxy-based estimates have large disagreements between each other (up to 50 mW m−2 in West Antarctica). To ascertain accuracy, I test the proxy-based estimates using a basal heat flux constraint (BHFC) assuming melting at the base of the ice sheet. In the presence of subglacial lakes, regions where proxy-based estimates should exceed this constraint. I find that while results show a subtle relation between lake and regions of elevated heat flux, a large number of lakes are in regions of insufficient heat flux to generate melting. These results indicates that current proxy models currently underestimate geothermal heat flux. Since there is a relation between heat flux and lake locations, the proxy-based estimates can be combined with other maps of Antarctic surface temperature, ice thickness, bedrock elevation, crustal thickness, bedrock slope and ice velocity to predict lake melt source regions. Three methods are tested, comparative property analysis, principal component analysis and machine learning method using a Subspace KNN classifier. The comparative analysis shows the properties of surface temperature, ice thickness and ice velocity to have the greatest disparity between sub-glacial lakes and Antarctica but are unable to make a clear prediction about the melt source for subglacial lakes. The PCA, while shown not to be a good predictive map, is excellent at identifying regions of Antarctica as either containing active lakes (with current water infill/outflow) or stable lakes (in which water levels remain constant). The Subspace KNN classifier meanwhile, is able to both identify lake melt sources and type of lakes generated from those sources. This work improves our ability to accurately map the geothermal heat flux at the base of the Antarctic ice sheet by giving proxy modellers by showing the importance of bedrock thermal conductivity as well as mapping it over a large section of Antarctica. We also show future avenues of research that can improve upon or use these geothermal heat flux models including mapping the melt sources of subglacial lakes.