Towards improved modelling of the high southern latitudes

Numerical Weather Prediction (NWP) faces significant challenges when forecasting over the high southern latitudes. The Antarctic system is a unique integration of atmospheric, oceanic and cryospheric dynamics where many NWP models, especially global models, struggle to adequately represent the governing physical processes of the region, leading to unreliable or inaccurate forecasts. Accurate, timely and reliable NWP forecast guidance is therefore crucial for logistical operations in the Antarctic, where conditions can quickly become hazardous to personnel and equipment in the region.
Antarctica and the Southern Ocean is intrinsically teleconnected to the global climate system, with sea ice playing a critical role in the global meridional overturning circulation (MOC). Deep convection driven by Antarctic bottom-water formation allows the MOC to transport heat, salt and other key elements between hemispheres and between ocean basins to contribute to the global climate equilibrium. The high albedo of sea ice and the frozen Antarctic continent reflect incoming solar radiation, reducing the amount of heat absorbed by the Earth through a cooling of the polar regions. Similarly, the Antarctic is also influenced by other parts of the global system through interactions with dominant atmospheric modes or the El Ni˜no Southern Osciallation (ENSO). However, more immediately applicable is the impact of Antarctic conditions on weather systems in the region or which reach population centres at lower latitudes on synoptic time scales.
The challenges faced by high-latitude NWP are numerous. An incomplete understanding of the physical processes which govern the Antarctic system is a field of continued and active research for the polar community. This thesis presents an overview of the key challenges currently faced by polar NWP researchers, with examples of the state of the art from the current generation of operational models. Key elements of Antarctic weather forecasting are discussed, as is an overview of the complexities surrounding observation and verification in such a harsh and remote environment.
It is well-known that the skill of global NWP models diminishes towards the poles, however, they still remain in use by several centres. Through a model case study of the Australian NWP model (ACCESS-G), this thesis has identified analysis biases in a global model which indeed increase towards the poles, leading to forecast biases similarly growing in magnitude more quickly poleward proportional to forecast length. This behaviour is evident across a range of verification and skill metrics. Further, this study found that ACCESS-G exhibited a warm surface temperature bias and associated slow-modelled winds, as well as a correspondence between surface pressure and geopotential height biases o↵ continent.
Through an intercomparison of operational NWP models during the Year of Polar Prediction Austral Summer Special Observing Period 2018/2019, the importance of atmosphere-ocean-sea ice coupling is illustrated, whereby forecast errors can effectively be de-correlated from analysis errors through coupling to a sea ice model. Further, the association between surface pressure biases and biases in geopotential height identified through the model case study of ACCESS-G were found to be systemic across several operational models, with the strong positive error correlation between the two parameters apparent off-continent at sufficient spatial scale. Positive surface temperature biases were found to be systemic across multiple models, with errors growing to 3.5°K by 48hrs and diverse correlations were shown across a range of other surface parameters. To test the sensitivity of the Antarctic atmospheric system to surface parameter uncertainty, an ensemble Bred Vector (BV) experiment was conducted to observe the fast-growing dynamical modes resulting from finite perturbations to the atmospheric state vector over synoptic timescales. This experiment characterises the atmospheric response to these surface perturbations and investigates the utility of a BV ensemble as a forecasting tool for the high southern latitudes. In particular, a BV approach was found to project onto the fast-growing synoptic features of the region, resolve smaller vortex structures near the Antarctic Peninsula and project a compensatory effect on regional wave-like disturbances in the control forecast error in the vertical. The results of this thesis highlight key areas for future Antarctic NWP model development and evaluation in order to produce more reliable model guidance to forecasters and researchers alike.