Proposed PhD Research Outline: Investigating Potential Predictability of the Southern Hemispheric Westerly-Wave and its Impact on Southern African Seasonal Climate on Decadal Timescales.


In the sections below I provide a brief motivation for considering the features and questions that I plan to investigate further in this project. In \ref{Intro:Wave}, some background to studies of the dynamics and predictability of the Southern Hemisphere circumglobal Westerly Wave. This is followed in \ref{Intro:Predict} by a general discussion on the importance of considering initial condition (IC) predictability and uncertainty in climate change and variability assessments.

Dynamics of and projected changes in the Southern Hemisphere Westerly Wave


Indications are that greenhouse gas (GHG) related warming will tend to result in a poleward shift in the SH MLC tracks (IPCC, 2013)—in part because the zones of subtropical stability are expected to expand zonally, thus extending the tropical Hadley Cell circulation (Lu et al., 2007). However, this influence is likely to be counteracted, to some extent, by stratospheric ozone recovery (Fogt et al., 2009; Thompson et al., 2011), implying that the net change remains somewhat uncertain (IPCC, 2013). Additionally, as in the Northern Hemisphere (NH), low frequency internal variability may dominate any predictable forced signal in storm track shift for some decades to come (IPCC, 2013). Additionally, the IPCC (2013) conclude that blocking frequency in both hemispheres will likely remain relatively unchanged.

The IPCC (2013) suggests that there is substantial evidence that summer monsoon systems will strengthen in response to anthropogenic climate change. If the hypothesised coupling between boreal summer monsoon systems in the Northern Hemisphere and the subsidence associated with the South Atlantic High Pressure system (Lee et al., 2013) indeed operates—and continues to do so under rapidly evolving climate forcings—this may further enhance shifts in the Westerly Wave over the Southern African sector .

Evidence has been presented that there is a significant influence exerted on Southern Hemisphere (SH) mid-latitude cyclone (MLC) track—and consequently winter rainfall over south-western South Africa and summer synoptic patterns associated with the Westerly Wave—by sea ice distribution and concentration around Antarctica (Blamey et al., 2007; Raphael et al., 2010; Hudson et al., 2001). Hudson (1998) finds that the winter time response to a change in Antarctic sea ice distribution in the South Atlantic sector is opposite to parts of the Pacific sector, suggesting a complex, intricate relationship between slowly varying climate state variables and atmospheric response.

Importance of understanding initial condition uncertainty and predictability


Stainforth et al. (2007) suggest that uncertainty in future climate change can be regarded as consisting of the following five components:

  1. 1.

    ”forcing uncertainty”—uncertainty regarding the state and evolution of climate drivers external to Earth’s climate system;

  2. 2.

    ”model uncertainty”—uncertainty regarding the optimal set-up and formulation of various climate processes in existing climate models;

  3. 3.

    ”model inadequacy”—uncertainty arising from the unavoidable shortcomings of models in simulating Earth’s climate, the impact of which on model fidelity is unknown and difficult to quantify;

  4. 4.

    ”microscopic IC uncertainty”—IC uncertainty a reduction in which does not significantly further constrain future climate variable distributions (see also Lorenz, 1969); and

  5. 5.

    ”macroscopic IC uncertainty”—IC uncertainty related to slowly varying components of the climate system, which, if better observed, could potentially help better constrain future climate distributions.

Quantifying both levels of IC uncertainty and their influence on subsequent climate system trajectories is of importance (Palmer, 2000; Stainforth et al., 2007; Daron et al., 2013; Daron et al., 2015; Hawkins et al., 2015). The importance of microscopic IC uncertainty lies therein that it is an intrinsic, fundamentally irreducible uncertainty; it represents an absolute lower bound to total uncertainty in climate projection (Slingo et al., 2011; Smith, 2000; Smith, 2000a; Smith, 2002). No possible future improvement in observations, model resolution, model set-up and parameterisation, or progress or in any other aspect of climate model development could significantly alter the magnitude of microscopic IC uncertainty. Macroscopic IC uncertainty is important to consider, because constraining the present state of slowly varying components of the climate system could help allow one to constrain future climatic probability distributions (Stainforth et al., 2007; Daron et al., 2013; Daron et al., 2015). Hence, understanding IC uncertainty is necessary for meaningful quantification of total uncertainty in future projections.

Both microscopic and macroscopic IC uncertainty relate to internal climate variability (climate system variability with characteristic time scales beyond the weather scale, such that predictability of second—rather than first—kind is of relevance (Lorenz, 1975)), and consequently also to climate predictability (Daron et al., 2013; Daron et al., 2015). Unpredictable internal variability can be quantified using IC ensembles, produced by choosing a single starting climate system state, to which small perturbations are applied. Such ensembles allow one to establish the spread of possible states consistent with a given forcing trajectory and macroscopic initial state in a given model set-up. Kay et al. (2015) and Deser et al. (2012) demonstrate using large initial condition ensembles that—especially on regional and ocean-basin scales—internal variability is responsible for a substantial proportion of the spread in multi-model ensemble climate change response, and this may persist for multiple decades (Hawkins et al., 2015). Using multiple ensembles with ICs corresponding to different states of slowly varying climate system components (such as sea ice state, or ocean temperatures at intermediate depth), allows one to quantify the associated potential predictability for a given ”model world” (see, e.g., Hawkins et al., 2015; Daron et al., 2015; Conradie, 2015).

Working Aim

To explore the relative contribution of initial condition predictability to total potential predictability of quantifiable characteristics of the Southern Hemispheric westerly wave over the southern African and neighbouring sectors of the Southern subtropical- and mid-latitudes—and consequently to southern African climate variables which are partially driven at the seasonal scale by wave state and behaviour.

Research Questions and Proposed Directions

  1. 1.

    Explore and evaluate a range of possible metrics of the SH westerly wave state and behaviour, focussing on wave amplitude, stationarity, zonal or meridional orientation, latitudinal position and upper-tropospheric flow speeds.

  2. 2.

    Assess the observed climatologies and inter-annual to inter-decadal variability and evolution of metrics that appear promising, focussing on the regions around southern Africa and possible links to wave behaviour around South America (upstream from South Africa).

  3. 3.

    Investigate the wave patterns and state, as evaluated using the proposed metrics discussed in (1) above, occurring during the highly anomalous years of 2015-2016 ((citation not found: Englebrecht2016); along with informal assessments of collected daily weather bulletin output, the Agricultural Research Council (ARC)’s monthly climate bulletin, Umlindi, and personal correspondence with Maximiliano Herrera) I plan to assess this in greater depth, by looking at a combination of SAWS and ARC station data (if available), reanalysis products and blended observational data sets), especially during MAM and ASO. In doing so, can we assess whether anomalous wave patterns contributed significantly to the anomalous observed conditions?

  4. 4.

    Produce a multi-centennial pre-industrial coupled control simulation (significant thought would need to go into other model characteristics that would be required for the intended purpose) from which to assess patterns of variability of the wave characteristics on various time scales; and

  5. 5.

    to use as a basis for initial condition ensemble simulations with (a) constant forcing and (b) transient forcing, using a set of different background starting states, chosen to cover a set of different sea ice states.

  6. 6.

    Use these simulations to compare the relative contributions of various influences to total potential predictability of the chosen wave metrics.