Within this project an algorithm-based approach was developed to objectively detect mid-latitude atmospheric waveguides in gridded atmospheric data. This algorithm was run on observation-based gridded data (re-analysis data)) for 1980-2019 to analyse the climatology of waveguide frequency. Consistent with our knowledge of the atmospheric jets, and theory of waveguides, we find local maxima in waveguide frequency around the region of the atmospheric jets: this is shown in the attached figure for summer months (June-July-August, JJA), with the frequency of each location being part a waveguide in colours (the frequency varies for waveguides of different length-scales, measured by wavenumber, k; in this case k=6 is shown), with the climatological summertime jet shown in black contours.
Having successfully identified waveguides in re-analysis data, this project then explored the association of such waveguides with quasi-stationary wave activity. Quasi-stationary waves are Rossby waves that stay in one place (with a near-zero phase speed) for an extended period of time (multiple days to weeks), and have previously been associated with many extreme temperature events. The project found that atmospheric waveguides are associated with a co-located increased likelihood of quasi-stationary wave activity. We also found a connection between extreme temperature events over Europe, and a greater likelihood of atmospheric waveguides present over the Atlantic, providing further evidence for the hypothesised connection between waveguides and extreme events.
The project investigated the sub-seasonal to seasonal prediction skill of heat waves, atmospheric jets, and atmospheric waveguides. In collaboration with experts in heatwaves, a metric of ‘seasonal regional heatwave propensity’ was developed. This metric takes into account the fact that it is physically unlikely for S2S prediction skill to exist for exactly when and where a heatwave may occur, but S2S forecasts may, based on the connections to larger-scale atmospheric circulation shown above, have skill in predicting the probability of heatwaves occurring over a season over a wider region. Our results found that operational seasonal forecasts from the ECMWF and MeteoFrance centers have significant skill in predicting the summer heatwave propensity for Europe during the period 1981-2016, particularly for forecasts initialized at the beginning of June. The skill provided by the modelling is significantly higher than that provided by estimating the impact of a linear warming trend, suggesting that the models provide some added skill, likely based, at least in part, on large-scale atmospheric dynamics such as atmospheric waveguides and their impact on Rossby waves. This research is currently being prepared for publication in a peer-reviewed journal.
We also studied waveguides in the S2S forecasts. The models studied reproduce the observed climatology and variability of waveguide frequency relatively well, although there is typically an underestimation of the frequency of waveguide occurrence, and slight biases in the latitude of the waveguides. We also find significant correlation between observed and predicted waveguide frequency. For waveguide frequency averaged over the northern hemisphere extratropics (across all longitudes) we find significant correlation prediction skill several months ahead. Prediction skill is often high in August, regardless of the initialization date of the forecast (i.e. forecasts initialized in May show significant correlation between predicted and observed August waveguide frequency). This was true across two different S2S operational forecast systems from different centers. Study of different regions shows that there is higher prediction skill of the Pacific waveguides than those over the Atlantic regions. The project also found similar S2S prediction skill of longitudinally averaged atmospheric jet strength, consistent with the prediction skill of the waveguides.
Despite the relatively high correlation between predicted and observed waveguide frequency, typically measures of “skill” of the forecasts are low, because the inter-annual variability of the ensemble-mean waveguide frequency is biased very low. Thus, the model is able to predict which summers are likely to have more frequent waveguides, but will underestimate the absolute frequency. Since waveguides are connected with extreme events, this may translate into an underestimation of seasonal heatwave frequency. This is valuable information as it means we may be able to apply bias-correction to the seasonal forecast models to obtain a more accurate prediction. This is the subject of on-going work. This work is currently being prepared for publication in a peer-reviewed journal.
