Real-time identification of equatorial waves and evaluation of waves in global forecasts

A novel technique is developed to identify equatorial waves in analyses and forecasts. In a real-time operational context, it is not possible to apply a frequency filter based on a wide centred time-window due to the lack of future data. Therefore, equatorial wave identification is performed based primarily on spatial projection onto wave mode horizontal structures. Spatial projection alone cannot distinguish eastward from westward-moving waves, so a broad-band frequency filter is also applied. The novelty in the real-time technique is to off-centre the time-window needed for frequency filtering, using forecasts to extend the window beyond the current analysis. The quality of this equatorial wave diagnosis is evaluated. Firstly, the “edge effect” arising because the analysis is near the end of the filter time-window is assessed. Secondly, the impact of using forecasts to extend the window beyond the current date is quantified. Both impacts are shown to be small referenced to wave diagnosis based on a centred time-window of re-analysis data. The technique is used to evaluate the skill of the Met Office forecast system in 2015-2018. Global forecasts exhibit substantial skill (correlation > 0.6) in equatorial waves, to at least day 4 for Kelvin waves and day 6 for Westward Mixed Rossby-Gravity (WMRG), and meridional mode number n=1 and n=2 Rossby waves. A local wave phase diagram is introduced that is useful to visualise and validate wave forecasts. It shows that in the model Kelvin waves systematically propagate too fast and there is a 25% underestimate of amplitude in Kelvin and WMRG waves over the Central Pacific