Errors in the simulation of storms characteristics, and their tracks, in weather and climate global models Pier Luigi Vidale(1), Malcolm J. Roberts(2), Kevin Hodges(1), Alexander J. Baker(1), Benoit Vanniere(1), Reinhard Schiemann(1), Marie-Estelle Demory(1), Matthew S. Mizielinski(2), Jane Strachan(1,2), (1)National
Joint Weather & Climate Research Programme
Centre for Atmospheric Science, University of Reading, UK. (2)Met Office Hadley Centre, UK. (3) ECMWF, Reading, UK.
1. Introduction
5. The Mediterranean storm track
Atmospheric storms are active agents of global transport and pose substantial risk to human activities, e.g. in the form of storm surge, extreme winds and precipitation. A number of routine and observational campaigns in the tropics focus on cataloguing storms according to intensity, track and landfall characteristics, so as to enable the computation of risk. No equivalently comprehensive resource exists for the extratropics: numerical models (e.g. used in re-analyses) are complementary to observations in this remit, but suffer from systematic errors in the simulation of storms; these errors interact, including with the large scale circulation, and limit overall model predictive skill. The main question: can enhanced model resolution reduce these systematic errors?
The annual results show that, as we increase model resolution, the strengthening and elongation of the NA storm track are robust features.
There is also a pattern of response in NW Europe vs. SE Europe, as well as an enhanced Mediterranean track, stronger in DJF.
2. TRACK (Hodges et al. 2002-2017) • Feature-tracking package based on a variety of fields, e.g. vorticity. • Suitable for many systems, e.g. tropical, extra-tropical, mesocyclones (polar lows, Medicanes), ocean eddies. • We have tracked several re-analyses and GCMs at various resolutions (ERA-I, MERRA-1,2, JRA-25,55, NCEP) In this study we have focussed on extra-tropical cyclones and used: 1. 6-hourly 850hPa relative vorticity data. 2. Spectral filtering, choice can depend on field: a) T42 truncation for vorticity. b) Planetary wave filtering, n≤5. c) Tapering of spectral coefficients. 3. Intensity threshold: 1.*10-5 s-1 for vorticity. 4. Tracking, minimize a cost function for track smoothness subject to adaptive constraints on displacement distance and track smoothness. 5. Post tracking filtering, 2 day, 1000km, focus on mobile systems. • Reference data set: derived by tracking ERA-Interim, overlapping period (1985-2012)
Fig. 3: model response to the change of resolution from N96 (130km) to N512 (25km). Left: annual changes in the storm track density. Right: the Southern European and Mediterranean storm track resolution sensitivity in DJF (top) and JJA (bottom).
6. Some mechanisms 6.1 Meridional eddy heat flux Motivation: L. Novak et al. 2014 “Our results suggest that high heat flux is conducive to a northward deflection of the jet, whereas low heat flux is conducive to a more zonal jet”. Question: is high-resolution providing increased meridional eddy heat flux? For JJA (and MAM), yes, but not for DJF. What other mechanisms change the storm track as we increase resolution?
Scaled track density on T42 grid from TRACK MERRA2 – 1980-2015 JRA55 – 1959-2014 ERAIN – 1979-2015 Fig. 4: simulated meridional eddy heat flux (N512 resolution, top) and its resolution dependence (bottom, N512 vs N96). DJF climatology (left) and JJA climatology (right).
3. AGCM simulations HadGEM3-GA3 • Project UPSCALE (Mizielinski et al. 2014) • 85 levels • OSTIA SSTs • Different resolutions, otherwise very similar • Simulation period: 1985-2012
In JJA we should rather track Medicanes…
6.2 Baroclinicity
3x26 years
5x26 years N96 (135 km)
N216 (60 km)
5x26 years N512 (25 km)
Enhanced model resolution produces a more baroclinic environment in the NA sector, which strengthens and shifts the storm track to the NE. This is more important in DJF.
Fig. 5: the Eady parameter as represented at various resolutions, from N96 (130km) to N512 (25km), and its resolution dependence. DJF climatology (top) and JJA climatology (bottom).
4. Mean track density errors against ERA-I The overall track density bias in DJF is reduced, particularly in the NE Atlantic sector, near the jet exit.
In JJA there is a similar increase in track density over NE Atlantic and NW Europe, as well as a decrease to the SouthSouth-East of the Alps. Both reduce errors. Fig. 2: resolution sensitivity in simulating track density over the North Atlantic (NA). DJF (top) and JJA (bottom). Fig. 1: errors in the simulation of track density for the HadGEM3-GA3 model at three model resolutions. DJF (top) and JJA (bottom).
hrcm.ceda.ac.uk
Track density errors are mitigated with enhanced resolution, particularly in the NE Atlantic in JJA. It is unclear whether the use of (low-resolution) ERA-I is suitable as reference, particularly for JJA in the tropics.
6.3 The role of orography 130km resolution orography (typical of climate models over last 10-15 years)
25km resolution orography, more typical of weather models
Question: is there secondary cyclogenesis in the Mediterranean storm track, and is a correct simulation dependent on model resolution?
Fig. 6: orography of the Alps at various HadGEM3 resolutions
Fig. 7: genesis density and its resolution dependency in MAM (top) and SON (bottom)
6. Conclusions and outlook HadGEM3-GA3 simulates the NA storm track more realistically as the resolution is increased from ~135 to ~25 km. In HadGEM3-GA3, resolution sensitivity is associated with stronger baroclinicity (and associated jet elongation) on the NE flank; the roles of the meridional eddy heat flux on the SW flank of the storm track and of enhanced orography require further investigation, via process studies. Higher-resolution re-analyses, particularly MERRA-2 and ERA-5, will help to interpret current errors. The forthcoming CMIP6 HighResMIP experiment, a deliverable of the EU-Horizon 2020 PRIMAVERA project, will produce a multi-model, multi-resolution ensemble of both atmosphere-only and coupled simulations. Detailed process-based analyses in PRIMAVERA and joint CLIVAR projects will allow us to assess the robustness of these findings, and implications for projections of future storm risks.
*contact:
[email protected]
