1. Using a novel coupled-model framework to reduce tropical rainfall biases Nicholas Klingaman Steve Woolnough, Linda Hirons National Centre for Atmospheric Science-Climate Department of Meteorology, University of Reading

2. Outline Introduction to tropical biases in the MetUM Biases in mean rainfall Biases in intra-seasonal variability (Madden-Julian Oscillation) Introduction to the MetUM-GOML atmosphere—ocean-mixed-layer model Using MetUM-GOML to explore the sensitivity of tropical biases to coupling in the tropics and extra-tropics Summary and conclusions

3. Tropical rainfall biases in the MetUM MetUM DJFM precipitation TRMM DJFM precipitation MetUM bias (MetUM-TRMM) The MetUM produces a strong and southward displaced ITCZ in DJFM.

4. Tropical rainfall biases in the MetUM MetUM JJAS precipitation TRMM JJAS precipitation MetUM bias (MetUM-TRMM) Large wet biases over tropical oceans in JJAS, with dry biases over tropical land, particularly India.

5. ● The MJO is the dominant mode of sub-seasonal (30-60 day) variability in the tropics. ● Events often form in the Indian Ocean, before propagating east to the Maritime Continent and the West Pacific. ● Suppressed convection precedes and follows each active event. ● In DJF (JJA) the MJO modulates the Australian (Asian, African) monsoon. Composites of observed OLR (shading) and 850 hPa winds (vectors) for each MJO phase (Source: Wheeler and Hendon, 2004) Madden-Julian oscillation

6. Tropical rainfall biases in the MetUM NOAA CIRES observations MetUM Control Regressions of 20-100 day filtered OLR on a base point at 70 ° E. The MetUM produces no eastward propagation in tropical convection, with very weak anomalies on sub-seasonal timescales.

7. Effects of convective entrainment NOAA CIRES observations MetUM Control MetUM 1.5x entrainment Increasing entrainment and detrainment by 50% in the MetUM produces some eastward propagation. Klingaman and Woolnough (2014a, QJRMS)

8. Effects of convective entrainment MetUM 1.5x entrainment JJAS 1.5x entrainment minus control 1.5x entrainment bias against TRMM Control bias against TRMM See also Klingaman and Woolnough (2014a, QJRMS) and Bush et al. (2014, QJRMS).

9. MetUM-GOML (Ocean Mixed Layer) model MC-KPP one- dimensional ocean model MetUM Atmospheric GCM OASIS coupler (sub-daily) Key advantages: Cheap: < 5% of the cost of the atmosphere, allowing high (1 metre) ocean vertical resolution. Controllable: Easily constrainable to any desired ocean state (small SST biases). Flexible: Air-sea coupling can be applied selectively in space and time to explore the role of coupling in a range of phenomena. Adaptable: Works easily with any GCM configuration or grid. Climatological three-dimensional heat and salt tendencies are applied to represent (a) the mean advection in the ocean (b) corrections for biases in atmospheric surface fluxes

10. MetUM-GC (NEMO 3D ocean)MetUM-GOML (KPP 1D ocean) By using climatological heat and salt corrections, MetUM-GOML produces much smaller mean SST biases than a fully coupled GCM. Biases are typically smaller than +/- 0.5K. The disadvantage is the lack of interactive ocean dynamics in MetUM-GOML, which are important for certain applications (e.g., ENSO) MetUM-GOML (Ocean Mixed Layer) model

11. (1) Indian Ocean (2) Warm Pool (3) Tropics- Wide (4) 50N-50S (5) Near-global Because the MC-KPP columns do not communicate, there is complete flexibility (except near sea ice) in where the atmosphere and ocean are coupled. MetUM-GOML (Ocean Mixed Layer) model

12. Experiments ExperimentCoupled?Coupling region or SST specification A-OBSNoClimatological observed SSTs K-30Yes30N–30S and 0–360E K-50Yes50N–50S and 0–360E A-K50(clim)NoClimatological SSTs from K50 A-K50(15day)No15-day smoothed SSTs from K50 All experiments use MetUM GA3.0 atmosphere with +50% to entrainment and detrainment for deep and mid-level convection. At least 25 years of data are analysed for each experiment.

13. Impacts on the Madden-Julian oscillation NOAAA-OBS K-30K-50

14. Impacts on the Madden-Julian oscillation NOAAK-50 A-50(clim)A-50(15day)

15. Effect of tropical coupling K-30 JJAS precipitation K-30 minus A-OBS (tropical coupling) K-30 bias against TRMM A-OBS bias against TRMM

16. K-50 JJAS precipitation K-50 minus K-30 K-50 bias against TRMM K-30 bias against TRMM Effect of extra-tropical coupling

17. Effect of coupling 50N-50S (vs. clim SST) K-50 JJAS precipitation K-50 minus A-K50(clim) K-50 bias against TRMM A-K50(clim) bias against TRMM

18. Effect of coupling 50N-50S (vs. 15-day SST) K-50 JJAS precipitation K-50 minus A-K50(15 day) K-50 bias against TRMM A-K50(15 day) bias against TRMM

19. Mean seasonal cycle of all-India rainfall Distributions of JJAS-mean all-India rainfall Extra-tropical coupling (red, orange) consistently produces stronger monsoons than tropical coupling (brown) or atmosphere-only (purple, blue). The strongest increases in rainfall come shortly after monsoon onset in June. All other configurations have a delayed and weak monsoon onset.

20. K-50 minus A-K50(15 day) JJAS means of: Colours: Precipitation Vectors: 850-hPa winds Contours: MSLP A-K50(15 day) bias against TRMM and ERA-Int A-K50 has a strong cyclonic bias with low MSLP in the West Pacific, both of which are reduced considerably in K-50.

21. April means of: Colours: Precipitation Vectors: 850-hPa winds Contours: MSLP K-50 minus A-K50(15 day) A-K50(15 day) bias against TRMM and ERA-Int In April, coupling prevents the erroneous northward migration of the Pacific ITCZ, focusing convection instead on the equator.

22. K-50 minus A-K50(15 day) A-K50(15 day) bias against TRMM and ERA-Int May means of: Colours: Precipitation Vectors: 850-hPa winds Contours: MSLP In May, coupling produces a strong extra-tropical Pacific jet and inhibits development of strong rainfall and cyclonic anomalies near the Philippines.

23. A-K50(15 day) “Wet India” compositeA-K50(15 day) “Wet minus dry” K-50 “Wet India” compositeK-50 “Wet minus dry”

24. Summary and conclusions MetUM-GOML is an ideal framework for examining sensitivities to tropical and extra-tropical coupling. Heat and salt corrections maintain the observed mean SST. Allows effects of regional coupling to be isolated in a model with very small SST biases. The MJO improves with tropical coupling, through better SST—surface-flux—rainfall relationships. Tropical rainfall biases reduce with extra-tropical coupling, through strengthening the subtropical high and delaying the seasonal progression of the ITCZ.