1. MJO simulations under a dry environment Marcela Ulate M Advisor: Chidong Zhang (… in a Nudging World)

2. Motivation The Madden Julian Oscillation (MJO) is a convective coupled wave that dominates the intraseasonal variability in the Tropics. It is characterized by westerly wind anomalies preceding the convection and easterly wind anomalies ahead of the convective center in lower levels (Madden and Julian 1994; Zhang 2005). We want to study the roles of convective cloud populations at initiation stages of MJO events, BUT Global Climate Models (GCM) and Regional Numerical Models (RNM) fail to reproduce the MJO (Slingo 1996; Zhang 2005; Zhang et al. 2006; Gustafson and Weare 2004, Monier et al. 2010). Some numerical models are not able to reproduce the MJO at all, and some others present decoupling between the wind and the precipitation.

3. The case of study LON TIME mm/day TRMM m/s U850 Anom. ERA-Interim

4. Model Description WRF v3.2 1º x 1º horizontal resolution. 28 vertical levels. Tropical channel domain: Periodic Boundary conditions in the east-west direction. Boundary Conditions form ERA-Interim data

5. Wind and Precipitation decoupling This was the best of our model simulations …. TIME LON

6. The dry bias WRF RH – ERAI RH P (hPa) Days P (hPa) Lat Lon P (hPa)

7. Hypothesis The model has a dry bias in the mid troposphere i.e. it dries the atmosphere faster than the adjustment time between the wind convergence at lower levels and the rain production by the cumulus scheme. The net effect is to inhibit convection. Therefore, the convective component of the MJO cannot fully develop in the model environment

8. Approach to the problem: Humidity Nudging Four-Dimensional Data Assimilation (FDDA) or nudging is the process where the model is set to converge at a desired rate to the analysis or observations. The process adds an extra tendency term to the model equations proportional to the difference between the model simulation and the analysis value at every grid point, forcing the simulation closer to the analysis value.

9. Humidity Nudging From Skamarock et al. 2008. Model forcing termsNudging Tendency term : nudging factor, : four dimensional weight function, :analysis field value

10. Spectral Nudging of Humidity Spatially filter the data (minimun x,y wavelengh) Analysis data (ERA-Interim) Nudging Tendency

11. Grid Nudging vs Spectral Nudging Dependency of the small scales (~100Km) BUT, Do we need the large scales ….? (See Future Work)

12. TIME LON x100y500

13. WRF RH – ERAI RH P (hPa) Days P (hPa) Lat Lon P (hPa) Reduction of the dry bias

15. Humidity Tendency due to nudging Humidity Tendency due to cumulus scheme MJO-like precipitation simulation Humidity Tendency due to cumulus scheme - CONTROL P (hPa) g/Kg day -1

16. Heating Tendency - Control Heating Tendency – MJO-like precipitation simulation P (hPa) K/day

17. Super MJO … ?? How much nudging is too much nudging? What if G a =1 ?

18. Conclusions Spectral and grid Nudging of water vapor mixing ratio reduces the model dry-bias and allows the model to produce an improved MJO-like precipitation pattern and wind signal. The moisture at mid levels of the troposphere is crucial in order to reproduce the convective signal associated with the MJO. The dependencies of the meridional and zonal scales close to ~100Km are important in order to obtain a better MJO-like simulations.

19. More Conclusions The prediction of the first MJO event improves when nudging is preformed, while the initiation of the second event is not. This suggests that improving the humidity field is one component of the problem, and we need to investigate further on this matter. Without nudging, the cumulus scheme remains relative inactive i.e. lack of precipitation during the MJO event. This translates to a weak heating profile. When the MJO precipitation pattern improves, the heating profile resembles the results of other studies more closely.

20. Future work Study more in depth the relationship between the large scale’s moisture role in MJO simulations. Investigate the general characteristics of the transition between 2 MJO events and the inability of the model to capture abrupt changes in precipitation patterns. Simulate the MJO using a cloud- resolving version of the WRF (25Km,8Km,3Km resolution) with YOTC data as boundary conditions for the same case of study and the addition of the second MJO event.

21. Thanks