Abstract
This work encompasses the lessons learned when studying the Madden Julian Oscillation (MJO) under a regional modeling framework. Regional modeling is widely used by the meteorological community; however, few studies outline the limitations of regional models and the challenges one must face when modeling particular tropical phenomena. This is the focus of this research. This study is divided into four parts, the first part is the development of a technique that can be used in order to discern between successful and unsuccessful simulations of MJO cases of study. This technique, referred here as The Tracking Method, consists of tracking the various eastward–moving features during the simulations and obtaining an optimal track amplitude, speed, and start date, which will represent the most prominent eastward-moving feature on different variable fields. This method was applied to long records of data such that climatological values of amplitude and speed were obtained. The tacking method is applied to each simulation and these are rated against the observation’s tracking values within the climatological deviations. The application of this procedure to the long-term precipitation and wind data show a level of decoupling between the convection and the circulation patterns within in the MJO. The next three parts of this work are dedicated to the investigation of different aspects that limit the regional models in reproducing the MJO: model physics, environmental moisture patterns, and initial, bottom, and lateral boundary conditions. The study of the influence of the choice of physics in MJO initiation under a regional model setting was focused on the cumulus and planetary boundary layer schemes. Both play an important role in the moisture re-distribution and convection. The effect that the choice of cumulus and planetary boundary layer scheme has in a regional model affects the numerical simulations so greatly that it may change the water cycle balance during the simulation, in as little as 5 days. We found that the non-linearity of the processes within the model physics play a key role in the entire forecast. Without their mutual improvement, we will not be able to accurately represent the MJO, with broader implications to tropical meteorology in general. The large-scale moisture patterns and their role during the MJO initiation were studied by using zonal spectral nudging of water vapor mixing ratio. Grid nudging of other variables such as temperature and wind was also performed. Our results show that the correction of moisture in the low and middle vertical levels is sufficient for a realistic MJO simulation. Moreover, the correction of fields such as temperature, wind, and their combination does not result in a successful simulation. Specifically, the error correction of the moisture mean and planetary zonal wavenumbers 1 to 3 will improve an MJO simulation greatly. Beyond these three wavenumbers together, the additional correction of higher wavenumbers, a specific wavenumber alone, temporal, or spatial means are not sufficient for a realistic MJO simulation. The moisture spatial patterns are so important that it is found to be possible to reproduce an MJO precipitation pattern during a period without MJO only by nudging the water vapor mixing ratio. The study of the influence of initial, bottom and lateral boundary conditions during MJO simulation in a regional model shows that the precipitation amplitude is dominated by the model physics rather than the initial and lateral boundary conditions configuration. However, the timing of precipitation-triggering and the start of the eastward-moving features are associated with the sea surface temperature updates. Moreover, our results show that updating the SSTs in a non-coupled simulation will lead to the premature triggering of the precipitation on top of the warm SST anomalies associated with the MJO. This differs from the real precipitation and SST anomalies observed during MJO cases, where the warm SSTs are ahead of the convection and cooler SSTs related to cloudiness and cooling by fresh water are below the rain. Lastly, regardless of the lateral boundary conditions setup (time depended vs time independent) in a regional model setting, there is dominance of the physics errors over the information from the lateral boundary conditions, which are overshadowed. Furthermore, the differences between simulations with different lateral boundary conditions will reach the center of the domain faster through the parameterized processes within the model than one might otherwise expect through advection, and the differences originally closer to the boundaries in the regional model will be amplified under convective situations. Regardless of the fact that regional modeling is a widely used framework due to its major advantage of increased resolution over a particular area of interest, it is necessary to take into consideration its many shortcomings, especially the strong dependence on the parameterized physics rather than the information in the boundaries in some cases. Going forward, it is also critical to take into account that improved forecasts in the tropics will require further development and improvement of the physics processes that need to be first fully understood by the scientific community, and then mathematically incorporated as a parameterization in any numerical model. This will be an enormous task for the meteorological community in the years to come.