TL;DR
This paper investigates the ability of the MM5 mesoscale model to simulate tropical convective systems using different resolutions (2 km, 15 km, and 60 km). The study finds that a 15-km resolution with parameterized convection can reproduce many observed features of tropical convection and its interaction with the large-scale environment, offering a computationally efficient alternative to higher-resolution, fully explicit simulations.
What Problem Does This Paper Address?
Accurately simulating tropical convection is a long-standing challenge in atmospheric science. Convective systems are crucial for the Earth's energy balance and climate, but their complex, multi-scale nature makes them difficult to represent in models. General Circulation Models (GCMs) often struggle with parameterizing these processes, while high-resolution Cloud Resolving Models (CRMs) are computationally expensive and limited to small domains, preventing them from capturing the organization of larger convective systems observed in the tropics. This paper aims to bridge this gap by using a mesoscale model (MM5) at resolutions that can explicitly resolve mesoscale organization while parameterizing sub-grid scale convection, allowing for larger domain simulations.
Methods
The researchers employed the nonhydrostatic version of the fifth-generation Pennsylvania State University/National Center for Atmospheric Research mesoscale model (MM5). They conducted three simulations with varying horizontal grid spacings: 2 km (fully resolving convection), 15 km (parameterizing convection), and 60 km (parameterizing convection). All simulations were driven by a time-varying "IFA mean forcing" derived from observed advective tendencies of temperature and moisture over the Tropical Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TOGA COARE) Intensive Flux Array (IFA). Newtonian relaxation was used to keep domain-averaged horizontal winds close to observations, and periodic lateral boundary conditions were applied. A Kain-Fritsch cumulus parameterization scheme was used for the 15-km and 60-km simulations, with modifications to better link convection to resolved-scale motion. A simple "explicit moisture" scheme was used for grid-scale microphysics in all simulations.
Datasets
The study utilized several observational datasets for comparison:
- TOGA COARE sounding data: Processed by Ciesielski et al. (1997) for IFA mean soundings.
- Rainfall data: Obtained from the Colorado State University TOGA COARE IFA data site (Lin and Johnson 1996a, b).
- Surface heat fluxes: Derived from the Colorado State University TOGA COARE IFA data site (Lin and Johnson 1996a, b).
- Infrared (IR) satellite data: From the Japanese Geosynchronous Meteorological Satellite (GMS) for cloud-top temperature analysis.
- ISCCP data: Cloud cover and radiative flux data derived from the International Satellite Cloud Climatology Project observations (Rossow and Zhang 1995; Zhang et al. 1995).
- MIT radar data: Collected on the R/V Vickers for radar reflectivity measurements (Leary 1997).
- IMET buoy measurements: For surface air temperature, water vapor mixing ratio, and equivalent potential temperature.
Key Insights and Ideas
- 15-km resolution as a sweet spot: The 15-km resolution with parameterized convection proved to be a valuable tool, capturing many observed features of tropical convection and its interaction with the large-scale environment, offering a good balance between resolution and computational cost.
- Model-observation agreement on variability: Despite differences in resolution and convection treatment, all simulations showed similar temporal variability in domain-averaged temperature and relative humidity profiles, closely matching observed patterns.
- Persistent cold bias: A consistent 2 K colder temperature bias was observed across all simulations, a common issue in many models simulating this period, suggesting potential issues with either the forcing data, model physics, or the technique of imposing external forcing.
- Overestimation of rainfall coverage during suppressed periods: During suppressed convective periods, the models tended to produce broader rainfall areas and more cirrus anvil clouds than observed, indicating a potential oversensitivity or misrepresentation of these quiescent phases.
- Surface flux overestimation: Surface sensible and latent heat fluxes were significantly overestimated (50%-100%), likely due to limitations in the Blackadar PBL parameterization scheme used in the MM5.
- Underestimation of solar flux: The simple shortwave radiation scheme led to an underestimation of downwelling solar flux at the surface.
Key Results
- Domain-averaged properties: The 2-km and 15-km simulations produced very similar domain-averaged temperature and relative humidity profiles, closely resembling observations in terms of temporal variability.
- Cloud and precipitation: Domain-averaged cloud amount and precipitation agreed well with observations.
- Temperature bias: All simulations exhibited a mean temperature drift of about 2 K colder than observed throughout the troposphere.
- Relative humidity bias: Models showed a relative humidity bias of 10%-20% higher than observed, particularly in the lower troposphere.
- Rainfall coverage: While overall rainfall amount and fractional coverage matched observations, models showed greater areal coverage of rainfall and more cirrus anvil clouds during suppressed periods.
- Cloud cluster size: The model-simulated cloud cluster size was generally smaller than observed.
- Surface fluxes: Surface sensible and latent heat fluxes were overestimated by 50%-100%.
- Solar flux: Downwelling solar flux at the surface was underestimated by about 20%.
- Radar reflectivity: The 15-km model slightly overpredicted radar reflectivity, while the 2-km model showed too high and temporally homogeneous reflectivities.
Limitations and Future Work
The authors acknowledge several limitations and suggest future research directions:
- Cold bias: The persistent cold bias, common across multiple models, suggests potential issues with the specified IFA mean forcing, the model's physical parameterizations, or the methodology of imposing large-scale forcing. Further investigation into the thermodynamic consistency of the forcing is recommended.
- Cirrus clouds: The excessive and persistent cirrus clouds during suppressed convection periods remain unexplained and point to potential issues with the microphysical parameterization of cirrus anvils.
- Surface flux algorithm: The overestimation of surface heat fluxes highlights the need for further investigation into the Blackadar PBL parameterization scheme and potentially exploring alternative flux algorithms.
- Radiation scheme: The underestimation of downwelling solar flux suggests a need to improve the shortwave radiation scheme in the MM5.
- Domain size and convective organization: The limited domain size of the 15-km simulation may have suppressed the development of larger convective systems. Future work could explore larger domains or weaker forcing to better understand convective organization.
- Model physics: The discrepancies in relative humidity, surface fluxes, cloud-top temperature, radar reflectivity, and cloud cluster size indicate room for improvement in the MM5's physical parameterizations, particularly microphysics.
- Case studies: The authors suggest using the MM5 to simulate well-observed tropical mesoscale convective systems (e.g., the TOGA COARE 22 February 1993 squall line) to further validate and improve the model.
Summary updated