What can Cloud-Resolving Models Tell us About Critical Phenomena in Atmospheric Precipitation?
Recent work suggests that observations of tropical precipitation conform to properties associated with critical phenomena of other systems (Peters and Neelin 2006). The measurements are averages over 25-km by 25- km areas and are snapshots in time, and therefore unable to reveal the underlying, smaller-scale physical processes. We are using a 3D cloud-resolving model (CRM) to resolve these processes in space and time, and thereby allow us to investigate the underlying physics in detail. The model is being run over a large domain (1000 km by 1000 km) for a long time (many days) in order to adequately sample the rare events. In addition, we are using results from a global climate model that is based on the multi-scale modeling framework (MMF). Whereas conventional parameterizations are based on statistical theories involving uncertain closure assumptions, MMFs represent cloud processes on their native scales, by embedding a 2D CRM with a 4-km horizontal grid size in each climate model grid column. We are analyzing the model results following the methodology of Peters and Neelin. We are using MMF results to produce rainfall rates conditioned on column water vapor and column temperature over the Tropical oceans. We are doing the same with 3D CRM results. Furthermore, we are comparing 2D and 3D CRM results and examining the impact of CRM horizontal grid size. We are also analyzing additional statistical aspects of Tropical convection in the 3D CRM simulations that are related to critical behavior, such as size distributions and other geometric properties of mesoscale convective systems, identified as clusters of adjacent pixels exceeding a precipitation threshold. And to evaluate the realism of the statistical properties of deep convection simulated by the 3D CRM, we are comparing its vertical velocity statistics and rainfall rate PDFs to observations from aircraft and precipitation radars, respectively.
Year of Tropical Convection (YOTC)
Tropical convection and the multi-scale organization of precipitating convection are associated with scale interactions that are fundamental to the atmospheric circulation and its interaction with the ocean. The realistic representation of tropical convection and its multi-scale organization is a long-standing challenge for numerical weather prediction and climate models. Incomplete knowledge and practical issues disadvantage the representation of important phenomena and processes in global models, such as the ITCZ, monsoons, MJO, and easterly waves and tropical cyclones. The tropical-extratropical interactions of tropical convection are key aspects of the Predictability and Dynamical Processes of THORPEX. The WCRP and WWRP/THORPEX are jointly coordinating a year of observing, modeling, and forecasting with a focus on the multi-scale organization of tropical convection, prediction, and predictability: Year of Tropical Convection (YOTC). Satellite, in-situ, and field-campaign measurements (e.g., TPARC), operational prediction, and cloud-system resolving models will be utilized. The temporal scales addressed, up to seasonal, enables the above phenomena to be modeled at high resolution, and seamless prediction issues at the intersection of weather and climate addressed. The 'Year', the period 1 May 2008 - 31 October 2009, began with the archiving of ECMWF T799 (i.e., 25 km) products: i) complete global analysis; ii) deterministic forecasts; and iii) special diagnostics. Plans are underway to obtain similar NCEP and NASA GEOS-5 data, and to integrate various multi-sensor satellite products. The YOTC Science Plan, which is available at http://www.wmo.int/pages/prog/arep/wwrp/new/documents/ YOTC_Science_Plan.pdf, has been published as a WMO Technical Document. The YOTC Implementation Plan, presently being drafted, will be discussed and finalized at an international workshop in July 2009. This talk summarizes programmatic aspects; science issues involving the multiscale organization of precipitating convection; progress with MJO studies; steps towards implementation; and future directions of the YOTC project.
Multi-resolution Cascade Simulations of an MJO Event
As part of the UK Cascade project, a recent enhanced-phase MJO event over the tropical Indian Ocean and Maritime Continent is studied using limited-area UM simulations at both ~50 km and 12 km grid spacing. The two simulations are compared to each other and to observations in order to investigate processes leading to convective organization and propagation, including moisture-convection feedbacks, vertical heating profiles, and vertical momentum transports. The effects of better-resolved topography and smaller available convective scales (though convection is still mainly parameterised in both runs) are analysed, as well as the degree to which lateral boundary conditions forced by model analyses incorporating observations may affect conditions throughout the domain. These findings will aid in the setup and analysis of higher-resolution runs in the near future.
Simulations of Convective Storms in Low CAPE, High Shear Environments
A large set of three-dimensional cloud-resolving simulations is used to explore convective storm behavior in
low CAPE, high shear environments, similar to those associated with tropical cyclones. While storms in these
low CAPE regimes expectedly produce less hail than those in moderate and high CAPE, surprisingly, storms
produce similar amounts of liquid precipitation regardless of the ambient CAPE. As highlighted in recent
research, the survivability of storms in low CAPE (and high shear) is critically dependent on the low-level lapse
rates, with steeper lapse rates preferred for storm persistence. Also, as environmental precipitable water (PW)
is reduced, mid-level storm updrafts become stronger, and near-surface vertical vorticity increases, all other
factors held equal. In tropical, high PW environments, it is found that reversible CAPE is superior to
pseudoadiabatic CAPE in predictions of simulated updraft velocity using parcel theory. These findings
highlight some of the unique difficulties in understanding deep convection in tropical environments.
Dynamical processes on top of penetrative tropical deep convective clouds
The Impact of Multi-scale Representation of Tropical Convection on the Simulation of Tropical Variability
Tropical variability on intraseasonal-to-interannual time-scales is investigated using a state-of-the-art version of the Community Climate System Model (CCSM) that includes in the atmospheric component an explicit cloud physics scheme through the multi-scale modeling framework (MMF) and is referred to as the SP-CCSM. In the SP-CCSM, a 2D cloud-resolving model (CRM) is embedded in each grid column of the atmospheric general circulation model (AGCM). MMF allows simultaneous realization of the atmospheric circulation on the coarse resolution grid of the GCM and physical processes such as convection and large-scale condensation on the fine resolution grid of the CRM. Comparisons with observations and the version of the model with parameterized cloud physics processes show a significant improvement in the simulation of tropical variability by the SP-CCSM. On interannual time-scales, the SP-CCSM produces an El Niño-Southern Oscillation (ENSO) with a power spectrum similar to the observations. The amplitude and the width of the spectral peak are realistic; the frequency of the events is irregular. Unlike the observations, the dominant period is about 30 months. In the SP-CCSM, the ENSO global teleconnections show a broadening of the SST response in the eastern Pacific compared with the control run, but a similar weak response in the Indian Ocean. On seasonal time-scales, the SP-CCSM simulates a realistic relationship between the Indian summer monsoon and the sea surface temperature (SST) anomaly during the winter season following the monsoon especially for the eastern Pacific. In the western Pacific, SP-CCSM shows a reversed correlation between the SST anomaly and monsoon. On intraseasonal time-scales, SP-CCSM captures the 20-100 day bandpass variability of precipitation and atmospheric circulation. Compared to the control simulation, the SP-CCSM produces a much more robust and realistic Madden-Julian Oscillation (MJO). The amplitude and the geographic distribution of the variance more closely resemble the observed.
Convection in a parametrized and super-parametrized model and its role in the representation of the MJO
We compare the behaviour of convection and the MJO in two simulations from the same GCM but with two very different treatments of convection: one has a conventional parameterization of moist processes (CAM) and the other replaces the parameterization with a cloud resolving model, the so-called super parameterization (SP- CAM). There is a strong MJO in SP-CAM, but the MJO is absent in CAM simulation. We conclude that the following features of convection are likely affecting the difference in MJO simulations in the two models: 1) In SP-CAM, rainfall exhibits a strongly exponential increase with increasing column integrated relative humidity, albeit with peak rainfall occurring in too moist an atmosphere as compared to observations. This sensitivity of rainfall to column integrated humidity is consistent with the strongest convection occurring in the most humid conditions. In the CAM, it tends to rain at too low a saturation fraction and high saturation fraction/high rainfall rates are not achieved. This indicates that the CAM model does not require high column humidity to precipitate. 2) In SP-CAM, heavy precipitation is associated with a positive/negative temperature anomaly couplet during heavy rainfall possibly indicative of the presence of a stratiform diabatic heating profile. In contrast, in the CAM simulation, the temperature profile is more indicative of a convective-dominated heating profile that will project onto faster modes. 3) For observations and the SP-CAM, strong rainfall events (based on unfiltered grid-point rainfall) are associated with increased latent heat flux, whereas for the CAM, strong rainfall is associated with decreased latent heat flux.
Statistical Evaluation of CRM-Simulated Cloud and Precipitation Structures Using Multi- sensor TRMM Measurements and Retrievals
Cloud resolving models are typically used to examine the characteristics of clouds and precipitation and their relationship to radiation and the large-scale circulation. As such, they are not required to reproduce the exact location of each observed convective system, much less each individual cloud. Some of the most relevant information about clouds and precipitation is provided by instruments located on polar-orbiting satellite platforms, but these observations are intermittent "snapshots" in time, making assessment of model performance challenging. In contrast to direct comparison, model results can be evaluated statistically. This avoids the requirement for the model to reproduce the observed systems, while returning valuable information on the performance of the model in a climate-relevant sense. The focus of this talk is a model evaluation study, in which updates to the microphysics scheme used in a three-dimensional version of the Goddard Cumulus Ensemble (GCE) model are evaluated using statistics of observed clouds, precipitation, and radiation. We present the results of multiday (non-equilibrium) simulations of organized deep convection using single- and double-moment versions of a the model's cloud microphysical scheme. Statistics of TRMM multi-sensor derived clouds, precipitation, and radiative fluxes are used to evaluate the GCE results, as are simulated TRMM measurements obtained using a sophisticated instrument simulator suite. We present advantages and disadvantages of performing model comparisons in retrieval and measurement space and conclude by motivating the use of data assimilation techniques for analyzing and improving model parameterizations.