4.1 Moist atmospheric convection

Understanding moist atmospheric convection and its influence on climate variability is central to empirical and process studies at CDC. It links small-scale turbulent motions to global circulations through cloud formation and precipitation. It connects the slowly varying conditions at the Earth's surface to the faster atmospheric responses, providing much of the long-term predictability of atmospheric circulations. Furthermore, it is the proximate cause of many climate impacts on humanity (e.g., drought, flood, and severe weather) and contributes significantly to systematic errors in climate and forecast models. In light of the complexity of convective processes and the wide range of space and time scales involved, a diverse and opportunistic research strategy has been adopted. This research draws on a wide range of models and observations, from cloud-resolving to planet-spanning, from highly idealized to highly realistic.

CDC scientists have developed a new analysis technique called cylindrical binning that facilitates statistical studies using large Doppler radar data sets. Making effective use of existing data, the relationship between rainrate and horizontal wind divergence (Fig. 4.1), which is fundamental to the interaction of mesoscale convection and large-scale circulations, is explored. Here, wind divergence (the line integral of Doppler velocity around a circle centered on the radar) at every level in the atmosphere is regressed onto hourly area-averaged surface rain rate (estimated from reflectivity). Color indicates the size of the circle over which averages are considered. All the profiles exhibit the expected low-level convergence and upper-level divergence. Other interesting features are present that indicate, among other things, the spatial scale of the convective systems. However, the statistical significance and physical interpretation of those features remains uncertain and is a subject of further study.

Vertical profiles of regression coefficient between horizontal wind divergence and near-surface rainrate

Fig. 4.1 Profiles of the regression coefficient between horizontal wind divergence estimates from Doppler velocity data and near-surface rainrate estimated from radar reflectivity, for 3 month-long radar deployments at sea. Profiles are shown from three approximately month-long deployments of shipborne Doppler radars in the Indian ocean (May 1999), west Pacific (Dec. 1992-Jan. 1993), and east Pacific ITCZ (July 1997).

A more detailed and quantitative, if synthetic, source of data about convection is cloud-resolving models. With increasing computer power, ambitious computations are being performed around the world, and some of the resulting data sets are being analyzed at CDC. For example, Fig.4.2 shows a vertical velocity field at 1500m altitude in a 1064x32 doubly-periodic cloud-resolving model Convective updrafts and downdrafts (white patches) are clustered in certain preferred areas, all embedded in the context of a complex gravity wave field. Complete quantitative data about thermodynamics and microphysics as well as motion fields are available-far beyond the capabilities of observations. The simplified context of statistically steady states, periodic domains, and known governing equations allows parameterization hypotheses to be developed, tested, and refined in a tractable context.

Vertical velocity at 1500 m altitude in a double-periodic cloud-resolving model run

Fig. 4.2 Vertical velocity at 1500 m altitude in a doubly-periodic cloud-resolving model run. White patches (off the ends of the color scale) are convective updrafts (inside red regions) and downdrafts (inside blue regions).

More realistic model studies of convection are being pursued with a nested-grid strategy in a mesoscale model (the MM5). The finest grid can resolve convection, while coarser grids require a cumulus parameterization scheme, so this project spans the interface between resolved and parameterized convection. The information flow in the model is very complex, rendering interpretation challenging, but the influence of complex, realistic lower boundary conditions on convection can be studied in detail.

Figure 4.3 shows simulated 3-hour rain accumulations over western Colombia and the adjacent eastern Pacific ocean on 3-4 September, 1998. This region is particularly interesting because it has both steep topography and a strong sea surface temperature gradient offshore, north of the equatorial cold tongue.

MM5 estimate of rainfall rate on 3-4 September 1998

Fig. 4.3 Rainfall rate on 3-4 September 1998 in a nested domain of the MM5 model at 2 km grid spacing. The coasts of Colombia and Panama and 1 km and 2 km terrain elevations are indicated with open contours.

In the afternoon/evening, convection occurs over land, fueled by the accumulated moisture from many hours of strong surface fluxes and locally forced by a well-defined sea breeze front near longitude -76.5. In the late night and morning, by contrast, convection erupts in a mesoscale region offshore. Runs without topography suggest that this is due to a mountain-lowland breeze, not a thermal land breeze per se.

The larger, coarser domains of the same model illustrate the interaction of parameterized convection, atmospheric dynamics, and a state-of-the-art land surface model of the Amazon basin at a 72 km grid spacing. Figure 4.4 shows time-longitude sections of rainfall (color) over the Amazon basin (8S-Equator) during 28 August-7 September, 1998, from the MM5 (left) and satellite observations (right). Rainbands sweep across the basin, from eastern Brazil to the Andes, at 10 degrees per day.

Model and satellite-observed rainbands over the Amazon

Fig. 4.4 Model (left) and satellite-observed (right) rainbands over the Amazon. Open contours are model surface flux, with contours at 200, 400, and 600 W m-2.

While the surface fluxes are spatially coherent across the basin (driven by the solar input), the convection occurs in traveling bands, exhibiting some enhancement in the afternoon. The result of this is that the long-term climatology of convection has a striped structure. If we think of the ultimate driver of convection as being the heating of the atmosphere from below, Fig. 4.4 suggests that convection may be out of equilibrium with this driving by at least half a day, even over a hot moist continent.

These findings suggest that equilibrium parameterizations of convection, as are used in many prediction models, may be inadequate. To explore the consequences of cumulus parameterization assumptions, it is helpful to first consider more idealized large-scale models with parameterized convection. Figure 4.5 shows rainfall patterns from one run of an idealized model on an earth-sized planet. A warm sea-surface temperature anomaly is specified in the middle of the picture, and the time-mean rainfall enhancement depends on the dynamics of the propagating transient activity, which in turn depends on aspects of the cumulus parameterization.

Rainfall rate in an idealized model

Fig. 4.5 Rainfall rate in an idealized model (Mapes 2000) with a sea surface temperature anomaly in domain center.

The broad perspective on convective variability afforded by the multi-model approach outlined above is being used to develop convective parameterization schemes whose behavior in models and whose spatial correlation statistics more closely mimic the observations. Models on the various spatial scales can then be used as testbeds for different cumulus parametrization schemes and hypotheses.

Back | Forward