December 2013 - A collaborative project to study the representation of convective storms over the southern UK.
The DYMECS (Dynamical and Microphysical Evolution of Convective Storms) project is a collaboration between Met Office scientists based in the University of Reading Department of Meteorology and University scientists to improve the Met Office forecast model representation of convective storms over the UK.
Accurate forecasting of convection is very important for society as it can result in a range of weather from light showers to damaging thunderstorms with flooding and snow in winter. Convection arises when warm, moist air rises in an unstable atmosphere and forecasting it is a difficult problem due to the small size of convective clouds. The Met Office now routinely runs a UK model (called the UKV model) with a gridlength of 1.5km, which is sufficient to represent convective clouds although there are concerns that the representation is still quite crude. The clouds are often around 10km in size but some important structures inside the clouds may be much smaller, specifically regions of rapidly ascending air. In order to address this problem it is important to compare the model with observations of convective clouds and storms.
Met Office scientists located at MetOffice@Reading (the Met Office unit based in the University of Reading) are working on this problem with experts on cloud processes and radar meteorology at the University of Reading Department of Meteorology. The project is based around the Chilbolton research radar which is used to scan convective clouds to obtain details of their structure. Radar works by sending radio waves into the clouds and detecting the reflected waves coming back. From the properties of these reflected waves a good deal can be deduced about the clouds and rain, for example the rain rate and the complex motion of air within the clouds.
Figure 1: 3D representation of a convective cloud from the radar observations, from the UKV model and from a 200m research version of the same model. The cutaway coloured cross sections show the radar reflectivity (a measure of the density of cloud ice or water) from 0-55 dBZ, the grey shading gives an idea of the extent of the cloud and the plane below gives the surface rainrate. The horizontal scale is in km.
A key (and novel) aspect of DYMECS is to take a statistical approach so it has been necessary to scan a large number of convective clouds so the statistics of observed clouds can be compared to those in the model. An algorithm has been developed to take the Met Office surface rainfall radar image (the rainfall map sometimes seen on weather forecasts) in real time and automatically identify and track convective clouds and automatically steer the Chilbolton radar to scan their full 3D structure. In addition to steering the radar the statistics of convective clouds from the tracking algorithm are useful in themselves for comparison to the model. Figure 1 shows a comparison of a 3D representation of the cloud and precipitation for a convective storm from the radar data with an equivalent representation of the output from the UKV and also a higher resolution 200m version of the same model. As expected the storm in the UKV model, with its relatively long gridlength, is too smooth and features (such as the area of precipitation) are too large. The 200m model shows more short range structure as would be hoped but incorrectly breaks the storm up into a number of smaller clouds with their own regions of ascending air.
Figure 2: Median equivalent radii of a number of the deepest storms on 25 August 2012 as seen by the radar and the model with 1.5km and 200m gridlength. The colours are radar reflectivity in units of dBZ and the grey shading is the 75th percentile of the 0 dBZ reflectivity.
The example shown in Fig. 1 is interesting but a major drawback is that it shows data from just one particular cloud and we have no way of knowing how representative these features are of convective clouds in general. Figure 2 shows an example of the DYMECS statistical approach which enables us to address this problem. This shows a representation of the average radius of a number of storms as a function of height. The different colours show different reflectivities which correspond to different densities of cloud or rain. This figure demonstrates that even when statistics over a number of storms are considered the message is similar to that in Fig. 1, namely that the UKV model produces storms that are too wide but the 200m model produces storms that are too narrow. This statistical approach is complemented by detailed physical process studies of the kind undertaken in COPE.
Another key parameter which controls the behaviour of the convection is the vertical motion of the air. This is very difficult to measure in practice because it tends to be much smaller than the horizontal winds. DYMECS is pioneering a new technique to obtain statistics of vertical velocities. The radar can measure the radial velocity (i.e. towards and away from the radar) using the Doppler effect and, with assumptions, the vertical motion can be estimated. This will greatly help understanding the performance of the models.
Last updated: 14 August 2014