Tracking the three-dimensional evolution of convective storms in radar observations and high-resolution models

Forecast models nowadays run routinely at convection-permitting resolution, specifically in the United Kingdom, where the Met Office UK 1.5km model (referred to as the UKV) is run four times a day to provide a 36-hour forecast to the public. At this resolution, individual convective storms are expected to be resolved, but comparison with surface rain rate observations from the Met Office radar network still reveals substantial errors in the timing and character of these storms. The discrepancies between modeled and observed storms requires the evaluation of three-dimensional structure of storms in models, using more detailed radar observations of the cloud and precipitation structures throughout storm life cycles. These results will help improve the microphysics and sub-grid mixing schemes, as well as determine whether even higher resolution (down to 100m) models provide a notably better representation of the three-dimensional evolution of convective storms.

A data set of high-resolution radar observations for a large number of days with convective storms is used to evaluate such storms in the UKV model for the DYMECS project (Dynamical and Microphysical Evolution of Convective Storms). The 3 GHz Chilbolton radar has been set up to automatically track convective storms in real-time through a scan-scheduling algorithm linked to a database of storms identified in the Met Office rainfall radar network. The radar is scheduled to obtain volume scans, generated from sets of scans at different elevation angles stretching across the entire storms horizontally, as well as sequences of vertical scans through the locations of most intense rainfall.

In this paper, the life cycle of a number of storms is presented in terms of the maximum height of occurrence and total area of different reflectivity values representative of cloud and precipitation processes. These measures of three-dimensional evolution are then related to the life cycle of the storms as measured by the intensity and coverage of the storm surface precipitation. A similar approach is performed on storms in the UKV so that the modeled evolution of microphysical processes can be evaluated statistically, where reflectivities are forward modeled from the UKV cloud and precipitation fields to allow for a like-with-like comparison. Due to the model's inability to represent the smallest storms, a minimum size threshold of 40 square kilometers is imposed on both the modeled and the observed storms used for the evaluation.

In terms of surface rainfall characteristics, the UKV storms show little variability in their life cycles, with a peak mean rainfall after the first quarter of their lifetime, which is sustained until about halfway through the life cycle, at which the modeled storms have reached their largest extent in surface rainfall area. This is in contrast with the observed storms, which tend to reach peak mean rainfall within the first quarter of their life cycle, after which storms reach their largest extent before halfway through the cycle.

In terms of the detailed three-dimensional structure observed with the Chilbolton radar, the height of the storms identified with different reflectivity thresholds is higher than such heights obtained from the UKV storms. For instance, the model generally fails to achieve reflectivities above 40 dBZ above the melting layer, suggesting that the microphysics scheme should include graupel to better represent the microphysical processes in convective cores. In addition, the rate at which storms grow and decay in both height and area of different reflectivity thresholds is used to identify key periods in the convective storm life cycle during which the model requires further processes to be represented by the microphysical and sub-grid mixing schemes.