A variational approach to retrieve 3D radar reflectivity composites

Radar networks allow improving Quantitative Precipitation Estimates (QPE) by enlarging the total covered area and providing multiple redundant radar measurements in the overlapping areas. Well-known problems affecting radar QPE such as beam blocking, attenuation by intense precipitation or beam broadening with distance can be mitigated when measurements from multiple radars are available for the same area. Usually radar reflectivity composites are built by selecting the observation from one of the radars at each grid point based on criteria such as the maximum observed value (as an attempt to compensate for strong attenuation and beam blockage), or the observation from the closest radar (which considers the distance to the radar as the main error-driving factor). These classic methods are based on simplified assumptions that may be useful for qualitative uses of radar measurements, but may not get the full benefit of the available network's information for more quantitative purposes.

This study proposes an alternative methodology to obtain high-resolution radar reflectivity composites based on a variational approach. The methodology retrieves the 3-dimensional precipitation field most compatible with the observations from the different radars of the network. With this aim, the methodology uses a model that simulates the radar sampling of the atmosphere. The model settings are different for each radar and considers features such as the radar location, hardware parameters (beam width, pulse length,…) or scan strategy. The methodology follows the concept of an inverse method based on the minimization of a cost function that penalizes discrepancies between the simulated and actual observations for each radar of the network.

The methodology has been applied on the network of radars in the vicinity of Barcelona, Spain. The retrievals have been obtained on a variety of rainfall situations (convective and stratiform rain events) and the analysis focuses on the horizontal and vertical structure of the retrieved precipitation fields.