Pyrad: a Real-Time Weather Radar Data Processing Framework Based on Py-ART

Pyrad is a real-time data processing framework developed by MeteoSwiss. The framework is aimed at processing and visualizing data from individual Swiss weather radars both off-line and in real time. It is written in the Python language. The framework is version controlled and automatic documentation is generated based on doc-strings. It is capable of ingesting data from all the weather radars in Switzerland, namely the operational MeteoSwiss C-band rad4alp radar network, the MeteoSwiss X-band DX50 radar and the EPFL MXPol radar.

The processing flow is controlled by 3 simple configuration files. Multiple levels of processing can be performed. At each level new datasets (i.e. attenuation corrected reflectivity) are created which can be stored in a file and/or used in the next processing level (for example, creating a rainfall rate dataset from the corrected reflectivity). Multiple products can be generated from each dataset (i.e PPI, RHI images, histograms, etc.). In the off-line mode, data from multiple radars can be ingested in order to obtain products such as the inter-comparison of reflectivity values at co-located range gates.

The framework is able to ingest polarimetric and Doppler radar moments as well as auxiliary data such as numerical weather prediction parameters (i.e. temperature, wind speed, etc.), DEM-based visibility and data used in the generation of the products such as rain gauge measurements, disdrometer measurements, solar flux, etc.

The signal processing and part of the data visualization is performed by a MeteoSwiss developed version of the Py-ART radar toolkit which contains enhanced features. MeteoSwiss regularly contributes back to the main Py-ART branch once a new functionality has been thoroughly tested and it is considered of interest for the broad weather radar community.

The capabilities of the processing framework include various forms of echo classification and filtering, differential phase and specific differential phase estimation, attenuation correction, data quality monitoring, multiple rainfall rate algorithms, etc. In addition time series of data in points, regions or trajectories of interest can be extracted and comparisons can be performed with other sensors. This is particularly useful when performing measurement campaigns where remote sensing retrievals are validated with in-situ airplane or ground-based measurements.