Procedures for improved weather radar data quality control

Resumen

Weather radar data and its downstream products are essential elements in weather surveillance and key parameters in the initialisation and validation of hydrological and meteorological models, among other downstream applications. Following the quality standards established by the European and global weather radar networking referents, the present thesis aims for the improvement of the base data quality control in the regional weather radar network operated by the Meteorological Service of Catalonia, the XRAD. This objective is accomplished through the analysis, development and implementation of new or existing procedures and algorithms for radar data quality assessment and improvement. Attending to the current radar technology and to the already implemented quality control procedures for the XRAD, the work is focused on the continuous evaluation of the radar system calibration status and on the correction of Doppler velocity data. The quality control algorithms and recommendations presented are easily translatable to any other operative weather radar networking environment. A Sun-based, fully automatic procedure for online monitoring the antenna alignment and the receiver chain calibration is adapted and operationally implemented for the XRAD. This Sun-monitoring technique was developed at the Royal Netherlands and Finnish Meteorological Institutes and is included in the quality control flow of numerous weather radar networks around the world. The method is modified for a robust detection and characterisation of solar interferences in raw data at all scan elevations, even when only data at relatively short ranges is available. The modified detection algorithm is also suitable for detecting interferences from wireless devices, which are stored for monitoring their incidence in the XRAD. The solar interferences detected, in turn, are input observations for the inversion of a two-dimensional Gaussian model that yields estimates of the calibration parameters of interest. A complete theoretical derivation of the model establishes its validity limits and provides analytical estimates of the effective solar widths directly from radar parameters. Results of application of this Sun-monitoring methodology to XRAD data reveal its ability to determine the accuracy of the antenna pointing and to detect changes in receiver calibration and radar system operation status. In order to facilitate the usage of the Sun-monitoring technique and the interpretation of its estimates, the methodology is reproduced under controlled conditions based on the distributions of solar observations collected by two of the XRAD radars. The analysis shows that the accuracy of the estimated calibration parameters is conditioned by the precision, number and distribution of the solar observations which constitute key variables that need to be controlled to ensure reliable estimates. In addition, the Sun-monitoring technique is compared under actual operative conditions with two other common techniques for quantifying the antenna azimuth and elevation pointing offsets. Pointing bias estimates gathered in a dedicated short-term campaign are studied in a direct inter-comparison of the methods that reflects the advantages and limitations in each case. The analysis of the bias estimates reported by the methods in the course of a one-year period reveals that the performance of the techniques depends on the antenna position at the time of the measurement. After this study, a reanalysis of the Sun-monitoring method results is proposed, which allows to additionally quantify the antenna pedestal levelling error. Finally, a post-processing, spatial image filtering algorithm for identification and correction of unfolding errors in dual-PRF Doppler velocity data is proposed. The correction of these errors benefits the usage of radar velocity data in downstream applications such as wind-shear and mesocyclone detection algorithms or assimilation in numerical weather prediction models. The main strengths of the proposed algorithm, in comparison with existing correction techniques, are its robustness to the presence of clustered unfolding errors and that it can be employed independently of post-processing dealiasing algorithms. By means of simulated dual-PRF velocity fields, the correction ability of the algorithm is quantitatively analysed and discussed with particular emphasis on the correction of clustered errors. The quality improvement in real dual-PRF data brought out by the new algorithm is illustrated through application to three selected severe weather events registered by the XRAD.