Extended Abstract (148K)P6B.10
Parameterisation of radar precipitation quality index scheme on raingauge data
Jan Szturc, Institute of Meteorology and Water Management, Katowice, Poland; and K. Osrodka and A. Jurczyk
Proposed Quality Index (QI) scheme is designed to evaluate the quality of radar precipitation data after standard corrections (ground clutter removal, VPR correction, raingauge adjustment, etc.). The data are generated by Polish POLRAD network (8 weather C-band Doppler radars) then processed by British NIMROD system.
In literature reviews of weather radar data errors and methods for their estimation are available. Due to the problem complexity it is impossible to obtain sufficient information and compute quantitatively impact of each error. It seems suitable to be able to evaluate the radar-based data quality not by giving definition and estimation of all radar errors, but analysing the data properties, among others statistical.
The selected quality parameters have been divided into particular groups. One of them is connected to measurement geometry, like distance to the nearest radar, digital elevation map, minimal height of radar visibility (altitude of the lower scan that is more significant at longer distances to radar and in mountainous areas as a result of radar beam blocking, shielding etc.), and volume of radar voxel (volumetric pixel) that depends on scan strategy. Second group of parameters is associated with structure of precipitation field, like spatial and temporal variability that are connected with each other and precipitation rate. The higher the parameters are, the bigger the uncertainty of the fields is. Last parameters are resulting from data aggregation, e.g. number of rain-rate maps (products) that compose 1-hour accumulation.
In the proposed QI scheme quality index for each quality parameter X is calculated. Next all individual QIx fields are summarised to an averaged QI field using appropriate weights. The scheme should be calibrated to get objective information about data quality. It requires procedure of the scheme parameterisation that involves determining some quantities: weights and lower X0 and upper X1 thresholds for each parameter X if linear relationship between quality parameter and quality index is assumed. Beyond the thresholds the quality index QIx values equal 0 or 1 (for quite poor and ideal quality respectively).
A quality reference is needed in order to perform such a calibration on historical dataset. It is assumed that raingauge data are exact in their locations therefore information about differences between raingauge (R) and radar (G) observations in these points can be a measure of quality. The relationship between the variance of raingauge-radar differences (R – G) and given particular quality parameter X is determined. After linear interpolation of these empirical relationships all QIx values can be determined. Next weights of all QIx are optimised using variance of differences (R – G) as a quality criterion.
Finally averaged QI is computed applying the optimal weights and particular QIx fields. The quality information field obtained in this way is attached to the radar-based precipitation product.
Poster Session P6B, Quantitative Precipitation Estimation Forecasting and Hydrological Applications
Tuesday, 7 August 2007, 1:30 PM-3:30 PM, Halls C & D
Previous paper Next paper