Session 13R.3 Real-time comparisons of VPR-corrected daily rainfall estimates with a gauge mesonet

Saturday, 29 October 2005: 9:00 AM
Alvarado ABCD (Hotel Albuquerque at Old Town)
Aldo Bellon, J.S. Marshall Radar Observatory/McGill Univ., Ste-Anne-de-Bellevue, QC, Canada; and G. Lee, A. Kilambi, and I. Zawadzki

Presentation PDF (1.4 MB)

The RAPID (Radar data Analysis, Processing and Interactive Display) system of the J. S. Marshall Radar Observatory of McGill University generates hourly and daily rainfall accumulations with different methods of correction for the vertical profile of reflectivity (VPR). After removing all non-precipitation echoes from the 5-min volume scans, four 1-h rainfall estimates are produced: 1) VPR uncorrected, (C0). 2) VPR-correct the 1-h C0 accumulations, (C1), 3) Using the observed space-time averaged VPR, correct the lowest (~1.5 km) CAPPI map every 5 minutes, but excluding the convective pixels, and integrate to obtain the 1-h accumulations, (C2), 4) As in C2, but corrected using a reflectivity-dependent climatological VPR, (Cclim). Since late 2004, the various 24-h rainfall accumulation estimates computed at 0000 UTC of each day are automatically compared with the measurements from a gauge mesonet (~40 gauges within 130 km from the radar). The results emphasize the importance of applying a VPR correction scheme to radar-derived precipitation estimates particularly in a climatic regime where most of the liquid precipitation falls from stratiform echoes. The bright band can be at a height at any time of the year in Montreal such that it affects rainfall estimates at some range interval of the radar coverage. It is particularly essential to apply a correction in cases of low bright band, since it is more intense and is observed with higher resolution at near ranges. The raingauge comparisons have revealed that in such “worse-case scenarios” uncorrected biases in excess of a factor 3 in rainfall (or an absolute difference AD > 250 %) have been reduced to 1.2 and 50% respectively by the C2 technique. The C1 and Cclim methods are generally less skillful. However, when the radar already underestimates liquid precipitation due to a miscalibration or an improper R-Z relationship, any VPR correction that further reduces the estimates appears to be worse, emphasizing the importance of separating the different sources of error when assessing the skill of various radar rainfall estimates techniques. The assumption of a homogeneous VPR throughout the radar coverage limits the skill of all correction methods, particularly in our region, where the melting layer can rapidly vary in space and time, especially in early spring and late fall. The results from specific cases will be discussed as well as the overall results that will be tabulated during this upcoming summer.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner