Underestimation of QPE in a Flash Flood Situation Due to Partial Beam Blocking and Attenuation: Correction Using the BREAM model

A historic flash flood event occurred in the northern suburbs of Birmingham, Alabama on 7 May 2003. Rainfall totals of 20 to 28 cm (8 to 11 inches) were common in northeastern Jefferson County near the Huffman and Trussville communities, with much of the rain falling in a very short time period. Rain gauge observations at the Birmingham Airport (KBHM) indicate that rain rates reached 61 mm hr-1 (about 2.4 inches per hour), and that the storm total rainfall, over about a 4 hour period, was 13 cm (5.14 inches). The Cahaba River at Trussville rose from 2.37 ft. at 07/1800 UTC to 19 ft. at 18/0000 UTC, well above flood stage. Streamflows in that normally small part of the river reached 208 m3 s-1 (7,352 cubic feet s-1).

The Birmingham, AL (BMX) WSR-88D radar is located almost directly south of KBHM, at a distance of only 42 km. However, the radar is only at 212 m MSL, while the foothills of the Appalachian Mountains, including Shades Mountain and Double Oak Mountain, rise as high as 438 m MSL between the radar and some of the densely populated parts of the Birmingham metropolitan area, including the airport, Huffman, and Trussville. This produces severe beam blockage, at some azimuths as high as 50% (3 dB), over highly populated areas. At high rainfall rates like those that occurred on 7 May 2003, a 3 dB error may produce a rain rate error of 50 mm hr-1 (2 inches per hour).

A numerical model developed at UAH, the Blocking of Radar by Elevation and Azimuth Model (BREAM), uses digital terrain data and radar propagation characteristics (based on proximity or model soundings), to determine the percentage of beam blockage occurring at any azimuth/range point in a radar's domain, and then adjusts the reflectivity and radar QPE accordingly.

In the 7 May 2003 case, the BMX radar estimated storm total rainfall of only 8.9 cm at KBHM, and 16.5 cm at Roebuck. This is compared to the observed rainfall of 13 cm at KBHM and 28 cm at Roebuck. In these areas of beam blockage, the nearby radar underestimated the storm total rainfall by 32% and 41%, respectively. With BREAM model corrections, radar rainfall estimates were improved significantly. Without real-time gauge measurements, forecasters in similar situations may be forced to make warning decisions based on incorrect information.