Areal Mean Basin Estimated Rainfall (AMBER) Research
send any comments or questions about this research to firstname.lastname@example.org
AMBER is a software application that uses high resolution rainfall estimates from the Weather Surveillance Radar, 1988 Doppler (WSR-88D) to determine where flash flooding is likely. In order to evaluate the flash flood threat, AMBER computes the Average Basin Rainfall (ABR) for 2,427 streams every 5-6 minutes for the Pittsburgh Weather Forecast Office (WFO) area of responsibility. The likelihood of flooding is established by comparing the ABR with Flash Flood Guidance (FFG); i.e., the ABR needed to bring a stream to bankfull as in Sweeney, (1991). In addition to determining flash flood potential for streams, AMBER calculates the ABR for a full range of hydrologic scales, from river forecast areas to urban areas.
2. HISTORY OF AMBER
AMBER has evolved from 13 years of operational development at the Pittsburgh WFO. From 1983 to 1992, rainfall estimates from the Radar Data Processor II (RADAP-II) were used operationally to detect where flash flooding was likely. Several publications resulted from this research, including Davis and Rossi (1985), Davis and Drake (1988), and Davis (1993). The software was upgraded in 1993 to take advantage of the higher resolution rainfall estimates from the WSR-88D radar. The one degree by one kilometer grid of the WSR-88D has nearly four times the resolution of the RADAP-II polar grid.
Previous research has revealed several important factors, which have shaped AMBER's current structure. First, the time scale of the one-hour rainfall product of RADAP-II was too large to detect rapidly developing flash floods. RADAP's ten minute rainfall product provided much better guidance. Second, flash floods typically occur in small watersheds (5 to 20 mi2). The forecaster must have access to these small watershed boundaries to determine flash flood potential. Third, forecasters need to estimate the volume of rainfall (ABR) within these small watersheds to effectively determine the flash flood threat. Finally, digital WSR-88D radar rainfall can provide sufficient time and space resolution to detect flash flood potential.
3. THE HYDROLOGIC DATABASE
The AMBER flash flood hydrologic area analysis starts with the Mean Areal Precipitation (MAP) areas used by the RFCs. Figure 1 shows the outline of the Dashield's (DSSP1) MAP area near Pittsburgh. The AMBER database includes 86 MAP areas.
Each MAP area is divided into primary stream (<200 mi2) or major stream ((200 mi2) basins. Chartiers Creek at Carnegie in Figure 1 is a major stream (257 mi2) that covers two-thirds of the entire Dashield's MAP area. Since the response time of major streams is relatively slow, flood warnings, rather than flash flood warnings, are typically issued for the larger streams. The AMBER database includes 35 major streams.
Figure 1. The Dashield's MAP area.
Flash flood streams in the AMBER database are either primary streams or their subdivisions (i.e., tributaries). Primary streams are less than 200 mi2 in area, but the great majority are less than 50 mi2. Figure 2, shows the 34 primary streams defined within the Dashield's MAP area. The AMBER database contains 1163 primary streams.
Chartiers Run is a primary stream (1646 in Figure 2) that flows into the city of Houston, PA. A flash flood occurred within this basin on September 7, 1990. AMBER estimated 4.6" of rain in three hours, allowing forecasters to issue a flash flood warning with almost an hour of lead time.
Stream subdivisions are smaller tributaries contained within primary streams or other stream subdivisions. Figure 3 shows the 40 stream subdivisions contained within the Dashield's MAP area. The current AMBER database contains 1,264 subdivisions.
Plum run (5155 in Figure 3) is a subdivision of the Chartiers Run watershed. A traveling circus troop was camped along Plum Run the night of the flash flood in 1990. The warning lead time of almost an hour, provided by AMBER, allowed their evacuation to Main Street in Houston, where elephants and camels were tied to the telephone poles, creating some very unique storm data pictures.
Urban areas can experience dangerous flash flooding when storm sewers back up from intense rainfall. Many of these urban areas are not directly impacted by stream flooding. AMBER calculates average rainfall in these urban areas to alert the forecasters to possible urban flooding. The AMBER database includes 141 urban areas.
Figure 2. Primary streams in the Dashield's MAP area.
4. ABR COMPUTATIONS
AMBER uses the WSR-88D Digital Hybrid Scan Reflectivity product (DHR) as input for estimating rainfall. The DHR provides one degree by one kilometer reflectivity data from the Hybrid Scan produced by the Precipitation Preprocessing algorithms as described in NOAA (1991). Rainfall accumulations are computed from successive DHR products, following as closely as possible the current Precipitation Rate and Precipitation Accumulation algorithms. However, data are output at 1 km resolution, instead of the 2 km data associated with the current Precipitation Rate algorithm.
Figure 3. Subdivisions in the Dashield's MAP area.
The computed ABR values represent an area weighted average accumulation of all radar rainfall estimates in a watershed, including zero accumulations. The 5-6 minute ABR values are summed to create rainfall accumulations for 0.5, 1,2,3 and 6 hours, and then compared with the RFC's 1, 3 and 12 hour FFG.
5. AMBER OUTPUT
At the completion of each radar volume scan, any stream or urban area whose ABR exceeds 60% of FFG is output to the computer screen. An output display table contains the stream basin number, state and county of the stream location, and the ABR and FFG for each time period. A yellow alert level is reached when ABR is 80% of FFG, and a red alert is reached when ABR exceeds 100% of FFG. By viewing the output ABR vs. Time display, the forecaster can quickly assess total accumulation, rate of accumulation, rainfall duration, and comparison to FFG over the past 6 hours.
Figure 4. RADAP-II ABR Output.
Figure 4 illustrates RADAP-II AMBER output for the Etna flash flood event of May 30, 1986. During this event, a 10 foot wall of water inundated a road during rush hour, causing 9 fatalities. Notice that the slope of the ABR curve (the rainfall rate) is steeper than the FFG slope from the onset of the event. Furthermore the yellow alert level was exceeded 37 minutes before the actual flood wave occurred. While the red alert level was 28 minutes before the flood event. This demonstrates that AMBER can provide significant lead time in serious short-fused flood events.
A maximum rainfall product can be printed by the forecaster, indicating the 20 primary and 20 stream subdivisions with the highest ABR totals for the previous 1, 2, or 3 hours. The output includes the stream number and name, county and state, area of the stream in square miles, ABR, FFG, and minutes of rainfall duration. This direct comparison with FFG provides a quick summary of all streams with significant flash flood potential.
A rainfall rate product can also be printed, showing the 20 primary and 20 subdivisions with the highest observed ABR rainfall rates during the past 1 hour. This product is useful for detecting potentially heavy rainfall events early in the accumulation phase of the storm.
6. OPERATIONAL APPLICATIONS
The ABR output from AMBER supports the entire range of hydrologic functions of the forecast office, from flash flooding to large scale river flooding.
AMBER output for primary streams, subdivisions, and urban areas provides reliable guidance for issuing flash flood warnings. Before AMBER, forecasters relied primarily on point rainfall amounts from rain gages, or maximum radar estimates to make flash flood warning decisions. Using the ABR calculations from AMBER, forecasters can obtain a better estimate of the actual volume of rainfall occurring within a stream basin, which is directly related to the volume of water in the flood hydrograph.
Although designed specifically for detection of flash floods, AMBER can be used to detect stream flooding that occurs over time spans of more than 6 hours. AMBER can compare ABR with 6 to 12-hour FFG to determine the risk of stream flooding. In areas not prone to frequent flash flooding, such as the midwest of the United States, AMBER can also be used to determine the flood potential for low lying areas prone to significant ponding of water.
The MAP computations in AMBER can be used at the forecast office to provide quick updates to river forecasts. The MAP estimates can be compared with the Quantitative Precipitation Forecasts (QPF) that were used to produce the river forecasts. The Pittsburgh WFO frequently uses the radar rainfall to update river forecasts based on comparison with QPF or by requesting new forecast runs from the RFC.
Davis, R. S., and T. Rossi, 1985, AScheme for Flash Flood Forecasting using RADAP-II and ICRAD, Preprints, 6th Conference on Hydrometeorology, AMS, 107-111.
Davis, R. S., and T. Drake, 1988, The Operational Effectiveness of RADAP-II During the 1987 Severe Weather Season in Western Pennsylvania, Preprints, 15th Conference on Severe Local Storms, AMS, Baltimore, MD, 194-197.
Davis, R. S., 1993, AMBER, A Prototype Flash Flood Warning System, 13th Conference on Weather Analysis and Forecasting Preprints, AMS, Vienna, Virginia, August 2-6, 1993, 379-383.
NOAA, 1991, Federal Meteorological Handbook No. 11, Doppler Radar Meteorological Observations, Part C, WSR-88D Products and Algorithms, U.S. Department of Commerce, Washington, D.C., 3-1 to 3-58.
Sweeney, T. L., 1991, Modernized Areal Flash Flood Guidance, NOAA Technical Memorandum 44, Office of Hydrology, Silver Spring, MD,1-17.