DETECTING TIME DURATION OF RAINFALL: A CONTROLLING FACTOR OF FLASH FLOOD INTENSITY
Author address: Pittsburgh NWS, 192 Shafer Road, Moon Twp., PA 15108; e-mail:Robert.Davis@noaa.gov.
Photo (right) shows Dirty Camp Run as it normally is flowing through the community of Pitcairn PA.
The severity of flash flooding in a watershed is determined by the average rainfall rate occurring across the watershed and the time duration of that rainfall rate. The rainfall rate combined with the time duration determine the volume of water that falls in the watershed, and is directly related to the runoff of water into the stream. The portion of the rainfall volume converted to runoff is modulated by several hydrologic factors including evaporation, infiltration, initial surface retention, the distribution of rainfall in the basin, and the intensity/duration of rainfall in the basin.
These hydrologic factors are quantified for the forecaster by flash flood guidance (FFG). Forecasters must fully understand the assumptions and limitations of FFG in order to effectively apply the guidance. A description of the assumptions used to compute FFG can be found in Sweeney (1991). Adherence to FFG without this knowledge can lead to serious problems. For example, there are situations where the county FFG will mask the flash flood potential across a small watershed, especially during intense convective events. This paper will address some important assumptions that determine FFG in order to illustrate the correct application of FFG to the flash flood warning process. Output from the Areal Mean Basin Estimated Rainfall (AMBER) program (Davis and Jendrowski, 1996) will be used to examine the Pitcairn, PA, flash flood of 1 July 1997. It will be shown that intense convective rainfall occurred over very small watersheds in a short period of time. This paper will show how the forecaster can use AMBER output, compared to FFG, to determine the flash flood threat well before flooding begins.
2. FLASH FLOOD GUIDANCE FACTORS
FFG is defined as the average basin rainfall (ABR) required over a specified time interval to begin flooding. In the northeast United States, FFG is produced for each county for a 1, 3, 6, 12 and 24 hour time duration. The FFG is updated once per day, based on the average rainfall in a Mean Areal Precipitation (MAP) area. Notice in Figure 1, the Turtle Creek basin, that contains the city of Pitcairn, covers only about 32% of the Braddock MAP area.
Figure 1. Braddock MAP area with 24-hour rainfall amount from rain gages used in the FFG computation. The solid line is the border of the Braddock MAP area, the dashed line in the southern border of the Turtle Creek basin, and the dotted lines are county borders.
Urban FFG is also produced for large urban areas. The urban FFG is generally lower than the county FFG containing the urban area because less rainfall can penetrate the urban ground. Therefore, the urban FFG is based on the percentage of impermeable ground found in the urban area. Table 1 shows the FFG for several counties in southwest Pennsylvania and the Pittsburgh urban FFG for 01 July 1997.
FFG must be compared with average basin rainfall (ABR). Therefore, a single point rain gage measurement of rainfall is not a valid comparison with FFG. Moreover, ABR can only be computed for defined watershed areas. Consequently, the FFG values in Table 1 represent the ABR needed to start flooding for any specific watershed in a county or urban area. It is important to note that FFG values will provide an accurate depiction of the hydrologic conditions of a watershed only if the following conditions are met:
* The observed ABR is uniformly distributed in the watershed in both time and space.
* The stream in the watershed is at low flow levels before the onset of the rainfall.
* Significant rainfall has not occurred in the watershed since the rainfall input cutoff time of 1200 UTC.
The daily update of FFG is created by using the average rainfall in a MAP area. MAP areas are large (200 to 1000 mi 2) compared to typical flash flood watersheds (1 to 100 mi 2). As a result, localized heavy rainfall can be masked, causing large discrepancies between the watershed and MAP area ABRs. This leads to a deceptively high county FFG, especially during the summer convective season. For example, if the observed ABR in a flash flood watershed for the past 48 hours ending at 1200 UTC was 3.50 inches, and the MAP area containing this watershed had an observed ABR of 0.50 inches, the county FFG issued at 1200 UTC will be too high for the flash flood watershed. This illustrates the importance of knowing of the recent ABR history of all flash flood watersheds.
3. COMPARING ABR WITH FFG
The Areal Mean Basin Estimated Rainfall (AMBER) program (Davis and Jendrowski, 1996) can be used to examine many of the factors that determine the accuracy of FFG, including ABR history plus the spatial and temporal variations of ABR. AMBER uses the Digital Hybrid Scan Reflectivity (DHR) product from the Weather Surveillance Radar - 1988 Doppler (WSR-88D) to compute radar rainfall estimates in flash flood watersheds. The DHR product uses a polar grid of 1 o x 1 km for all radar azimuths from 0 o to 359 o and for all radar ranges from 1 to 230 km. A single rainfall estimate is computed for each 1 km range bin. A rainfall amount is calculated every 5-6 minutes for each range bin. All range bins, whose center point falls in a stream watershed are averaged to compute the ABR for that watershed. The small DHR rainfall grid allows AMBER to compute ABR in watersheds as small as 1 mi 2 in area.
The observed 6 minute ABR is multiplied by 10 to compute the ABR rate, giving a normalized hourly rainfall rate. This ABR hourly rate provides critical information for evaluating the flash flood threat, and signals any watersheds that may flood, before flooding begins. For example, if the ABR hourly rate remains below FFG, flash flooding is not likely. Conversely, if an ABR rate is higher than the 1-hour FFG, and that rate is maintained for one hour, flash flooding is possible. Of course, the forecaster must still determine if the heavy rainfall rate will continue long enough to cause flooding.
Figure 2. Distribution of ABR in the Turtle Creek subdivisions. Each watershed is identified by a 4 digit number and the ABR for each watershed from 2200 UTC 1 July until 0015 UTC 2 July are labeled in parentheses.
The correct application of FFG requires that rainfall be uniformly distributed across the watershed. The forecaster can help satisfy this FFG factor by examining ABR in progressively smaller watersheds. The 2 hour rainfall that produced the Pitcairn flash flood is shown in the AMBER ABR analysis of the subdivisions of Turtle Creek (Fig. 2). Dirty Camp Run (subdivision #5323), which contains the city of Pitcairn, had the highest observed ABR during this two hour period. Dirty Camp Run and Aber's Creek (Fig. 3), experienced the majority of the flash flood damage. A detailed
Figure 3. Severely flooded subdivisions of the Turtle Creek watershed. The numbers in parentheses are the ABR for each watershed. Dots show rain gage locations. Hatched areas are cities.
analysis of the AMBER output for these two basins is presented. These basins are in the highly urbanized eastern suburbs of Pittsburgh. As a result, all ABR comparisons are made with the Pittsburgh urban FFG. The subdivisions of these two small watersheds will be examined to see if the requirement of uniform ABR in both space and time has been met, allowing for valid FFG comparisons with ABR.
3.1. The Aber's Creek Watershed
AMBER ABR (solid line) and ABR rate (dashed line) for the 10 mi 2 Aber's Creek watershed is shown in Fig. 4. If the radar estimate of rainfall is accurate and the FFG is valid, Aber's Creek should not flood, since the ABR (1.49 inches) is below the 1 hour FFG of 1.7 inches (Table 1). However, this was a convective event, resulting in a wide variation of rainfall amounts across the Aber's Creek basin.
Figure 4. ABR rate plot for Aber's Creek for 1 July between 2210 and 2340 UTC. The solid line is accumulated ABR and the dashed line is ABR rate.
Figure 5. ABR distribution in the subdivisions of Aber's Creek for 2200 UTC 1 July until 0000 UTC 2 July. The ABR is labeled for each subdivision in parentheses.
Figure 5 illustrates the observed ABR for the tributaries of Aber's Creek. A wide variation of ABR is shown across the Aber's Creek basin. Recall, FFG will represent the hydrologic conditions of a watershed only if the observed ABR is uniformly distributed across the basin in both time and space. The headwaters of Aber's Creek (#6308) received only 0.39 inches of rain, while Thompson Run (#6306) had an ABR of 3.21 inches! During the same time, the Holiday Park IFLOWS gage indicated 1.26 inches of rain while the Gateway Middle School gage reported 3.30 inches of rain (Fig. 5). Figure 6 shows the ABR and ABR rate plot for Thompson Run. It is noteworthy that the ABR rate reached 4 in hr -1
Figure 6. ABR rate plot for Thompson Run for 1 July between 2210 and 2340 UTC 1 July.
around 2250 UTC, and the associated ABR values were 1 inch over FFG by 2310 UTC. Flash flooding was reported to the National Weather Service at 2320 UTC along PA Route 22, about 4 km west of Murrysville. No homes are located immediately along Thompson run and flash flood problems in this watershed were primarily road flooding. Mud slides were reported at the intersection of I-76 and PA-22 (Fig. 5)
This case illustrates the value of examining the ABR and ABR rate of small watersheds (the Thompson Run watershed is 2.4 mi 2). By reducing the size of the watershed until the assumption of nearly uniform rainfall was possible, a valid comparison of FFG and ABR resulted. The ABR exceeded FFG by over one inch and significant flood occurred in the watershed. Thompson run was the only subdivision in the Aber's Creek watershed that reported any flooding.
The rainfall history of the previous 48 hours was not a factor in this flash flood event. No rainfall was observed in the Dirty Camp Run or Aber's Creek basins from 1200 UTC 29 June to 1200 UTC 1 July 1997. Thompson Run did receive 0.20 inches of ABR from 1200 UTC to 1600 UTC on 1 July 1997 with little impact on FFG.
3.2. The Dirty Camp Run Watershed
An estimated 10 million dollars of damage occurred to homes, businesses and cars in the city of Pitcairn on 1 July 1997. Pitcairn is contained in the southern portion of the Dirty Camp Run watershed (Fig. 3). Figure 7 highlights the ABR and ABR rate for the Dirty Camp Run basin (3.3 mi 2). Notice that ABR rates greater than 2 in hr -1 lasted for almost 40
Figure 7. ABR rate plot for Dirty Camp Run between 2215 UTC 1 July and 0015 UTC 2 July. The lines are the same as in Fig. 3. Note the county 1 hr FFG labeled at 1.7 inches.
minutes, from 2255 to 2335 UTC. As a result, the FFG was exceeded by 2325 UTC. Comparing Fig. 6 and Fig. 7, more significant flash flooding would be indicated in Thompson run than in Dirty Camp Run. ABR is much higher and the ABR fell in a shorter time duration in Thompson Run. Both of these factors should act to increase the flood wave in Thompson Run. However, the observed flooding was more intense in Dirty Camp Run. The following discussion shows how several other hydrologic factors combined to magnify the flood wave in Dirty Camp Run.
Despite the small size of Dirty Camp Run, significant variation of ABR occurred across the basin in both time and space. Figure 8 shows the ABR variation from the northern basin (#6305) to the southern subdivision (#6304) of Dirty Camp Run. Most of this watershed is highly urbanized, more urbanized than the Thompson Run watershed. The stream bed of Dirty Camp Run, within the city of Pitcairn, flows through a channel with walls made of stone block and mortar. The channel is about 16 feet wide and 9 feet deep. Just south of Eleanor street the stone channel goes directly under an elementary school. During the flood, water was flowing through the windows of the school.
Photo (right) shows another view of Dirty Camp Run as it flows through Pitcairn.
Photo (bottom ) shows how the local school building is built directly over Dirty Camp Run. During the flooding, water was said to be flowing through the windows of the school.
Figure 8.ABR distribution in the subdivisions of Dirty Camp Run for 2200 UTC 1 July until 0015 UTC 2 July.
Figure 9 indicates that the ABR rate for the northern basin (#6305) peaked at 3.7 in hr -1 at about 2300 UTC. This water moved quickly down stream and joined with the accumulating
Figure. 9. ABR rate plot for the Dirty Camp Run (2), #6305 in Fig. 8, subdivision between 2215 UTC 1 July and 0015 UTC 2 July.
rainfall in the southern basin (Fig. 10), where the ABR rate peaked at 2330 UTC. The phasing of the runoff in the northern basin with the rainfall in the southern basin significantly increased the flood wave in Pitcairn. The elevation of the Dirty Camp Run basin falls 460 feet in a distance of 5.5 km from the headwaters to the city of Pitcairn. Residents in Pitcairn reported that Dirty Camp Run was rising rapidly before the rain started falling in the city.
Figure 10. ABR rate plot for the Dirty Camp Run (1), #6304 in Fig. 8, subdivision between 2215 UTC 1 July and 0015 UTC 2 July.
The examination of ABR in these two very small watersheds provided important clues to the flash flood potential for the city of Pitcairn. The one-half hour time separation of the rainfall in the north and south enhanced the flood wave in Pitcairn. The northern basin received a burst of 2 in hr -1 rates from 2250 to 2320 UTC, while the southern basin had a similar burst of 1.5 in hr -1 rates from 2310 to 2355 UTC. These two bursts of heavy rainfall produced independent flood waves that joined forces in the city of Pitcairn between 01 July 1997 at 2330 UTC to 02 July 1997 at 0030 UTC. Only by examining the ABR in watersheds less than 2 mi 2 in area, and with WSR-88D rainfall data in 6 minute time intervals, did these important details become apparent to the forecaster.
Dirty Camp Run did receive 0.40 inches of ABR from 1200 UTC to 1600 UTC on 1 July 1997. This previous rainfall may have slightly reduced the FFG from the values listed in Table 1, perhaps by 0.2 inches. If the 1 hour FFG was reduced to 1.5 inches, the additional runoff would help to explain the enhanced flooding in Pitcairn.
Remember that ABR is a WSR-88D radar estimate. The forecaster needs to be assured that the radar is estimating the rainfall with some accuracy. AMBER computes an estimate of rainfall at the location of telemetered rain gages so the forecaster may estimate the accuracy of the ABR. Figure 11 shows a plot of the radar estimate for the two rain gages in Fig. 5. The radar estimates at the rain gage locations are very close to the rain gage measurements, indicating the WSR-88D radar estimates were very good .
Figure 11. AMBER ABR rainfall estimate for the Holiday Park IFLOWS and Pitcairn Middle School rain gages. Solid line is radar estimate. Dashed line is rain gage measurement.
Valuable comparisons of ABR with FFG can be accomplished if stream analysis is done for flash flood streams as small as 2 mi 2. Detailed WSR-88D data in the form of the DHR product can be used to successfully calculate the ABR data in these small watersheds.
The ABR analysis must be carried out in short time steps of 5-6 minutes, so that the details of the temporal distribution of ABR can be found. The short time duration rainfall analysis also allows the computation of ABR rate, which can alert the forecaster to developing flash floods well before flooding begins. The National Weather Service in Pittsburgh uses an alert threshold of 1 in hr -1 ABR rates, printing all streams that reach this threshold every 6 minutes. Notice in Fig. 6 that Thompson Run reached 1 in hr -1 at 2217 UTC and flooding began at 2255 UTC when ABR reached FFG. The forecaster was alerted about 40 minutes before flooding actually began, and over one hour before the first flooding report was received at 2320 UTC.
The National Weather Service flash flood warning program can be greatly enhanced by including within the warning text the streams and urban areas that will be impacted by the flash flooding. The detailed stream analysis that is found in the AMBER stream database allows the forecaster to easily include this type of warning call to action in real time. Including this detailed information on the location of the flooding allows the public to make a more accurate assessment of their level of risk.
The forecaster must be familiar with the assumptions that go into the formulation of FFG, to effectively apply that guidance to flash flood warning operations. Using the ABR data produced by AMBER, the forecaster can determine the relative merits of the current FFG.
The author would like to thank Tom Salem, Phil Manuel and Josh Korotky for their excellent reviews and suggestions.
Davis, R. S., and P. Jendrowski, 1996: The Operational Areal Mean Basin Estimated Rainfall (AMBER) Module, Preprints, 15th Conference on Weather Analysis and Forecasting, Amer. Meteo. Soc., Norfolk, VA, 332-335.
Sweeney, T. L., 1991: Modernized Areal Flash Flood Guidance, NOAA Technical Memorandum 44, National Weather
Send any comments or questions about this research to firstname.lastname@example.org