QPF and Flooding on the Meramec
Robert J. Wavrin
North Central River Forecast Center
The North Central River Forecast Center (NCRFC) began inputting 12 hours of Quantitative Precipitation Forecasts (QPF) into their river forecast model runs in June 1994. As a part of the evaluation process, the NCRFC is carrying out a series of studies on the utility of incorporating the QPF forecasts. This study examines the impact of the QPF on the river stage predictions issued for an event over the Meramec Basin in east central Missouri (Figure 1) during mid-January 1995.
Before June 1994, all released river stage forecasts issued from the NCRFC were based on precipitation up through 1200 UTC of the day the forecasts were issued. No future precipitation was considered, even if it was occurring at the time of the forecast. This often led to obvious errors in the forecasts and caused confusion with potential users. For example, a forecast may have indicated falling river trends when in reality the trends were upward.
Each morning several Weather Service Forecast Offices (WSFO) send 24 hours of QPF to the NCRFC in six hour increments. The NCRFC incorporated only the first 12 hours (first two increments) of expected precipitation in their river stage forecast model runs. This limit was based on skill levels shown during a risk reduction exercise in Wisconsin for 1993/94.
This study focuses on the events that occurred on January 13, 1995. Rain had fallen across the Meramec Basin in the days preceding, allowing the soil in the area to be saturated. The National Meteorological Center's (NMC) QPF (AFOS graphics' 94Q and 98Q) issued on January 12 had been highlighting rain across the southeast quarter of Missouri. In particular, the Day 2 forecast valid for the period 1200 UTC 13 January to 1200 UTC 14 January had between 0.25 and 0.50 inches of rain over the basin. As shown in Figure 2, the heaviest amounts (up to 2 inches) were forecast from western Mississippi up to southern Illinois. When updated on January 13, NMC had the 24-hour 1 inch forecast over a good portion of the Meramec Basin (Figure 3). The six hour amounts predicted by NMC on January 13 (AFOS graphics' 92E and 93E) were between 0.25 and 0.50 inches for each period with 1.5 inches maximum in each period. Thus, 12 hour totals over 1.0 inches were possible (Figure 4 and Figure 5).
The St. Louis Forecast Office was somewhat conservative in their precipitation forecasts, particularly during the second period. The WSFO predicted a half inch the first period ending at 1800 UTC 13 January for all basin segments and between a third and a half inch during the second period ending at 0000 UTC 14 January.
Actual rainfall amounts of 0.25 to 0.50 inches were common during each six hour period from 1200 UTC 13 January to 0000 UTC 14 January (Figure 6 and Figure 7). The actual amounts showed a reasonable correlation with the amounts forecast by the WSFO.
For the first 6-hour period 0.50 inches had been forecast for all segments. Actual amounts ranged from 0.30 over the eastern segments to 0.50 over the western segments. The corresponding NMC product had 0.50 or more with a 1.50 inch maximum.
During the second period a similar distribution occurred with 0.30 inches over the eastern segments and up to 0.50 inches over the western segments. This matched very closely with the WSFO forecasts for the same period. Again the NMC forecast had an even larger area of 0.50 inches or more with another 1.50 inch maximum just to the east (less than 0.25 inches fell in the area of expected maximum activity). The accumulated precipitation values for the first two periods ranged from just over 0.50 inches over the southeast segments to 1.00 inches over western basin segments.
A similar pattern continued the following two periods, although they were not included in the river stage forecasts. The 24-hour total from 1200 UTC 13 January to 1200 UTC 14 January ranged from around an inch in the southeast portion of the basin to an inch and three quarters along the western tier of segments (Figure 8). Although NMC moved the area of precipitation northwest as forecasts were updated, it still had the band of heaviest rain to the southeast of where it actually fell.
Several model runs were completed on January 13, 1995. Figures 9, 10, 11, and 12 display hydrographs for four selected sites in the Meramec Basin. Each Figure shows a hydrograph with no QPF, one with 12 hours of QPF and a hydrograph of what actually happened.
These hydrographs represent the eight stations in the Meramec Basin. By inputting future precipitation into the model, every station had an increased stage forecast. Forecast crest stages increased 2 to 6 feet with an average crest increase of 4.4 feet. By using QPF two additional stations, making a total of four, had forecast crests above flood stage.
Forecasts for the Meramec Basin were sent out beginning on the morning of January 13, 1995. As more data were received during the flooding event, forecasts were updated. Table 1 depicts the first forecast issued for each site in the basin. These forecasts were based on the model run using QPF and also other model runs using runtime changes such as modifying soil moisture, adjusting base flow in the river and adjusting the amount of runoff from earlier rainfall that the model didn't fully react too. These changes were made to match the forecast hydrograph, in the days before the January 13 event, to the actual hydrograph (this can be seen in Figures 9 through 12 in the period up to 1200 UTC 13 January, as all three hydrographs are the same). Then when QPF was input into the model, the resulting forecast hydrograph was as accurate as possible.
|RIVER||FORECAST CREST||OBSERVED CREST|
|Station (Flood Stage)||Stage (Feet)||Date||Stage (Feet)||Date|
|Union (15)||Near Flood Stage||1/15/95||19.5||1/16/95|
|Byrnesville (16)||18 to 19||1/14/95||Below Flood Stage|
|Steeleville (12)||Moderate within bank rise+||1st 10.8
|Sullivan (15)||17 to 18||1/14/95||16.6||1/15/95|
|Pacific (15)||Near Flood Stage||1/16/95||18.2||1/17/95|
|Eureka (18)||18 to 19||1/16/95||19.5||1/17/95|
|Valley Park (16)||Near Flood Stage||1/16/95||18.3||1/17/95|
|Arnold (24)||21 to 22||1/16/95||21.5||1/17/95|
Although the forecasts were not perfect, most of them improved on the forecast made without QPF. Of the eight forecasts, only one (Byrnesville, Figure 10) over predicted the actual crest, but was still below flood stage. This may have been the result of a gaging problem since the Byrnesville, Missouri hydrograph peaked at the onset of precipitation then oscillated up and down as rain fell for 24 hours (Figure 10). Most of the observed crests that reached or exceeded the forecast crest were also greater than flood stage. The higher observed crests can partially be attributed to continued rainfall past the 0000 UTC cutoff for the QPF input into the model and partially on model the model not being perfect.
How well the river forecasts would have looked if all 24 hours of QPF sent by WSFO St. Louis was used, is unknown. However, it is likely that the forecast crests would have been higher had the QPF been used.
On June 1, 1995, the NCRFC began inputting 24 hours of QPF into the river model. This will give the HAS forecasters more opportunity to evaluate how QPF contributes to river forecasts as nocturnal convection will be accounted for now.
The author would like to thank HAS forecasters Holly Reckel and Steve Kruckenberg at the NCRFC for gathering the bulk of the data. Thanks also go to Dick Felch who so kindly edited drafts of the paper.