During the weekend of October 31 to November 1 1998, 150 to 250 mm (6 to 10 inches) of rain drenched south-central Kansas (Figure 1). The excessive rainfall induced record flooding on the Arkansas, Cottonwood, Walnut, and Whitewater Rivers, as well as major flooding on the Chikaskia, Cottonwood, and Ninnescah Rivers. The flooding extended to the tributaries as well, most notably the Cowskin Creek which flooded much of west Wichita. The historic flooding resulted in one fatality, two injuries, the evacuation of 5,300 people and estimated $32 million in damage to crops, highways, businesses and private property.
The impacted areas received ample warning of the impending flood. The quantitative precipitation forecast (QPF) issued by the Wichita National Weather Service (NWS) office at 0900 UTC 31 October led to the issuance of river forecasts by the Arkansas-Red Basin River Forecast Center (ABRFC) Tulsa, Oklahoma that indicated record flooding on three rivers across south-central Kansas.
This paper will present an overview of the event as well as illustrate the importance of the Quantitative Precipitation Forecast (QPF) in the assessment of flood potential. In this case, the QPF made the difference between a timely forecast that ultimately saved lives versus a delayed forecast that would have provided the public little, if any, time to respond.
At 0000 UTC 31 October 1998, an 850-hPa low was positioned over southeast Arizona with an apparent warm front at that level draped east across New Mexico, the Texas Panhandle and central Oklahoma (Figures 2a-2d). Abundant 850-hPa moisture advected north across Texas and Oklahoma with dew points of 10-12 degrees Celsius (°C) pooling along and south of the frontal boundary. Analyses of the radiosonde data at 500-hPa and 250-hPa revealed an almost vertically stacked low pressure system centered over southern Arizona with a strong ridge over the central Plains (Figures 3a-3b).
Figure 2. 850-hPa analysis at:
a) 0000 UTC 31 October 1998,
b) 1200 UTC 31 October 1998,
c) 0000 UTC 1 November 1998
d) 1200 UTC 1 November 1998. Broad arrows indicate the location and magnitude of the low-level jet. Small arrows indicate 850-hPa flow. Dashed lines represent isotherms (C).
Figure 3. Upper-level analysis at:
a) 0000 UTC 31 October 1998
b) 1200 UTC 31 October 1998. Large arrows indicate upper-level flow. Dashed area indicates region of upper-level diffluence. Contours represent regions of 40 m/s speed maxima.
By 1200 UTC 31 October, the 850-hPa low had moved east across southeastern New Mexico and the 850-hPa warm front lifted north across northwest Oklahoma and extreme southern Kansas. Radiosonde data from Dodge City, Kansas (DDC) at 1200 UTC (Figure 4) indicated nearly saturated conditions thru at least 500-hPa across much of the central and southern Plains. As the 850-hPa low approached the Texas Panhandle, the thermal gradient and winds at that level increased markedly over Kansas and Oklahoma. Increased isentropic upglide (Carlson 1998), combined with increased upper-level diffluence ahead of a southern New Mexico jet streak resulted in rainfall that increased in both coverage and intensity. Between 0000 UTC and 1200 UTC, 25-to-50mm (one to two inch) rainfalls had soaked south central Kansas.
Between 1200 UTC 31 October and 0000 UTC 1 November, the 850-hPa low deepened as it moved east-southeast over west Texas. During this period, cold advection over the Texas Panhandle and warm, moist advection over Arkansas and eastern Oklahoma induced cyclonic pivoting of the 850-hPa warm front. The front assumed a nearly north-south orientation by 1200 UTC 1 November. As a result of this gradual pivoting, the portion of the front over south-central Kansas remained nearly stationary. Despite lacking radiosonde data across south-central Kansas, data from a privately owned wind profiler at Whitewater, approximately 50 km northeast of Wichita, confirmed the presence of the 850-hPa boundary over the area. Winds 1.5 km above ground level (AGL) near Whitewater became more southerly by 1800 UTC 31 October (Figure 5).
Figure 5. Wind profiler data for October 31, 1998, from Whitewater, Kansas, 50 km northeast of Wichita. Wind speeds in knots.
Strong isentropic lift induced widespread rainfalls of at least 50mm (two inches) over the southern half of Kansas and northern Oklahoma between 1200 UTC 31 October and 0000 UTC 1 November. Furthermore, the thermal gradient strengthened along the 850-hPa front across southern Kansas from 3°C per 100 km at 0000 UTC 31 October to 7°C per 100 km at 0000 UTC 01 November. The resulting ageostrophic frontal circulation (Bluestein 1993) focused a band of 75-to-150mm (three to six inch) rainfalls along the 850-hPa front over Wichita and along the Kansas Turnpike (which runs from the Oklahoma border north thru Wichita then northeast thru El Dorado) for the 12-hour period ending 0000 UTC 1 November. Flash flood warnings were issued for Harper, Sumner and Sedgwick Counties that included the Wichita metropolitan area. Many roads in Wichita and Hutchinson were closed due to flooding produced by the heavy rainfall. Rain continued to fall during 1 November, but the intensity decreased as the upper low slowly exited to the east. However, the worst was yet to come for southern Kansas.
The QPF issued by the Wichita NWS at 0900 UTC 31 October 1998, prompted a coordination call from ABRFC Tulsa to discuss the forecast. The focus of the discussion was that such high 24-hour basin average QPF of 38 to 63mm (1.5 to 2.5 inches) across much of south-central Kansas would induce record flooding on three rivers in the Wichita Hydrologic Service Area (HSA). Upon reassessing the situation which included an examination of the most recent rainfall reports, a return call was made to the ABRFC shortly thereafter stating that no changes would be made to the QPF that would go into effect at 1200 UTC. ABRFC then issued river forecasts indicating record flooding on the Whitewater, Arkansas, and Walnut Rivers with the flood stages to be reached 24 to 36 hours later.
Three factors weighed heavily in the Wichita NWS QPF strategy:
Therefore, the Eta forecast model was the model of choice for the WFO Wichita staff that morning.
As a result of the WFO Wichita QPF, the ABRFC issued record flood forecasts that provided law enforcement and emergency preparedness officials ample lead time to commence evacuation procedures. The tremendous rainfall that resulted had a profound impact on area rivers and creeks in Wichita and surrounding communities (Table 1). The most impressive rises occurred on the Walnut, Arkansas, and Whitewater Rivers (Figures 6-8). For the Walnut River at Arkansas City, a crest of 32.45 feet (9.9 meters) was reached at 0713 UTC 3 November. This crest was a staggering 14.45 feet (4.4 meters) above the flood stage of 18.0 feet (5.5 meters) and shattered the record crest of 29.22 feet (8.9 meters) set on June 11, 1995. On the Arkansas River, Arkansas City reached 28.89 feet (8.8 meters) at 1000 UTC 3 November, nearly 12 feet (3.6 meters) above the flood stage of 17.0 feet (5.2 meters). Record flooding on the Whitewater River also occurred at Augusta where the river rose to a crest of 34.95 feet (10.66 meters) around 2200 UTC 1 November. This record crest was nearly 14 feet (4.27 meters) above the flood stage of 21.0 (6.4 meters), surpassing the previous mark of 32.70 feet (10 meters) set on June 8, 1979. The most serious flooding occurred in and around Augusta where one river (the Whitewater) was in record flood with a second (the Walnut) in major flood. Approximately 1,800 people were evacuated and 565 homes, 230 mobile homes and 100 businesses sustained damage. In Towanda, the Whitewater River crested at 30.53 feet (9.3 meters) which was 8.53 feet (2.6 meters) above flood stage.
All told, an estimated $10 million in damage occurred in Butler County. Southward neighboring Cowley County suffered an estimated $8 million damage that included 479 structures, both business and residential. The majority of this damage occurred in and around Arkansas City where 3,000 people were evacuated. On the west side of Wichita, where an estimated 100 people were evacuated, the flooding of the Cowskin Creek caused $4 million damage to homes, businesses, roads and crops. In Chase County, $2 million occurred and another 200 people were evacuated.
|Flood Stage||Crest||Date||Old Record||Date|
|Towanda||22.00 ft.||30.53 ft.||11/01/98||29.73 ft.||06/08/79|
|21.00 ft.||34.95 ft.||11/01/98||32.70 ft.||06/08/79|
|Derby||12.00 ft.||16.60 ft.||11/01/98||16.19 ft.||07/15/93|
|Arkansas City||17.00 ft.||28.89 ft.||11/03/98||27.62 ft.||05/11/93|
|18.00 ft.||32.45 ft.||11/03/98||29.22 ft.||06/11/95|
|32.00 ft.||36.77 ft.||11/01/98||35.70 ft.||06/05/65|
Figure 6. Time series forecast for the Arkansas River at Arkansas City, Kansas. QPF versus no-QPF, issued at 0300 CST 31 October 1998, with stage in feet and time in Central Standard Time (CST).
Figure 7. Time series forecast for Walnut River (Arkansas City, Kansas), QPF versus no-QPF issued at 0300 CST 31 October 1998 with stage in feet and time in Central Standard Time (CST).
Figure 8. Time series forecast for Whitewater River (Towanda, Kansas), QPF versus no-QPF, issued at 0300 CST 31 October 1998, with stage in feet and time in Central Standard Time (CST).
The Halloween Flood dramatically illustrated the importance of the Quantitative Precipitation Forecast (QPF). Hydrologic models run at ABRFC, using the 38 to 63mm (1.5 to 2.5 inch) basin average QPF issued at 0900 UTC 31 October, correctly indicated that record flooding would occur on the Arkansas, Walnut and Whitewater Rivers. In contrast, when the QPF was not incorporated into the models, the resultant river level forecasts based solely on the overnight 25 to 30mm (one-two inch) rainfall were much lower. In some cases, the forecasts indicated the river in question would not even reach flood stage.
On the Arkansas River at Arkansas City, a river forecast without QPF brought the river to the flood stage of 17.0 feet (5.2 meters) near midday on November 1 with a crest of 18.0 feet (5.5 meters) forecast early that evening. The no-QPF forecast kept the river minimally above flood stage for about 24 hours. When QPF was included, flood stage was expected around 1000 UTC 1 November, about eight hours sooner than the no-QPF generated forecast. A broad crest of 24.0 feet (7.3 meters) was forecast for around 1800 UTC 2 November, with the river remaining in flood for just over three days.
On the Walnut River at Arkansas City, the difference between the no-QPF and QPF- generated river forecasts was glaring. The no-QPF generated river forecast indicated the river would not come close to approaching the flood stage of 18.0 feet (5.5 meters) with a crest of onl 11.6 feet (3.54 meters) forecast during the early morning of November 3.
Meanwhile, the QPF-generated river forecast indicated the river would rise rapidly during November 1, with flood stage reached later that evening. Moreover, it predicted a crest of just over 24.0 feet (7.3 feet) around 1800 UTC 2 November, six feet (1.83 meters) above the flood stage and 13 feet (3.96 meters) higher than the no-QPF generated crest. Additionally, the QPF-based river forecast kept the Walnut above flood stage for nearly three days.
The Whitewater River at Towanda, provided additional landmark evidence as to the importance of QPF in the flood warning program. According to the no-QPF generated forecast, the river was expected to crest around 18.5 feet (5.64 meters) around 1800 UTC 1 November or 3.5 feet (1.07 meters) below the flood stage of 22.0 feet (6.7 meters).
However, when QPF was figured into the river model, the Whitewater was forecast to reach flood stage around 1000 UTC 1 November, with a crest of 25.5 feet (7.8 meters) predicted around 0600 UTC 2 November. The QPF-generated forecast crest was 3.5 feet (1.07 meters) above flood stage and seven feet (2.14 meters) above the no-QPF generated crest. The QPF-generated forecast kept the Whitewater above the flood stage for nearly two days.
During the Halloween weekend of 1998, a slow-moving upper low crossed the southern Plains. The combination of factors such as strong isentropic lift and upper-level diffluence, led to a widespread heavy rainfall event. Frontogenetic forcing along a nearly stationary 850-hPa boundary led to a band of excessive rainfall along the boundary. The rainfall induced record flooding on several rivers in southern Kansas. Although the QPF issued at 0900 UTC 31 October under-forecast the resultant rainfalls that occurred and subsequently led to under-forecast crests of the rivers in Wichita's Hydrologic Service Area, the QPF proved invaluable nonetheless. The QPF-generated river forecasts correctly warned of an impending major flood, while the no-QPF generated forecasts indicated minor flooding at best. Without the QPF-generated forecast, river flood warnings would have been delayed until after the rainfall occurred, substantially decreasing critical lead times needed to save lives and property.
This case clearly illustrates the importance of QPFs in the river flood warning process. River flood warning issuances with the lead time needed to save life and property depend on accurate QPFs and the incorporation of the QPFs into the models used to forecast river levels. The historic flood of Halloween weekend 1998, will no doubt haunt the residents of south-central Kansas for years to come. However, it will also be remembered as an event that clearly showed the value of the QPF in the Hydrologic program of the National Weather Service.
The authors wish to acknowledge the valued assistance of Larry Lowe of the ABRFC in the procurement of the time series for the Arkansas, Walnut and Whitewater Rivers. The authors also greatly appreciated the assistance of Marian Baker, Service Hydrologist, WFO Wichita, in securing specific crest information.
Bluestein, H.B., 1993: Synoptic-Dynamic Meteorology in Mid-latitudes. Volume II, Observations and Theory of Weather Systems. Oxford University Press, 594 pp.
Carlson, T.N., 1998: Mid-latitude Weather Systems. American Meteorological Society, 507 pp.