SERVICE DELIVERY

Flood Preparedness Coordination Effective

The NCRFC and NWSO FGF recognized the potential for dangerous floods in the Red River of the North in 1997 well in advance and made a concerted effort to advise the public of the impending hazard. The NCRFC treated the flood as a non-typical event by issuing early narrative and numerical outlooks. Coordination was vigorously pursued by NWSO FGF and the NCRFC with local agencies throughout the Red River of the North Basin, and meetings were better attended than usual. For example, a meeting on February 27, 1997, was an added "pre-flood" coordination meeting with about 50 participants set up by the Service Hydrologist (SH) at NWSO FGF. An early numerical outlook issued by the NWS on February 28, 1997, was used as justification by USACE to obtain $8 million in early flood-fighting monies. Early use of news conferences by the Eastern North Dakota Forecast Office in Grand Forks to highlight flood outlooks raised public and intergovernmental attention to the likelihood of Red River of the North flooding. In a National press conference on March 18, the Director of the NWS was quoted as predicting record-breaking floods on the Red River of the North in very strong language, "...highest floods in 150 years... You're going to see hundreds of square miles under water..." In all, the advance information was very effective in mobilizing flood-fighting efforts throughout the basin.

All NWS offices should be encouraged to vigorously pursue notification of serious flood threats at the earliest possible time. Every NWS river forecast center should consider earlier release of flooding outlooks when conditions suggest a significant potential for spring flooding and when available data on snowpack conditions give high confidence that a serious flood threat exists. Decision-making on product release should also include consideration of how NWS outlook products are used and any intergovernmental actions which are triggered primarily by the outlook products' content.

Forecast Service Overview

During the spring flood of 1997, 163 forecasts for above-record stage were issued by the NCRFC for the Red River of the North and its tributaries. The overall quality of NCRFC outlooks and forecasts was excellent although there were difficulties with the forecast services at three locations: East Grand Forks, Pembina, and Drayton.

Stage forecasts at East Grand Forks were raised from the 50-foot forecast value issued on April 14 in increments of 0.5 to 1.0 foot until the final crest forecast of 54 feet was issued on the evening of April 18 (see details in Appendix A). This "stair-stepping" contributed to decisions by those engaged in the flood fight to gear up to add "just one more foot" to the emergency levees until it was not possible to reallocate resources effectively to save critical facilities or higher elevation neighborhoods. The NWS raised these forecasts NOT because of revisions in the discharge forecast but because of the growing realization that the rating curve was inaccurate as the previous record stage was approached, then exceeded (see Finding 1).

There has also been some criticism of NWS forecasts at Drayton and Pembina (downstream of East Grand Forks). Although these forecasts were raised in part as a reaction to the difficulties in the forecast situation at East Grand Forks, the increases primarily reflected a concern for the possibility of wind-induced waves. Wind had little effect on the crest stages at Drayton or Pembina (although such effects did occur downstream in Canada) so that ultimately these forecasts turned out to be too high.

NWS does not intentionally "stair-step" its flood forecasts, so a recommendation to simply avoid this would be useless. In a similar vein, the effects of wind on river stage are difficult to predict, so the addition of an allowance for wind-induced waves in a subjective fashion can be a reasonable forecast practice.

Staffing Considerations for Hydrologic Operations

The hydrology program at NWSO FGF has been developed effectively as a fully integrated station function even though the office is a recent spinup HSA office. NWSO FGF stringently applied the Central Region policy that all program manager positions (Warning Coordination Meteorologist (WCM), Service Hydrologist (SH), Meteorologist in Charge (MIC), Science and Operations Officer (SOO), and Data Acquisition Program Manager (DAPM)) are to be used primarily to meet their program management responsibilities. The MIC has a clear expectation that on-duty personnel should be able to handle duties every shift or utilize the standard call-back list if overwhelmed. As a result, all appropriate staff are trained and capable of gathering hydrologic data and issuing flood products. In this flood event, the MIC continued to abide by this policy for the most part, although the staff worked about 200 hours of overtime and/or compensatory time at FGF between March 31, 1997, and May 10, 1997. At the NCRFC, the duty time for the primary Red River of the North RFC forecasters was extended to include all the weekend periods through the flood.

NWS managers should give careful consideration to extending the duty hours of Service Hydrologists (SH) during major flooding episodes. This needs to balance multiple objectives including the need to have hydrologic services be a 24-hour station function at every WFO (i.e., not just a job for the SH), the desire to take advantage of the most highly skilled staff available during critical forecasting situations, the realistic limits on the number of hours an individual SH can be effective while on duty, the severity and duration of the flood event itself, and other operational demands. The survey team believes that MICs should consider extending the duty time of SHs when the supporting RFC goes into 24-hour operations as a result of flooding within their service area but considers the final judgement to be in the hands of the MIC, as is the case under current policies.

Diverse Interpretations of NWS Products

Overall, the NWS provided forecasts that were used successfully by communities in the Red River of the North Basin. It is clear, however, that forecast users, including the public, had very different interpretations of NWS products, especially as regards the uncertainty in those products.

In the interviews conducted by the Red River of the North survey team, different people interpreted the flood outlooks and forecasts of flood crests in different ways. Many users do not differentiate between outlooks and forecasts and view them as the same products, conveying similar levels of uncertainty.

NWS outlooks include two outlook peaks -- one (lower) outlook peak assuming no future precipitation and the other (higher) assuming normal precipitation. Both the language qualifying the higher outlook value and the historical experiences with past NWS outlooks indicate that the higher value is approximately a median; i.e., it has about 50 percent chance of being equaled or exceeded. (At East Grand Forks, the higher outlook value was equaled or exceeded in 6 of the 12 outlooks produced in 1980, 1982, 1984-1987, 1989, and 1993-1997.) In spite of the qualifying language and the historical performance of the outlooks themselves, many users interpreted the two flood crest levels issued in the outlook as a range; i.e., the two were viewed as minimum and maximum levels.

Moving from outlooks to forecasts, some viewed the flood crest forecast issued by the NWS as a maximum; i.e., a value that would not be exceeded. Others viewed the number as certain; still others viewed it as somewhat uncertain, with interviews revealing that the uncertainty in the forecasts is viewed to range from 2-10 percent of the predicted flood crest. The manner in which decision-makers interpreted the flood crest forecast affected the particular actions that they took in preparing for the flood. Consequently, it became obvious to the team that both the SENDING and RECEIVING of forecast stage information plays a critical role in preparing for and responding to floods, particularly floods at near-record or record levels where the stakes are high. It is also apparent to the team that the use and value of the EXISTING flood forecast products are not well understood, much less the potential improved usefulness and value that might be attained through adjusting the content of the products and manner in which they are delivered.

No common theme emerged in response to questions by the survey team regarding potential changes to NWS flood forecast products. Some users would like to be given a product that includes a range of possible crests; others want a single "best" estimate of the flood crest. Some would like a range but said they would only use the highest value. Some would like a range, but only if it were narrow; e.g., two-feet difference between the high and low value. Some would like a probability distribution; i.e., multiple values at various chances of exceedance.

Given the diversity of opinions within the survey team itself and the diversity of responses to questions regarding the format and content of NWS products, a fairly conservative approach to immediate changes in NWS services is indicated. The only specific change that can be recommended at this time is that NCRFC products should include both stage and discharge (not just stage) as an aid to improved coordination between the RFC and those users able to interpret discharge information (see Recommendations 7A and 7B).

Confidence and Uncertainty in NWS Products

NWS products and discussions by NWS staff generally included qualifications intended to convey the uncertainty inherent in NWS outlooks and forecasts. Nevertheless, many users developed a false sense of precision in NWS products. A number of factors helped contribute to this:

• Repetition of the outlook value (e.g., 49 feet at Grand Forks) contributed to an impression of certainty in that value. Due to the possibility of record flooding, the timing of the outlook issuance was changed by the NWS to include an earlier (February) numerical outlook. The earlier outlook had the positive impact of early mobilization of flood-fighting efforts, but it also extended the length of time that the same outlook value was cited. The March outlook value was identical (in value) to the February value, again extending the period when a single value was repeated. Finally, the blizzard event on April 5-6 delayed the start of the forecast production at East Grand Forks (the first forecast was issued on April 14 for a crest of 50 feet at East Grand Forks). During the period after the blizzard and before the April 14 forecast, the NWS continued to refer to the outlook value which had been provided in March. Again, repetition of the same outlook value contributed to a "locking in" by users on the outlook value of 49 feet at East Grand Forks.
  
  
• The numerical outlooks for Wahpeton, Fargo, and East Grand Forks were respectively 0.5, 0.16, and 0.2 foot above the historical record crest. In interviews with users, at least in the Grand Forks area, rather than elevating concern, the 49-foot outlook actually created a sense of complacency since it was only 0.2 foot over the record flood of 1979, which the city had survived. Many felt they were ready for the flood as forecast by the NWS. The high river stage in 1996 and subsequent successful flood fight (no damages) also contributed to this sense of confidence in 1997.
  
  
• The period after the blizzard of April 5 and 6 and before the issuance of the first operational forecasts on April 14 produced considerable confusion among NWS customers, including the general public. By repeatedly citing the flood potential outlook values produced before the blizzard, the NWS conveyed to many a high confidence in these numbers. Users expected that the forecast crest would increase immediately after the blizzard and generally did not understand that the period from April 6-14 was required to assess the impact of this blizzard on NWS forecasts.
  

NWS needs to study the methods used to convey the uncertain nature of its flood forecasts and outlooks. An analysis of the range of observed stages at particular discharge values based on historical discharge observations can help to establish a sense of the "inherent" uncertainty in stage forecasts at specific locations. Possibly, the addition of narrative information describing forecast difficulties; such as, the impact of the early April blizzard on data collection, may help users to assess forecast accuracy. The NWS needs to be as diligent as possible in guarding against over-confidence in the face of flood threats (or any other natural hazard) and be aware of the unintended message that repetition conveys increased confidence. A proactive approach to getting information out regarding the limits of forecasting accuracy needs to involve many forums: brochures, fact sheets, NWS products, flood or public information statements, contacts with the media, etc.

All of these concerns regarding conveying uncertainty in NWS forecasts are especially important for record and near-record events since there is little or no information available in the historical record regarding the actual behavior of the river / stream for the forecast.

The NWS has conducted an Advanced Hydrologic Prediction System (AHPS) demonstration in the Des Moines area that can provide helpful information on forecast uncertainty. There are two main features of the AHPS demonstration that are important: The first is that AHPS technology offers the best means for the NWS to explicitly and objectively convey the uncertainty in NWS hydrologic forecasts. The second is that the interactive approach taken in the AHPS demonstration of surveying NWS customers for desired product changes, then, evaluating user acceptance of the new information is essential.