Case Study · Riverine Flood · 1997
April 1997. North Dakota. The National Weather Service forecast the Red River cresting at 49 feet. Grand Forks built its dikes to 49 feet. The river rose to 54. When 50,000 people were evacuated — the largest American civilian evacuation since the Civil War — downtown Grand Forks was on fire. The fire hydrants were underwater. The city that had prepared for everything it had been told to prepare for couldn't fight a fire because the streets were four feet deep.
Grand Forks, North Dakota · April 1997
The winter of 1996–97 delivered record snowfall to the Red River basin. An unusually wet fall had preceded it. By the time the spring melt began in April, the basin was as saturated as it had ever been in recorded history. The Red River of the North flows north — unlike most rivers in the continental United States — which means spring melt in the south arrives at the river before melt in the north, and the flood-prone reach is often blocked by ice downstream. The combination was severe. The National Weather Service forecast: the river would crest at 49 feet, matching the 1979 historic record.
The city of Grand Forks took the forecast seriously. Thousands of residents, along with National Guard members and Air Force personnel from the nearby Air Force Base, worked to build sandbag dikes to the forecast level. They built to 49 feet. The Clio historical account of the flood documents what happened next: "The cities had built their dikes to this level, but the river continued to rise past it, to the astonishment of the NWS (which did not upgrade its forecast until April 16, the day the river actually reached 49 feet)." When the river hit the forecast level and kept climbing, the NWS upgraded its forecast — but the dikes were already at capacity. The river would eventually crest at 54.35 feet at Fargo-Moorhead and 54.11 feet at East Grand Forks. The gap between what the forecast said and what the river did was 5 feet. Everything built to protect the city had been calibrated to the wrong number.
April 1997
Date
54.35 ft
River Crest (forecast: 49)
50,000+
Evacuated
11 buildings
Lost to Fire
$1–2B
Damage
On April 18, with water four feet deep in downtown Grand Forks, Mayor Pat Owens ordered the evacuation of more than 50,000 people — the entire city, essentially. The Clio account records her own reflection on the decision: "If I evacuate this city and nothing happens they're gonna impeach me." She gave the order. The MNHS account documents what followed: sirens woke residents in the early morning hours, and 90% of the city's 52,000 residents were eventually displaced to Grand Forks Air Force Base, to motels, and to homes in nearby towns. The following day — April 19 — a fire broke out in the Security Building in downtown Grand Forks. The city's fire department could not get trucks through the flooded streets. The fire hydrants were underwater. The city's water supply had been exhausted. The only firefighting option was aerial: U.S. Forest Service tanker aircraft dumping chemical fire retardant on burning rooftops, and helicopters dropping water. Eleven buildings and sixty apartment units burned before the fire was contained. The image that defined the 1997 flood — fire rising above dark floodwater in a city whose streets could not be reached — was not a weather event. It was a compound disaster: the flood made the fire impossible to fight.
The Science
Think of a river crest forecast as a statistical extrapolation from historical data: previous floods of known magnitude produced known crests, and the current conditions are compared against those historical analogs to project the likely crest. The Red River's 1979 historic crest of 49 feet was the highest on record and became the reference point for the 1997 forecast. The problem the University of Colorado science policy analysis of the 1997 flood identifies: the 1996–97 winter produced conditions — record snowpack, exceptionally saturated soils from a wet fall — that had no direct historical analog in the flood record. The statistical model was extrapolating into conditions it had never seen. The forecast of 49 feet was, within the model's assumptions, a reasonable estimate. The assumptions were wrong because the conditions were historically unprecedented.
The Red River of the North presents flood management challenges unlike most U.S. rivers. It flows north into Canada through a flat former glacial lakebed — the bed of ancient Lake Agassiz. The gradient is extremely gentle: the river drops only a few feet over hundreds of miles. This means floodwaters spread extremely broadly rather than flowing swiftly through a confined channel. Spring snowmelt in the southern part of the basin arrives as the river is still frozen in the north, backing up flows. Ice jams form and fail, producing sudden crest spikes. The gentle terrain means that a 5-foot difference in crest level may correspond to miles of additional lateral inundation — in 1997, nearly every home in East Grand Forks was flooded. A five-foot forecast error on the Red River is not a small miss. It is the difference between protected and underwater for an entire city's footprint.
The Grand Forks fire during the flood is the clearest modern example of a compound disaster: two simultaneous events whose interaction makes each one more dangerous than it would be alone. Fire departments respond to fires with water from hydrants, with truck access to the affected area, and with the ability to deploy personnel close to the fire. All three capabilities were eliminated by the flood: hydrants were underwater, streets were impassable to trucks, and firefighter safety in four feet of floodwater was untenable. The only firefighting modality available — aerial tankers — is less precise, less effective, and limited in the volume it can deliver compared to ground-based suppression. The CBS4 account is direct: "We can't get trucks in there, the hydrants are underwater, we have no pressure." A flood that was being managed by evacuating 50,000 people became simultaneously a fire disaster that no fire department was equipped to fight.
Timeline
01
Fall 1996–April 1997: Record winter snowfall on top of a wet fall. Unprecedented saturation across the Red River basin. Spring melt begins in April. NWS forecasts the river cresting at 49 feet — the 1979 historic record. Grand Forks mobilizes. Thousands of residents, Guard members, and Air Force personnel build sandbag dikes to exactly 49 feet. The city is prepared for 49 feet.
02
April 16: The river reaches 49 feet. The NWS upgrades its forecast. The dikes are at capacity. April 18: Dikes in Grand Forks' low-lying Lincoln Drive neighborhood fail first. Other dikes fail throughout the day and the next. Water is 4 feet deep in downtown Grand Forks. Mayor Pat Owens orders evacuation of 50,000+ people at 4:00 AM with sirens — the largest U.S. civilian evacuation since Atlanta during the Civil War.
03
April 19–20: Fire breaks out in the Security Building, downtown Grand Forks. Streets are flooded. Trucks can't get through. Hydrants are underwater. City water supply is exhausted. Firefighters: "We can't get trucks in there, the hydrants are underwater, we have no pressure." U.S. Forest Service aerial tankers and helicopters fight the fire from above. 11 buildings and 60 apartment units are destroyed. The lasting image: flames above dark floodwater.
04
April 22: President Clinton tours the area. Red River crests at 54.11 feet at East Grand Forks. 90% of residents in East Grand Forks have their homes damaged. Recovery: 850 properties voluntarily bought out; schools resume in rented facilities; Grand Forks Herald keeps publishing from a school in Manvel, ND; new elementary school named "Phoenix." Grand Forks rebuilds with a new flood protection system designed well above the 1997 crest level.
Human Decisions
What worked
The Clio account documents that Mayor Owens ordered the evacuation of 50,000+ people in the early morning hours of April 18, using sirens to wake residents. Her own reflection on the decision — "if I evacuate this city and nothing happens they're gonna impeach me" — documents the genuine difficulty of ordering a mass evacuation before conditions were visually undeniable. She gave the order anyway. The evacuation proceeded effectively, moving most of the city to the Air Force Base and surrounding communities. The evacuation saved lives that would have been lost to the fire and to the flood conditions that followed.
The CBS4 account notes that the Grand Forks Herald — its own office destroyed by both flood and fire — kept publishing, first from a makeshift newsroom in a school in Manvel, North Dakota, and later printing each edition at the St. Paul Pioneer Press. The paper won the Pulitzer Prize for Public Service for its coverage of the flood. It is a specific instance of a community institution refusing to abandon its function in the worst conditions it had ever faced — and continuing to serve the community's need for information during a displacement that lasted months.
What the flood documented
The University of Colorado science policy analysis of the 1997 flood examines the central problem: "the city prepared for such an enormous flood" was "difficult" because the NWS forecast said 49 feet and the city built to 49 feet. The actual crest was 54 feet. The East Grand Forks mayor's comment — "They missed it, and they not only missed it, they blew it big" — captures the public response to a forecast that was off by 5 feet in a situation where the community had invested enormous effort in preparing to exactly the forecast level. The science policy paper raises the deeper question: what is the appropriate level of caution when communities build protection to a forecast rather than a design margin above it?
The compound disaster — fire during flood — was not in any emergency plan for Grand Forks. No emergency plan for a major city flood anticipates simultaneously fighting a major downtown fire with no truck access, no hydrant pressure, and no water supply. The CBS4 account's firefighter quote — "We can't get trucks in there, the hydrants are underwater, we have no pressure" — documents a scenario that was not envisioned in the standard emergency management playbook. Eleven buildings burned because a flood made fighting fire physically impossible. Compound disasters — events that interact to make each worse than it would be alone — are the least-prepared-for category in most community emergency plans.
The cascade lesson
Grand Forks in 1997 is the case study for two distinct lessons. The first: flood forecasts are predictions, not guarantees. Communities that build their protection infrastructure to exactly the forecast level have accepted the entire risk of forecast error. A 5-foot margin above the forecast — the design standard that many modern flood protection systems incorporate — would have changed the outcome in Grand Forks dramatically. The second lesson is harder: even perfect preparation for a known flood scenario doesn't protect a community from the compound disaster that flooding creates. A city underwater cannot fight a fire. A city with no water pressure cannot use its hydrants. The CBS NWS hydrologist's quote from the 20-year retrospective applies: "One thing we've learned over the years is you can always expect the unexpected. Mother Nature has a way of teaching us new lessons every year." Grand Forks learned both lessons simultaneously. The school that replaced two of those lost in the flood was named Phoenix — a name that captures exactly what the community believed it was doing, and what it was.
What You Can Do Now
The 1997 flood's lessons go beyond flood preparedness to the harder problem of compound disasters — events that interact in ways that standard preparation doesn't address.
The Grand Forks flood is the clearest case study in U.S. history for the danger of preparing to exactly the forecast level. Flood forecasts carry uncertainty ranges. A forecast crest of 49 feet in a record snowmelt year means the actual crest could be 49 feet, or could be higher. If you are sandbagging a property, stacking flood barriers, or making decisions about whether to evacuate, don't optimize for the exact forecast number. Build to it and then add a margin. The NWS may revise its forecast upward as conditions develop — and in 1997, that revision came after the city had already built to the previous estimate.
Flood preparation guideThe Grand Forks evacuation of 50,000 people was ordered at 4:00 AM with sirens. The decision to issue the order at that hour — rather than waiting for daylight and visual confirmation of the breach — allowed the evacuation to proceed before conditions made it impossible. If a mandatory evacuation is issued for a flood event, the earlier you leave, the better your conditions: less traffic, more time, and more routes available before roads begin flooding. In a slow-rise river flood, you typically have hours; in a rapidly developing situation like 1997, the window can close faster than expected.
Flood evacuation guideThe Grand Forks fire documented a scenario that most people don't consider: a city that is flooded cannot use its standard emergency response infrastructure. Fire hydrants need water pressure. Trucks need passable roads. 911 communications need functional infrastructure. In a major flood event, multiple emergency response systems may be degraded simultaneously. Your household's independence from the emergency response system — through supply buffers, communication plans, and knowledge of what to do when professional help cannot reach you — matters most in exactly these compound scenarios.
72-hour self-sufficiency guideMayor Owens reflected that she feared being "impeached" if she ordered the evacuation and nothing happened. This is the decision-maker's dilemma: waiting for certainty may mean waiting until it's too late to act effectively. At the household level, the equivalent is having a pre-made evacuation plan so that "should we leave?" becomes "conditions X and Y mean we leave" — a predetermined trigger rather than a real-time judgment call under pressure. Know your trigger for leaving before the event and commit to it in advance.
Evacuation planning guideThe Grand Forks flood produced a buyout program: the city purchased approximately 850 properties. The residents who participated and those who pursued insurance claims both needed documentation: photos of their property before the flood, insurance policy numbers, mortgage information, records of personal property. The Grand Forks City timeline documents the recovery proceeding over years. Having your essential documents — insurance policies, property records, identification — in a waterproof go-bag or in cloud storage means the extended displacement that follows a major flood doesn't compound the physical loss with an administrative disaster.
Document preparedness guideFlood case study series
1927 Mississippi covers federal disaster response origins. Big Thompson 1976 covers flash flood survival. Midwest 1993 covers levee false security. Buffalo Creek 1972 covers industrial dam failures. Together, they cover every major flood failure mode documented in the U.S. record.
Full flood case study seriesSources