A 1.3 billion-gallon dam failure at 5:12 AM in a rural county whose PSAP learned about it from 911 callers, not the EAP.
At 5:12 on a December morning in 2005, the upper reservoir of AmerenUE's Taum Sauk pumped-storage hydroelectric plant on top of Proffit Mountain failed catastrophically. 1.3 billion gallons of water — enough to fill roughly 2,000 Olympic swimming pools — cascaded down the mountain in 12 minutes.
The wall of water hit Johnson's Shut-Ins State Park below with a 20-foot crest. It obliterated 281 acres of the park. It swept vehicles off Highway N into adjacent fields. It destroyed the home of state park superintendent Jerry Toops, washing him, his wife Lisa, and their three young children — including a 7-month-old — across the road and into a debris field in predawn darkness and December cold.
No one was killed. That outcome was not by design. It was December, and the park campground was empty. If it had been July, the death toll would likely have been catastrophic.
What the comm center saw, and when. Color coding indicates the operational dimension.
The Reynolds County Sheriff's Office and PSAP serve a county of roughly 6,000 people across 818 square miles of Ozark hill country. There is no municipal 911 center — emergency calls go to the sheriff's dispatch, which handles everything from traffic to search-and-rescue to dam failures it wasn't told were possible.
That morning, dispatchers woke up to a catastrophic flooding event the same way the public did: calls started coming in, and nobody in the warning chain had called them first. AmerenUE's Emergency Action Plan for the Taum Sauk facility — required by FERC — had notification procedures. The gap this exercise explores is what happened between the failure of that private-sector EAP and the moment Reynolds County dispatchers picked up their first 911 call.
Dispatchers initially responded to a flooding event and a missing family, not a dam failure — until information developed on scene. That time gap between "unknown-cause flooding" and "dam breach" shaped every decision in the first hour: what to brief responders on, what to tell downstream agencies, and whether to issue evacuations without knowing whether the lower reservoir would hold.
The EAP question. AmerenUE was required by FERC to maintain an Emergency Action Plan for Taum Sauk. That plan had notification procedures. The Reynolds County PSAP was not in the practical notification chain in a way that reached a dispatcher who could act on it before 911 calls started.
Acting before you know the cause. Call #1 described a destroyed house and missing family. Without knowing a dam had failed, dispatchers were treating a flood. With that context, everything changes — responder hazard briefing, downstream notification, scope of evacuation. Dispatching resources immediately while asking questions that characterize the hazard is the trainable disposition.
Downstream threat under uncertainty. The lower reservoir absorbed the bulk of the water and held. Dispatchers didn't know that. When a dam fails and downstream containment status is unknown, the conservative posture is to treat every downstream community as threatened until information says otherwise — voluntary evacuation advisories, downstream PSAP notification, and getting someone on the phone with whoever can confirm containment.
Responder safety when the hazard type is unknown. A flash flood from rain runoff is a different responder risk than a dam-breach wave. The peak flow had passed before responders arrived — but dispatchers at the time of dispatch didn't know that. The training point is the questions that accelerate understanding of hazard type: "Is the water still rising?" "What was the water level before it hit?" "Did anyone see where it came from?"
Normalization of deviance. AmerenUE had known the sensor housings were unreliable since October 2005. Operators created a workaround. The system kept running. Nothing bad happened — until one morning at 5 AM, the workaround failed at the worst possible moment. In comm centers, the parallel is procedures that haven't been reviewed, CAD notes that get ignored, systems running on deferred maintenance, training gaps that persist because "nothing's gone wrong yet."
No right answers. Tap a question to expand the analysis. Use one or all — whatever fits your time.
This is a cascading information problem. Call #1 describes a structural collapse and missing persons — it warrants immediate dispatch of whatever resources Reynolds County has available, including notifying state patrol. The challenge is that without knowing a dam has failed, you're treating a flood event.
Hold the call type loosely. Dispatch the resources the first call demands. Keep asking characterizing questions — "what else just happened?" "how big was the water?" "did you see where it came from?" When a second caller adds "the whole mountain came down," everything changes: responder hazard briefing, downstream notification, scope of evacuation.
The trainable disposition is parallel action. "We don't wait to respond — we respond and gather information simultaneously." Waiting for information before dispatching costs lives. Continuing to gather after dispatching is what protects them.
This is a structural gap that exists in many jurisdictions. FERC, EPA, and state agencies require EAPs for dams, chemical facilities, and industrial infrastructure — but those plans are written between the facility and the regulatory body. Your PSAP may or may not be in the notification chain, and even if it is, the practical question is: does your dispatcher on the 3 AM shift know who calls them and what that call means?
Discussion starter for supervisors: pull your county's dam inventory and ask who holds the EAPs. Are any of those EAPs current? Does any notification procedure route to your center in a way a dispatcher would actually recognize? This is a pre-incident planning gap, not an ops failure — but it becomes an ops failure the morning the dam breaks.
In many jurisdictions, NWS warnings flow to Emergency Management, not to the 911 center directly. Your PSAP may be downstream of a chain that runs: event → NWS → county EM → public alert → 911 calls start coming in. That's backwards from a dispatch perspective. By the time you're receiving 911 calls about a flash flood, the event is already happening to someone.
In the Taum Sauk case, NWS was the functional public notification mechanism — but they also weren't notified until after the failure was underway. Explore: does your county EM have a direct line to your PSAP for NWS-level events? Do you receive Wireless Emergency Alerts simultaneously with the public, or after? Who in your county is the first call for the NWS warning coordination officer at 3 AM?
Without knowing the source of the flooding, dispatchers can't effectively brief responding units on the hazard they're entering. A flash flood from rain runoff is a different responder risk profile than a dam-breach wave that's already emptied. The peak flow had passed by the time responders arrived — but dispatchers at the time of dispatch didn't know that.
You can only brief on what you know. The training point is what questions to ask that would accelerate your understanding of the hazard type. "Is the water still rising?" "What was the water level before it hit?" "Did anyone see where it came from?" These aren't just good customer service — they're responder safety questions.
When a dam fails and you don't know downstream containment status, the conservative approach is to treat every downstream community as threatened until you have information otherwise. That means voluntary evacuation advisories, downstream PSAP notification, and getting someone on the phone with whoever can tell you what the lower reservoir's status is.
The challenge: Reynolds County dispatchers had the most immediate victim situation (the Toops family rescue) plus a potential catastrophic downstream threat — simultaneously. How do you triage? This is a supervisor decision as much as a dispatcher decision. When do you call in off-duty staff, when do you call for mutual aid dispatch support, and who makes that call at 5:30 AM?
This is normalization of deviance. AmerenUE knew the sensors were unreliable. They created a workaround. The system kept running. Nothing bad happened. So the workaround became the norm. Then one morning at 5 AM, the workaround failed at the worst possible moment.
In comm centers, this shows up as: procedures that haven't been reviewed because "we've always done it this way," CAD notes that get ignored because "that caller always calls," systems running on deferred maintenance because "it works fine," and training gaps that persist because "nothing's gone wrong yet." The question isn't whether your center has a Taum Sauk — it's which one it is.
Five questions. Answer, then submit for inline feedback. Progress saves locally.