Dr. Holly Widen is a Product Manager in the Model Product Management team, focusing on the U.S. Flood suite of products. She joined RMS in 2016 upon completion of her Doctorate in Geography from Florida State University, where she studied tornado risk and vulnerability using applied spatial statistics. She has co-authored over ten peer-reviewed journal articles and is a member of the American Meteorological Society and the American Association of Geographers.
It was off to London’s Savoy Hotel for members of the RMS London team last Thursday, for the Eleventh Trading Risk Awards. And apart from the great hospitality, and the flowing conversation from colleagues and industry peers alike, RMS was also recognized by the award judges, receiving the “Initiative of the Year” award for the RMS U.S. Inland Flood HD model.
Without sounding like an Oscar acceptance speech, on behalf of the team that worked on the model, I would like to thank the judging panel made up of representatives from the media and the industry for selecting our entry. Released last October, the flood model is designed to help the private insurance market seize the opportunities presented by this peril, and to also ultimately help accelerate flood insurance take up in the U.S.
From major wildfires just over four months ago, and now major flooding, Northern California seems to leap from one perilous state to another. This time, rainfall from a “potent atmospheric river”, as described by the National Weather Service, caused flooding to over 3,000 properties in Sonoma County. This atmospheric river – a flowing column of condensed water vapor pumped up from the Tropics which can be up to 375 miles (603 kilometers) wide – started delivering rain and snow into the region late on Sunday, February 24.
The small town of Guerneville (pop. ~4,500) fared worst, reporting nearly 21 inches (529 millimeters) of rainfall in just 72 hours by 5 p.m. local time on Wednesday, February 27. The source of the town’s flooding was the Russian River, which flows from Mendocino County through to Sonoma County, reaching a maximum level of 45.5 feet (13.9 meters) at Johnson’s Beach, near Guerneville. This exceeded the defined 40 feet (12.1 meters) threshold for a major flood at this point, with local media reports stating that this is the worst flooding since New Year’s Day in 1997, when the river rose to 45 feet (13.7 meters). The nearby Napa River also crested at 26 feet (7.92 meters), one foot above the flood stage.
The town of Guerneville, which was originally built on a meander in the river, on February 27 was declared by the Sonoma County Sheriff’s Office “… [as] officially an island …” as all roads in an out of the town were flooded. 4,000 residents in both Guerneville and Monte Rio (pop. ~1,200) were under evacuation orders until Friday, March 1.
RMS modeling reveals the wider risk of U.S. flooding, and a significant protection gap
A combination of heavy rainfall and melting snow had filled Lake Oroville in northern California near to its capacity. Dam operators released water through the main spillway to control the reservoir level, but a 300-foot hole unexpectedly emerged, and the surrounding soil was eroded by the water gushing out. Spillway outflows were reduced to stop the erosion.
Oroville Dam after heavy rainfall in mid-February (Source: California Dept of Water Resource)
But this made the problem with rising reservoir levels worse, as the water then began to flow over the emergency spillway. On February 12, 2017, at least 188,000 residents were told to evacuate, while trucks and helicopters dumped over 1000 tons of material per hour on the weakened structure to prevent a more significant breach. As water levels in the reservoir subsided, the risk reduced.
With no massive discharge or flooding, insurance losses are expected to be limited to coverage for business interruption, loss of use, or additional living expenses (ALE) incurred by the evacuees.
It could have been far worse if the dam had completely failed – this was a worrying near-miss.
Dam failure, though rare, is not a negligible risk. In fact, a similar near-miss situation occurred during the 1971 San Fernando earthquake where the lower Van Norman Dam was a near-breach and forced the evacuation of 80,000 people. Our modeling teams decided to model a counter-factual version of February’s Oroville Dam breach, in which either (i) the dam continues to disintegrate in a controlled manner or, (ii) in the worst-case scenario, it collapses.
Our flood modelers, Ye Tian and Sonja Jankowfsky, simulated these ‘what if’ scenarios over a 72-hour cycle, using reservoir water level data from the California Data Exchange Center’s Department of Water Resources, and their own estimation of the lake’s bathymetry (contours of the lake bed) and capacity.
The worst-case scenario is modeled assuming the height of the dam wall becomes ‘zero’ instantaneously – for example, as might happen if there was an explosion due to sabotage. For the controlled breach scenario, the dam wall is gradually lowered at two meters per hour to simulate what could have happened had the erosion of the dam continued.
In our modeling, the amount of discharged water would be expected to overwhelm local flood control measures over 100 miles downstream, as was evidenced by the widespread flooding in the simulations. We estimate that under either scenario, about $21.8 billion of building value would be at risk, which, for the communities near Oroville, would be a huge problem because much of it is uninsured.
Insurance coverage is one of the most effective ways of ensuring a region re-bounds quickly after disaster, but in the Oroville region flood insurance penetration is fairly low. Most property owners rely on the National Flood Insurance Program (NFIP), which is most commonly sold to homes in the 100-year flood plain along the Feather River.
FEMA 100-year flood zones
Potential inundation depths simulated for a complete and instantaneous collapse of the Oroville Dam
This flood plain is shown as blue areas in this map. Compare that to the map below.
This map (right) shows modeling from the RMS ‘what if’ scenario of the Oroville Dam breach. The 100-year flood plain (map above) covers significantly less area than that which would be inundated if the dam breached (map right).
So, why does this matter? Residents living outside the 100-year flood zone are not required to purchase flood insurance, and therefore most do not. These areas include the towns of Biggs, Gridley, Live Oak, Oroville, South Oroville, Thermalito, and Yuba City. And yet as RMS modeling shows, many of those communities would have experienced major flooding if the dam had breached completely.
The Protection Gap
It is obvious that NFIP flood zone maps do not include dam failure scenarios, and yet these failures typically inundate a much wider area beyond the naturally occurring flood plain, because the volume of water and speed of release overwhelms natural and man-made defenses.
FEMA cannot quickly nor easily change the flood maps to incorporate this type of risk explicitly. At RMS, we are developing tools to quantify these kinds of extended risks to allow the private flood insurance market to step in and fill the current gaps in coverage.
Thankfully, our ‘what if’ scenario didn’t become a reality. But it highlights the risk of aging infrastructure which may not be able to withstand extreme weather events. Nationwide, the National Inventory of Dams indicates that of the 90,580 situated across the U.S., over 30% exhibit significant to high hazard potential due to structural deterioration.
[Note: clients can obtain modeling files for the Oroville Dam analysis from RMS account managers. This blog has been edited to provide further detail on the initial dam failure.]