Time is Now to Abandon ‘Bathtub Modeling,’ Researchers Argue

Media reports often assert that increasingly intense flooding will “swallow millions of U.S. acres within decades” and “erase cities by 2050,” among other alarming predictions. However, many such articles cover modeling studies that make an outdated assumption, argue authors of a recent research letter published in the journal Earth’s Future: that floodwaters spread out over coastal and inland areas at a level, uniform depth regardless of topographical features or flood-control infrastructure.

This approach, known as bathtub modeling, is useful for its ability to estimate a basic picture of potential flood vulnerabilities across large areas quickly and cheaply. However, while researchers typically note the practice’s inherent uncertainties and biases in their studies, these nuances often are lost in translation by reporters, leading to sensationalized headlines that inspire panic and undermine public trust in climate science when they prove inaccurate, commentators write.

In their literature review, letter authors representing University of California, Irvine (UC Irvine) and the University of Bristol (England) explore the extent of bathtub modeling’s inaccuracy based on real-world flood events. They use these data to urge the scientific community to abandon bathtub modeling in favor of physics-based modeling approaches, which they say have become affordable and accessible worldwide thanks to advances in computing power.

“Accurate maps of areas at risk of flooding are of paramount importance for everyone from home and business owners to insurers, banks, and governments,” said Paul Bates, letter co-author and University of Bristol hydrologist, in a release. “We all have a role to play in reducing flood losses, but it all starts with trustworthy information.”

Less Accurate Than a Chimpanzee

The researchers reviewed findings from six studies published within the last 20 years, each of which directly compares the accuracy of bathtub modeling against actual observations of historical flooding events. The scope of these studies encompasses multiple continents, various scales, as well as both inland and coastal settings. Each study judges flood-extent accuracy using a measure called the critical success index (CSI) — a scale in which a score of 1 indicates a perfect match and a score of 0 indicates complete inaccuracy.

According to the research letter, the consensus among modeling experts holds that CSI scores below 0.65 indicate that a model is insufficiently accurate for useful and locally relevant decision-making. Each of the studies cited in the letter computed CSI values below 0.5 for bathtub models at all scales.

“CSIs under 0.5 indicate that these models are worse than a random classification,” said Oliver Wing, University of Bristol honorary research fellow and study co-author. “In other words, a chimpanzee has more skill than a bathtub model in delineating flood hazard areas.”

Digging deeper within these studies, the researchers also identified an array of specific ways in which bathtub modeling fails to predict flooding behavior accurately. For example, bathtub modeling likely will flag low-lying areas just behind built flood barriers as flood-prone even when there is no hydraulic connection to these areas. For another, in cases of permanent inundation, bathtub modeling sometimes underestimates true flood coverage because it does not account for a specific area’s hydrologic budget — important for modeling phenomena such as emerging groundwater and inland ponding. They found that bathtub modeling is particularly inaccurate in urban areas that tend to have greater flood-defense coverage and pumping capabilities as well as finer-scale topographic features such as elevated roadways.

All told, letter authors list six key factors of flood prediction that bathtub modeling fails to consider. They include

• flood attenuation from the effects of event dynamics and friction on flood spreading;
• tidal amplification associated with the resonance of ocean tides within coastal embayments;
• flood defenses such as levees and floodwalls that may overtop during an extreme event but still restrain the degree of inland flooding;
• shoaling of the groundwater table;
• surfacing groundwater from the combined influence of rising sea levels and changing hydrologic budgets; and
• pumping of groundwater within lands below sea level to mitigate inundation by rising groundwater.

Grounded Forecasts Promote Trust in Science

In the past, dynamic models capable of quickly solving and applying physics-based equations to provide locally relevant estimates of flooding extent suffered from poor spatial resolution and high computational complexity. However, researchers note that in recent years, needs within the insurance and financial services industries have propelled advances in physics-based modeling that have driven down costs and made these models more flexible for different applications.

Today, dynamic models exist that allow for feasible deployment at scales as small as individual properties and as large as entire continents, commentators write. The researchers assert that these models have reached a point of accessibility that eliminates the need for bathtub modeling — typically regarded as more practical for studies in which finer-scale estimates are unnecessary or available data is minimal — altogether.

In their letter, authors note that several previous studies have established that residents in flood-prone areas are unlikely to trust projections of future flooding that do not align with their lived experiences. In other words, sensational projections can have a direct and lasting effect on residents’ flood-preparation decisions. By forgoing bathtub modeling, academics can downplay risks of media reports misrepresenting their work, which in turn can give the public more confidence in scientific assertions and protect people and property, the researchers write.

“Projections of flooding need to make sense to people, not only for building understanding of what’s at risk but also for deciding upon the investments and policies that will be made to manage it,” said Brett Sanders, UC Irvine Professor of Civil and Environmental Engineering and letter co-author. “Research studies that oversimplify flooding and don’t represent real-world data pose a threat to transformative action.”

Read the full letter, “Flooding Is Not Like Filling a Bath,” in Earth’s Future.

Top image courtesy of Markus Distelrath/Pixabay