Thoughts From Engineers: A Moving Target: The Challenge of Predicting Flood Risk
The Aug. 20, 2018, storms that hit Dane County, Wis., near Madison, unloaded in some locations nearly 15.33 inches of rain in a 24-hour period. Flood stage on Black Earth Creek climbed from 2 feet to 9.7 feet within a few hours, and the two-day rain event firmly shattered all precipitation records for the region.
For decades, the 100-year floodplain has guided the delineation of the country’s special flood hazard zones, restricting development in and around areas of high flood risk, and serving as the backbone for property loss claims under the National Flood Insurance Program (NFIP).
But storm events of recent years frequently eclipse this historically high flood threshold. Houston, for example, experienced three 500-year floods three years in a row. In 2016, four 1,000-year floods hit the Northeast and the South. In the face of extreme weather and storms, the 100-year flood standard falls short in the face of a new reality. We could well ask whether the standard is enough to protect communities and guide infrastructure design in the years ahead. From a climatological perspective, more is at play and more is required of us as engineers as we build and interpret hydrologic models.
Deriving the 100-Year Flood Standard
Back in the 1960s, the United States decided to use the 1 percent Annual Exceedance Probability (AEP) flood as the standard by which to identify zones that posed a flood risk to human life and property. Annual peak stream flows collected from stream gages located across the country and monitored by the U.S. Geological Survey (USGS) form the basis for the AEP for floods of different magnitudes.
The 1-percent AEP flood (or “100-year flood”) is another way to express the average recurrence interval for a particular flood event. The rub is that we’re entering a new era in which historic measures of rainfall and stream flows aren’t necessarily predictive of future storm events or risk.
The Complexities of Floodplain Modeling in the Current Age
Engineers have always worked under data limitations, but in the face of changing climatic factors, nonexistent, incomplete or flawed data make flood-risk analysis much more difficult. The USGS operates 7,500 stream gages across the country, but this still leaves many U.S. waterways without baseline data. Engineers have extrapolated data from other sources or made assumptions about specific locations based on similar conditions elsewhere. Incomplete data compound the uncertainties of modeling future flood scenarios.
The other limitation is that although the 100-year flood gives us some indication of risk, it’s not the outlier event it once was. In the last 10 years, significant changes in terms of amounts of precipitation and frequency have been observed across the United States. The National Oceanic and Atmospheric Administration (NOAA) recently updated rainfall frequency values for Texas and reclassified the 100-year rainfall amount as an average 25-year event.
In many parts of the country, simplified or erroneous assumptions continue to guide decision making. Municipalities may elect to move forward with more-rigorous design standards (employing 500-year flood levels, for example), but these initiatives are still rare. Diminished building areas translates into a smaller tax base for a flood hazard that may or may not materialize. Due to many competing interests, the vast majority of localities are still following a “wait and see” approach.
Other models have been developed to correct for perceived over-reliance on historical data. In June 2020, the nonprofit First Street released Flood Factor, which consists of a numeric value assigned to each property that presumably reflects the property’s level of flood risk. Flood Factor is based on a model that integrates multiple environmental factors such as changing air and ocean temperatures as well as other metrics into the flood analysis. Shortly after the release, the Association of State Floodplain Management stated that the information in Flood Factor could help bring additional awareness to property owners about flood risks but shouldn’t be used to replace FEMA flood maps. The verdict is still out on how well this model performs.
Moving Away from Default Values
Hydrologic models, based on evidence-based inputs and long-established mathematical relationships, are simulations of natural processes. Unlike certain fields of engineering such as structural engineering, which is based largely on constants (e.g., mass of building materials, structural loads, force of gravity, etc.), this is less true of hydrology. A hydrologic model’s results depend on many environmental variables, often unknown and in flux, which is why the work of a hydrologist is particularly challenging. Even so, a model—if thoughtfully built and expertly applied—can provide accurate answers.
Modeling for future storms means knowing the capabilities of your tools and maximizing the model’s ability to predict risk. Engineering companies often are hired by clients who want to maximize the value on an investment (increase the number of buildable lots, for example), which can translate to a short-sighted analysis based on flawed assumptions or yesterday’s baselines.
Engineers may fall back on default values in HEC-HMS and HEC-RAS (the U.S. Army Corps of Engineers hydrologic and flood modeling software) instead of taking time to choose values that more accurately reflect conditions on the ground. With small adjustments to the infiltration rates, curve numbers, lag times and Manning’s values—all critical coefficients in defining a hydrologic flood model—you suddenly have a very different scenario predicting very different flood results. For all the unknowns inherent in a particular set of data, the computations that model the dynamics of stormwater runoff and floodwater conditions are mathematically sound. But details matter; an engineer’s ability to enter reasonable input data is key to a computer simulation’s overall performance.
In the Wake of the Flood
Until we come up with a new way of gauging risk, the construct of the 100-year flood will carry us forward—for better or worse. I suspect that for the foreseeable future, government will continue to manage flood risk (and the crises that inevitably result) retroactively, taking action only when floodwaters recede. The remedies and strategies to avert future damage will be as unique as the municipalities themselves.
Dane County, site of the historic flooding referred to in the beginning of this column, is setting up an alert system for thousands of properties located outside regulated floodplains. The City of Houston, having successfully passed a $2.5 billion bond in 2018, now hustles to buyout property and make passage for managing future floods.
For hydrologists and engineers tasked with modeling flood risk for projects large and small, we can “up our game” by knowing the capabilities and limitations of our modeling tools, taking the time to represent risk with well-chosen coefficients, and being forthright—even to unreceptive clients—about projected flood hazards now and in the future.