Long before COVID-19 fixated the world on the issue of how viruses spread, Texas A&M University researchers investigating ways to improve the modeling of urban flooding asked themselves the question: Would it be possible to predict, in near real time, the behavior of flooding using an approach that epidemiologists and others rely on to predict the spread of viruses in social networks? Somewhat to their surprise, the researchers found that the answer was yes. In a recently published paper, they propose a new “contagion” model of flooding in urban areas that could complement existing hydraulic and hydrologic models. This contagion model could also help emergency managers and first responders better understand and respond to the spatial extent of flooding during a high-water event.
The standard way for assessing flood propagation and flood risk in cities involves the use of physics-based hydraulic and hydrologic models, says Ali Mostafavi, Ph.D., Aff.M.ASCE, an associate professor in the Zachry Department of Civil and Environmental Engineering at Texas A&M. Such models “are very powerful and they are very helpful,” Mostafavi says. “But the main use of this method is before an event happens, for planning purposes.” Once urban flooding has begun, the models “are not very helpful” at predicting how and where flood waters will spread or recede during short periods of time, he notes.
The beatuy of the model is its simplicity.
“That was one of the motivations of our work, a model that could be used during emergency response to provide in real time predictions of where the water will flow,” Mostafavi says. About 18 months ago, he and two graduate students — Chao Fan, in the Department of Civil and Environmental Engineering, and Xiangqi “Alex” Jiang, in the Department of Computer Science and Engineering — began investigating approaches to simulate flooding by means of mathematical modeling. They hit upon the idea of modeling flooding in much the same way that existing models are used to predict the spread of viruses and diseases in social networks.
“In the case of flooding, we can think of floodwater as the virus and the street networks as the social networks in which people come into contact with each other,” Mostafavi says. In this analogy, he says, “when two streets are close to each other and are connected with each other, if one is flooded, the chance that the other one gets flooded increases.”
The three presented the results of their findings in the article “A Network Percolation-Based Contagion Model of Flood Propagation and Recession in Urban Road Networks,” which was published online by the journal Scientific Reports on Aug. 10. In the article, the authors summarize the mathematical model they developed to describe the spatial and temporal spread and recession of floodwaters in urban road networks.
In part, the researchers based their contagion model on the mathematical approach for analyzing the spread of infectious diseases known as the Susceptible-Exposed-Infected-Recovered model. For their purposes, the researchers label as “susceptible” a road segment that is in a floodplain or other low-lying area. An “exposed” road is one that is receiving significant amounts of rainfall or stormwater runoff, while an “infected” road is one that no longer can be used because of flooding. Finally, a “recovered” road is one from which floodwaters have receded. Essentially, Mostafavi and his team combined the SEIR model with an approach for predicting the spread and recession of flooding on a road network based on the degree to which nearby road segments are flooded.
In the article, the researchers also show how they verified the results of their model using 2017 data regarding flooded roads from Harris County, Texas, when the Houston metropolitan area and much of southeast Texas experienced severe flooding as a result of Hurricane Harvey. “The results show that the model can monitor and predict the fraction of flooded roads over time,” the article states. “Additionally, the proposed model can achieve 90 percent precision and recall for the spatial spread of the flooded roads at the majority of tested time intervals.”
A major benefit of the model is its simplicity. The contagion model relies on a “system of differential equations rather than very computationally expensive data-driven models or hydraulic and hydrology models,” Mostafavi says. Despite its straightforward nature, the contagion model can “provide really powerful and accurate predictions.”
At the same time, the contagion model also can indicate “in which road segments the flooding will recede faster than other road segments,” Chao says. Existing flood models usually do not show when water will recede from a given roadway, he notes.
However, a key limitation of the model is that it cannot predict where flooding will begin. “But as soon as we have that (information), it can predict the propagation,” Mostafavi says.
Among its other potential benefits, the proposed contagion flood model could be used by emergency management officials to inform first responders and residents of the likelihood of particular roads flooding, for example, in the coming four to six hours. Such predictions could assist emergency response endeavors as well as the actions residents take to evacuate. “That could be very important information,”
This article first appeared in the November 2020 issue of Civil Engineering.