Amir Behzadan, Texas A&M University – Artificial Intelligence Helping Communities in Floods

Floods are devastating to communities, but help could be on the way.

Amir H. Behzadan, associate professor in the department of construction science at Texas A&M University, discusses how A.I. could play a role in flood management.

Since Fall 2017, I have been an Associate Professor of Construction Science at Texas A&M University (TAMU). Prior to joining TAMU, I served as Associate Professor of Construction Management at Missouri State University (MSU) (2015-17), Assistant Professor of Civil Engineering (Construction) at the University of Central Florida (UCF) (2009-15), and Assistant Professor of Construction Technology at the New York City College of Technology (2008-9) of the City University of New York (CUNY).

I obtained my Ph.D. degree (2008) in Civil Engineering, and my M.S. degree (2005) in Construction Engineering and Management both from the University of Michigan, Ann Arbor. I also hold a B.S. degree in Civil Engineering (2002) from Sharif University of Technology. I have several years of construction industry experience working as field engineer and quality inspector in projects ranging from commercial to high-rise buildings.

My research interests encompass artificial intelligence (AI), adaptive simulation, and advanced sensing and visualization with applications in urban informatics, smart health and ergonomics, disaster resiliency, and safety and productivity. My work has been funded by the U.S. National Science Foundation (NSF), Texas Sea Grant program of the National Oceanic and Atmospheric Administration (NOAA), Florida Department of Transportation (FDOT), Engineering Information Foundation (EiF), Hinkley Center for Solid and Hazardous Waste Management (HCSHWM), and the industry.

My teaching interests include core construction and civil engineering topics. At the undergraduate level, I have taught in areas such as surveying and instrumentation, construction methods, construction contracts, structural design, soils and foundations, mechanics of materials, and statics. At the graduate level, I have developed and taught courses in research methods, project management, decision-making under uncertainty, data mining, and process simulation. I have received multiple teaching awards including the ASEE Outstanding New Teaching Award (2012), Distinguished New Faculty Award from the International Academy for the Scholarship of Learning Technology (2012), UCF Teaching Incentive Program (TIP) Award (2015), and Excellence in Undergraduate Teaching Award (2014) from UCF College of Engineering and Computer Science. In 2009, I attended and completed the requirements of the ASCE ExCEEd Workshop in West Point, NY.

I have also served the professional community in multiple capacities. Most notably, I am currently the Chair of the Visualization, Information Modeling, and Simulation (VIMS) Committee under the ASCE Computing Division. I also serve on the editorial board of two technical journals, namely Construction Engineering and Management (JCEM), and Smart and Sustainable Built Environment (SASBE). At MSU, I served as the faculty coach of the ASC Heavy Civil student team. Between 2010 and 2015, I served as the faculty advisor of the ASCE-UCF student chapter. During this time, I received several recognitions for my contributions to the profession including three consecutive Faculty Advisor of the Year Awards (2013, 2014, 2015) from the ASCE FL Section, and the Outstanding Faculty Advisor Award from ASCE Region 5 (Florida, Georgia, Alabama, Mississippi, Louisiana) in 2014.

Artificial Intelligence Helping Communities in Floods

Floods are the most frequent type of natural disasters in the U.S. There have been 5,000 flood events in the past 20 years, which took the lives of more than 2,000 people. Flood recovery is also very expensive! Only 1” of floodwater can cause $27,000 in damages to an average home, as little as 6” of water can reach the bottom of your car, and all it takes is 1-2 ft of water to float and carry away many vehicles.

Floodwater does not permeate asphalt or concrete, which can cause water run-offs, hamper evacuation and movement of goods and services, and of course, disrupt traffic.

FEMA flood maps are widely used to determine flood risk, vulnerability, and insurance cost. But the process of creating and updating these maps is very complex. 75% of these maps are older than 5 years, leaving many U.S. communities with no or outdated flood maps.

My research uses citizen science and artificial intelligence (or AI) to measure flood risk from photos of traffic signs that have standard dimensions and are easy to find. A good example is stop signs. There are roughly 1 billion of these signs in the U.S., or 250 signs per square mile. By comparison, the USGS operates a little more than 9,500 flood gauges nationwide.

In my work, we deploy computer vision and AI to detect submerged stop signs in crowdsourced street photos, and use them as benchmark to estimate floodwater depth, and create local flood risk maps. Crowdsourcing also allows us to work closely with flood-prone neighborhoods, and engage with local residents especially from underserved communities to make AI solutions that work for the public good.