It began as a low, haunting sound building in the distance. It grew to a thunderous roar that drowned everything else. The sky changed to an unnatural shade, then went black. The wind slammed trees and buildings with brutal force. Sirens wailed. Windows and buildings exploded.
In the spring of 2011, Joplin, Missouri was devastated by a EF5 tornado estimated to have winds exceeding 200 mph. The storm killed 161 people, injured more than 1,000, and destroyed and damaged 8,000 homes and business. The tornado left behind miles of splintered debris and caused over $2 billion worth of damage.
Tornadoes can produce winds that are stronger than the design limits for most residential and commercial structures. The traditional methods of assessing the damage after a catastrophe can take weeks or months, delaying emergency responses, insurance claims, and long-term reconstruction efforts.
A new research from Texas A&M University could change that. Researchers at Texas A&M led by Dr. Maria Koliou (associate professor and Zachry Career Development Prof II in the Zachry Department of Civil and Environmental Engineering) have developed a method that combines deep learning, remote sensing and restoration models to accelerate building damage assessments and forecast recovery times after tornadoes. Once images from the event are available, this model can produce damage estimates and recovery forecasts within an hour. The researchers published Sustainable Cities and Society with their model
.
“Manual field inspections are labor-intensive and time-consuming, often delaying critical response efforts,” Abdullah Braik is a civil engineering student at Texas A&M and coauthor of the model. “Our method uses high-resolution sensing imagery and deep learning algorithms to generate damage assessments within hours, immediately providing first responders and policymakers with actionable intelligence.”
It helps predict repair costs, and estimate recovery time. Researchers can estimate these timelines and costs for different situations using deep learning technology. Deep learning is a type artificial intelligence.
“We aim to provide decision-makers with near-instantaneous damage assessment and probabilistic recovery forecasts, ensuring that resources are allocated efficiently and equitably, particularly for the most vulnerable communities,” Braik said. How It Works
Researchers used three tools to create this model: deep learning, remote sensing and restoration modeling.
Remote Sensing uses high-resolution images from sources like NOAA or satellites to show the extent and size of damage in large areas.
“These images are crucial because they offer a macro-scale view of the affected area, allowing for rapid, large-scale damage detection,” Braik said.
Deep Learning analyzes images automatically to accurately identify the damage. The AI is trained in advance of disasters by analyzing images of past events. It learns to recognize visible damage signs such as missing walls, collapsed roofs and scattered debris. The model then categorizes each building as having no damage, moderate damages, major damages, or destroyed.
Restoration modelling uses past recovery data, details of the building and infrastructure and community factors – like income levels or resources – to estimate how long homes and neighborhoods might take to recover under different funding and policy conditions.
By combining these three tools, the model can quickly assess damage and predict short and long-term recovery times for communities affected by natural disasters.
“Ultimately, this research bridges the gap between rapid disaster assessment and strategic long-term recovery planning, offering a risk-informed yet practical framework for enhancing post-tornado resilience,” Braik said.
Test The Model
Koliou & Braik used the data from 2011 Joplin tornado for their model testing due to its size, intensity, and availability of high quality post-disaster info. The tornado destroyed thousands buildings, resulting in a diverse dataset which allowed the model to train and be tested at various levels of structural damages. The detailed ground-level damage assessments were a reliable benchmark for assessing how accurately the model classified the severity of damage.
“One of the most interesting findings was that, in addition to detecting damage with high accuracy, we could also estimate the tornado’s track,” Braik said. Future Directions “By analyzing the damage data, we could reconstruct the tornado’s path, which closely matched the historical records, offering valuable information about the event itself.”
Researchers are working to use this model for other types disasters such as hurricanes and quakes, as long as satellites detect damage patterns.
“The key to the model’s generalizability lies in training it to use past images from specific hazards, allowing it to learn the unique damage patterns associated with each event,” Braik said. “We have already tested the model on hurricane data, and the results have shown promising potential for adapting to other hazards.”
According to the research team, their model could help communities recover more quickly and efficiently in future disasters. The team wants to extend their model beyond damage assessments to include real-time updates and tracking recovery progress over time. He said. The technology could transform the way emergency officials, insurers, and policymakers respond to storms in the critical hours and days following a storm. It can provide near-instant assessments and projections of recovery. The National Science Foundationfunded this research.
