Researchers use deep-learning frameworks and advanced data analysis to help better prepare for disasters to save lives and minimize damage.
FREMONT, CA: In a world where natural disasters can strike unexpectedly and cause devastating impacts, reducing their effects in advance may seem impossible. These unpredictability and widespread consequences make it difficult to anticipate and respond appropriately, leaving individuals and communities vulnerable to their destructive forces. Urban cities heavily rely on road networks to facilitate the movement of goods, information, and people. However, during times of disaster, such as floods, these road networks can be damaged or blocked, severely impacting access to crucial services like hospitals, shelters, and grocery stores. Furthermore, floods in urban areas often lead to vehicle accidents due to driving on flooded roads, making them a significant cause of fatalities and underscoring the failures of road infrastructure.
To address the challenges of flooding events, researchers have developed a deep-learning framework capable of effectively predicting the near-future flooding of roads. The framework was rigorously tested using three models, and the results demonstrated an impressive accuracy in predicting the flooding status of roads, with precision values of 98 percent and recall values of 96 percent. Validation of the models using data from the 2017 Hurricane Harvey flooding further reinforced their reliability.
Knowing the flooding status of roads in advance can help affected communities avoid flooded streets and enable emergency management agencies to plan efficient evacuations and resource allocation. After a natural disaster, assessing damages to homes and buildings can be time-consuming, often taking months.
In another study, researchers delved into the dynamics of post-disaster recovery, specifically during the short-term period after a hurricane. The focus was on understanding the time it takes for people to return home after evacuating and how long they can remain in their homes without having to evacuate again. Location-based data were employed to analyze the speed at which individuals in different areas undertake these actions.
The findings revealed that more than half of the census tracts in Harris County returned from evacuation within five days after Hurricane Harvey's landfall and ceased evacuating after six weeks. However, recovery times varied across different areas and among various population groups.
The researchers also investigated how people of different socio-demographic statuses, such as income or rental status, responded to flooding. The team gained insights into diverse recovery patterns by studying different subpopulations' evacuation and relocation patterns. While the study indicated that it took longer for residents in flooded areas to return compared to non-flooded areas, there was no significant difference in evacuation and relocation times for low-income populations between the two types of areas.
It was uncovered that high-income census tracts took longer to return than low-income ones when faced with flooding. This finding suggests the vulnerability of low-income populations in evacuating and relocating during disasters. Additionally, the study revealed that areas with shorter return durations may not necessarily be more resilient to disasters, as they may signal challenges faced by low-income and minority populations, necessitating additional assistance and resources.
