FAMU-FSU Engineering Study Advocates for Hybrid Flood Modeling to Bridge Gap Between Speed and Physical Accuracy

FAMU-FSU Engineering researchers call for hybrid flood models that combine data-driven speed with physics-based accuracy to improve global disaster prediction.

By: AXL Media

Published: Mar 13, 2026, 5:17 AM EDT

Source: Information for this report was sourced from Florida State University

The Divergence of Modern Predictive Simulations

As the technology used to forecast storm impacts has evolved, the scientific community has witnessed a significant split in modeling methodologies. Advanced digital simulations are currently used to guide billion-dollar decisions in infrastructure design, land-use planning, and emergency response. However, Assistant Professor Ebrahim Ahmadisharaf noted that these methods have diverged into narrow applications, often being used beyond their original intended scope. This isolation creates a missed opportunity to combine the strengths of different modeling paradigms, which is essential for improving the accuracy of predictions across diverse environmental domains.

Balancing Computational Efficiency with Physical Reality

Flood models are generally categorized into four types: physics-based, data-driven, observational, and conceptual. Currently, there is a growing trend toward using data-driven methods because they are easier to implement and require less computational power than traditional physics-based systems. While these newer models are excellent for identifying complex patterns in existing data, they often fail when predicting events that fall outside their training data. Ahmadisharaf explained that because these models lack strong physical constraints, their reliability in operational forecasting and regulatory hazard analysis is limited when compared to computational models that understand the fundamental physics of water movement.

The Limitations of Siloed Scientific Paradigms

When new modeling methods are developed in isolation from older frameworks, their improvements remain siloed within specific academic or industrial domains. This fragmentation hinders the ability of engineers and emergency managers to prevent flood events effectively. The research suggests that while data-driven models are useful for comparing the relationship between flooding and other variables, they trade reliability for efficiency. By not taking full advantage of high-performance computing resources to run more complex, physics-based simulations, the field risks overextending simplified systems beyond their actual capabilities during extreme weather events.