This episode of Making Waves features Kevin Nebiolo, a senior data scientist at Kleinschmidt Associates, as he discusses the development of Emergent, an innovative agent-based modeling software designed to simulate fish behavior in hydropower environments.
Key Highlights of Emergent & the Project
- What is Emergent? (5:51): It is an agent-based modeling tool that simulates fish behavior from the bottom up. Instead of averaging data, it models individual fish as autonomous agents following simple rules—like avoiding collisions or sensing fatigue—resulting in complex, realistic emergent behaviors like schooling and delays.
- Project Origin (7:28): The software was developed for a hydroelectric licensing project in Alaska, helping a rural cooperative evaluate how a proposed diversion would impact fish passage and population sustainability.
- Technological Breakthroughs (10:29): Kevin and his team shifted to GPU-accelerated processing using Numba and CUDA, which allowed them to simulate thousands of interacting agents in near real-time, drastically improving efficiency compared to initial CPU-based runs.
- Validation and Stakeholder Reception (19:12): The team successfully validated the model by comparing simulation outputs side-by-side with actual drone imagery of fish behavior in the river, which proved highly effective in building trust with stakeholders and agencies.
- The Future of Coding (21:50): Kevin shares that integrating AI-assisted coding tools has completely revolutionized his workflow, allowing him to build complex software in a fraction of the time and shifting his focus toward creativity and problem-solving over manual coding.
Takeaways for the Industry
- Embrace Complexity (28:47): Kevin emphasizes that oversimplifying behavioral systems often leads to inaccurate predictions. By embracing the “chaos” of collective behavior, scientists can achieve a much deeper and more realistic understanding of how fish navigate infrastructure.
- Interdisciplinary Teamwork (14:58): The project highlights the importance of combining data science with senior biological expertise, as having a fish biologist on the team was critical for ensuring that the model’s parameters remained biologically accurate and defensible.