Introduction
We are thrilled to announce the winner of the GAMSPy Student Competition for Fall 2025! This season’s first-place prize goes to Nikhil Trivedi, a dual-major undergraduate student in Computer Science and Atmospheric & Oceanic Sciences at the University of Wisconsin-Madison.
Nikhil’s winning project, titled “Optimal Weather Radar Network Design for Texas,” tackles a critical meteorological challenge using advanced optimization techniques. By transforming a complex environmental problem into a mathematical model, Nikhil demonstrated how GAMSPy can be used to improve early warning systems for severe weather.
A big shout out also goes to our 2nd and 3rd place winners: Emilie Lesinski, also from the University of Wisconsin-Madison and Koushik Malli Jayachandran, from RWTH Aachen University. We had many impressive submissions, but these three projects stood out for their creative use of modeling methods, combined with a clever pipeline integration in Python, making the best use of GAMSPy capabilities.
The Challenge: Closing the Gap in Weather Monitoring
Texas is notorious for its severe weather -ranging from tornadoes and supercells to flash floods, hurricanes and even snowstorms. Detecting these events at low altitudes is essential to provide the public with accurate and timely weather warnings. However, creating a perfect radar network that can spot these phenomena is a challenge.
For starters, the resources to build NEXRAD radars are limited, and these have to be installed in candidate airport sites that provide the required supporting infrastructure -meaning they have to be strategically positioned. Plus, one has to account for terrain obstructions and the curvature of Earth itself, which create “coverage gaps” where the radar signal can’t detect and monitor certain events.
Nikhil decided to frame this challenge as a Facility Location Problem (FLP). The goal was simple: to determine the optimal locations for a finite number of radars among a set of candidate airport sites, and maximize “quality-weighted population coverage” (ensuring the maximum number of people are covered by high-resolution radar data).
The Methodology: Physics Meet Optimization
What set this project apart was the integration of atmospheric science into the optimization model. Nikhil didn’t just treat radar coverage as a simple “yes/no” binary. Instead, he developed a continuous Coverage Quality Metric based on three physical factors:
-
Beam Broadening: As a radar beam travels farther, it widens, losing the resolution needed to see small features like tornado vortices.
-
Earth Curvature: Because the Earth is curved, a straight radar beam eventually rises too high above the ground to detect low-level weather phenomena.
-
Terrain Blockage: Using Digital Elevation Models (DEM), the model identifies if mountains or hills physically block the radar’s line of sight.
Nikhil implemented his model using GAMSPy, leveraging the CPLEX solver through our free academic license to find the global optimum. By using sparse matrix construction to filter out physically unviable connections (where distances were too great or beams were blocked), he was also able to significantly improve the model’s computational efficiency.
The Result
After validating his model on a synthetic dataset, Nikhil applied it to a real-world scenario in Texas, using:
- 6,896 Census Tracts from the 2020 US Census, representing the population.
- 127 Candidate Airport Sites from the Texas Department of Transportation records, for infrastructure support.
- 1-km Resolution Terrain Data from the United States Geological Service (USGS).
The model successfully placed an optimal network of radars that maximized high-quality data coverage for the most densely populated areas. In the final Texas simulation, the model achieved a 91.44% quality-weighted coverage for the state’s population.
A Word from the Winner
Nikhil was introduced to the world of optimization through Professor Michael Ferris’s course at UW-Madison. Reflecting on his experience with GAMSPy, he shared:
“What I enjoyed most about working with GAMSPy was its ability to transform complex problems into a solvable and displayable solution. Being able to integrate powerful solvers like CPLEX with Python’s geospatial data analysis and plotting capabilities was incredibly useful for my project. As long as I built my mathematical model correctly, GAMSPy made the implementation straightforward, allowing the solver to do the heavy lifting.”
Conclusion
Congratulations once again to Nikhil Trivedi for his outstanding contribution to the intersection of computer and atmospheric sciences. Our jury highlighted the project’s clear pipeline integration, good code quality, and its easy-to-follow presentation. Together with our runner ups Emilie Lesinski and Koushik Malli Jayachandran, this year’s winners continued raising the bar for what’s possible when using GAMSPy to its full potential.
In its second ever edition, our GAMSPy student competition demonstrated again the results of integrating advanced optimization techniques with creative problem solving. We are very proud to support an academic community that keeps pushing the boundaries and using our tools at the cutting edge of science and technology.