ASCENT

ASCENT helps predict and analyze the environmental impact of rocket launches by modeling the dispersion of key pollutants (CO2, NOx, and Al2O3) based on launch parameters and weather conditions.

Comment

ASCENT: Aerospace System for Chemical Emissions & Numerical Tracking

Inspiration

I've always been captivated by spaceflight: the deafening roar of engines, the plumes of exhaust expanding in the sky, and the thrill of watching humanity push the boundaries of exploration. Week after week, I tuned in to SpaceX missions, fascinated by the technical marvel of rocket launches.

But one day, I found myself asking a question no livestream ever answered: What kind of environmental impact do these launches have on our atmosphere? While we celebrate each mission to orbit, the lingering chemical footprint of every launch is invisible to the naked eye. That curiosity sent me down a path of research, leading to the development of ASCENT—a system designed to predict and visualize the dispersion of pollutants released by rocket launches.

With ASCENT, we can quantify what’s happening in the atmosphere after each launch, offering scientific insights that can shape the future of sustainable space exploration.

What It Does

ASCENT is an AI-powered simulation tool that predicts how rocket emissions spread in the atmosphere. Users input launch parameters (Launch Latitude, Launch Longitude, Payload Mass (kg), Fuel Type, and Rocket Type) and with these, ASCENT simulates atmospheric dispersion of three different pollutants (CO2, NOx, and Al2O3) using a combination of:

  • Machine Learning (Random Forest Regressor): Predicts pollutant concentration and spread.
  • Gaussian Dispersion Modeling: Simulates how emissions behave in various atmospheric conditions.
  • Weather Data Integration: Incorporates real-time wind, pressure, and humidity data for enhanced accuracy.
  • Interactive UI (Streamlit): Visualizes the dispersion as a heatmap overlaying a geographical map.

How We Built It

ASCENT is powered by a modular and data-driven architecture:

  • Backend AI Model: Trained using historical emissions data, employing Random Forest Regression for pollutant concentration predictions.
  • Gaussian Dispersion Simulation: Models how emissions spread based on atmospheric variables like wind, temperature, and pressure.
  • Weather Data Integration: Fetches live weather data via APIs to ensure accurate real-time simulations.
  • Streamlit Frontend: A user-friendly web app where users input launch data and see dispersion heatmaps.
  • Folium Mapping: Creates an interactive geospatial visualization overlaying emission patterns on real-world maps.

Data Challenges and Breakthroughs

One of our biggest challenges was acquiring and processing reliable emissions data. Unlike car pollution, rocket emissions aren’t as widely studied. We sourced data from:

  • NASA & ESA atmospheric studies on launch emissions.
  • Scientific research papers modeling the impact of various propellants.
  • Historical weather archives to validate our model against real-world launches.
  • SpaceX’s Previous Launch Data to get information about how successful launches were under what conditions.

Ensuring the model balanced accuracy and computational efficiency was another hurdle. By optimizing our Random Forest implementation, we improved prediction speed without sacrificing precision.

Challenges We Ran Into

Developing ASCENT wasn’t without obstacles:

  • Dynamic Weather Data Handling: Real-time API calls needed error handling and data smoothing to avoid inconsistencies.
  • Finding Reliable Emission Profiles: Each rocket type (Falcon 9, Starship, SLS) has unique emissions—standardizing the data took effort.
  • Balancing Model Complexity vs. Speed: We had to optimize computational efficiency for near-instant heatmap updates.
  • Seamless Integration: Bridging machine learning, physics-based models, and web-based visualization was a major engineering challenge. All the parts could work perfectly when separate, but combining them together would lead to issues at times.

Accomplishments We’re Proud Of

Despite these challenges, ASCENT achieved several milestones:

  • Built a functional AI-driven prediction model capable of estimating atmospheric dispersion with high accuracy.
  • Created an intuitive dashboard where users, researchers, and space agencies can interactively analyze pollution effects.
  • Validated our system with historical launches, ensuring reliable predictions based on past weather data.
  • Optimized for efficiency, enabling real-time pollutant heatmap rendering.

ASCENT is more than a tool: it’s a step toward making space exploration more sustainable.

What We Learned

This project was a deep dive into atmospheric modeling, machine learning, and environmental science. Key takeaways include:

  • Atmospheric dispersion is incredibly complex. Factors like wind shear, humidity, and temperature inversions drastically affect pollutant spread.
  • Machine learning is powerful in environmental science, but with scarce data in very niche areas such as the dispersion of pollutants from a rocket launch throughout an atmosphere means that we have to think more abstractly about how to get this data we need in the most accurate fashion.
  • Scientific models must be accessible. ASCENT’s user-friendly interface makes environmental data more understandable for non-experts.

What’s Next for ASCENT?

ASCENT is just the beginning. We envision a future where:

  • More pollutants (e.g., water vapor, aluminum oxide) are included in our models.
  • Displaying HeatMaps over a long period of time such that we can see the change in dispersion of these chemicals as time moves along.

I still remember seeing my very first rocket launch, in awe of its beauty and power. That same curiosity led to ASCENT, being able to solve those questions I had earlier using data and computer science. Now, every launch represents a chance: not only to venture into space, but also to more fully comprehend and safeguard our planet

References:

  • Stafoggia, M., et al. (2020). "A Random Forest Approach to Estimate Daily Particulate Matter, Nitrogen Dioxide, and Ozone at Fine Spatial Resolution in Sweden." Atmosphere, 11(3), 239.
  • Biau, G. (2012). "Analysis of a Random Forests Model." Journal of Machine Learning Research, 13, 1063-1095.
  • Holzworth, G. C. (1974). "Climatology of Atmospheric Dispersion." NOAA Technical Report.
  • Stockie, J. M. (2011). "The Mathematics of Atmospheric Dispersion Modeling." SIAM Review, 53(2), 349-372.
  • "A comprehensive review of Gaussian atmospheric dispersion models." Modeling Earth Systems and Environment, 2023.

Built With

Share this project:

Updates