Product · 2025.10 (48-hour hackathon)
Meteor Madness — NASA Space Apps 2025
An asteroid-impact dashboard built in 48 hours for the NASA Space Apps Challenge: live NASA feeds (near-Earth objects, close approaches, fireballs, natural events) behind a plain Java server, and a browser that turns a diameter, a speed and a point on a Three.js globe into crater, blast, seismic, tsunami, population and cost estimates, plus a kinetic-impactor deflection simulation.
- Role
- Two-person team — the 3D globe, the 2D map and the impact calculations; backend and API built together
- Period
- 2025.10 (48-hour hackathon)
- Stack
- Java 17 · com.sun HttpServer + Jackson · Vanilla JavaScript · Three.js · Leaflet · NASA NeoWs · CNEOS CAD & Fireball · EONET
- Scale
- 48 hours · 44 commits · 2 people · 7,425 lines: 6,205 JS / HTML / CSS, 1,220 Java · 4 NASA sources · 8 endpoints · 7 tabs · Run on Oct 11, 2026: 200 close approaches, 429 fireballs
- Links
- GitHub ↗
Exhibits
NASA Space Apps Challenge, 4–5 October 2025, two-person team. Exhibits 2–4 come from running the project on 11 October 2026: the physics sweep calls the app's own functions, the two data charts are one pull of NASA data through the project's Java backend.
- Hackathon
- 48 h44 commits · 2 people · Oct 4–5, 2025
- Code
- 7,425 lines6,205 JS / HTML / CSS · 1,220 Java
- Data
- 4 → 84 NASA sources behind 8 endpoints
- Pull, Oct 11 2026
- 200close approaches · 12 inside the Moon's orbit
From four NASA feeds to a dashboard
- NeoWs feedobjects passing in a ±2-day window · size, hazard flag
- CNEOS close approachesdate · distance · velocity · brightness
- CNEOS fireballssince 2015 · energy · position · speed
- EONET natural eventswildfires, storms, volcanoes
- Java HTTP server8 endpoints · 3 backup API keys
- Browser dashboardvanilla JS · 7 tabs · Leaflet maps
- Overviewcounts · averages · date ranges
- Impact predictionmass → energy → crater · blast · seismic · tsunami · people · cost
- 3D globeThree.js · click to set the impact point
- Deflectionkinetic impactor Δv · Monte Carlo hit probability
What the app's physics says, from 10 m to 2 km
stony body (3,000 kg/m³) at 20 km/s, 45° · computed by calling the app's own functions · both axes logarithmic
Close approaches the dashboard pulled: how near, when
200 objects within 0.05 AU, Sep 11 – Oct 14, 2026 · y = distance in lunar distances, logarithmic · filled = already passed on the pull date, hollow = upcoming
Fireballs recorded by CNEOS since 2015, by year
429 events · bars = events per year · lower panel = summed impact energy in kilotons · 2026 is a partial year
Problem
The Meteor Madness challenge asked for a tool that makes NASA’s near-Earth-object data understandable and lets people explore what an impact would do and whether it could be deflected. The raw material is four separate NASA feeds with different shapes, rate-limited demo keys, and physics that most visitors have never seen. We had 48 hours and two people.
Decisions
Put a thin server between NASA and the browser. A Java HttpServer with no framework fetches the NeoWs feed, CNEOS close approaches and fireballs, and EONET events, flattens each to simple JSON and serves eight endpoints (Exhibit 1). When the demo key hits its quota the server rotates through three backups instead of failing the page.
Do the physics in the browser. Mass, energy, crater, blast radius, seismic magnitude, tsunami height, exposed population and economic loss are computed in JavaScript from the user’s inputs, so the globe, the 2D map and the outcome cards update instantly. The formulas are simple power laws chosen to be readable in a hackathon, not research-grade (Exhibit 2).
Make the impact point a click. My part of the front end was the Three.js Earth, textured with NASA Blue Marble, synchronised with a Leaflet map so a click on either sets the latitude and longitude that the prediction uses. I also reworked the calculation methods and fixed the 2D map to use real-world coordinates. Part of what I built in those two days was not kept in the final submission.
Show uncertainty and the way out. Every outcome carries a range (±20% on crater size, ±0.5 on magnitude), and the Simulation tab runs a Monte Carlo over size, speed, Δv and lead time to report how a kinetic impactor changes the hit probability.
Result
A working dashboard with seven tabs, submitted on time. Run a year later, it still answers from live data: on 11 October 2026 the close-approach tab pulled 200 objects within 0.05 AU, twelve of them inside the Moon’s orbit (Exhibit 3), and the fireball tab 429 events since 2015, led by the 49 kt Bering Sea airburst of 2018 (Exhibit 4). A 300 m stony body at 20 km/s over New York comes out as a 4.4 km crater and 116 million people exposed.
Limits
Hackathon physics: a 10 m body is given a 340 m crater and seismic magnitude stays at zero below about 200 m, which real impact models would not do. Population exposure comes from a short built-in list of major cities, not a population grid. The NASA key lives in a properties file, so the rate limit is shared by everyone who runs it. The code was written in two days with AI assistance, which the repository discloses, and has no tests.
04
Gallery
Overview: the four NASA feeds reduced to counts, averages and date ranges, refreshed from the Java server on load. Close approaches: a date window, a maximum distance in AU, and the objects plotted on a Leaflet map above the table. Fireballs: atmospheric entries from the CNEOS catalogue, mapped by their reported position. Impact prediction: set diameter, density, speed and angle, then click the Three.js globe or type a latitude and longitude. The outcome cards for a 300 m stony body at 20 km/s over New York, with the nearest major cities and their exposure. Deflection: a Monte Carlo over size, speed, Δv and lead time, reporting the Earth-hit probability before and after a kinetic impactor.