⚙️ How-To Meteorites & fireballs Data sources & downloads Notebook By

How to Read NASA's Fireball and Bolide Reports in Python

On this page

Problem statement

When a small asteroid, a metre or a few metres across, hits the atmosphere and explodes as a bright fireball, US government sensors in orbit usually see it. NASA's Center for Near Earth Object Studies (CNEOS) at JPL publishes those detections through its Fireball Data API: 1,074 events from 1988 to October 2026, each with a time, the energy it radiated, an estimated impact energy, and, when known, the position, altitude and speed at peak brightness. The API returns numbers as strings, positions as unsigned values with N/S and E/W letters, and many blanks. This page turns it into a usable table and shows what it covers.

Quick answer

Convert the string columns to numbers and apply the direction letters as signs:

num = ["energy", "impact-e", "lat", "lon", "alt", "vel", "vx", "vy", "vz"]
fb[num] = fb[num].apply(pd.to_numeric)
fb["date"] = pd.to_datetime(fb.date)
fb["lat"] = fb.lat.where(fb["lat-dir"] == "N", -fb.lat)
fb["lon"] = fb.lon.where(fb["lon-dir"] == "E", -fb.lon)
print(fb[["date", "energy", "impact-e", "lat", "lon", "alt", "vel"]].head(4).to_string(index=False))
               date  energy  impact-e   lat    lon  alt  vel
2026-10-04 03:14:45     5.4     0.180 -41.8 -173.1 26.9 13.8
2026-09-15 11:26:13     2.2     0.079 -37.6 -161.6 37.0  NaN
2026-09-11 10:18:03     6.1     0.200  -0.2 -129.6 70.0  NaN
2026-09-11 01:12:11    24.6     0.670  54.4 -100.1 38.0 20.7

The rows come newest first. The latest, on 4 October 2026, exploded at 26.9 km over the Pacific east of New Zealand at 13.8 km/s, with 0.18 kilotons of impact energy. Without the signs it would plot in the North Pacific east of Japan. energy is the light radiated, in units of 10¹⁰ joules, and impact-e the estimated total energy in kilotons of TNT.

World map of 888 located fireballs sized by impact energy, 648 over sea in blue and 240 over land in red, with the large Chelyabinsk circle over Russia. World map of 888 located fireballs sized by impact energy, 648 over sea in blue and 240 over land in red, with the large Chelyabinsk circle over Russia.
Fireballs detected since 1988, sized by impact energy.

What the columns hold

1. What each column holds

print(fb[num].notna().sum().to_string())
print(f"from {fb.date.min():%Y-%m-%d} to {fb.date.max():%Y-%m-%d}")
energy      1074
impact-e    1074
lat          888
lon          888
alt          602
vel          363
vx           363
vy           363
vz           363
from 1988-04-15 to 2026-10-04

Every event has both energies, because they are what the sensors measure. The position is known for 888, the altitude of peak brightness for 602, and the speed with its three components for only 363. The table starts with a single event in April 1988 and runs to 4 October 2026. A map can use 888 events, a study of entry speeds a third of the table.

Horizontal bars of how many of the 1,074 fireballs have each value: both energies for all, a position for 888, an altitude for 602 and a speed for 363. Horizontal bars of how many of the 1,074 fireballs have each value: both energies for all, a position for 888, an altitude for 602 and a speed for 363.
What the fireball reports record.

2. Over land or sea

land = gpd.read_file("data/ne_50m_land.zip")
pts = gpd.GeoDataFrame(fb.dropna(subset=["lat"]), geometry=gpd.points_from_xy(fb.dropna(subset=["lat"]).lon, fb.dropna(subset=["lat"]).lat), crs=4326)
pts["over_land"] = pts.within(land.union_all())
print(f"located: {len(pts):,}; over land: {pts.over_land.sum()} ({pts.over_land.mean():.1%})")
located: 888; over land: 240 (27.0%)

240 of the 888 located fireballs, 27.0%, exploded over land, close to the share of the Earth's surface that is land. That is what arrivals spread evenly over the globe would produce, and it is why so few of these events are witnessed or yield meteorites: most end over open ocean.

3. Velocity and its components

v = fb.dropna(subset=["vx"])
mag = np.sqrt(v.vx**2 + v.vy**2 + v.vz**2)
print(f"with components: {len(v)}; |v| - vel: median {(mag - v.vel).median():.3f} km/s, worst {(mag - v.vel).abs().max():.1f}")
print(v.assign(norm=mag.round(1)).loc[(mag - v.vel).abs() > 1, ["date", "vel", "norm", "vx", "vy", "vz"]].to_string(index=False))
with components: 363; |v| - vel: median 0.002 km/s, worst 18.4
               date  vel  norm   vx    vy    vz
2026-10-04 03:14:45 13.8  15.2 14.8   2.9   2.2
2026-08-14 07:48:36 12.2  22.9  5.2  22.3   0.2
2023-06-21 19:39:12 16.5  18.1  7.3  -2.0 -16.4
2018-12-18 23:48:18 13.6  32.0  6.3  -3.0 -31.2
2014-08-23 06:29:41 16.2  17.6 -2.3   5.7  16.5
2014-06-28 02:40:05 12.4  16.3 12.0   3.5 -10.5
2009-10-08 02:57:14 19.2  22.1 14.0 -16.0  -6.0

The speed vel should be the length of the vector (vx, vy, vz), given in an Earth-centred frame, and for most of the 363 events it is, to a few metres per second. Seven are not, and in all seven the components give the higher speed. The 2018 Bering Sea fireball is listed at 13.6 km/s, but its components make 32.0 km/s; the newest event's make 15.2 km/s against 13.8. The table does not say which is right, so a speed study should flag these rows rather than trust either value.

Reports over time and the largest events

The data used on this page

NASA/JPL's Fireball Data API from the Center for Near Earth Object Studies, every record with velocity components (vel-comp=true), downloaded on 28 October 2026 and saved on this site, because new events are added as they are reported. Natural Earth's 1:50m land, public domain, decides over land or sea.

from pathlib import Path
import json
import zipfile

import geopandas as gpd
import numpy as np
import pandas as pd
import requests

Path("data").mkdir(exist_ok=True)
# Every fireball in NASA/JPL's Fireball Data API (CNEOS), with NASA's Meteorite Landings table and Natural Earth's
# countries and land. Saved 28 October 2026: new fireballs are added as they are reported.
# Live: https://ssd-api.jpl.nasa.gov/fireball.api?vel-comp=true
ARCHIVE = "nasa-meteorites-2026-10-28.zip"
if not Path(f"data/{ARCHIVE}").exists():
    Path(f"data/{ARCHIVE}").write_bytes(requests.get(f"https://data.spatialworkflow.io/{ARCHIVE}", timeout=600).content)
Z = zipfile.ZipFile(f"data/{ARCHIVE}")
raw = json.loads(Z.read("fireball.json"))
fb = pd.DataFrame(raw["data"], columns=raw["fields"])
for f in ["ne_110m_admin_0_countries.zip", "ne_50m_land.zip"]:
    Path(f"data/{f}").write_bytes(Z.read(f))
print(f"{raw['signature']['source']}: {len(fb):,} fireballs")
NASA/JPL Fireball Data API: 1,074 fireballs

Example 1: reports per year

per = fb.groupby(fb.date.dt.year).size()
print(per.loc[1994:2025].describe().round(1).loc[["mean", "min", "max"]].to_string())
print(per.loc[[1994, 1995, 1996, 2005, 2015, 2019, 2022, 2025]].to_string())
mean    32.7
min     13.0
max     45.0
date
1994    13
1995    21
1996    33
2005    45
2015    45
2019    42
2022    42
2025    33

From 1994 on, reports run at a mean of 32.7 a year, from 13 in 1994 to 45 in 2005 and 2015. Before 1994 there are only five events in six years, which reflects what was released rather than what arrived. The count has no trend since the mid-1990s, which fits a steady rate of arrivals seen by sensors of roughly constant sensitivity. The series is a record of detections above a threshold, and small events are still missed.

Bars of fireballs reported per year from 1988 to 2026, a handful before 1994 and then between 13 and 45 a year around a mean of 32.7. Bars of fireballs reported per year from 1988 to 2026, a handful before 1994 and then between 13 and 45 a year around a mean of 32.7.
Fireballs reported per year.

Example 2: the largest

print(fb.nlargest(5, "impact-e")[["date", "impact-e", "lat", "lon", "alt", "vel"]].to_string(index=False))
               date  impact-e   lat   lon  alt  vel
2013-02-15 03:20:26     441.0  54.8  61.1 23.3 18.6
2018-12-18 23:48:18      49.0  56.9 172.4 26.0 13.6
2009-10-08 02:57:14      33.0  -4.2 120.6 19.1 19.2
1994-02-01 22:38:09      30.0   2.7 164.1  NaN  NaN
2009-11-21 20:53:29      20.0 -22.0  29.2 38.0 32.1

The largest event by far is Chelyabinsk, on 15 February 2013, at 441 kilotons: it exploded 23.3 km above the Russian city at 18.6 km/s, and its shock wave injured more than a thousand people. Next come the Bering Sea fireball of December 2018 at 49 kt, over the sea off Kamchatka, and an event over Indonesia's Bone Gulf in October 2009 at 33 kt. The 1994 event over the western Pacific, at 30 kt, has no altitude or speed.