Live public-orbit data.
EmberSat train illustrative.
Drag to rotate.

Super-Early Wildfire Intelligence

Every wildfire
starts small enough
to stop.

EmberSat turns satellite observations into wildfire alerts crews can use. Our software has already run autonomously for nine days across 800,000 km². We are improving and deploying that software now. Dedicated sensing hardware can extend its speed and coverage as future funding allows.

See the Field Proof Run a Pilot

2nd place XPRIZE Wildfire · Space-Based Track
Team Snuffed · $250,000 award

9 days Fully autonomous over NSW · Proven

800,000 km² Monitored on public data · Proven

70% Finals events detected first, before public hotspot feeds · Proven

The Problem

Wildfires are easiest to stop when they are hardest to see.

Wildfire costs the United States on the order of $500 billion a year, all-in. Yet the moment when a fire is cheapest and safest to contain is still the hardest to see across large areas. Past roughly five acres, the odds of stopping a fire collapse below ten percent. About a quarter of fires escape initial attack, and those drive more than 85% of all area burned.

Today's satellites make operators choose. Geostationary weather satellites refresh in minutes but see large pixels. Polar imagers see finer detail but revisit less often. EmberSat is built around the number that matters at the fireground: minutes from image capture to a usable alert.

These figures are drawn from U.S. economic-impact and initial-attack research. Request the source memo.

What We Build

Our software already runs—and keeps improving.

EmberSat's software began as part of Team Snuffed's XPRIZE Wildfire entry. Alexander Shenkin then founded EmberSat, and the software now runs autonomously on public satellite data.

EmberSat NSW

Proven

Second Place in XPRIZE Wildfire

EmberSat was Team Snuffed's platform for the Space-Based Detection & Intelligence Track finals. The team took second place and received $250,000 as the Methodological Innovation Awardee. Read the XPRIZE announcement.

NAU reports an initial field of more than 300 teams.

During finals testing with the NSW Rural Fire Service in April 2026, EmberSat ran fully autonomously for nine days over 800,000 km² of New South Wales, entirely on public satellite data: no human in the loop, sixty active incidents on screen at once.

~14 min Median, image capture to alert · Proven

70.7% Finals events detected first, before public hotspot feeds · Proven

88 of 742 Candidate alerts emailed after AI review · Proven

Figures reported by the team in its final report to XPRIZE. The 14 minutes measures speed only; it does not measure accuracy. The 70.7% is the share of events EmberSat detected before the public VIIRS, MODIS and Sentinel hotspot feeds; it is not a lead time. Paid pilots will measure missed fires and false alerts with agency partners.

Download the press release (PDF)
September 23, 2026 · 2 pages

EmberSat NAZ

Deploying

Deploying Over Northern Arizona

The same detection stack, rebuilt for GOES-West and the Flagstaff fire corridor: continuous monitoring, automated alerting, and a dispatcher portal, again on public satellite data.

Dedicated Sensing Layer

Mission development

Designing a Satellite Constellation for Campfire-Scale Detection

We are working with a consortium based in Flagstaff to design a satellite constellation for high-risk fire corridors. The planned constellation would combine thermal-plus-RGB sensing with EmberSat's software, targeting campfire-scale detection and actionable alerts within ten minutes. We are now in discussions with spacecraft bus providers to define the path to flight. These are design targets, not current performance.

Who It Protects

See beyond the reach of ground systems.

Fire and emergency agencies

Find new ignitions sooner, assess them automatically, and send dispatchers information they can use.

Utilities and rail corridors

Watch long, ignition-prone corridors and terrain that camera networks do not cover.

Forests and remote landscapes

Monitor vast, remote landscapes without building a dense ground network.

How It Works

Software turns public satellite data into alerts today.

EmberSat continuously ingests live public satellite feeds, identifies and prioritizes likely ignitions, and delivers usable information to dispatchers. We are improving that operational chain now. Dedicated sensors and onboard triage are a future acceleration path if hardware funding becomes available.

Proven Autonomous detection on public feeds. EmberSat's software has already operated without a human in the loop across 800,000 km². That is the current product foundation.
Proven An AI agent checks every alert. Before a dispatcher gets an email, an AI agent reviews each candidate fire: its pixel history, nearby events, known heat sources, assets nearby, and the weather. In the finals it sent 88 emails from 742 candidates. It declined or held the other 654, and it wrote down its reason each time.
Proven Events close on evidence. EmberSat records every satellite look at a fire, including passes that saw no heat. An event closes when those observations show the fire is out.
Deploying Built for operational delivery. The Northern Arizona deployment adds continuous monitoring, automated alerting, and a dispatcher portal to the proven detection stack.
Funding-dependent Hardware can shorten the loop. A future dedicated sensor layer could add focused observations and onboard triage, targeting under five minutes from image capture to a crew-usable alert. That is a design target, not current performance.

9 days Autonomous software run · Proven

<5 min Future hardware target^†

How EmberSat finds a fire

Each pixel is compared with the heat expected for it at that hour. Many independent models then combine into one fire probability, and the AI agent reviews the result.

Real Himawari-9 data for one pixel near Drake, New South Wales, 9 to 11 April 2026. Fire probability is shown from 04:40 UTC on 11 April, when stored values begin. The VIIRS overlay is schematic, and the model inputs and review checks are simplified. 742 and 88 are finals totals.

What a dispatcher sees

Open an event, check its probability and heat-output history, and read every observation behind it in the detection log.

The EmberSat dispatcher view, replayed from archived records of BENDEMEER 106, New South Wales, 15 April 2026. Map: OpenStreetMap contributors.

What a dispatcher receives

The AI agent writes each alert in plain language. It says what it is sure of, what it is not, and what to do next.

From [email protected]
Subject Firealert_Snuffed_0416_14:17_380012

Bushfire detected ~10 km ENE of Deniliquin (Edward River LGA), ~0.8 km from the Deniliquin 132kV to Coleambally transmission line. Multi-sensor confirmation. Activity may be diminishing in the most recent frame. Recommend verifying line status.

First detected: 14:17 AEST (NOAA-20 VIIRS)
MODIS confirmation: 15:44 AEST (peak FRP 36.57 MW, high confidence)
Latest detection: 16:40 AEST (4 adjacent Himawari pixels)
Latest Himawari frame at 16:50 AEST: no detections. Fire may be cooling, but this is a single-frame change.

Threatened assets:
Deniliquin 132 to Coleambally (132kV transmission line): 0.84 km from leading edge, ETA 46 min (lower bound 26 min)
Nearest residential addresses: 8.6 km (Deniliquin township)

Weather: FFDI 10 (low-moderate), NW wind ~17 km/h.

RFS: no matching incident record. RFS may not be aware yet.

What we're confident about: this is a real fire (VIIRS + MODIS + multi-pixel AHI agree). A 132kV transmission line is within 1 km of the fire footprint.
Less certain: whether the fire is still actively spreading. Worth a phone call to TransGrid/Essential Energy to confirm line status.

An alert the AI agent sent during the XPRIZE Wildfire finals, 16 April 2026. Shortened and lightly repunctuated; recipients and links removed. Three independent satellite systems saw this fire.

Future Hardware Path

Funding-dependent

The mission would monitor the fire corridor.

Public satellites either cover a wide area at coarse resolution or provide detailed images only a few times a day. A dedicated thermal and RGB sensor over the fire corridor would provide detailed images more often. The target is to detect a fire at campfire scale and alert a crew within five minutes of the image being taken.

Katalyst Space Technologies, a Flagstaff mission-engineering firm, completed the orbital and cost baseline for the mission design.

A consortium is forming to build the hardware.

Team

Fire science. Satellite engineering. Program delivery.

Alexander Shenkin

Founder & CEO

Remote sensing ecologist at Northern Arizona University, where he runs the Ecosystem Science and Innovation Lab. Engineering physics and electrical engineering at UCSD; engineering at CU Boulder; satellite communications R&D at Lockheed Martin; then ecology — Yale, a doctorate at Florida, a postdoc at Oxford. Nature co-author.

Amy Wolkowinsky

Program Director

Directs programs and operations across EmberSat, working full-time with Arizona's emergency-management and forestry agencies.

Partners

Flagstaff puts space engineering and the fireground in the same room.

Katalyst Space Technologies

Flagstaff-based mission-engineering partner. Performed the orbital, mission, and cost baseline study behind our constellation figures.

Arizona Department of Forestry & Fire Management

Years of collaboration on remote sensing of vegetation stress through the Forest Health Protection program.

USFS Region 3 Forest Health Protection

Flagstaff-based forest-monitoring overflights, using EmberSat data to prioritize where to look next.