About

Operational decision support for wildfire smoke.

Mission

PlumeSentinel AI transforms wildfire smoke science into real-time, AI-powered decision support. The project is led by NSF NCAR / ACOM (National Science Foundation National Center for Atmospheric Research, Atmospheric Chemistry Observations and Modeling Laboratory).

By reducing the path from plume detection to decision-ready briefing from hours of manual synthesis to under fifteen minutes, PlumeSentinel AI addresses the critical gap between frontier science and timely public health guidance during fast-moving wildfire smoke events.

Why now

Wildfire smoke is one of the fastest-growing public health threats in the United States. Recent severe events — including the January 2025 Los Angeles fires that resulted in 31 direct civilian fatalities and destroyed over 16,000 structures, and the summer 2023 Canadian wildfire plumes that blanketed the U.S. East Coast for weeks and reached 100 million people — highlight the urgency.

At the same time, the technical ingredients for an integrated decision-support system are now mature: federal operational smoke forecast products, satellite observations from polar-orbiting and geostationary platforms, independent forecast-evaluation frameworks, and frontier AI capable of reasoning over multimodal scientific data at operational tempo.

Approach

PlumeSentinel AI maintains a continuously updated incident state by combining federal smoke forecasts, satellite observations, and partner inputs into a single operational picture. It runs chemistry-aware and aerosol-aware forecasts and scenarios, ranks response options, and generates incident wildfire smoke briefings and public protective messaging drafts.

The system is designed as a decision-support tool, not an autonomous oracle. It includes human-in-the-loop safety gates for every public-facing message and high-stakes recommendation. AI agents generate options and track confidence; humans provide the authority and policy discussion.

For the layered architecture and the role of each component, see Architecture.

Standards alignment

PlumeSentinel AI is built around emerging open standards for global meteorological telemetry. The system architecture uses MQTT (Message Queuing Telemetry Transport) — a lightweight publish-subscribe messaging protocol — and is natively compatible with WMO WIS 2.0 (World Meteorological Organization Information System 2.0), the next-generation data-sharing infrastructure for WMO's 193 Member States.

WIS 2.0 entered operational status on 1 January 2025, replacing the legacy GTS (Global Telecommunications System) that had been in operation since 1971. The U.S. NOAA (National Oceanic and Atmospheric Administration), through its NWS (National Weather Service), is actively deploying WIS 2.0 Global Broker and Global Cache services — establishing MQTT as the standard protocol for real-time atmospheric data distribution.

Building on MQTT from the outset means PlumeSentinel AI can subscribe to upstream forecast notifications today and, in later phases, publish plume intelligence on the same global network — supporting international scaling without custom integration.

Lead institution

NSF NCAR is a federally funded research and development center sponsored by the National Science Foundation and managed by UCAR (the University Corporation for Atmospheric Research). NCAR provides the atmospheric and Earth system science community with research tools, facilities, and observational data.

The project is housed within ACOM, NSF NCAR's atmospheric chemistry laboratory. ACOM brings decades of expertise in fire-smoke emissions, plume chemistry, atmospheric transport, and the satellite observing systems used to verify operational forecasts.

Recent news

All news →