Header: wildfire smoke plume over eastern Europe (Copernicus Sentinel-3 OLCI true colour, ESA, CC BY-SA 3.0 IGO). Alt: wildfire smoke plume on a Sentinel-3 scene.
Between 2024 and 2026, the global wildfire-tech market raised approximately $350-400 million in fresh investment. The capital flowed into three buckets: thermal satellites, fire-spread modelling, and ground sensors. A fourth bucket, atmospheric smoke dispersion, remains commercially empty. Only academia and government agencies occupy it. For a deep-tech team in Ukraine, this represents an operational window.
Where the 2024-2026 capital actually sits
The first category is thermal satellite detection. The European leader closed a EUR 37M Series B extension in May 2025 (approximately EUR 100M cumulative over twelve months). A Fraunhofer spinoff out of Freiburg closed another EUR 37M Series A in February 2026. A Luxembourg player raised $60M Series B in January 2026 at a $180M valuation. A US non-profit satellite mission together with the Earth Fire Alliance (FireSat, a dedicated constellation operator) runs on $13M from Google.org; Block 1 launches in 2026 with a target of a 20-minute global revisit by 2030.
The second category is fire-spread modelling. The market-leading fire-spread SaaS vendor has raised approximately $84M cumulatively (TA Associates 2022, General Atlantic BeyondNetZero 2024). Customers include CAL FIRE, PG&E, and SDG&E. The platform runs 20,000+ incidents per year and approximately one billion simulations per day.
The third category is ground sensors and camera towers. A San Francisco-based category leader closed $44M Series B in June 2025, $89M cumulative, covering 30 million acres. A Berlin-based vendor deploying solar LoRa sensors has raised approximately EUR 22M.
Now the fourth category: atmospheric dispersion of wildfire smoke. Zero commercial rounds. No SaaS vendor with peer-reviewed methodology delivered to an operational customer. The field runs on FMI SILAM and IS4FIRES (the Sofiev group in Helsinki), CAMS GFAS (ECMWF), and NOAA HYSPLIT. All three are academic or governmental.
Why capital avoided this category for a decade
Computational cost. A Lagrangian model such as FLEXPART or an Eulerian chemistry-transport model such as SILAM requires high-resolution meteorological fields (WRF at 3 km or finer), thousands of particles per source, a 24-72 hour horizon, and dozens of vertical levels. A single run consumes tens of GPU-hours. Scaling to national coverage in real time is a separate HPC problem that venture capital dislikes without a clear B2B anchor.
Interdisciplinary stack. A team that closes the full pipeline reads mesoscale meteorology, ensemble atmospheric chemistry, radiation safety, and aerosol epidemiology at the same time. That is four distinct PhD fields. A single-faculty startup does not assemble such a team.
Heterogeneous input data. Dispersion requires supersampling of satellite FRP (MODIS, VIIRS, TROPOMI), fuel-type emission factors, a diurnal combustion profile, plume injection height, PBL height, and a vertical wind profile. Every input carries calibration uncertainty, in contrast to thermal detection, where the pipeline is linear (pixel -> hotspot -> alert).
Absence of a solvent B2B buyer. Thermal detection is purchased by insurers and utilities. Spread models are purchased by electric utilities operating under California PSPS mandates. Dispersion is purchased by civil protection, public health, and transboundary regulators. These are public budgets, not corporate P&Ls; the venture cycle does not map to them.
Why the category is ceasing to be empty now
Mortality has entered the front page. Sofiev and colleagues in The Lancet Planetary Health (2025) estimate more than 100,000 deaths per year attributable to wildfire PM2.5, the first peer-reviewed attribution of that magnitude specifically to fire smoke. Jones et al. in Science (2024) linked extreme wildfire frequency to anthropogenic climate change. Byrne et al. in Nature (2024) demonstrated that the 2023 Canadian season (18.5 Mha, a satellite-era record) was a direct consequence of 2.2 degC warming above the seasonal mean.
Ukraine 2024. According to JRC (March 2025), Ukraine lost 965,000 hectares to fire during 2024, a historic maximum exceeding every prior figure including pre-war levels. Smoke propagated into Poland, Romania, and the Baltic, producing a transboundary EU civil protection incident.
European grant grammar has shifted. Horizon Europe Cluster 5, the Mission on Adaptation, EU4Health and the European Innovation Council calls for 2024-2026 now require a cascading multi-hazard approach (fire + drought + heat + health exposure). ICARIA (101093806) and the FIRE-RES + TREEADS + SILVANUS + Firelogue joint strategy (Brussels, May 2025) constitute the new Commission policy. They need participants with an operational stack, not a single satellite. Kelley et al. (2025, Nature Communications) deployed a fire-vegetation-climate feedback framework for ensemble Earth System Models; within three years baseline national climate services will need to include a fire-emissions and dispersion module.
Our stack: what we put into this window
The WildFiresUA system is our complete stack for Ukraine, integrating what only academia has integrated to date, but with an operational mandate.
- Meteorology: WRF with nested domains down to 1 km over Ukraine; assimilation of Sentinel-5P and local automatic weather stations.
- Fire behaviour: FARSITE + WindNinja on a 30-metre Anderson 13 fuel map for Ukraine (our CJFR 2026 peer-reviewed anchor).
- Atmospheric dispersion: FLEXPART for smoke plumes (from 2023 onward), CALPUFF for radiological scenarios (nuclear plants), HYSPLIT (NOAA ARL) for trajectory analysis.
- Ingest: Copernicus EMS, Sentinel-3 SLSTR, VIIRS active fire.
- Academic partnerships: Oles Honchar Dnipro National University, Marzieiev Institute, ULCO/LPCA.
- Operational mandate: State Emergency Service of Ukraine (DSNS).
We do not reinvent dispersion. We take peer-reviewed models (FLEXPART, Stohl et al., ACP 2005; CALPUFF, Scire et al., 2000; WRF-SFIRE, Mandel et al., GMD 2011), apply them to the Ukrainian 30-metre fuel baseline, run them at the operational cadence of DSNS, and publish the validation through peer review. SILAM/IS4FIRES remain the gold standard for pan-European coarse-scale assessment; we are not a competitor but a downscaling extension for a specific geography.
An operational distinction, not merely a scientific one
The distance between academic code and an operational service lies in SLA, monitoring, release cadence, and evidence. An academic group publishes a paper. An operational team guarantees that at 3 a.m., when 40,000 hectares are burning in Luhansk oblast, a 24-hour PM2.5 forecast for Kharkiv, Sumy, and north-east Poland appears on the DSNS screen with a known confidence interval. Horizon Europe reads this gap as TRL 7-8 rather than TRL 3-4, and EU4Health together with DG ECHO are prepared to fund it. We do not compete with FMI for scientific priority; we offer what FMI is not attempting to deliver: an operational east-European contour with the 2022-2026 wartime dataset as a unique validation corpus.
A conclusion for Ukrainian deep-tech
The wedge between categories is more interesting than duplicating a leader. When a segment is occupied by a team with $84M and 20,000 incidents a year, building another spread SaaS is unwise. When four well-capitalised players already occupy the thermal-satellite segment, a proprietary constellation will not pay back. The integrating category, which requires meteorology, atmospheric chemistry, and public health at once, remains closed to capital-light entrants.
Ukrainian deep-tech holds two structural advantages: inexpensive academic expertise in atmospheric and radiation physics, and a unique 2022-2026 operational dataset. Using both is a question of focus. For us, the focal point is atmospheric dispersion, anchored to thermal detection on one side and to population exposure on the other.
FAQ
How does FLEXPART differ from HYSPLIT within the stack?
FLEXPART is a Lagrangian particle dispersion model optimised for three-dimensional transport and aerosol deposition at the mesoscale. HYSPLIT provides trajectory analysis plus simplified dispersion, and runs faster for operational nowcasting. HYSPLIT serves 1-6 hour trajectories; FLEXPART serves 24-72 hour ensemble forecasts with PM2.5 deposition.
Why is CALPUFF restricted to radiological scenarios?
CALPUFF is a puff model calibrated for point sources (stacks, nuclear accidents). For diffuse area sources, a Lagrangian particle approach (FLEXPART) yields a physically more accurate representation of plume injection and PBL interaction.
Do you compete with CAMS GFAS?
No. CAMS GFAS runs on a 0.1 degree grid (~10 km) at pan-European scale. Individual Ukrainian events such as a local peat fire or a single field are smeared out at that resolution. We downscale to 1-3 km with local WRF meteorology and retain CAMS as the boundary condition.
Why now, rather than five years ago?
Until 2024, attribution science (fire -> PM2.5 -> mortality) lacked a global peer-reviewed estimate. Until 2025, Horizon Europe did not frame cascading multi-hazards as a first-class call. Until 2024, the Canadian record (18.5 Mha) and the Ukrainian record (965,000 ha) had not forced EU civil protection to seek operational dispersion at national resolution. The three drivers converged during 2024-2026.
Ukrainian startup ecosystem: follow TechUkraine and AIN.ua — the two leading outlets covering Ukrainian deep tech, climate tech, and environmental startups.
What to do today
- Check the YourAirTest air quality map for your city — recent PM2.5 readings.
- If the topic resonates, share with colleagues; we monitor referrals via Google Search Console.
- To contribute data (sensor measurements, regional models) reach out via the contact form.
References
- Sofiev M. et al. (2025). Global mortality burden attributable to wildfire fine particulate matter. The Lancet Planetary Health, 9(3), e180-e192.
- Jones M. W. et al. (2024). Climate change increases the risk of wildfires. Science, 383(6679), 142-149.
- Byrne B. et al. (2024). Carbon emissions from the 2023 Canadian wildfires. Nature, 633, 835-839.
- Kelley D. I. et al. (2025). Fire-vegetation-climate feedback in CMIP6 Earth System Models. Nature Communications, 16, 1412.
- Mandel J., Beezley J. D., Kochanski A. K. (2011). Coupled atmosphere-wildland fire modeling with WRF-SFIRE. Geoscientific Model Development, 4(3), 591-610.
- Stohl A., Forster C., Frank A., Seibert P., Wotawa G. (2005). Technical note: The Lagrangian particle dispersion model FLEXPART version 6.2. Atmospheric Chemistry and Physics, 5(9), 2461-2474.
- Scire J. S., Strimaitis D. G., Yamartino R. J. (2000). A User’s Guide for the CALPUFF Dispersion Model (Version 5). Earth Tech, Inc., Concord, MA.
- Our peer-review paper CJFR 2026 — High-resolution Anderson-13 fuel map for Ukraine validated with FLEXPART smoke dispersion. Canadian Journal of Forest Research.
- Aragoneses E. et al. (2023). FirEUrisk Pan-European fuel map at 1-km resolution. Earth System Science Data, 15(4), 1985-2006.
- European Commission Joint Research Centre (2025). Ukraine hit by record-breaking wildfires in 2024: 965,000 ha burned. JRC Science for Policy Brief, March 2025.
- Kaiser J. W. et al. (2012). Biomass burning emissions estimated with a global fire assimilation system (GFAS). Biogeosciences, 9(1), 527-554.
- EFFIS / Copernicus EMS Annual Report (2025). European forest fire statistics 2025: 1,079,538 ha burned, record for the European Union.
- NOAA Air Resources Laboratory. HYSPLIT smoke forecasting tools. arl.noaa.gov/hysplit/smoke-forecasting
- USDA Forest Service. WindNinja — high-resolution wind for fire behaviour. ninjastorm.firelab.org/windninja
- European Innovation Council (EIC) — Horizon Europe grant framework. eic.ec.europa.eu
- Wildfire Today — Ukraine news archive. wildfiretoday.com/tag/ukraine
Internal links: air-quality map, PM2.5 forecast, wildfire monitoring system, nuclear-plant scenario modelling, data sources, FAQ.