evidoria

← Back to browse

Good practice Imported

DHMZ's AI-Based Three-Day Air Quality Forecast — Croatia Adapts Belgium's Machine-Learning ATMOSYS System

Croatia · Zagreb · See the Croatia profile · See the Zagreb profile

Evidence: Descriptive / self-reported Top 47% 60/100 · Ask Evidence Copilot about this practice

Croatia's DHMZ launched a nationwide 3-day AI/ML air-quality forecast (PM2.5, PM10, NO2, O3) on 17 Jan 2025, adapting Belgium's EU-funded ATMOSYS system. 85% EU-financed under the AIRQ project; the underlying model is documented to struggle with peak pollution events.

DHMZ's AI-Based Three-Day Air Quality Forecast — Croatia Adapts Belgium's Machine-Learning ATMOSYS System

Details

Maturity
Established
Promoter
Croatian Meteorological and Hydrological Service (DHMZ), with VITO / IRCEL-CELINE (Belgium)
Period
Launched 17 January 2025 (EU AIRQ project); underlying ATMOSYS system built 2010-2013
Keywords
environment, public health, meteorology, machine learning

Context

Croatia's national weather service DHMZ launched a public three-day air-quality forecast on 17 January 2025, covering PM2.5, PM10, nitrogen dioxide and ground-level ozone nationwide, running on Croatia's adaptation of ATMOSYS, built under the EU-cofinanced AIRQ project.

Activities

ATMOSYS originates in Belgium, where VITO, IRCEL-CELINE and VMM built it 2010-2013 under an EU LIFE+ grant, continuing to develop it into an operational forecasting system using gated recurrent unit neural networks for multi-day pollution prediction.

Conclusions

Croatia's adoption of the same modelling lineage for a national public forecast is a direct case of cross-border technology transfer, though neither operator has published country-specific accuracy figures, and the forecast chain 'has difficulties forecasting certain peak pollution events'.

Implementation
Indicative cost
Low (< €50k) — AIRQ project financed 85% by EU funds, remainder national; the underlying ATMOSYS system was originally built under a EUR 2.3 million EU LIFE+ grant (2010-2013).
Time to results
Medium (1–3 years) — Nationwide 3-day forecast launched 17 January 2025, adapting a system operational in Belgium since 2013.
Staffing & skills
DHMZ meteorological analysts, VITO/IRCEL-CELINE technical support (Belgium)

Conditions for success

  • cross-border adoption of a proven ATMOSYS modelling lineage
  • 85% EU co-financing via the AIRQ project

Common failure modes

  • documented difficulty forecasting peak pollution events

Commonly funded by

LIFE Programme National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful