Name in original language
Софтверска платформа за предикцију и превенцију поплава заснована на методама вештачке интелигенције и IoT уређајима - VodostAI
Initiative overview
The initiative addresses the problem of late and unreliable flood warnings caused by limited monitoring coverage, fragmented data, and slow analysis. It leverages the opportunity to combine low-cost river-level sensors with AI to produce faster, more consistent risk assessments. Its objectives are to:
- Improve flood prediction accuracy and lead time
- Provide clear, actionable alerts and dashboards for water authorities and emergency services
- Support prevention planning, including critical locations, thresholds, and response protocols
- Create a shared, auditable data and analytics layer that different institutions can rely on.
In the next phase, the platform will be implemented through pilot deployments with river basin authorities and municipalities, then embedded into routine monitoring and emergency procedures via defined roles, SOPs, and data-sharing agreements. It is expected to scale by adding more sensor locations, integrating additional data sources, including weather forecasts, radar, and historical events, and standardising deployment as a repeatable package comprising hardware, software and training.
Over time, VodostAI can be expanded to national coverage and adapted for related risks such as flash floods, landslides, and drought monitoring.
Results, outcomes and impacts
The main results are a working VodostAI prototype, comprising sensor data ingestion, AI forecasting and an alert dashboard, validated on historical flood events. Performance has been measured through back-testing against observed river levels, tracking forecast error and warning lead time versus baseline approaches.
In future pilots, we expect earlier warnings, with a target of 2–12 additional hours of lead time where conditions allow, fewer false alarms, and faster response coordination. Impact will be monitored via operational logs, including alerts issued and acted on, event reviews, and reductions in reported damage and disruption in covered areas.
