Dashboard

The FloodTags dashboard is simple but powerful, designed for flood and disaster management professionals. It gives access to real-time and historical media data, delivered within seconds of being posted. Users can explore current situations, search by topic or location, filter and tag events, and easily zoom in on specific patterns for deeper insight.

Data sources

FloodTags gathers data from a wide range of sources, including BlueSky, X, YouTube, Facebook, Instagram, and over 150,000 news sites, blogs, and forums worldwide. Users can also submit new content via X or Telegram, and specific WhatsApp groups can be added by agreement. We combine official and informal sources using broad keyword strategies to ensure timely, comprehensive event coverage.

Event types

We started with flood monitoring and have expanded to now include floods, landslides, mudflows, windstorms, snow, hail, and wildfires. These are all weather impact events that start suddenly and affect large areas. Especially in such situations, social media data is extremely valuable, offering early signals and ground-level observations that are often unavailable through traditional channels.

Event detection

We detect new events by identifying anomalies using domain-specific models. The system updates every ten minutes, evaluating incoming media for signs of floods or other weather impacts. Alerts are shared via the dashboard or email. With visual tools like maps, timelines, and filters, users can monitor where and when events occur—enabling fast, informed action.

Location detection

FloodTags extracts event locations from media posts with up to 97% accuracy, using AI tools such as ChatGPT. We focus on identifying the location of the event itself, not the user, ensuring relevance without privacy concerns. This allows for accurate geographic mapping of weather impacts and supports timely analysis, response planning, and validation of affected areas.

Information extraction

We extract detailed information from text on event causes, impacts, and responses—such as hydrologic load, system failures, fatalities, evacuations, infrastructure damage, and public sentiment. This includes both confirmed facts and reported perceptions. To process large volumes of unstructured content efficiently and accurately, we increasingly make use of large language models (LLMs) such as ChatGPT.

Image analysis

We analyze photos to detect whether a flood is present—distinguishing floodwaters from regular water bodies. We also estimate water depth and recognize features such as vehicles, motorcycles, and cattle. This allows users to quickly verify whether an event is ongoing and better understand conditions on the ground using real-world visual evidence.

Data entry and crowdsourcing

Citizens can submit reports through X and Telegram. On X, a designated hashtag encourages public reporting. On Telegram, our chatbot guides users through a few simple questions. This feature can also be used by local team members to submit structured observations. All reports are automatically added to the FloodTags dashboard, supporting shared situational awareness across teams.

Ongoing research

We collaborate with universities and institutes to advance weather impact monitoring. For example, with Deltares we combine media and hydrometeorological data to validate models. Through the InDeep programme, eight PhDs improve information extraction using interpretable AI. And with the British Geological Survey, we are building on FloodTags for landslide detection. These efforts bring research directly into practice.

API Integration

FloodTags’ API integrates with operational systems to ingest, display, and process real-time media data. It connects with platforms such as Delft-FEWS, MIKE, Tygron, Lizard, or any web-based environment. The API enables real-time model calibration, forecasting, and enriched situational awareness within your own tools—helping organizations streamline workflows and make timely, data-informed decisions.

Open to collaborating and developing new features

If you have ideas to improve your use of FloodTags, we’re happy to think along. Shared feature requests that benefit multiple users, are added to our backlog at no cost, with priority based on demand. Client-specific features are also possible through a development agreement. We welcome suggestions that help tailor the platform to real-world needs.