Twitter and Real-Time Flood Maps

In a joint study, two Dutch organisations, Deltares and Floodtags, have developed a way to derive real-time flood maps using a combination of flood modelling and data mining from social media.

“a better view of what is actually happening during a flood”

Deltares say that, when implemented in an operational warning system, the method will create real-time maps based on tweets that people have sent a minute previously.

“This new method will eventually give crisis managers a better view of what is actually happening during a flood so they can make more effective decisions: the right measures at the right time, in the right place,” said Dirk Eilander, flood expert at Deltares.

Currently, standard flood-extent maps are derived from a limited number of sources, such as satellite images, areal images, ground observations, hydrodynamic models and post-flooding flood marks. Much of this information is usually supplied after the event. Traditional data sources don’t offer the capacity to create accurate flood-extent maps in real-time.

Social Media – New Data Sources

The emergence of social media allows access to new data sources that contain large numbers of real-time observations from local people.

Twitter has long been considered a rich potential source of real-time information on disasters and severe weather events such as floods. According to Deltares, in the city of Jakarta, Indonesia, the intensity of unique flood-related tweets during a flood peaked at almost 900 tweets a minute during floods in February 2015. That amount of real-time data is too good an opportunity to overlook.

But data from Tweets and other social media sources may be unreliable, or at least unfocused. In order to be of use to civil protection and disaster management agencies such as BPBD in Jakarta, it needs to be filtered and validated in some way.

Combining the data processing and flood modelling expertise of Deltares with the expertise in data mining and social media of Floodtags, the joint study of the two organisations developed a procedure to use the thousands of observations generated by the social media to create reliable, real-time flood-extent maps.

Pilot in Jakarta

The study was piloted in Jakarta, during the floods of February 2015. According to Deltares, a significant number of 900 flood-related tweets per minute mentioned above included information about water depth and location. However, uncertainties arose because observations were generally rough estimates.

Deltares say that if disaster managers are to use this cloud of observations, the data needs to be filtered, enriched, validated and transformed into easily interpretable flood-extent maps.

Deltares and Floodtags therefore used what they describe as “hydrodynamic corrected Digital Elevation Maps to create real-time flood-extent maps for Jakarta”. The real-time flood-extent maps provided a good comparison with ground-truth photographs in most neighbourhoods in Jakarta.

When implemented in an operational warning system, the method will create real-time maps based on tweets that people have sent a minute previously.

Better still, the process can be used in other locations. Deltares say that this method can be scaled easily for any place in the world with enough Twitter activity.

According to Deltares, the maps are also useful in the post-flood phase for the calibration of hydrodynamic flood models and for insurance companies to obtain rapid information about areas where damage has occurred.

The research was presented at the General Assembly of the European Geosciences Union in Vienna on 14 April 2015.

For more information, see Deltares and Floodtags.