Concept
The Global Flood Monitor (GFM) of Flootagsmaps the cities around the world which are being hit by a flood in real-time based on twitter messages (see Figure 1, left). The GLObal Flood Forecasting Information System (GLOFFIS) is a global hydrological forecasting system that forecasts river runoff for 1 to 14 days in advance based on external forecasts of precipitation for the entire world, with a resolution of 1 degree (see Figure 1, right). By linking both systems, we can create flood alerts based on real-time ground-truth information about flooding combined with forecasted runoff which will give a idea of how much more flooding is to come. At the same time we can learn from ground-truth flood information about the threshold in simulated runoff at which riverine flooding occurs.
...
The cities of for which tweets are being monitored in GFM are based on the populated places (cities) database from the Natural Earth Data. We classified these cities according to their exposure to different types of flooding in: riverine flooding, coastal flooding, or riverine and coastal flooding, based on their location. For those cities exposed to riverine flooding, we paired a flow measuring station from the Global Runoff Data Centre (GRDC) in the same floodplain, with the city. For these measuring stations we have real-time and forecasted flow from GLOFFIS, and measured runoff from the GRDC dataset. The runoff data is normalized per station to make the data more informative. The normalized flow conditions are expressed as the percentage of time over the last 30 years during which the modeled runoff was lower than current modeled runoff. In this way, for each city susceptible to riverine flooding (based on the paired station) we can get real-time and forecasted normalized flow conditions. This information can be presented in the global flood monitor together with ground truth flood observations.
Case study: Louisiana, March 2016
In order to test this approach, we analyzed a flood that took place in March 2016 in Louisiana, Southern United States. The flood was widely reported in the news and had several mentions in the GFM (see Figure 2).
...
Figure 4. Simulated flow for Ouachita river at Monroe station (GLOFFIS results)
Next steps
After this proof of concept we will now automate the link between social media flood observations and modeled runoff. The results will be presneted together in a webviewer. A first mockup of how this could look like is shown in Figure 5
...