Using Big Data to detect and predict natural hazards better and faster: lessons learned with hurricanes, earthquakes, floods

Simone Sala Blog

Simone-Sala-image-04

Data–Pop Alliance has been conducting ongoing research on Big Data, climate change and environmental resilience. With funding from the UK’s Department for International Development (DfID), we published a synthesis report evaluating the opportunities, challenges and required steps for leveraging the new ecosystem of Big Data and its potential applications and implications for climate change and disaster resilience. This report will feed into the World Humanitarian Summit to be organized in Istanbul in May 2016.

This is the first in a series of companion pieces that offer insights from the synthesis report. The authors of the series attempt to go “beyond the buzz” to lay out what we actually know about Big Data’s existing utility for disaster science and for building practical resilience.


Every day, geological, biological, hydrological, and climatic factors produce natural hazards, which in some cases result in natural disasters that can have a devastating impact on ecosystems and human societies. Hazards can be geophysical (e.g. earthquakes, cyclonic storms), biological (e.g. infestation), or generated by a combination of different factors (e.g. floods, wildfires, etc).

Big Data technologies can play a role in:

  • monitoring hazards
  • determining the exposure of human societies to disaster risk
  • tracking impacts of disasters and monitoring recovery efforts
  • mitigating vulnerabilities; and
  • strengthening resilience of communities.

Particularly interesting is the role of Big Data for detecting earthquakes, floods, hurricanes, as well as forecasting future occurrence of such hazards.

Earthquakes

Even if science is clear about the impossibility of predicting earthquakes, the detection of such events increasingly leverages data from sensors and digital social data.

Earthquake epicenters (1963-1998)
Source: NASA, Digital Tectonic Activity Map (DTAM) project

The combined application of accelerometers in mobile phones and computers with cloud computing can help faster detection of their occurrence. A group of scientists (Cochran, Lawrence, Christensen, and Jakka) employed this approach to develop the Quake-Catcher Network (QCN), a seismic network leveraging distributed/volunteer computing to gain critical insights on an earthquake by bridging traditional seismic stations with innovative data sources. In 2009 the group demonstrated that it is possible to detect small earthquakes through a global network of computers connected via the Internet, highlighting the existing capacity to develop rapid earthquake early warning systems at relatively low cost thanks to distributed data collected from the Internet.

In 2014 a group of scientists (Musaev, Wang, and Pu) developed LITMUS, a model to detect landslides following earthquakes by integrating multiple data sources. By integrating social sensors (Twitter, Instagram, and YouTube) and physical sensors (USGS seismometers and TRMM satellite), the model scored better than traditional techniques employed by USGS for real-time hazard mapping.

Digital social data from relevant organizations are being integrated to detect when crises happen. For example, the USGS monitors Tweets mentioning earthquakes worldwide with magnitudes of 5.5 and above as a means of detecting them and issuing alerts more broadly through their Twitter Earthquake Dispatch (@USGSted).

Floods

Big Data also allow the early detection of floods. By combining information related to flooding from Twitter and satellite observations, a group of scientists (De Groeve, Kugler, and Brakenridge) built a real-time map of location, timing, and impact of floods. The map, constantly updated, can be accessed online.

Global flood maps from 1985 to 2007
Source: Dartmouth Flood Observatory Global Active Archive of Large Flood Events

Social media enables qualitative situational analysis before, during, and after disasters. Floodtags (a social media analytics platform) was employed to extract information from Twitter, enabling the filtering, visualization, and mapping of social media content based on location and keywords. Satellite data came from the Global Flood Detection System (GFDS), which provides a service for rapid identification of inundated areas through daily passive microwave satellite observations. The approach was tested in two case studies, respectively in the Philippines and in Pakistan, proving to be particularly appropriate for monitoring large floods in densely populated areas.

Twitter pattern linked to 2014 floods in the Philippines
Source: Jongman et al., 2015

In the Netherlands, where the vast majority of the population lives in flood-prone areas, the government has started experimenting with how machine learning may help strengthen preparedness to future floods. In Australia, the New South Wales State Emergency Service developed an early warning system able to perform predictive analysis of floods in the region based on the integration of the Bureau of Meteorology’s external data and additional datasets (e.g. data from flood plain, historical data information from various databases).

Storms

Big Data has also proven to be helpful in monitoring and assessing the impacts of storms, whether they be hurricanes, typhoons or cyclones (such distinction depending only on the location in which the storm happens). Indeed, Big Data technologies allow an unprecedented capacity to crunch data from distributed datasets that will help gain innovative insights on the weather system. The Government of South Korea, for example, upgraded the simulation capacity of its meteorological office by 1,000% – providing it with the most capable storage system of the country. Not only weather data, but also social data as well as data from mobile telecommunication operators can be useful for mapping and analyzing meteorological hazards. In Bangladesh, the Mobile Data, Environmental Extremes and Population (MDEEP) project investigated how data from the national telecommunication operator Grameenphone could have provided insights on the effectiveness of early warning systems during the occurrence of cyclone Mahasen in 2013.

Big Data-powered visualization tools seem particularly promising for helping real-time sensemaking of the weather system as well as for raising awareness of natural hazards among citizens. Earth, an open source animated map integrating data from NOAA’s global forecast system and OSCAR’s ocean currents, among other sources, is a clear example of such potentialities.

Cyclone Pam in the proximity of Vanuatu (March 13, 2015) as portrayed in earth
Explore the visualization on the platform

Works cited

Cochran, E. S., Lawrence, J. F., Christensen, C., & Jakka, R. S. (2009). The Quake-Catcher Network: Citizen Science Expanding Seismic Horizons. Seismological Research Letters, 80(1), 26–30. http://doi.org/10.1785/gssrl.80.1.26
De Groeve, T., Kugler, Z., & Brakenridge, G. R. (2007). Near Real Time Flood Alerting for the Global Disaster Alert and Coordination System. In Proceedings of the 4th International ISCRAM Conference. Delft, the Netherlands.
Musaev, A., Wang, D., & Pu, C. (2015). LITMUS: A Multi-Service Composition System for Landslide Detection. IEEE Transactions on Services Computing, 8(5), 715–726. http://doi.org/10.1109/TSC.2014.2376558