In this digital age, social media is playing an increasingly important role in helping authorities streamline their preparedness, as well as their ability to monitor and respond to events such as natural disasters. At the Lab, we have been developing CycloMon, a real-time analysis and visualisation platform that generates insights on behaviour patterns before, during and after cyclones using social media big data.
As recently seen in Barbuda, Puerto Rico and other affected islands in the Caribbean that just experienced devastating hurricanes, these weather conditions often result in widespread damages. Just last year in the Pacific, Tropical Cyclone Winston ravaged Fiji, with damages estimated at 2 billion USD or 2.5 billion AUD (which is just under half of Fiji’s GDP). An even more dire view, as evidenced by NOAA and World Bank, posits that global warming and climate change cause more cyclones, more frequently — and in an unpredictable way. Compounded with other societal or natural crises, the negative impacts from these weather disasters could be worse. In Madagascar this year for instance, Cyclone Enawo in March was preceded by a drought in the latter part of 2016.
Numerous attempts have been made to better protect vulnerable populations and provide timely humanitarian aids to people affected. These include implementations that utilise innovative technologies and approaches, such as: the Artificial Intelligence for Disaster Response (AIDR) platform, a satellite imagery analysis tool that quantitatively assessed in real-time the damages during Hurricane Harvey; the deployment of drones after Tropical Cyclone Pam in Vanuatu to capture aerial footage; and the use of financial records to map the levels of persons’ preparedness and recovery. Collective intelligence via crowdsourcing has also emerged as a serviceable way to help people in flooded areas. The U-Flood platform, in particular, was launched during Hurricane Harvey to enable people in affected communities to register and monitor streets that were flooded.
A mutual friend in all of these weather disasters nevertheless is social media. As a platform, Facebook’s Crisis Response has several functions. Some of which include sharing one’s safety status with loved ones (Safety Check), matching the supply and the demand of aids (Community Help), and raising funds. As a data source, social media has the potential to provide rich information that can fill information gaps, especially during instances of sudden disasters. At the Lab, we have demonstrated how social media can be used to better inform peatland fire and haze disaster management units, and our study has since manifested into a publicised analytics and visualisation platform known as Haze Gazer.
Extending Haze Gazer, we have developed a cyclone monitoring platform nicknamed CycloMon, based on its primary function. This platform took shape after several discussions with cyclone-related organisations and our data science team. We were particularly interested in how a series of automatic processes (combining advanced data analytics and data mining) could 1) collect and analyse social media data and 2) generate near real-time insights both from social media and baseline information, to better inform a cyclone disaster.
Here, we would like to share a few fundamental functions of CycloMon, illustrated with the three screenshots below. Aiming to demonstrate how this platform could be beneficial for cyclone management units in the region and beyond, we’ve tested its feasibility using Tropical Cyclone Winston which occurred in the Pacific.
- As a cyclone can affect multiple countries simultaneously (or sequentially), we conceptualised this platform as a tool that would be useful throughout the region, possibly to monitor 14 countries in the Pacific as shown in Figure 1.
- Once a warning is issued, the platform evolves to an emergency mode (as seen in Figure 2), which presents not only detail information on a particular cyclone, such as its trajectory and essential meteorological information, but also a close monitoring on social media signals from potentially affected countries.
- The platform allows for the exploration of rich, country-specific information mined from various sources of social media, such as text-, image-, and video-based contents, as presented in Figure 3. By the way, this is exactly how Haze Gazer functions.
Next month, Pulse Lab Jakarta will open a public domain of CycloMon in specific geographical regions using a set of disaster-related taxonomies in various languages (which we previously collected using Translator Gator, a tool for crowdsourcing translations in multiple languages with some gamification elements and incentive schemes). We look forward to discussing the potential and challenges of such an approach, and exploring further collaboration opportunities with both academia and practitioners.
Pulse Lab Jakarta is grateful for the generous support from the Government of Australia.