Integrating Big Data Sources to Better Respond to Natural Disasters
In communities across the world that are prone to natural disasters, having an effective disaster response strategy in place is vital to save lives and support affected communities. To successfully coordinate what are often complex disaster relief efforts, governments, disaster response authorities and humanitarian agencies need useful, up-to-date information that can be easily accessed.
Building on our cyclone monitoring prototype, our team spent the past couple months developing an integrated, big data analytics prototype to provide timely insights that can aid logistics planning and information management after natural disasters, including cyclones, earthquakes, hurricanes and floods in Indonesia and countries in the Pacific region. In a few weeks, we’ll be launching it for testing with end users to better shape its future development. Here we take a look at the main features and the prototype’s potential for global impact.
Nicknamed DisasterMon, the current prototype features three data insight layers related to logistics planning, social media communication and socio-economic variables, which are based on multiple data sources (including open data platforms, national statistics and social media) that we hope to add to over the coming months. Albeit plenty of data sources exist that can aid in humanitarian assistance and disaster relief efforts, many of them are scattered across different databases and institutions.
With benefits for different stakeholders, such as local disaster authorities (insights provided by the tool can be paired with on-the-ground information to support analysis and decision making); humanitarian agencies (registered users can export their own data set to a custom layer to generate easy-to-interpret map visualisations); and the general public (citizens can view the latest information about areas affected and ongoing activities from Twitter verified news-sharing accounts), DisasterMon seeks to enhance natural disaster response at the local, national and international levels using big data. It also has potential to influence future development of automated real-time disaster monitoring systems.
Against the backdrop of the 2018 Earthquake in Palu, Sulawesi, Indonesia, below we explore what each feature offers:
Optimising Planning Logistics
Transporting valuable resources to disaster-affected areas or planning the best evacuation route are often hampered by several factors, for instance lack of knowledge about local road conditions, the length of time the journey may take and even topography, all of which can influence one’s choice of route. The logistics layer of DisasterMon provides information on road type, level of elevation, distance between points and estimated length of journey.
For example, one can see from the screenshot below that getting from Kasiguncu Airport (a domestic airport) to SMK PGRI Palu (a school near the site of the earthquake) may take five hours of driving, and the elevation of the road peaks at more than 900 meters. Users can weigh their options by choosing alternative points of origin and destinations. This feature of the tool relies on a pair of OpenStreetMap and OpenRouteService APIs.
Taking into Account Socio-Economic Variables
In many instances, the people most affected by a natural disaster are the poor and vulnerable. Having an understanding about population density, GDP per capita and other socio-economic characteristics of a certain region can help inform authorities’ estimates of supplies and resources needed, as well as help to target their allocation at the most vulnerable areas. Visualising data from 1960 to 2017 below, the tool depicts information about the total number of people living in the region, GDP per capita and the poverty headcount ratio based on the Global Facility for Disaster Reduction and Recovery, World Bank data.
Getting Updates from Social Media
Many people rely on social media as an information source during and after natural disasters to get updates and reduce uncertainty. Wrapping your head around the deluge of photographs, captions and videos shared however can take some time. Applying data mining approaches, the tool provides frequent updates on cities affected, the number of people injured, the number of casualties, the number of people evacuated and information about shelter locations based on information shared by Twitter verified accounts. DisasterMon also extracts various patterns to provide relevant information, such as the number of tweets related to the Palu earthquake over several days, popular words and terms being used to describe the event, as well as most shared videos from verified accounts.
Visualising Data in a Custom Layer
Noted earlier, one of the benefits for registered users is the ability to export their own data set to be visualised by DisasterMon. The data fed into the tool is not saved long-term, but users can capture screenshots and conduct analysis based on quickly generated map visualisation of complex data sets. As of now, the tool can process files in GeoJSON and Marker formats.
Seen in the custom layer below, DisasterMon bounds the geographical area affected by the 2018 earthquake and tsunami in Sulawesi and highlights areas where disaster relief agencies are active — the darker green indicates a larger presence of disaster relief agencies — based on 4W data received by OCHA and AHA Centre from different cluster leads. A cluster is a sector-specific coordination group of humanitarian organisations, both UN and non-UN, focused on strengthening preparedness and technical capacity to respond to humanitarian emergencies.
In the coming weeks, the prototype will be launched for testing with tutorial materials available for governments, disaster authorities, humanitarian agencies and general members of the public. The team is also exploring the possibilities of developing other layers for volcano and wildfire disaster events, as well as evaluating how best to incorporate additional data sets such as mobile network data to provide insights. Data is the lifeblood of this tool and to reach its global potential, critical hazard-related national and global data is essential. DisasterMon relies on multiple data sets that are available globally through public APIs, in conjunction with other data sets that Pulse Lab Jakarta has access to through data partnerships.
The prototype is modular. If you have ideas of how it can be improved, perhaps more specific for your country, please get in touch with us: firstname.lastname@example.org
Pulse Lab Jakarta is grateful for the generous support from the Government of Australia.