Flowminder has been awarded a grant from the American people through the United States Agency for International Development (USAID) to enhance and scale its capacity to inform disaster management by providing estimates of population mobility.
Natural disasters, such as earthquakes, floods or hurricanes, often trigger large population movements and internal displacements. Therefore, humanitarian and relief agencies need data on people’s locations and movements rapidly, once a country is hit and later throughout the recovery period, to effectively support those in need and host communities. Data on mobility are also needed routinely to estimate population exposure to different possible hazards and prepare accordingly. Call Detail Records (CDR data) is a dataset automatically generated by mobile network operators for billing purposes which can be processed and analysed to extract estimates of human mobility. The need for timely estimates for disaster management, often on a large scale, makes it the most relevant use for these mobile operator datasets.
We have received a grant from the American people, through the United States Agency for International Development (USAID), to develop an automated mobility in crisis information system, improve our methods for generating mobility information from CDRs, collect survey data to correct for biases in population representativity, and ultimately better support humanitarian actors with disaster preparedness and emergency response decision making.
Since our team pioneered the use of CDR data to estimate population displacement during the Haiti earthquake response in 2010, we have spent the last decade developing, implementing, and testing our methods to respond to a variety of humanitarian and development needs in a bespoke manner. From these learnings, we are now ready to scale, through standardisation and automation, and this funding is taking this work to the next level.
A lot of the specifications for this work have already been determined via past projects (including FlowKit development and release), research, and user requirements consultations with end users, governments, and development or humanitarian actors. Flagship engagements and activities include for example our work with IOM-DTM where we have specified requirements in understanding mobility in crisis especially in Haiti. We also conducted various engagements with government institutions such as during our work in 2020-2022 with the Mozambique Disaster Management Institute (INGD) where we gathered user needs on disaster-driven displacement estimation, with continuous support from, and discussions with, Dr Beleza, Deputy Director of the National Emergency Operations Centre at INGD; and more recently the work we are doing with the National Disaster Management Organisation (NADMO) in Ghana, where we are producing maps of population exposure to hazards, on average, per hour, to better plan for potential crises and disasters depending on location and time of the day, and day of the week.
With this ‘Population Mobility Data for Disaster Management’ project, we aim to improve the production and usability of population mobility and internal displacement estimates from Call Detail Records for humanitarian use. This development project is built around four main components of our mobility in crisis information system:
Indicators of mobility for disaster management preparedness, mitigation, response, recovery:Our team will develop new and enhance existing code and methods for the production of crisis mobility indicators (which require different methods than routine mobility indicators), specific to each phase of a crisis and component of disaster management. The specification and intended use of these indicators have already been discussed with decision makers but they may also be refined through consultations during the project.
Survey work and correction of population representation biases in CDR-derived mobility estimates:This core organisation-wide development focus aims at producing mobility estimates (and the resulting changes in population distribution over time) that represent the entire population of the country of interest, instead of representing solely the frequently active subscribers of the participating operator whose data are being processed. This work involves the collection of survey data and the development of a mathematical model to combine them with CDR-derived mobility, and static, population estimates. The funding for this project will enable this work – which we previously conducted in the Democratic Republic of the Congo (DRC) and Ghana – to also be done in Haiti.
Standard reports on population mobility to inform disaster management:We will design and develop information products targeted at humanitarian actors. These reports will contain summarised and easy-to-digest statistics and narratives on population mobility to inform each phase of disaster management (preparedness, forecast and warnings where applicable, response, recovery). Knowing which indicator to share when, and having standard report templates ready to be populated when the data is ready will help us to send the information rapidly.
Build and automate a system disseminating mobility reports from CDRs:Building on our FlowKit platform (our CDR data processing and analytical software installed on the operator’s server), we will expand our library of automated quality checks conducted on the pseudonymised CDRs, introducing alerts for data quality issues, and where possible fixing problems automatically. We will automate the computation of a list of mobility indicators that can vary depending on the phase of a crisis and update regularly (every month or every day depending on needs), correct all indicators for representation biases on an ongoing basis, populate report templates automatically and send them to a pre-decided list of recipients after a short manual review. Automation will help get reliable and consistent mobility information to end users in a timely manner, and allow us to respond to more events.
In addition to these major development activities, we are also planning a contingency plan and live crisis response, if a crisis were to occur in Haiti during the phase of the project, regardless of the state of the information system. Outputs from the ‘Population Mobility Data for Disaster Management’ project will be developed and tested in Haiti, where we will also conduct new user consultations. Future ambition is to then secure further funding to roll out the system in all countries where we have FlowKit installed.