CDR-derived indicators have applications at all stages of a crisis, from the warning period ahead of a forecasted crisis to the crisis itself and to the recovery after.

During and in the aftermath of a crisis, the humanitarian response relies on understanding how many people have been affected and where affected people are moving to. Mobile operator data, and Call Detail Records (CDRs) particularly, are valuable in these situations because they can help generate near-real-time insights into the number of people in impacted areas and where people from impacted areas have been displaced to.

Forecast and warning period

Some crises, such as hurricanes, can be forecast. We can use CDR-derived indicators to support decision-makers in the period between a crisis event being forecast and the event occurring, by providing insights into the population which may be directly exposed and the response of the population to warnings.

Forecast period: estimating number of people who may be exposed

When it is possible to forecast the areas likely to be impacted by a crisis, CDR data can be used to derive indicators quantifying the number of people who would usually be either residing in or visiting these areas.

CDR data are generated in real-time and can capture the variation in the number of people in a given area and therefore take into account daily, weekly and seasonal changes in the distribution of the population. This up-to-date, granular picture of the number of people who may be exposed, either as residents of an area or visitors to the area, has important implications for the response to a crisis, such as the number of people who may require evacuation and the number of people who may be displaced from their homes by the crisis.

Warning period: evaluating population movements triggered by disaster warnings

CDR-derived indicators can also provide insights into how the people in an affected area respond to warnings of a forecast crisis. By  describing the number of people travelling into and out from the likely affected areas after the warning, relative to prior, historical movements, we can help decision-makers understand how warnings are affecting population movements, and whether people are following official advice or instructions, such as evacuation orders.

Similarly, we can produce indicators describing travel activity within the warning area, such as numbers of locations visited by subscribers and trip distance [LINK] relative to the period prior to the warning. These can provide insights about the response to warnings which may have important implications for decision-makers.

These different indicators can help us understand whether people are leaving an area, planning to shelter in places or are continuing their routine activity. In Bangladesh, ahead of the 2013 Cyclone Mahasen, indicators derived from CDR data showed large increases in the movement of mobile phone subscribers in Chittagong District, where the cyclone was forecast to make landfall, as people evacuated. In comparison, mobility in Barisal decreased ahead of the cyclone indicating that people were suspending their regular trips but were sheltering in place rather than evacuating.

Active crisis

During an active crisis, CDR-derived indicators can help determine the impact of the crisis on the human population and also on key infrastructure. These indicators can provide important insights for decision-makers planning the response.

 

Observing changes in subscriber presence

Information on the short-term distribution of the population can help estimate the number of people present in the area at the time of the disaster, and who may have been directly impacted by the crisis and need immediate aid. Similarly, longer-term indicators describing the number of people residing in impacted areas may support early insights into the number of people who may be affected or displaced by the crisis.

It is important to remember that CDR data do not only show people affected by a disaster, and that other actors and factors may come into play. CDRs can also provide indications of more people coming into the area, to respond or report on the crisis. In some crisis scenarios, we can observe from the CDR data an increase in subscribers in the affected areas, which can be due to the arrival of people involved in the response, such as humanitarian workers or journalists. This influx may confound changes in subscriber presence, with less pronounced decreases in the number of subscribers in the area than would otherwise be the case.

For example, looking at data from the 2021 Haiti earthquake, we notice that the travel distance for subscribers in the departments of Haiti decreased sharply following the earthquake and during the tropical storm that also affected the area shortly after, before increasing to above normal levels as people responded to the crisis. 

 

Estimating the number of people internally displaced 

We can also use CDR data to estimate the  number of people relocating from affected areas, therefore giving  important indications on the number of internally displaced persons (IDPs), where they have been displaced from and where they have been displaced to. This can provide up-to-date, granular information on the number of people who may need aid and how they are geographically distributed.

The changes in the geographic distribution of the population described by these indicators can also be used to identify vulnerable areas in need of support, both those with large decreases in population as people have been displaced, and those with large increases in population that need support hosting IDPs.

Flowminder conducted similar analyses following natural disasters in a range of LMICs, including Bangladesh, the Democratic Republic of the Congo, Haiti and Nepal. For example, during the eruption of Mount Nyiragongo near the city of  Goma, in the north east of the Democratic Republic of the Congo in 2021, Flowminder researchers analysed the numbers of mobile phone subscribers present in the areas surrounding the volcano, and the movements of subscribers between locations. They were able to show that departures from the affected areas only truly began on 27 May, five days after the eruption and the same day as the evacuation order was given for Goma. You can read more about this work here.

 

Impact on infrastructure & network

Furthermore, we can analyse CDR data to estimate  the impact of a disaster on infrastructure, particularly telecommunications and transportation.

Indicators relating to the numbers of trips between locations, the numbers of locations visited by subscribers and the distances they are travelling can help assess the disruption of transportation by a crisis, which may include damage to transport infrastructure.

However, changes to travel patterns during times of crisis are not necessarily due to transport issues only. Different factors such as reluctance to leave home and go too far; elder/younger members of family inability to travel; presence of extended family in non-affected areas nearby; or lack of financial resources to travel far ‒ to only name a few factors ‒ could contribute to shorter distance travelled during the crisis. 

CDR data can also be used to assess whether there are telecommunications network outages which may affect communications in affected areas and the coordination of the response. However, these outages may also affect the quality of the data and impact or even prevent the generation of indicators for some affected areas. Similarly, changes in the use of mobile devices in response to the crisis, either increasing or decreasing their activity, may impact indicators derived from CDR data and need to be adjusted for.