If we are to avert the worst of climate change impacts, we need better tools for identifying patterns of displacement and migration around climate extremes. In vulnerable developing countries, increasingly frequent and intense storms will likely exacerbate current patterns of displacement and permanent migration. Displacement often leads to humanitarian crises in the short term and can derail progress toward development in the long term. Because of this dangerous potential, displaced persons and migrants are a common focus in humanitarian responses. However disaster responders must often “fly blind” without the benefit of current, accurate information about the worst-affected populations. To better respond to the impacts of climate change as they unfold, we will need more rapid, cost-effective, and accurate methods for identifying patterns of displacement and migration.

Our knowledge of climate change is based on very large data sets that help distinguish historically “normal” patterns in Earth’s climate system from new patterns, such as rates of glacier recession, intensity of tropical cyclones, and persistence of drought. This notion that we can establish what is normal and then detect anomalies can be applied to the human impacts of climate change as well. To do so, however, we need large-scale data on human behaviors around weather extremes.

Mobile phone networks, which can provide detailed information about subscribers’ behavior, including how people move and spend money, are one of the most promising avenues for such empirical evidence. As we demonstrate in two articles for Climatic Change and Global Environmental Change, when subscribers use a mobile network they provide an excellent timeline of behavior that researchers can use to distinguish normal behavior from unusual behavior, like migration, as they react to climate change impacts.

Bangladesh Reacts to Cyclone Mahasen

Bangladesh’s low-lying southern coast is frequently cited as among the front lines of climate change. With roughly 30 million people living just above sea level, cyclones and extreme flooding have produced some of the most deadly disasters in human history. At the same time, roughly 89 percent of households have a mobile phone. When Cyclone Mahasen struck this area in May 2013, it became an ideal test case to study changing patterns of migration around weather extremes. We used anonymized data from 5.1 million Grameenphone subscribers in Barisal Division and Chittagong District to identify anomalous patterns of mobility before, during, and after the cyclone.

While we originally anticipated finding evidence of mass displacement in coastal areas in the weeks following the storm, we saw no evidence that this occurred. We did detect substantial anomalous patterns of mobility around the time that early warning messages were broadcast and during the storm’s landfall.

In Chittagong City, where the storm was originally forecast to make landfall, our findings suggest that widespread evacuation occurred the day before Mahasen made landfall. We found greater movement than normal throughout the Chittagong District as the storm approached. The national early warning system appears to have worked.

However in rural Barisal, a vulnerable district where heavy losses of human life have been concentrated in past cyclones, the analysis paints a different picture. We observed more people than normal moving from place to place as the most intense wind and rainfall moved into the area. Exactly why people were moving during the storm is unclear, but if the storm had been more powerful, more people may have perished. Fortunately, Cyclone Mahasen was relatively weak, resulting in far fewer fatalities than other recent storms. Nevertheless, this analysis shows where Bangladesh’s early warning system was effective in Chittagong, and identifies the places where it could improve in Barisal.

We also investigated long-term migration trends following the cyclone by comparing migration flows the year before the cyclone (2012) to the year the cyclone struck (2013). While we observed massive yearly flows of migration, contrary to expectations, the comparison provided no evidence of permanent migration due to Mahasen, but rather showed patterns that were nearly identical from year to year. In other words, while some people may have moved differently in the short-term, the population of Bangladesh appears to be resilient to the effects of low magnitude cyclones.

A Repeatable Method

Taken together, our analysis establishes a method for identifying anomalous patterns of evacuation, displacement, and long-term migration across very large spaces that can be repeated in nearly any context on Earth. Mobile phone usage rates are rising steadily in the developing world. As of 2015, there were roughly 4.7 billion subscribers to mobile phone networks worldwide – well over half the Earth’s population – and subscriber rates are increasing most rapidly in climate-vulnerable, low-income countries.

While our method should not replace current approaches to measuring cyclone impacts, it can serve as a complement to reduce the cost and improve the timeliness and accuracy of assessments. One of the keys to successfully reducing humanitarian crises is shortening the time between disaster and the arrival of first responders. The uptake of this method as part of official disaster preparedness and response policy could help shorten response times to deliver help to the people in greatest need.