The destruction of weather stations might seem like a footnote to a conflict that killed thousands and displaced more than 1 million people. But today, on Earth Day, as scientists work to understand changes to our climate and how those changes affect people’s lives, the loss has a particular sting.
In recently published research, we show the data loss in the Central African Republic is part of a wider trend. Violence and instability create gaps in knowledge that deprive affected communities — and scholars — of important information. Here is why this happens:
Weather might affect conflict — but how does conflict shape weather data?
There’s growing evidence that weather shocks, such as high temperatures and droughts, are associated with a higher risk of civil conflict — rebellions and insurgencies — particularly in vulnerable areas in Africa. Climate change will probably increase the frequency of these shocks, so scholars study this relationship to understand the implications for future conflicts.
But this research depends on having accurate weather data. We became concerned by the possibility that civil conflict could be affecting temperature observations — which, in turn, would affect scholars’ ability to study the relationship between climate and conflict.
How could this be?
Most temperature data come from readings taken at weather stations distributed unevenly worldwide. The tasks of establishing, staffing and maintaining these stations fall under the jurisdiction of national governments. This means unstable political conditions could influence the climate record in two ways.
First, the ability to establish and maintain weather stations could be related to a country’s governing and bureaucratic capacity, factors that also influence the risk of civil conflict.
Second, violence and instability could lead to the destruction of facilities, as was the case in the Central African Republic, or divert government resources from the collection of weather data.
Here is how we did our research.
Pursuing this hunch required looking beneath the hood of commonly used temperature data sets to learn about the actual weather station observations on which they are based. This meant poring over “metadata” — the data that describe the data — as well as tracking down leads with the World Meteorological Organization and the U.S. State Department.
The main goal was to develop a statistical analysis of the location and performance of weather stations. Along the way, we encountered anecdotes about paper records riddled with bullet holes or sold as wrapping paper at open-air fish markets.
Our results confirm that countries in sub-Saharan Africa that have experienced more years mired in civil conflicts have, on average, fewer weather stations contributing reports to their climate records. We also found that periods of civil violence lead to the failure of existing stations.
Why are closely spaced weather stations so important?
Why does this matter? If you follow climate news, you’ve seen colorful maps illustrating global temperatures. But we cannot actually measure temperature with a thermometer at every location on the earth’s surface — that would be impossible. Instead, a network of weather stations collects land temperatures. Scientists take these readings and fill in, or interpolate, the temperature in places without a weather station by using nearby observations.
Consider Palo Alto and Berkeley in California. Even though they are just 30 miles apart, the cities have different climates. You could drive from warm and sunny Palo Alto and arrive to find Berkeley cold and foggy. If it was a hotter-than-average month in Palo Alto, it was probably also a hotter-than-average month in Berkeley. So if you didn’t have a weather station in Berkeley, you could estimate its monthly temperature anomaly pretty accurately using data from Palo Alto.
The problem is that, as distance increases, the correlation goes down. Estimating the temperature anomaly in Berkeley from a weather station in Los Angeles would be less accurate.
This is why poor weather-station coverage can create data problems. As estimates of temperature in a given location rely on fewer and farther stations, more error creeps into the data. Areas that are most at risk of civil conflict tend to have more lapses and errors in weather data — and the problem is getting worse.
These errors present a challenge for those studying how temperature influences political and economic outcomes. Poorly measured variables often make it harder to find a relationship where one actually exists, meaning researchers could underestimate the effect of temperature on civil conflict risk.
By combining four temperature data sets in a way that is designed to reduce measurement error, our estimate of the effect of temperature on conflict risk in the 1960-2014 period almost doubles when compared with uncorrected estimates. The magnitude of the effect is relatively small, and researchers are still trying to understand the mechanisms at work, but the relationship between temperature and conflict is cause for worry in a warming world.
Technology might help alleviate the data gaps.
Climate scientists and meteorologists are aware of station loss, and there have been efforts to bolster capacity and recover data that might have been recorded but never passed on to international bodies. Even documenting the scale and consequences of data loss — the focus of one ongoing reporting project in Rwanda — is a crucial step.
These kinds of efforts are important to ensure those most vulnerable to the effects of global warming — including higher conflict risk — do not also lose their ability to document, understand and adapt to the changes taking place.
Kenneth Schultz is a professor of political science at Stanford University. His research seeks to understand the causes of international and civil conflict. Follow him on Twitter @kschultz3580.
Justin Mankin is an assistant professor of geography at Dartmouth College. As a climate scientist, he researches the causes and consequences of climate change and its variability. Follow him on Twitter @jsmankin.