Disclaimer: Below is an excerpt from this week's edition of our newsletter, Predictions, which can be found here.

In the second issue of Predictions, we reviewed a research paper published by Perry World House at the University of Pennsylvania which explored making aggregated forecasting platforms more useful to policymakers. We attempted to take the paper’s line of inquiry even further, and ventured to slot forecasting into the policymaking process ourselves. But in light of recent events, we now think it’s wise to both take a step back and get more specific as we think about how to leverage forecasting in high-impact areas. 

Late last Friday night, Morocco was hit with a 6.8 magnitude earthquake near Marrakech that has taken nearly 3,000 lives and injured over 5,000. A 4.9 magnitude aftershock followed. Emergency responses have been held up due to geopolitical disagreements and alleged domestic political neglect, only worsening the tragedy. Damages beyond the loss of human life do not yet seem to be calculated. 

Over the same weekend, Mediterranean Storm Daniel hit Libya, bringing with it heavy rainfall and flooding that resulted in large-scale destruction. Included in that destruction were two critical dams, which after breaking, led to increased flooding resulting in the loss of homes. As of Wednesday, more than 5,000 have been presumed dead with over 6,000 people still missing.

In a single weekend, nearly 20,000 lives were tragically affected by natural disasters around the Mediterranean Sea. And the rate of natural disasters occuring only seems to be increasing as the climate warms. As forecasters, we believe to our core that at least some of this devastation could have been avoided. Foreseen. 

There are two areas of natural disasters where forecasting could be exceedingly useful (that we’ve identified).

Can We Predict Natural Disasters?

First, most obviously, you can try to forecast the actual occurrence of a natural disaster. This is what we’ll call event-based forecasting. And within event-based forecasting for natural disasters, there is a range of success and accuracy. 

Predicting weather-related natural disasters seems to be a much more far-along endeavor, likely because weather-related natural disasters are within the scope of general meteorological services i.e. the National Weather Service. For example, hurricane forecasts feed a wide breadth of data from satellites, ships, radar feeds etc. into several predictive models to determine “a hurricane’s future track and intensity.” 

Those results are then evaluated by human forecasters who “use their experience and judgment to decide how to use the individual and ensemble model guidance to produce the best possible forecast.” And while this system isn’t perfect, we can anecdotally recall successful hurricane evacuation warnings being heeded and saving lives. But these processes aren’t available everywhere: “If there would have been a normal operating meteorological service, [Libya] could have issued the warnings,” World Meteorological Organization head Petteri Taalas told reporters in Geneva. “The emergency management authorities would have been able to carry out the evacuation.”

Forecasting earthquakes, however, is a completely different exercise. In this article published by the United States Geological Survey titled Can you predict earthquakes?, the first word is “No.” The article goes on to say that “Neither the USGS nor any other scientists have ever predicted a major earthquake. We do not know how, and we do not expect to know how any time in the foreseeable future.” The best we can do at this current moment, according to the article, is calculate “the probability that a significant earthquake will occur (shown on our hazard mapping) in a specific area within a certain number of years.” For more on earthquake forecasting check out this article – it won’t make you more optimistic. 

Even the CFTC addressed the effects natural disasters are having on grain farmers and their ability to transport their grain in a statement to the Agricultural Advisory Committee this summer: “Shifting weather patterns and more extreme weather threaten crops and livestock via hurricane, wildfire, and drought. They are also making it more difficult to get these products to market. That’s why I’m very interested to hear about developments in Mississippi River barge traffic.”

So here we have high-impact events that involve the preservation of human life as well as hundreds of millions of dollars in recovery efforts and reconstruction. And sometimes damages exceed the millions, with the U. S Department of Agriculture announcing last year that “commodity and specialty crop producers impacted by natural disaster events in 2020 and 2021 will soon begin receiving emergency relief payments totaling approximately $6 billion through the Farm Service Agency’s (FSA) new Emergency Relief Program (ERP) to offset crop yield and value losses.” 

Preparation Forecasting for Natural Disasters

Insurance costs, governmental resilience, human lives, all rest of the responsiveness of emergency aid. But the ability to actually predict when these events will occur and to what scale is varied and region-dependent. So where can forecasting provide added value? We think, preparation. 

Because we cannot yet accurately predict when a natural disaster will occur, demand for commodities after such a disaster is often irregular, difficult to communicate, and changing. These inefficiencies can lead to an even larger loss of life, or exacerbated damages to infrastructure. And researchers have been working on this issue for some time.

In 2009, researchers from the School of Management, University of Science and Technology of China published a paper titled Forecasting demand of commodities after natural disasters. The authors used a combined approach of an autoregressive moving average model (ARMA) and an empirical mode decomposition model (EMD) to more accurately predict demand for commodities after a natural disaster. 

And just last year, researchers from Taiwan published a report titled Multi-Commodity distribution under uncertainty in disaster response phase: Model, solution method, and an empirical study which suggests creating a “network flow model” where the resources of relief centers are connected by transportation webs that allow for a more efficient delivery of commodities during natural disasters.

While these areas of research might not address the issue at hand precisely, they do direct our attention into the right direction. Demand forecasting is critical to alleviate the damages created by less-predictable natural disasters. This summer the United States launched the Office of Pandemic Preparedness and Response Policy which coordinates the government's response to potential pandemics, advances pandemic preparedness, oversees scientific efforts, and provides periodic reports to Congress. Why not create an office like this for natural disasters? We have FEMA but…

Columbia University published research suggesting the prediction markets could be one potential salve to changing opinions on climate change and directing more focus towards climate-related issues, but if we’re using prediction markets to get people to care we might yet be too late :/

Future of Natural Disaster Forecasting

But let’s end on a bit of optimism. 

Welthungerhilfe, a German private aid agency uses an approach to aid called Forecast-based Action which leverages “forecasting models that can predict natural disasters and their effects on the local population.” The organization implements Early Action Protocols to determine the level of funds to be released in the case of a particular natural disaster, where that money will be used, and how it will be distributed. 

And the International Federation of Red Cross and Red Crescent Societies (IFRC) use an approach called Forecasting-Based Financing to anticipate natural disasters and take action before those disasters occur. They have a helpful YouTube video that explains their approach.

While first-order thinking would lead one to consider how to better forecast the occurrence of natural disasters, second-order thinking will lead you to forecasting their effects. And that is where we think more attention should be focused. Forecasting commodity demand after a natural disaster is not just about humanitarian aid, either. These events can significantly impact commodity markets, currency exchange rates, and geopolitics more broadly (think of how the Arab world responded to the Turkey-Syria 7.8 earthquake in February, especially Saudi Arabia). 

We believe that in addition to using forecasting to predict events, we can use forecasting to anticipate demand (forecasts conditional on event outcomes) to increase responsiveness of aid during natural disasters and potentially other areas of emergency service as well. And we could add natural disasters as a top geopolitical risk in the coming years, and these actor-less events and the aid they require have just as much influence over the shape of geopolitical relationships. If you want us to write more about demand forecasting in the future, let us know!

And finally, one useful tool that you can use to take action in light of the recent tragedies in Morocco and Libya is Charity Navigator. Charity Navigator recently acquired Causeway, a charitable giving startup run by effective altruists, and is trying to increase the impact of, and participation in philanthropic/charitable giving.