Projections vs. Forecasts

Source: https://www.lacan.upc.edu/admoreWeb/2018/05/all-models-are-wrong-but-some-are-useful-george-e-p-box/

Buried somewhat in today's Salt Lake Tribune is a great interview by Robert Gehrke with University of Utah Professor Lindsay Keegan, an epidemiologist in the Department for Internal Medicine.  Dr. Keegan makes a number of important points, including one that I could relate to as a meteorologist.  She immediately stated that what she is doing in collaboration with others is creating scenario-based projections.  I put that in italics because there is a difference between a projection and a forecast and that distinction is important, not well recognized, and important for how we interpret model results.

Projections vs. Forecasts

Although they are sometimes used interchangeably, there is a difference between a forecast and a projection.  As noted by MacCracken (2001), a projection is a statement of what will happen in the future if certain conditions develop.  In the study of climate, projections are used extensively.  We use models to produce estimates of future climate under selected emissions pathways.  For example, the graph below illustrates multi-model average changes in global mean surface temperature under four future greenhouse-gas emissions scenarios, including "low" (blue) and "high" (red) scenarios.  These are projections rather than forecasts because each assumes a future pathway for key forcing agents.

Source: http://www.ipcc.ch/report/ar5/wg1/

In contrast, a forecast is a prediction of what will happen in the future based on what is known today.  MacCracken (2001) actually distinguishes between a prediction and a forecast, but we won't get into those semantics here.  In a forecast, one does a best estimate of present conditions (and possibly future conditions) and then forecasts what will happen in the future.  This what is done every day in the weather forecasting business, including by our forecast models.

In her interview, Dr. Keegan correctly emphasizes that they are doing projections because they are making assumptions about what interventions or choices we might take to address the virus.  This is often swept under the rug in news articles which simply state that "the model from <insert credible group here> shows this."

My mom lives in Florida and when there is a hurricane forecast to affect the region I tell her repeatedly to monitor official forecasts and heed the recommendations of local emergency management officials.  I tell her to ignore the tweet that says model X says you will fine or another that says model Y says you will be wiped out.  Your best course of action is to listen to people who can bring their expertise to bear on the problem and do so in a collaborative fashion after examining all of the evidence and models.

Similarly, it's probably best if we non-experts don't read too much into the results of what Dr. Keegan describes as "public facing models."  Let's focus on what those in the know tell us to do.