Climate and social scientists study past conditions to establish links between environmental conditions and human behavior and health. Once these patterns are identified, the interest may shift toward understanding what they imply about the evolution of our health outcomes as conditions change in the future.
Of course, studying future outcomes is no easy task, as it’s impossible to make concrete observations of phenomena that haven’t happened yet! Still, we often have enough information to make informed judgments about what is likely to happen in the future.
This is the first in a series of posts that will discuss some of the complexities involved when attempting to anticipate future conditions. In this post, we’ll establish the foundational terminology that will be critical to understanding the concepts we introduce going forward.
Scenarios, predictions, forecasts: What’s the difference?
Climate and social scientists use several different terms when discussing possible future outcomes. Some of the most common are scenarios, forecasts, and predictions.
Scenarios attempt to describe the possible ways a system could evolve over time. These often include expert opinion on what time steps, thresholds, or drivers of different model output should be presented, but are still based on quantitative or numeric models.
Forecasts are usually short-term, numeric model estimates of physical phenomena which can have accuracy assessments embedded in them.
Predictions are used to describe estimates of future changes in the context of more complicated political, social, or biophysical changes. The term is often used as if it were synonymous with both scenarios and forecasts, but it typically emphasizes expert opinion.
While these terms refer to related or shared ideas, they have distinct meanings and will be used in different ways in the upcoming posts in this series. In the rest of this post, we’ll elaborate on these definitions and and tie these terms to real world examples.
Term | Dictionary Definition | Scientific Use | Example |
---|---|---|---|
Scenario | An imagined or projected sequence of events, especially any of several detailed plans or possibilities | Used to challenge assumptions, explore potential futures, and for strategic planning and risk management | Scenarios of different potential outcomes from the results of an election or policy change |
Forecast | To estimate or conjecture the course of events or future condition of things based on present indications | A quantitative estimate of a future event or condition based on past and current data, using statistical models and trends | The weather forecast is for warm and dry conditions for the next week. |
Prediction | The action of foretelling future events, a prophecy | Prediction often includes expert judgement and estimation of a future event or outcome. Can be based on information and expertise but does not necessarily include models or data. | Economists are predicting higher inflation due to unprecedented changes in trade policy. |
Scenarios
Scenarios are the result of thought experiments or models that incorporate information from a small subset of a multitude of events that could occur. A scenario is a postulated sequence of events, which considers the possible evolution of the events under consideration. For instance, a scenario may consider the likely actions or decisions made by governments or individuals within specific countries or organizations.
Notably, scenarios do not affirm that these actions or decisions will actually happen. Rather, they intend to illustrate the potential ramifications of risks and uncertainties and highlight the possible effects of government decisions and policy. Thus, scenarios can help decision makers think through the potential consequences of their decisions and consider alternatives that may improve outcomes.
For example, Lloyd’s of London is a market regulator for insurance providers and those that buy and sell insured risk in a global marketplace. Lloyd’s maintains a set of Realistic Disaster Scenarios (RDS) to stress test both individual syndicates and the market as a whole. The disasters within these scenarios include climate events (e.g., earthquakes, floods, and windstorms in very high value markets in the US, Japan, UK, and Europe) as well as terrorism and cyber attack events. If an earthquake occurs in the San Francisco area, for instance, Lloyd’s has estimated the losses for residential and commercial property insurance to be around $80 billion. The scenarios provide ways to ensure the stability of the reinsurance market and to reduce exposure by any one risk holder.
The United States government has a sea level rise tool, which provides a set of scenarios for the likely amount of sea level rise that will be experienced in different coastal areas. The five scenarios consider a variety of processes that could influence sea level across a wide range of future warming conditions—specifically, the amount of greenhouse gases emitted due to economic activity. They are defined by target values of Global Mean Sea Level Rise in 2100. These scenarios draw a direct connection between policy and action on climate mitigation and the amount of area along the US coastline which will likely be inundated by 2100.
The Famine Early Warning System Network (FEWS NET) used scenarios to estimate future food insecurity four months into the future through analysis of rainfall, harvests, commodity prices, seasonal labor demand, government policy, international aid, and other factors (see Figure 1). The scenarios relied on analysis of the current situation, assumptions about future changes, and the likely response to events by various actors. Updated every month, the scenarios provided members in the network guidance on appropriate responses to local shocks, like drought or conflict.

Population scientists can employ scenarios when estimating future population counts as well. Typically, population scenarios are constructed by assuming constant, higher, or lower rates of core population processes, like fertility, migration, and mortality. Our World in Data has a tool for users to create population projections based on fertility scenarios drafted by the United Nations.
Forecasts
Unlike scenarios, which are estimates based on assumptions about context and response to events in the future, forecasts are model predictions based on current and historical data. Forecasts are always evaluated for their accuracy and predictive ability using quantitative metrics. For example, the accuracy of a weather forecast can be evaluated by comparing the prediction to the observation dataset, allowing the forecast users to know whether the forecast was able to predict events that occurred.
Subseasonal to seasonal (S2S) forecasts extend short-term weather forecasts from two weeks to two years into the future. While raw forecasts from S2S models are informative, calibration is necessary to correct systematic biases and quantify forecast uncertainty to facilitate applications such as scheduling agricultural activities or planning outdoor events. Most S2S forecasts can predict that it will be warm in summer and cool in winter, but being able to forecast significantly above or below average temperatures in a particular month is far more challenging.
Population scientists create population forecasts—predicted population counts—using various methods depending on the amount of input information available. The process of creating a population forecast is called population projection. There is some debate on whether population projections should be considered true forecasts, however, and most population scientists refer to their projections as “estimates” to showcase their professional caution.1,2
Predictions
Predictions result from the expert analysis of information that integrates knowledge from multiple disciplines. The term “prediction” has the connotation of being less rigorous as it focuses on expert knowledge and is based on flexible assumptions. Predictions may also consider hypothetical conditions such as policy, politics or economic changes.
Forecasts, in contrast, often represent events that are likely to occur regardless of near-term government or individual actions, and are checked for accuracy once future conditions have been observed. However, there often is no direct way to determine the accuracy of a prediction due to incomplete information on the system and the lack of a specific model used in its creation.
For example, a government may decide to invest in subsidized fertilizer for smallholders, but this may not affect the price of commodities in the market in the current year given the impacts of broader macroeconomic forces and immediate supply and demand. Forecasts of drought impacts on crop production in the current year can’t be influenced by agricultural policies made by governments, since those investments take time to affect productivity. However, experts could still make predictions about the likely long-term impact of the investments.
Looking ahead
This blog post is the start to a series covering scenarios, forecasting, and population projections and how these concepts can be applied to climate and child health research. The next post will explain population projections in more detail and later posts will provide technical examples of these methodologies and how they can be used together.