The previous post mentioned the uncertainty involved in all climate projections. A discussion of that uncertainty might be useful. I’m not a climate scientist, but I’ve read a good deal. It’s a complex subject, and this post will be lengthy, but here’s how I understand the issues.
Sources of Uncertainty
First, the future is always uncertain. Nobody can precisely predict the future. But that doesn’t mean that people can’t estimate trends. A great deal of our economy depends on the fact that we can make reasonable estimates about the future, from how much gasoline the nation needs to have on hand to cover cover anticipated demand, to how much food we have to make for a party we’re hosting, to what time we need to leave for work in order to arrive on time.
Second, the climate is an immensely complex system. Our understanding of it is improving, but far from complete. Climate patterns are influenced by many seasonal or multi-year oscillations (like El Niño) that are poorly captured by climate models. Further, regional and local climate patterns are influenced by local topography that is also poorly captured by climate models (like the micro-climates of mountains, or the effects of the Great Lakes). Regional climate simulations suffer from these gaps in our knowledge.
Third, by definition climate involves periods 30 years or longer. It always represents an average of many years’ weather, and does not even try to capture the yearly variation that occurs. Thus, for any given year, the actual weather will almost never be the climatic average. Rather, each year will vary.
Fourth, future climate depends in large part on human behavior. At what rate do we continue to emit greenhouse gases? At what rate do we continue to cut down rainforests? Does population continue to grow, or does it level off, or even decline? In addition to policy, these trends depend on other factors. Will there be another global economic recession? If so, how will that affect funding to combat climate change? Will there be a global pandemic that reduces population? Will there be an unanticipated technological advance, or will anticipated advances fail to materialize? These factors cannot be predicted, and thus, their important effects on future climate cannot be predicted.
In the absence of the ability to make predictions, scientists develop scenarios describing many possible directions the future could take. Then they run their climate models for each scenario. But there is no way to choose which scenario is “right.”
Fifth, climate simulations are essentially large computer calculations in which the results for any given year depend on the results for the previous year, which depend on the results for the year before that, and so on. In any compounding calculation of this sort, small differences at the beginning result in large differences after many years. Climate calculations often extend to the end of this century or beyond. That’s a lot of compounding, and small differences in initial assumptions result in very large differences at the end.
And finally, perhaps you intuit that climate models integrate the effects of many subsystems that are themselves quite complex. What effects to clouds have? Does it matter if they are in the troposphere or the stratosphere? What effect does ocean ice have? What effect do trees have? The effects of these complex subsystems are represented in climate models by estimates. But estimates are themselves subject to error, and in many of the subsystems, gaps in the data increase the error. This error affects the results of the climate models. Because of the compounding discussed above, small differences have the potential to affect the results quite a lot.
For example, a recent report published in the Proceedings of the National Academy of Science suggests that estimates of United States emissions of methane might be 33-41% too low. Methane is a powerful greenhouse gas, so the difference would influence climate simulations and be compounded over the years. It’s too early to know if the report is correct, but it illustrates the uncertainty. (See the following post for more on this report.)
How Uncertainty Shows Up in Climate Studies
The climate scenarios for the Midwest discussed in the previous post are affected by all of these kinds of uncertainty. The report studied two main scenarios. The A2 scenario was developed to represent a future world in which greenhouse gas emissions continued to grow. The B1 scenario was developed to represent a future world in which the rate of GHG emission growth decreases, until by mid-century emissions begin to decrease worldwide.
These two scenarios were studied multiple times by different scientists, using different climate simulation models and techniques. The models differed slightly, the differences compounded, and the results varied.
The first graph at right exemplifies the results. The graph concerns the simulated change in temperature for the A2 scenario (high ghg emissions). Three date ranges are represented: 2021-2050, 2041-2070, and 2071-2099. For each date range, the graph presents the results of 15 simulations (the plus signs) and the mean of the simulations (the dot). Blue means winter, green spring, red summer, and orange fall.
Click on graph for larger view.
First, be sure to notice that even with all this uncertainty, none of the simulations show a decline in temperature. The simulations all show an increase.
Second, notice that for 2021-2050 the difference between the highest and lowest simulations is smaller than for the period from 2070 to 2099. The farther out we get, the more the results are influenced by small differences between simulations. This is conceptually attractive, as it seems logical that trends for the distant future should be more uncertain than those for the near future.
Third, notice that the results for some seasons are more scattered than for others. The results for summer seem to involve more uncertainty than do the other seasons.
And fourth, notice the relative size of the difference between the highest result and the lowest, and compare it to the mean change in temperature. For instance, in the 2021-2050 period, the highest simulated temperature increase for summer was about +5.2 degrees, while the lowest was about +1.5. This is a difference of 3.7 degrees, which is larger than the mean estimated increase.
The second graph at right shows similar data for changes in precipitation. (This graph shows results for 23 simulations.) The models disagree on whether precipitation will increase or decrease. The overall conclusion drawn in the report is that the change in precipitation is simulated to be small, whichever direction it goes.
Click on graph for a larger view.
Most climate change reports use a mean to represent their final “answer.” There is a rationale for this choice: a body of literature suggests that the judgment of a group is superior to that of an individual in making quantitative estimates. For instance, if you ask a group of people to estimate the weight of a person, many of the individual guesses are typically widely off the mark, but the mean of everybody’s guess is typically quite close. Some evidence suggests that a similar principal might hold for computer simulations.
So climate simulations are fraught with uncertainty. That does not mean that they should be thrown away, discounted, or ignored. These studies represent the best that can be done, and the alternative is to simply guess or make opinions based on bias. The simulations are certainly better than opinions based on ignorance and bias. However, they should be understood to involve a great deal of uncertainty. The future always does.
National Oceanic and Atmospheric Administration (2013). NOAA Technical Report NESDIS 142-4. Regional Climate Trends and Scenarios for the U.S. National Climate Assessment. Part 3. http://www.nesdis.noaa.gov/technical_reports/142_Climate_Scenarios.html. This report is one of several regional reports used to create a new National Climate Assessment report scheduled for publication in early 2014.
Miller Scot, Steven Wofsy, Anna Michalak, Eric Kort, Arlyn Andrews, Sebastien Biraud, Edward Dlugokencky, Janusz Elszkiewics, Marc Fischer, Greet Janssens-Maenhout, Ben Miller, John Miller, Stephen Montzka, Thomas Nehrkorn, & Colm Sweeney. 2013. “Antropogenic Emissions of Methane in the United States.” Proceedings of the National Academy of Science, 10.1073/pnas.1314392110, www.pnas.org/cgi/doi/10.1073/pnas.1314392110.
Surowiecki, James. (2004). The Wisdom of Crowds: Why the Many Are Smarter Than the Few and How Collective Wisdom Shapes Business, Economies, Societies, and Nations. New York: Doubleday, Anchor.