Guest essay by Michael Cochrane
Of the many important issues clamoring for the attention of world policy-makers and government officials, global climate change is among the most controversial. Its existence, likely effects on the environment, probable causes, and possible solutions have all been highly politicized. Many on the political left consider it the defining issue of our time, while many on the political right are highly critical of what they perceive as “junk” science undergirding the assertion of dangerous anthropogenic (human caused) global warming (AGW).
Policy solutions undertaken to deal with AGW, on the premise that it is dangerous (DAGW), will have massive monetary and non-monetary costs to society. Such high costs demand a correspondingly high degree of certainty regarding the likelihood of possible future scenarios. In particular, the scientific community must be able to show, with a very high level of confidence, that reducing or ceasing human activity associated with greenhouse gas emissions will stop or reverse global warming. If this is not possible, policymakers should assume that DAGW is essentially unstoppable and consider defensive measures (adaptation) as the most prudent policy approach.
One way to cut though the noise on climate change and help policy makers think clearly and rationally about governmental responses to the risks of global warming is to develop a risk or decision-scenario model. This can help senior policy makers explore alternative scenarios and decision outcomes before actually taking steps to commit national resources or enter into agreements with lasting economic effects.
Probability trees are valuable tools for modeling the potential outcomes of a succession of events. Each path along the tree from the origin to a terminal node—from root to tip of branches—describes one of potentially dozens of possible outcomes. Each node in the tree describes a possible event with two (or more) outcomes having a given probability of occurrence. A scenario is described by tracing a sequence of events along a path from the origin to a terminal node. The likelihood or probability associated with each path or scenario is found by multiplying the probabilities of all the event nodes along that path. The total probability for all possible paths described on the tree should sum to one.
Building a Climate Change Policy Decision Model
The following climate change policy model is designed to help us explore a range of scenarios associated with different answers to a sequence of questions fundamental to the issue of global warming and that must be addressed prior to proposing policy solutions. The questions are:
1. Is the earth actually warming? A summary of modern climate history suggests that from 1979 to the present there has been “a large disparity between surface thermometers, which show a fairly strong warming, and independent temperature readings of satellites and balloons, which show little warming trend” (Singer & Avery, 2007, p. xv).Rigorous data analysis (Heller, 2016) and application of statistical methods that control for heteroskedacity and autocorrelation (McKitrick & Vogelsang, 2014) suggest that apart from a step-wise change in global average temperature (GAT) around 1977 there has been no statistically significant warming trend since ~1958 or earlier.
2. If the earth is warming, is this actually a problem? There is some disagreement over this question, with global warming activists citing the potential for rising sea levels, more extreme weather events, famines and other catastrophes. Other scientists, however, argue that rising atmospheric CO2 levels and warming attributable to it may actually have net beneficial effects such as longer growing seasons, expanded growing ranges, increased plant growth, increased food production, and reduced morbidity and mortality from cold snaps. (Davis, 2005) (Singer & Avery, 2007)
3. If the earth is warming, and this is a problem, to what extent is human activity causing the warming? This question gets at the heart of the issue. Many environmental activists think human activity (i.e., burning of fossil fuels and other activities that generate CO2 and other so-called “greenhouse gases”) is the primary cause of global warming, while others believe warming is largely or wholly a natural, cyclical phenomenon primarily caused by solar cycles, ocean current cycles, and the eccentricities of the earth’s axial tilt and orbit. (Singer & Avery, 2007)
4. If human activity is the primary cause of global warming, will reducing this activity also reduce global warming? The assumption undergirding environmental policies such as the (now obsolete) Kyoto Protocol, the “Clean Power Plan” in the U.S., and the global climate agreement negotiated in Paris in late 2015 is that if anthropogenic CO2 causes warming, then reducing the output of anthropogenic CO2 will retard the warming trend. Many scientists reject the deterministic view of the relationship of anthropogenic CO2 and climate change assumed by such policies, arguing that there is a high degree of uncertainty associated with understanding the effects of changes in atmospheric CO2. (Posmentier & Soon, 2005)
5. If human activity is not the primary cause of global warming, is it still possible to stop it? Posing this question acknowledges the existence and the problem of climate change, but forces consideration of alternative solutions.
Each of the five questions is modeled in the probability tree (Figure 1) by an event node with two possible outcomes. There are six unique paths through the tree resulting in a total of six outcomes. Four of the outcomes require some type of policy response or proposed solution. Table 1 lists the model outcomes along with a suggested high-level policy approach or solution strategy.
The next step in preparing the probability tree model is to assign probabilities to each of the five events in the tree. This might best done by eliciting estimates from subject matter experts, who should be able to ground their estimates in an understanding of the current state of knowledge in their field and offer persuasive empirical evidence for any hypotheses or theories underlying their judgments. However, the beauty of a decision model is that one can still learn a great deal about appropriate policy responses by experimenting with a range of different probabilities and observing the degree to which the outcome values are sensitive to changes in the event probabilities.
Figure 2 shows the probabilities associated with the first two questions. Although satellite and radiosonde data show no statistically significant warming over the last twenty years, and the possibility that warming since ~1960 is limited largely to a stepwise rise around 1977 with no long-term trend either before or after (McKitrick & Vogelsang, 2014), much of the scientific community argues that the earth has generally been warming over the last several decades. We have, therefore, elected to assign a probability of 0.9—near certainty—that the earth is warming.
Although there is not a broad consensus regarding the degree to which global warming will be a problem for humanity, and such palaeoclimatologists and palaeoecologists as Hubert H. Lamb presented evidence that warmer periods were healthier for all kinds of life on Earth, including humans (Lamb, 1965), the prevailing opinion among environmentalists and climate activists today seems to be that, on balance, a warmer environment would be a net negative outcome for the world, so again, to be conservative in our estimates, we have assigned a probability of 0.9 that global warming is a problem. Figure 2 now shows events one and two in sequence, with their associated assumed probabilities.
The issue of whether global warming is blamed on what is called anthropogenic carbon dioxide forcing (i.e., human activities such as burning of fossil fuels) seems to be more of a political than a scientific one. Most scientists appear to acknowledge that human-based greenhouse gas emissions contribute something to climate change, but the key question is whether anthropogenic CO2 forcing is the primary cause. There is debate over whether there is a strong scientific consensus on this question. However, for purposes of this initial modeling effort, we will give the benefit of the doubt to anthropogenic CO2 forcing and assign it a high probability of 0.9. Figure 3 shows the first three events in the probability tree.
If human activity is the primary cause of global warming, the next to last event models the question, “Does reducing the level of human induced greenhouse gas emissions retard or reverse the advance of global warming?” If human activity is not the primary cause of global warming, the final event models the question, “Is it still possible to stop global warming?” Figure 4 illustrates how we would model these questions in our decision tree.
We assign a probability of 0.5 to event four to reflect the level of uncertainty and indifference in the scientific and policy community about whether this is actually possible. We will eventually perform a sensitivity analysis on this probability distribution.
If the primary causes of global warming have nothing to do with human activity, it is highly likely we will be unable to intervene effectively in the complex climate system to reverse a warming trend. Therefore we assign a very low probability of 0.1 to event five.
In the completed model, we multiply the probabilities for each event along each of the six paths through the tree. The resulting probability distribution at the six triangular terminal nodes reflects the relative significance of each of the outcomes.
Interpreting Model Results
Because we have already associated likely policy solutions to four of the six model outcomes, we can now associate the relative probabilities (or likelihood of occurrence) to the policy responses. In Table 2, we can see that, although current policy approaches and defensive measures both score the same if human activity is the primary cause of global warming, defensive measures are also an important response to the relatively small probability that we cannot stop global warming (given that it is not caused by human activity). Consequently, we combine these two scores for the Defensive Solutions policy response for a probability of 0.44. The most appropriate policy solution given this particular model would be those that direct national resources toward measures to defend the country against potentially devastating effects of climate change, e.g., rising sea levels or more frequent or severe droughts.
 The compound event probability distribution generated by the probability tree is based on the assumption that each of the event probabilities following the initial one is a conditional probability based on the prior event outcome.
The probability scores for the top two policy responses are fairly close. So one of the first questions we must ask is, “how would the event probabilities in the model have to change for these scores to be equal?” Since the model is formulated in Microsoft© Excel™ we can use the “goal seek” function in the data tools menu to explore this possible outcome. We set the terminal probability score for “current policy approaches” to 0.4 and let the goal seek algorithm find the corresponding probability distribution for event four (reducing human activity). The model outcome “reducing human activity fixes global warming problem” must have an assigned probability of 0.55 or higher for current policy approaches to equal or exceed the scores for defensive solutions (assuming all other probabilities in the model are held constant). This is clearly visible in Figure 6, which shows that the these two general policy approaches are highly sensitive to changes in the probability that reducing human carbon dioxide emitting activities also reduces global warming.
Figure 7 shows the degree of sensitivity of the two major policy approaches to changes in the probabilities for event five. There is no point across the range of probabilities for this event at which the probability score for defensive solutions drops below that of current policy approaches. This is because the defensive solutions policy is also the appropriate response for the outcome in which global warming cannot be reversed given that human activity is not a primary contributor.
Figure 8 addresses model sensitivity to the probability associated with event three: whether human activity is the primary cause of global warming. As with event five, the model is insensitive to changes in event probability across the range of possible probability settings (again assuming all other event probabilities are held constant).
Varying the probabilities of events one and two has no bearing on the relative movement of the outcome probabilities scores we have been observing. This is because any change in these probabilities is applied uniformly across all the event probabilities downstream from these nodes.
The degree to which reducing anthropogenic CO2 forcing correspondingly retards global warming is the major driver of policy solutions. Our model shows that it does not matter whether human activity is actually the primary cause of global warming (90% probability). The currently advocated global environmental policies exemplified by the Kyoto Protocol, Cap and Trade laws, and the recently negotiated Paris climate agreement assume a deterministic relationship between global warming and anthropogenic CO2 forcing that operates in both directions. In other words, if humans caused it by their activity, they can “un-cause” it by reducing or ceasing that activity. This may not be the case. Our modeling effort suggests that we must be at least 55% certain that reducing human activity will cause a corresponding reduction in global warming before we even consider the current regulatory policies advocated by the environmental movement.
It should also not be overlooked that, even assuming a model heavily weighted toward problematic global warming caused by human activity, there is still an almost twenty percent probability that warming either is not happening or won’t be a significant problem if it is.
The cost of policy solutions undertaken to deal with the threat of climate change will be massive and will entail both monetary and non-monetary costs to society. Any benefit to society from such policies must be weighed against those costs. Not wanting to overly complicate an initial policy model, we intend to explore such a cost/benefit analysis in future research.
The likely high costs of climate change mitigation policies demand a correspondingly high degree of certainty regarding the likelihood of possible future scenarios. In particular, the scientific community must be able to show with a very high level of confidence, that reducing or ceasing human activity associated with greenhouse gas emissions will reverse global warming. If this is not possible, policymakers should assume that global warming is essentially unstoppable and consider defense measures as the most prudent policy approach.
Michael Cochrane, Ph.D., Engineering Management and Systems Engineering, is Founder of Value Function Analytics, a consulting firm that helps clients achieve their objectives by helping them to think about values. Also a writer with World News Group, he has expertise in statistical modeling and analysis.
Davis, R.E. (2005). Climate Change and Human Health. In P.J. Michaels, Shattered Consensus: The True State of Global Warming (pp. 183–209). Lanham, MD: Rowman & Littlefield.
Heller, T. (2016). Evaluating the Integrity of Official Climate Records. Paper presented at the annual meeting of Doctors for Disaster Preparedness, online at http://realclimatescience.com/wp-content/uploads/2016/07/Evaluating-The-Integrity-Of-Official-Climate-Records-4.pdf.
McKitrick, R.R., & Vogelsang, T.J. (2014). HAC robust trend comparisons among climate series with possible level shifts. Environmetrics 25(7): 528–547.
Posmentier, E.S., & Soon, W. (2005). Limitations of Computer Predictions of the Effects of Carbon Dioxide on Global Climate. In P.J. Michaels, Shattered Consensus: The True State of Global Warming (pp. 241–281). Lanham, MD: Rowman & Littlefield.
Singer, S.F., & Avery, D.T. (2007). Unstoppable Global Warming: Every 1,500 Years. Lanham, MD: Rowman & Littlefield.