The Rightful Place of Science
Page 4
Let’s start with the IPCC’s definition of “climate”:
Climate in a narrow sense is usually defined as the average weather, or more rigorously, as the statistical description in terms of the mean and variability of relevant quantities over a period of time ranging from months to thousands or millions of years.[49]
“Weather” refers to “the conditions of the atmosphere at a certain place and time with reference to temperature, pressure, humidity, wind, and other key parameters (meteorological elements).” Extreme weather events include phenomena such as heat waves, winter storms, tropical cyclones, floods, and so on. Weather events occur over minutes, hours, days, and perhaps even weeks.
The IPCC also defines what it means by “climate change”:
Climate change refers to a change in the state of the climate that can be identified (e.g., by using statistical tests) by changes in the mean and/or the variability of its properties, and that persists for an extended period, typically decades or longer.
It is important to understand that the IPCC definition of “climate change” makes no reference to the cause of the observed change. It is simply an identifiable change in the statistical properties of climate over a fairly long time period, which the IPCC identifies as “decades or longer.” Because many extreme events are—by definition—rare events, the time scale for the detection of change will necessarily be longer than that for variables which are measured more frequently, such as daily weather, precipitation, or sea level rise.
Even in the absence of detectable “climate change,” the occurrence of weather events varies on all time scales from seconds to millennia. The IPCC calls this “climate variability”:
Climate variability refers to variations in the mean state and other statistics (such as standard deviations, the occurrence of extremes, etc.) of the climate at all spatial and temporal scales beyond that of individual weather events.[50]
The presence of climate variability is one of the most significant obstacles that scientists must overcome in detecting a change in climate. For instance, there is a broad consensus that there has been greater hurricane activity in the North Atlantic since 1970. However, there is also a broad consensus that the increase since 1970 falls within the variability observed in North Atlantic hurricanes observed since 1900.[51] Thus, “climate change” as defined by the IPCC has not been detected with respect to hurricanes.
It is also important to understand that no discernible change in particular extreme weather events does not mean that they or other climate metrics are not changing. A change may be underway, but will not be detectible until sometime in the future. For example, as noted above, under recent model projections assuming that greenhouse gas emissions will influence North Atlantic hurricane behavior, a change in the statistics of hurricanes is expected under some model projections, but it would not be detectable for many decades. The magnitude of the ongoing changes is simply too small in the context of existing variability to be detected in the near future.
Once a change in climate is detected, scientists then ask another difficult question: why has the observed change occurred?
The answering of this question is called “attribution”:
Attribution of causes of climate change is the process of establishing the most likely causes for the detected change with some defined level of confidence.[52]
For example, with respect to North Atlantic hurricanes, the fact that there has not been detection of a change in the statistics of storms since 1900 means that there is not a climate change signal to be attributed. This stands in contrast to the robust detection of an increase in global average surface temperatures since the 19th century, which the IPCC attributes with high levels of certainty to human causes.
Even within the IPCC definitions there is considerable room for debate and different perspectives. For instance, Kerry Emanuel, a climate scientist at the Massachusetts Institute of Technology, recently discussed the challenges of understanding hurricane behavior in the North Atlantic since 1970:
In the Atlantic, demonstrably hurricane power has increased over the last 30 years by a big factor, too. I don't profess to understand that. It's gone up hand in hand with the tropical Atlantic surface temperature in the summer time. It's a tiny piece of the globe. And maybe some of that is global warming. I don't honestly know. I don't want to try to give you the illusion that I understand this.[53]
In the public discussion of climate change there are often two types of confusion that show up.
One is confusion between climate variability and climate change. For instance, by some measures global average surface temperature has slowed its rate of increase or even paused over the past decade or so. Some point to this as evidence that climate change, as measured by the warming of global average surface temperature, has stopped. While such a slowdown might give scientists good reasons to ask hard questions of climate model projections, it has been going on for too short a time period to understand what, if anything, it might be telling us about changes in climate, which is only discernible on longer time scales. The IPCC explains that climate variability “diminishes the relevance of trends over periods as short as 10–15 years for long-term climate change.”[54]
With respect to extreme events, such confusion is common, even down to the level of individual weather events or seasons, which are often cited in isolation as evidence for or against the role of human influences on the climate system. For instance, extreme winters are pointed to by some as evidence counter to theories of human-caused climate change, while individual hurricanes or tornadoes are used as evidence of human-caused climate change. Of course, all of this is imbued with a heavy overlay of politics. The fact is that the shorter the time period, the less relevant it is to understanding longer-term climate change and its causes.
A second common confusion is to conflate detection and attribution. Observing a change in a weather variable is not the same thing as associating that change with a particular cause, such as changes in climate resulting from greenhouse gas emissions. In popular discourse this distinction is frequently lost, with trends in any variable automatically assumed to be caused by emissions of greenhouse gases. This is most clearly represented in the common usage of the phrase “climate change” as synonymous with “human-caused climate change,” despite the IPCC’s broader definition.
This second type of confusion can further be seen when the phrase “climate change” is used itself as a causal factor. For example, whenever there is an extreme event there are inevitably many media stories that ask “did climate change cause X?”[55]
This question is inherently nonsensical. “Climate change” is not a causal actor. It is a statistical property that reflects the consequences of causes.
Imagine a baseball player who steps up to the plate and knocks a pitch out of the park. Home run! It so happens that this season his batting average is an impressive 0.320, after a sub-par performance of 0.220 last year.
Would we say, “The home run was caused by batting-average change”? Of course not. That would be circular and empty.
His batting average is a measure of change in his hitting. That measure is not a reason why he hits better, but a description of that change. Maybe he practiced more, had laser surgery on his eyes, or is taking performance-enhancing drugs. Or maybe he is just lucky. These might all be causal explanations for his improved hitting. “Batting average change” is not.
Unfortunately, much discussion of climate change is also circular and empty in exactly the same manner. “Climate change” no more causes weather events than changes in batting averages cause home runs.
In the climate debate both types of confusion often occur simultaneously, with extreme events routinely associated with greenhouse gas emissions regardless of whether or not detection or attribution has been achieved. Sometimes this reflects clever political expediency. Sometimes it just reflects confusion. In either case, it is simply wrong.
The IPCC definitions and its fra
mework for detection and attribution overviewed here provide a logical approach to addressing the central question of this short volume:
Have disasters become more costly because of human-caused climate change?
In order for research to show that disasters have become more costly because of human-caused climate change, several criteria must be met.
There must be a detectable increase in either the frequency or intensity of weather events, on climate time scales, which are associated with the disasters.
The detected increase in frequency or intensity must be attributed to human causes, typically defined narrowly in terms of greenhouse gas emissions, but other causes are also possible.
The following page shows a flowchart which illustrates the necessary and sufficient conditions for the detection and attribution of a role for human-caused climate change in increasing disaster losses. The flow chart also incorporates the concept of “normalization” which was introduced in the previous chapter.
The next chapters will survey a selected number of studies focused at the global level, and then several case studies focused on individual phenomena, and will consider detection and attribution.
Figure 3.1: The Steps Necessary and Sufficient to Achieve Detection and Attribution, under the IPCC Framework, of a Role for Human-Caused Climate Change
4
A Global Perspective on Disasters and Climate Change
The IPCC offers a useful definition of a disaster:
Severe alterations in the normal functioning of a community or a society due to hazardous physical events interacting with vulnerable social conditions, leading to widespread adverse human, material, economic, or environmental effects that require immediate emergency response to satisfy critical human needs and that may require external support for recovery.[56]
For present purposes, there are two parts of this definition that are important to highlight.
First, a disaster results from the intersection of a physical event and a vulnerable society. This implies that to understand changes in weather-related disasters over time, we need to focus on both how physical events may have changed and how societal vulnerability may have changed. Such understandings are not just of academic interest, but they also can help to shape our thinking about policy options to better prepare for and respond to future extreme events.
A second important part of the definition is its recognition that the effects of a disaster can be measured in many different dimensions—such as loss of life, property damage, loss of social capital, or environmental impacts. The focus of this volume is on economic losses, which primarily occur through damage to physical property, both public and private, as a consequence of extreme weather events. The reason for this focus is not that other types of losses are unimportant, but because the capacity to measure economic losses with some precision gives us a better chance of determining empirically whether human-caused climate change has made disasters worse, which is the central focus of this volume.
The main extreme weather events which cause property damage are wind storms and floods. According to data kept by Munich Re, from 1980 to 2013 almost 80% of all events causing losses worldwide (as measured above a certain threshold) came from wind storms (tropical cyclones, winter storms, thunderstorms) and floods.[57] If we can understand what is driving increasing losses with respect to wind storms and floods, then we will have gone a long way to understanding what is driving disaster trends overall.
So far, there have been five peer-reviewed studies at the global scale looking to disentangle social and climate factors which may underlie loss trends. Four of these studies examined the Munich Re dataset and the fifth looked at a different dataset. All five reach consistent conclusions, despite using different approaches in their analyses.
The studies are:
(1) A 2006 expert workshop that I helped to organize with Munich Reinsurance;
(2) The 2008 RMS study that was supposed to be the source for the 2007 IPCC “mystery graph”;
(3) A 2011 study funded by Munich Re and conducted by scholars at the London School of Economics;
(4) A 2014 study that I collaborated on, which looked at the Munich Re global dataset in the context of studies of other more localized datasets; and
(5) A 2014 study which looked at a Belgian research center’s data on disaster losses, focusing on economic losses as well as other metrics.
Let’s briefly consider each study in turn.
1. The 2006 Hohenkammer Workshop
In 2006, as a contribution to the work of the IPCC Fourth Assessment Report (AR4), I helped to organize a workshop, sponsored by Munich Re and research bodies in the U.S., United Kingdom, and Germany. The workshop included experts from around the world, and sought to assess the literature and reach consensus on the state of the science on disasters and climate change.
We reached a number of unanimous conclusions, among them:
· Analyses of long-term records of disaster losses indicate that societal change and economic development are the principal factors responsible for the documented increasing losses to date.
· Because of issues related to data quality, the stochastic nature of extreme event impacts, length of time series, and various societal factors present in the disaster loss record, it is still not possible to determine the portion of the increase in damages that might be attributed to climate change due to greenhouse gas (GHG) emissions.
· In the near future the quantitative link (attribution) of trends in storm and flood losses to climate changes related to GHG emissions is unlikely to be answered unequivocally.[58]
We subsequently published these conclusions in Science in 2007.[59] As we saw in Chapter 1, the workshop results were selectively mis-cited by the IPCC AR4 report. In fact, the workshop arrived at conclusions exactly the opposite to the claims advanced by the IPCC.
2. The “Mystery Graph” Study
Another examination of the Munich Re dataset can be found in the denouement to the “mystery graph” episode from the 2007 IPCC AR4 report, which I discussed previously. Robert Muir-Wood and colleagues at RMS looked at the Munich Re data from 1980 to 2005 and reached the following conclusions when their paper was finally published in 2008:
[T]he large portion of the rising loss trend is explained by increases in values and exposure as well as by an increasing comprehensiveness of reporting global losses through time.
With respect to a climate signal in the loss record, they reached the following conclusions:
In sum, we found limited statistical evidence of an upward trend in normalized losses from 1970 through 2005 and insufficient evidence to claim a firm link between global warming and disaster losses. Our findings are highly sensitive to recent U.S. hurricane losses, large China flood losses, and interregional wealth differences. When these factors are accounted for, evidence for an upward trend and the relationship between losses and temperature weakens or disappears entirely.[60]
One interesting aspect of this study was that its final year of analysis was 2005, which saw Hurricane Katrina and its massive economic impacts. With the presence of a big loss year at the end of the dataset, it would enhance any trend. Even so, the authors find little or no evidence of a relationship of increasing global temperatures (“global warming”) and increasing disaster losses.
3. The Munich Re LSE Project
Soon thereafter Munich Re provided financial support to the London School of Economics for a large research project to re-examine the same dataset. The scholars at LSE applied two methods to their normalization, looking at trends in losses since 1980.
They concluded in a paper published in 2011:
Independently of the method used, we find no significant upward trend in normalized disaster loss. This holds true whether we include all disasters or take out the ones unlikely to be affected by a changing climate. It also holds true if we step away from a global analysis and look at specific regions or step away from pooling all disaster types and look at spe
cific types of disaster instead or combine these two sets of dis-aggregated analysis.
Much caution is required in correctly interpreting these findings. What the results tell us is that, based on historical data, there is no evidence so far that climate change has increased the normalized economic loss from natural disasters. More cannot be inferred from the data.[61]
In particular, they caution against using loss data, even after it is normalized, to reach conclusions about how specific types of weather events may or may not be changing. The general lesson, one widely accepted and as discussed above, is that if you want to look for changes in the frequency or intensity of extreme weather phenomena, it is always best to look at data on extreme weather phenomena.
4. Trend Reconciliation
In 2014 Shalini Mohleji of the American Meteorological Society (who had previously been a student of mine) and I published a paper which attempted to close the circle on this research by disaggregating the Munich Re dataset into its component parts, organized by phenomena and region. We then compared the trends in the disaggregated data from the Munich Re data set with independent analyses of losses for specific phenomena in various regions. In many cases data are available for specific phenomena that go much further back in time than 1980. For instance, disaster loss data on U.S. hurricanes go back to 1900. We wanted to assess the consistency between the Munich Re data and the broader literature.
We concluded:
To sum, at the regional level, analyses of normalized damage over time periods longer than but encompassing the data covered by Munich Re’s dataset, show no evidence of an anthropogenic climate change signal in economic loss trends for phenomena which account for 97% of the documented increase in losses 1980-2008.[62]