Changing Climate, More Damaging Weather

The weather varies, but climate change affects the frequencies with which particular weather occurs, including the frequencies of extreme weather, such as heavy storms, heat waves, and droughts. More frequent weather extremes will underlie the most serious physical and economic effects of climate change. Prudent programs to adapt to current and future climate change must take these changing probabilities into account when making risk assessments and devising adaptation measures.

Federal agencies and other bodies charged with estimating the probabilities of such extreme weather events have been deriving their estimates from historical frequency data that are assumed to reflect future probabilities as well. These estimates have not yet adequately factored in the effects of past and future climate change, despite strong evidence of a changing climate. They have relied on historical data stretching back as far as 50 or 100 years, which may be increasingly unrepresentative of future conditions. As a result, the risks of damage from climate change based on these estimates may be badly understated.

These backward-looking probability estimates may be understating future frequencies and risks, and this might affect assessments of possible adaptive investments. Our analysis shows that government and private organizations that use these probability assessments in designing programs and projects with long expected lifetimes may be investing too little to make existing and newly constructed infrastructure resistant to the effects of changing climate. New investments designed to historical risk standards may suffer excess damages and poor returns. We need to accelerate research programs that link climate change to future probabilities of extreme weather events and, despite remaining uncertainties, embody the findings in estimates disseminated to the public.

Sources of error

Over the past half-century, temperatures and precipitation in the United States have gradually increased, more of the precipitation has fallen in heavy storms, sea level and sea surface temperatures have risen, and other aspects of climate have also changed. A scientific consensus agrees that such changes will continue for many decades, whatever reductions of greenhouse gas emissions are achieved. But these gradual changes are not the most threatening. Organisms and ecosystems can tolerate a range of weather conditions, and buildings and infrastructure are designed to do so as well. Within this range of tolerance, weather variability causes little damage, and if change is sufficiently gradual, many systems can adapt or be adapted.

When weather varies outside this range of tolerance, however, damages increase very disproportionately. As floodwaters rise, damages are minimal as long as the levees hold, but when levees are overtopped, damages can be catastrophic. If roofs are constructed to withstand 80 mile per hour (mph) winds, a storm bringing 70 mph winds might damage only a few shingles, but if winds rose to 100 mph, roofs might come off and entire structures be destroyed. Plants can withstand a dry spell with little loss of yield, but a prolonged drought will destroy the entire crop. The most alarming risks of damage from climate change arise from an increasing likelihood of such extreme weather events, not from a gradual change in average conditions.

Unfortunately, even if weather conditions do not become more volatile as climate changes, which might happen, a shift in average conditions will also change the probability of weather events that are far removed from average conditions. For example, as more rain falls in heavy storms, the probability rises that deluges will occasionally occur that result in extreme flooding and disastrous damages. As average temperatures rise, the likelihood of an extreme heat wave rises too.

Weather-frequency estimates have yet not come to grips with the changing probabilities of extreme weather. The methodologies in use typically are backward-looking and conservative. The frequencies with which specific weather events occur are estimated from measurements in the historical record going back decades. These frequencies are then used to “fit” to the data a probability distribution with a similar mean, variance, and skewedness. The probability distribution can then be used to estimate the likelihood of extreme weather, even though there are few, if any, such events in the historical record.

Estimating the probability of extreme, very infrequent, weather events in this way is inherently difficult, because there are so few such events in the measured record. Extrapolating from the occurrence of rarely observed events to the probability of even more extreme events beyond the historical record is unavoidably uncertain.

When climate is changing, an even more serious problem lies in assuming that the future will be like the past and projecting probabilities estimated from historical data into the future. Not only are agencies charged with assessing weather distributions assuming that the estimated probability distributions are stationary, they are also ignoring measured trends in historical weather patterns.

They do so for two main reasons. The first is uncertainty about whether an apparent trend is real and long-lasting, a poorly understood cyclical phenomenon that will be reversed, or a string of random events. The second is the dilemma of giving more weight to recent observations, which might better represent current conditions but would provide less data with which to estimate a probability distribution representative of extreme and unlikely events.

Uncertainty about future climate conditions affecting particular localities and weather phenomena is the main reason why weather probability assessments estimates are still based on historical data, despite strong scientific and empirical evidence that the future will not be like the past. Conservative agencies retain methodologies and estimates that are likely to be erroneous rather than make use of scientific projections of future conditions that are still quite uncertain, especially at a regional or local geographic scale. The question bedeviling weather forecasters is “If the future will not be like the past, what will it be like?” Climate models are still unable to provide highly reliable answers to this question.

Nonetheless, weather probability estimates become increasingly outdated as time passes or when projected further into the future. They provide unreliable guidance for the design, placement, and construction of infrastructure that will be in place for many decades and vulnerable to extreme weather throughout its useful life. By producing underestimated future risks, they also provide unreliable guidance for investment and program decisions to make existing infrastructure and communities more resistant to extreme weather. As a result, according to the 2009 National Research Council report Informing Decisions in a Changing Climate, “Government agencies, private organizations, and individuals whose futures will be affected by climate change are unprepared, both conceptually and practically, for meeting the challenges and opportunities it presents. Many of their usual practices and decision rules—for building bridges, implementing zoning rules, using private motor vehicles, and so on—assume a stationary climate—a continuation of past climatic conditions, including similar patterns of variation and the same probabilities of extreme events. That assumption, fundamental to the ways people and organizations make their choices, is no longer valid.”

This is a problem of broad and significant scope. Among the public- and private-sector organizations that are exposed to increasing but underestimated risks are:

• Local, state, and federal disaster management agencies
• Local, state, and federal agencies that finance and build public infrastructure in vulnerable areas as well as those that own and operate vulnerable infrastructure
• Private investors and owners of vulnerable buildings and other physical property
• Property and casualty insurers
• Creditors holding vulnerable infrastructure directly or indirectly as collateral
• Vulnerable businesses and households

Clearly, this listing encompasses a large proportion of the U.S. economy, and the vulnerable regions extend over a large part of the country, including coastal regions subject to hurricanes, storm surges, and erosion; river basins subject to flooding; and agricultural areas subject to wind, storm, and drought damage.

These underestimated risks should not be neglected in any program of adaptation to climate change. Research is under way to address this problem but should be accelerated, and efforts to improve climate change forecasts at regional and local scales should be intensified. More emphasis should be placed on forecasts of the likelihood of extreme weather events. Even while these efforts are under way, however, agencies responsible for weather probability assessment should update their estimates, incorporating the best available scientific climate projections that provide guidance regarding future conditions. Uncertainties in these projected weather frequencies should be frankly acknowledged and explained. In addition to their best estimates, agencies should also present plausible uncertainty bands around those probabilities. Finally, critical agencies should be encouraged or directed to use these revised probability estimates in their risk assessments and investment planning as an important step toward anticipatory adaptation to climate change.

Devastating hurricane hits New York City

To put this statistical analysis in concrete terms, imagine the impact of a major hurricane hitting New York City. The New York metropolitan region extends across three states and encompasses an extraordinarily dense concentration of infrastructure, physical assets, and business activity. The value of just the insured coastal property in the New York, Connecticut, and New Jersey region was almost $3 trillion in 2006. A major hurricane reaching New York could produce storm surges of 18 to 24 feet. Low-lying regions, including Kennedy Airport and lower Manhattan, would flood. Subway and tunnel entrances would be submerged, as would many essential roads. High winds would do severe damage, partly by blowing dangerous debris through city streets. The risk is real enough that the city government in 2008 created the New York City Panel on Climate Change and the Climate Change Adaptation Task Force to develop a strategy for preparing for extreme weather events.

We have prepared a case study to illustrate to what extent hurricane probabilities may be underestimated, how economic damage risks may consequently also be underestimated, how these risk assessments can be updated and projected into the future based on relevant scientific information, and how these updated risk assessments might be used to improve decisions on investments in adaptation. The analysis is relevant not only to all coastal areas vulnerable to hurricanes but also to inland areas susceptible to floods, droughts, and other extreme weather.

The starting point is the probability assessment carried out by the National Hurricane Center (NHC), an office within the National Oceanic and Atmospheric Administration. The methodology used for New York City and other coastal regions counts the occurrence of hurricanes of specific intensities (defined in terms of maximum sustained wind speeds) striking within a 75-mile radius during the historical record of approximately 100 years. NHC scientists fitted a particular probability distribution, the Weibull distribution, to these observed frequencies, and the probabilities of hurricanes of various intensities were then calculated from the fitted probability distribution. There were no actual observations of the most severe hurricanes in the historical record for the New York region, so those probabilities were extrapolations based on the fitted distribution. The results, expressed as the expected return periods, are shown in Table I for various categories of hurricanes.

These probability estimates were constructed in 1999. It is questionable whether these estimates were valid in that year, because there has apparently been an upward trend in intense hurricanes in the North Atlantic over at least the past 35 years. The number of category 4 and 5 hurricanes in the North Atlantic increased from 16 during the period 1975–1989 to 25 from 1990–2004. Consequently, the earlier years in the historical record used to compute frequencies might not have been representative of the final years.

There is good reason to believe that this increasing frequency of stronger hurricanes in the North Atlantic is linked to climate change through the gradual rise in sea surface temperatures. Warming ocean waters provide the energy from which more intense hurricanes are developed and sustained. According to a recent study, a 3° Celsius (3°C) increase in sea surface temperature would raise maximum hurricane wind speeds by 15 to 20%.

Measurements throughout the oceans have found a rising trend in sea surface temperatures at a rate of approximately 0.14°C per decade. The rate of warming is apparently increasing, however, and the North Atlantic warming has been faster than the global average. According to a recent examination, in the 28-year period from 1981 to 2009, warming in the North Atlantic has averaged 0.264°C per decade, roughly twice the global average. Rising sea surface temperatures in the North Atlantic, the driving force behind the increasing frequency of intense hurricanes, explain why backward-looking historical probability estimates, such as those generated with the NHC’s approach, probably do not provide adequate guidance with respect to current and future risks.

This problem is compounded by the rising trend in sea level, itself partly the result of increasing ocean temperatures. Higher sea levels and tides raise the probability of flooding driven by hurricane-force winds. In the North Atlantic between New York and North Carolina, sea level has also risen more rapidly than the global average, at rates between 0.24 and 0.44 centimeters per decade.

These scientific findings and measurements can be used to project hurricane frequency estimates into the future. The trend in sea surface temperature, linked to the relationship between sea surface temperature and maximum wind speed, provides a way to forecast changes in the intensity of future hurricanes. High and low estimates can define a range of future probabilities. Though there are considerable uncertainties inherent in forecasts based on this approach, the results are arguably more useful than static estimates based on historical data that fail to incorporate any relevant information about the effects of climate change. At a minimum, this approach can provide a quantitative sensitivity analysis indicating by how much existing estimates may be underestimating future risks.

Table 2 displays some results, based on both the higher and lower estimates of sea surface temperature trends and the relationship between sea surface temperature and maximum wind speeds. The table shows the estimated return periods for hurricanes striking the region, based on the 1999 Weibull distribution estimated by the NHC return periods for the New York metropolitan region. (Figures differ slightly from those in Table 1 for less intense storms because of curve-fitting variances.) In addition, it presents return periods for 2010, 2020, and 2030, estimated by indexing the scale parameter of the probability distribution to a time trend based on the rate of temperature change and its effect on maximum wind speeds. The ranges shown for the decades from 2010 to 2030 are based on the high and low estimates of the rate of sea surface temperature increase.

The effects of climate change will increase the probability that New York will be struck by a hurricane, especially the more severe hurricanes. By 2030, the probabilities of category 4 and category 5 hurricanes striking the New York metropolitan region are likely to have increased by as much as 25 and 30%, respectively. These changing probabilities have dramatic economic implications.

Paying the price

A professional risk management consultancy recently estimated that a category 3 hurricane with a landfall in the New York metropolitan region would probably result in losses of approximately $200 billion in property damage, business losses, and other effects. According to the NHC’s 2000 estimates, there is only a 1.5% chance of that happening in any year. However, this may be a very misleading portrayal of the economic risk.

A more complete assessment makes use of a tool common in the insurance industry: the loss exceedance curve, which represents the annual probability of a loss equal to or greater than specified amounts. It summarizes the probabilities of hurricanes of various intensities and estimates of the damages they would create. A loss exceedance probability of $200 billion represents the chance that a hurricane loss of that amount or more, into the trillions of dollars, might occur. To construct such a loss exceedance function for the New York region, one needs not only the probabilities of category 1 to 5 hurricanes but also the damages that they would inflict.

A recent study by Yale economics professor William Nordhaus, based on hurricanes recorded throughout the United States, investigated the relationship between maximum wind speeds and resulting damages. Shockingly, this study found that damages increase as the eighth power of the wind speed: If a hurricane with wind speeds of 50 mph would cause $10 billion in damages, then one with maximum winds of 100 mph would cause not twice the damages but more than $2.5 trillion. The reasons for this dramatic escalation are threefold. First, higher winds will obviously do more damage to everything in their path; second, more intense hurricanes are likely to affect a wider area; and third, their winds are likely to persist at damaging speeds, although not at the maximum, for longer periods of time.

The loss exceedance curve implied by this relationship is plotted in Figure 1 for the year 2000 and for subsequent decades, using the higher estimate of sea surface temperature increase. On the horizontal axis, damages are marked in hundreds of billions of dollars. On the vertical axis are the probabilities of hurricane losses of those amounts or more. One striking feature that is immediately apparent is that the exceedance curve is “fat-tailed”: Probabilities decline slowly as heavy losses mount. As maximum wind speeds increase, damages mount very rapidly, offsetting the declining probability of the more intense storms. The probability of losses exceeding a trillion dollars is not half the probability of losses exceeding $500 billion, but substantially more than that. This illustrates how vulnerable to catastrophic hurricane damage the New York metropolitan region is now.

The second feature that Figure 1 illustrates is that the probabilities of large losses shift upward over time, as climate change makes intense hurricanes more likely. By 2030, the probability of hurricane damages exceeding amounts in the range of $100 billion to $500 billion could be 30 to 50% greater than current estimates assume. Rising sea surface temperatures and rising sea levels increase the economic risks to coastal cities. In the absence of effective adaptation measures, the risks of catastrophic losses will very likely continue to rise over coming decades. If New York could insure itself against these catastrophic damages, the actuarially fair annual premium would double over this period.

Risks to investors

Investors in infrastructure projects vulnerable to hurricane damage, whether buildings, roads, or other structures, face greater risks than they realize and are likely to experience rates of return from their investments that are dramatically below those that they anticipate. Infrastructure projects are designed and engineered to withstand extreme weather, so that it would take an extremely unlikely event to cause major damage. There is a tradeoff between an extra margin of safety and the additional cost required to achieve it. Civil engineers and planners are trained to estimate and base decisions on such tradeoffs, often going beyond what is strictly required by building codes and other regulations.

Unfortunately, in assessing these tradeoffs, civil engineers and planners are still relying on historical frequency estimates and are making the same assumptions that the future will be like the past, despite climate change. Thought leaders in the engineering profession have only recently begun weighing alternative approaches to climate change issues.

An infrastructure project with a 40-year lifetime expected to earn a 12% return on investment, if historical risks persisted, would earn only an expected return of 3.9% because of the rising risks of damage. Moreover, if past frequencies of extreme weather events are projected into the future without taking into account the effects of climate change, the economic value of investments in adaptation and prevention is dramatically underestimated. Imagine that at an additional investment cost, the project can be strengthened to withstand an additional 10 mph of maximum wind speed without any additional damage. The payoff from this adaptation investment would be a lower risk of hurricane damage and a higher expected income return. Suppose further that such an investment in adaptation would just break even if the historical hurricane frequencies were projected into the future, over the project’s anticipated lifetime. Under these assumptions, adaptation would be considered uneconomic, since it would yield no positive return on investment.

If the effects of climate change were taken into account by anticipating the increasing probabilities of more extreme storms striking the region, the economic advantage of investing in adaptation and prevention would appear much more attractive. The supposedly break-even adaptation project would earn an expected 58% return on the investment. Because, with few exceptions, private investors and public agencies at local, state, and federal levels are still relying on static, historically based probability estimates of extreme weather events and haven’t yet incorporated the effects of climate change into these probability estimates when evaluating the economics of adaptation investments, these agencies are grossly underestimating the economic case for investments in adaptation. This is one of the reasons why adaptation has lagged and is proceeding so slowly.

Facing up to the future

Every year the United States is hit with hurricanes, floods, droughts, and other weather-related disasters such as wild-fires and pest outbreaks. These cause many billions of dollars in damages, loss of life, and disruption or displacement of entire communities. Some of these losses can be avoided if preventive and anticipatory actions are taken. If the risks of extreme weather events are underestimated, however, the pace and extent of preventive activities will lag.

Ignoring the effects of climate change on future probabilities of extreme weather events could significantly underestimate future risks to vulnerable communities, infrastructure, and investments. Deriving such probabilities from historical records going back many decades, with no adjustment for changes in climate extending inevitably into future decades, is likely to produce faulty estimates for planning and investment decisions. Climate change will affect the frequency with which many forms of extreme weather will occur.

The effects of climate change on weather and storm patterns are still uncertain, particularly at local and regional geographical scales. Uncertainty does not justify paralysis. It should be incorporated into estimates of future risks by establishing plausible ranges for key variables and parameters. Adhering to estimates almost certain to be wrong while waiting for uncertainties to be resolved provides misleading information for current decisions. The resulting decision errors can be very costly.

Public- and private-sector agencies responsible for providing estimates of weather risks are now grappling with the problems of incorporating the effects of climate change, but progress is slow and the bias is toward conservatism: sticking to the historical record until an alternative is clearly established. Moreover, much of the current research into this issue is narrowly focused and is not connected to adaptation program planning. Leadership in the responsible agencies is needed to ensure that their frequency estimates, to the extent now possible, reflect current and future probabilities, not past historical conditions, and that their estimates are frequently updated to incorporate new information about climate change effects.