If Local Weather Was Our Only Indicator

Abstract

It is understandably difficult for individuals and communities to recognize the effect of gradual climate change when it occurs in the context of local weather patterns, which normally vary from year to year. This recognition difficulty delays discussion of the causes of climate change and forestalls adjustments in policy and action. In this article, the authors estimate the length of time it would take a majority of localities to simultaneously acknowledge climate change if the only source of information about climate change was local weather. They run computer simulations using U.S. weather station data from 1946 to 2005. Local weather is allowed to vary randomly around a constant mean (models assuming no climate change) or a rising mean (models assuming climate change). They run separate models for annual average temperature, annual maximum temperature, and annual variation in monthly precipitation, varying the definition of unusual weather from 0.5 to 2.5 standard deviations from the historic average and varying the number of consecutive years of unusual weather required to reject the belief of normal variation and accept the belief of climate change from 1 year to 5 years. When it is assumed that acknowledgment of climate change requires three consecutive years of weather, a full standard deviation or more above the historic mean, it requires, depending on which weather event is being modeled, an average of 21 years, 86 years, or 82 years for a majority of localities to be in a state of climate change belief.

Keywords

acknowledgment belief of normal variation climate change climate change belief delay experiential factors global climate trends local weather modeling perception precipitation predictors recognition simulation temperature unusual weather weather

It’s the same wherever you are: all weather, like politics, is local. Herein lies a major, but little discussed problem confronting those who want action on global warming: We can’t agree on whether to act, largely because we experience weather locally rather than globally.

London, 2009, para. 3

The human experience of nature is always a local phenomenon. Because human belief or disbelief is usually predicated on what people experience as real and tangible in their everyday lives, local conditions become powerful predictors of what people believe to be true about climate and global warming (Williams, 2001).

The role of local weather in beliefs about climate change is well documented. Kempton (1997) notes the difficulty that ordinary citizens have of differentiating between global climate trends and local weather patterns. Bostrom and Lashof (2006) observe that, over the past two decades, results from both national polls and in-depth studies repeatedly confirm that the general public routinely conflate weather and climate change. Leiserowitz (2006) and Marx et al. (2007) conclude that perception of climate change risk is more strongly driven by experiential factors than by rational analysis. Thus, for example, Rabe and Borick (2008) report results of a Virginia survey of climate change beliefs in which personal experience is the primary reason given by both climate change believers and nonbelievers for their beliefs. Zahran, Brody, Grover, and Vedlitz (2006) as well as Egan and Mullin (2009) find local weather the previous week significantly affecting survey responses about the existence of climate change. Li, Johnson, and Zaval (2011) find local weather the day of the survey significantly affected belief in climate change.

“Was last summer’s warm weather or this year’s unusual pattern of precipitation a result of climate change or just normal variation in weather?” The relatively slow pace of climate change and the large variations in local weather that we have come to see as normal are complicating the emergence and stability of belief in climate change (Hansen, 2010; National Aeronautics and Space Administration, 2005). It comes as no surprise, for example, that the climate change disbelievers in the Rabe and Borick (2008) study dismissed recent above-average temperatures as simply due to normal variation in weather.

Research Purpose

Because local weather is not a sensitive indicator of climate change, scientists and nonscientists have called on individuals to look beyond local weather and concentrate on regional or global changes in weather that more accurately point to climate change; they warn that complete dependence on local weather conditions will dangerously delay response to climate change (Abler, 2003; Kempton, 1997; Turner, 2011). However, what if, like politics, all climate is ultimately local? How long would it take under various decision rules for a majority of localities to simultaneously agree that climate change is occurring? That is the question that this article examines.

It is our purpose to estimate, under various decision rules, how long it would take to recognize that climate change is occurring if our only indicator was local weather, weather that from year to year randomly varies around a mean—a constant mean when climate change is not occurring, a rising mean when it is. We derive our estimates by means of computer modeling, making use of data on changes in local temperature and precipitation from U.S. weather stations.

This is not an attempt to reproduce actual decision-making patterns in the real world, where other factors in addition to local weather clearly affect conclusions about climate change. Rather it is an attempt to establish baseline estimates of time to majority belief in climate change under various requirements for belief. In reality, actual time to majority belief may be quicker than these baselines if the net effect of other influences tips belief toward climate change, or actual time to majority belief may be longer if the net effect of other influences shifts belief away from climate change.

Model Parameters

Although computer models of climate change are numerous, most model climate change itself and few model belief in climate change (Climate Change Science Program, 2008). Some that do are Gu et al. (1994), Bray and Shackley (2004), and Bleda and Shackley (2005).

Our computer simulation is distinct from others because it illustrates how just a single factor, local weather, may influence belief. It incorporates the normal variation in local weather that can, at least in the short term, mask climate change. Just how long local weather is likely to mask climate change, in other words, how long that short term is, will be the object of our inquiry.

One of the advantages of computer simulations is that they force the simulator to address assumptions that often go unrecognized in other types of analyses (Gilbert, 1997). Specifically, to create our models we identify

what local weather characteristics we thought local populations would be sensitive to (we chose three: annual mean temperature, annual high temperature, and annual variation in month-to-month precipitation);

how much of a change would be required in those weather characteristics for local populations to recognize that the weather was unusual (we choose a threshold point of one standard deviation above or below normal, but run models with other thresholds for comparison); and

how long must the unusual weather persist for local populations to dismiss the possibility that this was simply normal local variation (we chose three consecutive years, but again for comparison sake run models with shorter and longer requirements).

Weather Characteristics

We chose three different weather characteristics around which to build our models: average annual temperature, maximum annual temperature, and annual variation (as indicated by the standard deviation) in monthly precipitation. Consistent with the idea of global warming, the first two weather characteristics are temperature based. Our annual average temperature models assume people sense and react to a general warming of their locality. (“Hasn’t the weather seemed a little warmer this year?”) Our annual maximum temperature models assume people recognize and react to unusual weather extremes. (“It got so hot one day this summer . . . ”)

Our third weather characteristic is not temperature based. Climate change affects precipitation patterns. Rather than raising or lowering overall rainfall amounts, references to precipitation change cite increased variability in rainfall occurrences and amounts: more rain when it rains, but longer periods of little or no rain (Intergovernmental Panel on Climate Change [IPCC], 2001; Knapp et al., 2002; Madsen & Figdor, 2007). Our measure of annual variation in monthly precipitation seeks to capture this phenomenon. If rainfall is increasingly concentrated in certain times of the year, the variation in monthly precipitation amounts will increase.

We chose annual weather characteristics even though belief in climate change has been shown to respond to much shorter weather patterns—patterns as short as a week (Egan & Mullin, 2009; Zahran et al., 2006) or even a day (Li et al., 2011). We did this because beliefs based on such short time periods are likely to be particularly changeable given the normal fluctuation that occurs in local weather. Beliefs so quickly formed can be just as quickly changed and are unlikely to provide a basis for community consensus, perceived risk, and action. Furthermore, the use of annual weather statistics has a long tradition: Agricultural communities have long thought in terms of annual weather patterns (Vedwan & Rhoades, 2001; Wilber, 1881), weather services have long reported annual weather statistics (Spears, 1935; Ward, 1908), and climate change models have usually used annual weather characteristics (IPCC, 2001).

How aware of these annual weather conditions are members of a locality? While highest temperature all year along with extended dry periods punctuated by heavier than usual rains are probably noticed, annual average temperature is more difficult to gauge because any year will include both hot and cold spells over a variety of seasons. However, recognition of shifts in all three weather characteristics is aided by weather services reporting daily, monthly, and annual weather summaries and by practical clues such as changes in utility bill levels and increased or decreased health of local vegetation.

Magnitude and Duration of Unusual Weather

How different must local weather be for humans to notice the change? Because we are not particularly sensitive to small changes in weather, change must be large enough and last long enough for us to recognize that something is different (Williams, 2001). In our models, we operationalize this by requiring the following condition to be met before a locality moves from disbelief in climate change to belief: The weather must be a full standard deviation above the normal average, and it must be so for three consecutive years. We assume that deviations from normal averages of less than a full standard deviation are unlikely to be noticed and that even deviations of a standard deviation or more which last just 1 or 2 years followed by returns to normal weather are likely to be dismissed as just natural fluctuation in local weather.

How realistic is the assumption that weather just a single standard deviation from the historic mean will be noticed as unusual? While persons may recognize the change in temperature or precipitation patterns on their own (Griffiths & McIntyre, 1974; Schellen et al., 2008), as already noted, the reporting of weather statistics by weather services and media outlets along with utility and vegetation changes makes recognition of small weather shifts more likely. Nevertheless, because our selection of a 1 standard deviation threshold for unusual weather and a duration of 3 years, although seemingly realistic, are nevertheless arbitrary, we will for the sake of comparison also run models using smaller and larger thresholds and shorter and longer durations.

Models

We initially run simulations using six different models: two model annual average temperature, two model annual maximum temperature, and two model annual variation in monthly precipitation. For each pair of models, one assumes an underlying change in climate and the other does not. All models simulate weather in 100 units (localities) for as much as 5,000 cycles (years). Each year’s weather (average temperature, maximum temperature, or variation in precipitation) in each locality reflects random variation around a mean.¹ The random variation corresponds over time to a normal distribution.² In models that assume an underlying change in climate is occurring, the mean around which annual weather varies rises a small amount each year; in models assuming no climate change, the mean remains constant.

Figures 1 and 2 illustrate our modeling process as it looks for single localities. Both figures plot annual weather (e.g., annual average temperature) for a single locality for 50 years. The vertical axis indicates how far above or below the historic average was each year’s weather. In both figures, points falling within the shaded area represent annual weather within 1 standard deviation of the historic average. The points in both figures were randomly selected from normal distributions, normal distributions with identical standard deviations. In Figure 1, the means (represented by the heavy line) of the normal distributions from which each year’s weather value was selected remain constant and are equal to the historic average. In Figure 2, the mean of the normal distribution from which the weather value for Year 1 was selected is the historic average, but for each successive year, the mean of the normal distribution increases by a fixed amount over the mean from the previous year. Thus, the heavy line representing the means rises, and the annual weather values, although still being randomly drawn from a normal distribution, are generally higher in later years because they are being drawn from normal distributions with successively higher means.

Figure 1.

An example of random annual weather fluctuation around a constant average.

Figure 2.

An example of random annual weather fluctuation around a rising average.

All 100 localities begin in a condition of climate change disbelief. If a locality in a condition of disbelief experiences three consecutive years of weather a full standard deviation or more above the original mean (e.g., the locality portrayed in Figure 2 for Years 7, 8, and 9), the locality switches to a condition of climate change belief. If a locality in a condition of belief experiences three consecutive years of weather a full standard deviation or more below the original mean, it returns to a condition of disbelief. (Although the locality illustrated in Figure 1 would still be in its initial condition of disbelief, it does experience three consecutive years of weather a full standard deviation or more below the historic average in Years 41, 42, and 43.) A simulation continues for 5,000 years or until a majority of localities are simultaneously in a state of belief, whichever comes first.

Other than sharing the same mean, standard deviation, and, for climate change models, rate of change in the mean, weather in any locality is independent of weather in any other locality. Where a locality’s weather for a particular year falls within the normal distribution of possible values is entirely due to chance. Because of this, two runs of the same model are unlikely to reach majority belief at the same time. Therefore, we run each model 500 times and report for each model the average number of years it took to reach majority belief along with the minimum number of years, the maximum number of years, and the standard deviation in the number of years.

For each weather event (average temperature, maximum temperature, and variation in monthly precipitation), our models require an initial mean, a standard deviation,³ and, for models assuming climate change, an annual rate of change in the mean. We used parameters derived from data from the United States Historical Climatology Network (USHCN; 2009) on 1,218 weather stations in the continental United States.⁴ Initial means and standard deviations were based on annual average temperatures, annual maximum temperatures, and annual standard deviations in monthly precipitation for the years 1946 through 1975.⁵ These years represent a period not marked by sustained warming of global temperatures according to the National Oceanic and Atmospheric Administration (NOAA; 2009, “Temperature Trends” section, para. 1). For models assuming climate change, the annual rates of change in the mean annual average temperature, annual high temperature, and annual variation in monthly precipitation were the average annual rates of increase experienced by those same 1,218 weather stations, but for the years 1976 through 2005, a period described by NOAA (2009, “Temperature Trends” section, para. 1) as a time of increasing global temperature.⁶

Results

Parameters Derived From USHCN Data

Table 1 shows the average means, standard deviations, and rates of change for annual average temperature, annual high temperature, and annual monthly variation in precipitation for the 1,218 weather stations for 1946 through 1975 and for 1976 through 2005 as derived from the USHCN database.

Table 1.

Average Means, Standard Deviations, and Slopes for 1,218 Weather Stations for 1946-1975 and for 1976-2005.

	1946-1975	1976-2005
Weather event	Stable climate	Changing climate
Annual average temperature
M	52.246°F (11.248°C)	52.825°F (11.569°C)
SD	1.127°F (0.626°C)	1.320°F (0.733°C)
Slope (regressing average temperature on year)	−0.022 for °F (−0.012 for °C)	0.064 for °F (0.036 for °C)
Annual maximum temperature
M	87.173°F (30.652°C)	87.329°F (30.738°C)
SD	2.396°F (1.331°C)	2.573°F (1.429°C)
Slope (regressing maximum temperature on year)	−0.021 for °F (−0.012 for °C)	0.012 for °F (0.007 for °C)
Annual standard deviation in monthly precipitation
M	1.847″ (4.691 cm)	1.927″ (4.895 cm)
SD	0.551″ (1.400 cm)	0.574″ (1.458 cm)
Slope (regressing standard deviation of monthly precipitation on year)	0.002 for ″(0.005 for cm)	0.003 for ″(0.008 for cm)

Note: Italicized values are used as parameters in computer simulations.

The italicized values in Table 1 are used as parameters in our models. From 1946 through 1975, the USHCN weather stations had a mean annual average temperature of 52.246°F (11.248°C), a mean annual maximum temperature of 87.173°F (30.652°C), and a mean annual variation (standard deviation) in monthly precipitation of 1.847″ (4.691 cm). These values will be used as the constant means around which localities’ actual weather values fluctuate for models assuming no climate change and the starting means around which localities’ actual weather values fluctuate for models assuming climate change.

For models assuming climate change, the mean around which localities’ actual weather values fluctuate will increase by 0.064°F (0.036°C) per year for annual average temperature models, by 0.012°F (0.007°C) per year for annual maximum temperature models, and by 0.003″ (0.008 cm) per year for annual variation in monthly precipitation models. Each of these values from Table 1 represents the average of the 1,218 slopes calculated when the weather characteristic was regressed on year for each weather station for the years 1976 through 2005.

For all models, both those with and those without climate change, annual average temperatures will be assumed to fluctuate normally around the mean with a standard deviation of 1.127°F (0.626°C), annual maximum temperatures will be assumed to fluctuate normally around the mean with a standard deviation of 2.396°F (1.331°C), and annual variation in monthly precipitation will be assumed to fluctuate normally around the mean with a standard deviation of 0.551″ (1.400 cm). These values from Table 1 represent the averages of the standard deviations for the years 1946 through 1975 for the individual weather stations.

Thus, the parameters used for our models are an accurate description of no single weather station’s experiences. Rather, they represent a statistical average of the unique experiences of each of the 1,218 weather stations. In each version of our computer models, all 100 localities operate using identical parameters. Nevertheless, simulated weather at each of the 100 localities will differ because where within the normal distribution of weather possibilities each locality’s annual weather will fall depends on random chance.

The values not italicized in Table 1 are provided solely for comparison purposes. It is noteworthy that the differences in means and slopes between the 1946-1975 values and 1976-2005 values all point to warming temperatures and more varied monthly rain patterns during the latter period. Mean annual average temperature, mean annual maximum temperature, and mean annual standard deviation in monthly precipitation are all higher in the 1976-2005 period. The slopes when weather condition is regressed on year for the 1976-2005 period are all positive suggesting climbing temperatures and increasingly uneven monthly rainfalls, whereas the slopes for annual average temperature and annual maximum temperature were negative during the 1946-1975 period. The slope for annual standard deviation in monthly precipitation was positive for 1946-1975, but it becomes steeper for 1976-2005.⁷

Years to Majority Belief in Climate Change

With model parameters determined, simulations could then be run. For each of the weather conditions (annual average temperature, annual maximum temperature, and annual standard deviation in monthly precipitation), 500 simulations were initially run assuming no underlying climate change (a constant mean around which weather conditions fluctuate) and 500 simulations were run assuming underlying climate change (an increasing mean around which weather conditions fluctuate.) Each simulation generated weather conditions for 100 independent localities. All localities begin in a condition of disbelief about climate change. For these runs, three consecutive years of weather at least a full standard deviation above the historic mean are required for a locality to switch from disbelief to belief in climate change, and three consecutive years of weather at least a full standard deviation below the historic mean are required for a locality to switch from belief in climate change to disbelief. Each simulation runs for 5,000 years or until 51 of the 100 localities are simultaneously in a condition of belief. Table 2 shows the results of running each model 500 times.

Table 2.

Number of Years for a Majority of Localities to Simultaneously Believe in Climate Change.

Weather event	Models in which climate change not occurring (stable mean)	Models in which climate change occurring (increasing mean)
Annual mean temperature
M	470.61 years	21.34 years
Minimum	185	19
Maximum	1,217	24
SD	176.87	0.97
Annual maximum temperature
M	471.30 years	85.88 years
Minimum	198	68
Maximum	1,335	103
SD	183.42	6.35
Annual standard deviation in monthly precipitation
M	470.50 years	82.04 years
Minimum	178	67
Maximum	1,376	99
SD	180.91	5.70

Note: Threshold = 1 standard deviation, required duration = 3 years. Results based on 500 runs.

Could a majority of localities based on local weather conditions simultaneously believe in climate change when climate change is not in fact occurring? Yes, however, it is a very infrequent occurrence. While most of the simulations based on models without climate change did, in fact, reach a state where a majority of the localities believed in climate change even though the local weather conditions that caused the belief (three consecutive years of weather a full standard deviation or more above the normal mean) occurred entirely by chance, it took on average about 470 years. (The results of the three “no climate change” models in Table 2 appear very similar regardless of the weather event being modeled. This is not a coincidence. As long as models are using a constant mean and a constant standard deviation, the number of years it takes for majority belief in climate change becomes strictly a function of chance. The properties of a normal distribution remain the same regardless of what the values of the mean and standard deviation are.)

Climate change models reach majority belief much more quickly, but just how quickly that occurs depends on the weather event being modeled. For these models, the number of years to reach majority belief depends not only on chance, but also on the rate of climate change relative to the normal fluctuation in the weather condition. The larger the rate of climate change relative to normal fluctuation, the sooner does a majority in support of belief form.

The rate of increase in annual average temperature reported by the USHCN weather stations from 1976 to 2005 was large relative to the standard deviation in annual average temperature reported by those same stations from 1946 to 1975. As a result, it took the simulations an average of just 21.34 years to reach majority belief. At least one simulation reached majority belief after just 19 years, and all simulations reached majority belief by the 24th year.

For both annual maximum temperatures and annual standard deviations in monthly precipitation, the 1976 to 2005 rates of increase reported by the USHCN stations were small relative to the 1946 to 1975 standard deviation in those measures reported by the weather stations. As a result, it takes longer for the simulations basing belief on maximum temperatures and on variation in rainfall to reach majority belief: an average of 85.88 years in the case of maximum temperatures and an average of 82.04 years in the case of variation in rainfall. Some simulations reached majority belief in just 67 years, and all simulations reached majority belief by the 103rd year.

Alternate Magnitude and Duration Levels

What if localities are less sensitive to deviations from usual weather or, less likely, more sensitive? What if localities required longer durations of unusual weather in order to become convinced of climate change or, again less likely, shorter durations of unusual weather? Table 3 shows the mean number of years it took when climate change was not occurring for a majority of localities to simultaneously (and mistakenly) believe in climate change under five different thresholds ranging from 0.5 to 2.5 standard deviations from the historic mean and five different required durations ranging from 1 to 5 years. Each mean is again based on 500 runs of the model. As noted earlier, a single set of mean outcomes suffices for all three weather events because when the mean and standard deviation of the probability distribution remain stable, the probability of particular outcomes is simply a function of the normal distribution regardless of the actual mean and standard deviation.

Table 3.

All Weather Events: Mean Years to Majority Belief When Climate Change Is Not Occurring (Various Thresholds, Various Required Durations, Results Based on 500 Runs).

	Threshold for recognizing unusual weather
Required duration of unusual weather	0.5 SD	1.0 SD	1.5 SD	2.0 SD	2.5 SD
1 year	5.38	10.58	24.83	70.85	261.13
2 years	20.96	72.22	384.11	3074.96^a	5000.00^a
3 years	73.50	470.78	4555.27^a	5000.00^a	5000.00^a
4 years	248.96	2931.34^a	5000.00^a	5000.00^a	5000.00^a
5 years	807.13	5000.00^a	5000.00^a	5000.00^a	5000.00^a

Actual mean is higher, but some or all runs stopped after 5,000 years without reaching majority belief.

It can be observed from Table 3 that the likelihood of a majority of localities simultaneously believing in climate change is exceedingly unlikely unless one assumes localities are sensitive to changes as small as half a standard deviation from normal and that they require just one or two consecutive years of unusual weather to believe. We have been working with the assumption that localities require at least three consecutive years of weather at least a full standard deviation above the historic average to acknowledge climate change. If either of those assumptions is too low, then the average length of time before a majority of localities simultaneously believe in climate change when it is not occurring quickly increases to several thousand years.

Table 4 for annual mean temperature, Table 5 for annual high temperature, and Table 6 for annual variation in monthly precipitation each show the mean number of years it took when climate change was actually occurring for a majority of localities to simultaneously acknowledge it under those same five threshold levels and five required durations. Each mean is based on 500 runs of the model.

Table 4.

Annual Mean Temperature: Mean Years to Majority Belief When Climate Change Is Occurring (Various Thresholds, Various Required Durations, Results Based on 500 Runs).

	Threshold for recognizing unusual weather
Required duration of unusual weather	0.5 SD	1.0 SD	1.5 SD	2.0 SD	2.5 SD
1 year	3.36	5.22	8.39	13.74	21.28
2 years	7.87	13.29	21.10	29.77	38.54
3 years	13.47	21.34	30.06	38.84	47.66
4 years	18.86	27.36	36.19	44.99	53.78
5 years	23.29	32.05	40.86	49.71	58.42

Table 5.

Annual High Temperature: Mean Years to Majority Belief When Climate Change Is Occurring (Various Thresholds, Various Required Durations, Results Based on 500 Runs).

	Threshold for recognizing unusual weather
Required duration of unusual weather	0.5 SD	1.0 SD	1.5 SD	2.0 SD	2.5 SD
1 year	4.69	8.77	16.63	35.02	73.87
2 years	15.15	33.77	81.40	167.12	265.30
3 years	34.27	85.88	175.41	274.32	374.00
4 years	63.58	145.80	244.62	344.52	444.26
5 years	100.69	195.64	295.09	394.78	495.23

Table 6.

Annual Variation in Monthly Precipitation: Mean Years to Majority Belief When Climate Change Is Occurring (Various Thresholds, Various Required Durations, Results Based on 500 Runs).

	Threshold for recognizing unusual weather
Required duration of unusual weather	0.5 SD	1.0 SD	1.5 SD	2.0 SD	2.5 SD
1 year	4.51	8.38	16.84	34.01	71.92
2 years	14.87	33.05	78.32	157.30	248.23
3 years	33.14	82.04	165.03	257.19	348.37
4 years	61.29	138.50	229.42	320.31	412.54
5 years	95.93	183.45	275.50	367.22	458.58

These tables show that at every threshold/duration combination, the average number of years to majority belief is smallest for annual average temperature and greatest for annual high temperature. As noted earlier, this is because the annual increase in annual average temperature is large relative to the historic standard deviation in annual average temperature while the annual increases in annual high temperature and in annual monthly variation in precipitation are both small relative to their respective historic standard deviations.

Also, as expected, every increase in the minimum threshold for recognizing unusual weather extends the average length of time for a majority of localities to simultaneously believe in climate change as does every increase in the minimum duration of unusual weather required. With our original assumptions of a 1 standard deviation threshold and a minimum duration of three consecutive years, annual average temperature produces majority belief in approximately 21 years, annual high temperature in 86 years, and annual monthly variation in precipitation in 82 years. To achieve majority belief sooner requires what we would consider to be unrealistic assumptions that weather just one-half a standard deviation above normal is registered as unusual and just 1 or 2 years of unusual weather is enough to produce belief. If our assumptions of a 1 standard deviation threshold and three consecutive years duration are themselves unrealistically low, time to majority belief obviously increases.

Less obvious, however, is the rate at which years to majority belief changes as threshold and duration increase. When climate change is not occurring (Table 3), each increase in threshold or duration results in an exponential increase in years to majority belief. When climate change is occurring, however, as in Tables 4, 5, and 6, increases in threshold or duration do not produce exponential increase in years to majority belief. Increases in threshold or duration do increase the number of years to majority belief, but the increases are much smaller than when climate change is not occurring and even appear, when threshold and duration requirements are fairly high, to be approaching a constant or linear rate of increase. Thus, the probability that a majority belief in climate change will be caused by chance rather than by actual climate change (i.e., that a positive result will be a false positive) declines dramatically as the required threshold and duration for belief increase.

Discussion

We draw two conclusions from the results of our modeling. First, we are not only able to conclude that three consecutive years of exceptionally warm temperatures or exceptionally varied rainfalls are much more likely to occur under conditions of climate change than when no climate change is occurring, but we are also able to estimate just how much more likely. While a majority of localities might simultaneously mistake normal fluctuation in weather for climate change even in the absence of climate change, a sequence of three such exceptional years is likely to occur only once every 470 years or so. Under conditions of climate change, however, three consecutive years of weather a full standard deviation or more above the historic average will occur between 5 times sooner (in the case of annual high temperature and annual monthly variation in precipitation) and 20 times sooner (in the case of annual average temperature) than under conditions of no climate change. As the threshold for recognizing unusual weather and the duration of unusual weather needed to believe in climate change increase, majority belief takes longer, but the likelihood that chance occurrence is being mistaken for actual climate change lessens.

Second, even when climate change is occurring, however, majority recognition of its occurrence is far from immediate. This is because local weather normally fluctuates and because the annual rate of change in the underlying mean is small relative to the magnitude of normal fluctuation. Our climate change model in which belief is based on annual average temperature was the quickest to produce majority belief, but it still took an average of over 21 years. When climate change is occurring and belief is based on annual maximum temperatures or on variation in monthly temperatures, the simulations took an average of over 80 years to reach majority belief—longer than an average person’s lifetime. While majority belief can be reached sooner if the threshold at which weather is judged unusual is set lower and the duration unusual weather must last to cause belief is set shorter, such assumptions are unlikely to be realistic.

In fact, our estimates of time it takes to recognize climate change based solely on local weather conditions should be considered low-end estimates. Three assumptions built into the models make that so. First, we have assumed that weather a standard deviation above or below the normal average is recognized by the public as unusual. A 1 standard deviation difference in annual average temperature is a difference of 1.127° (0.626°C), a 1 standard deviation difference in annual maximum temperature is a difference of 2.396° (1.331°C), and a 1 standard deviation difference in annual standard deviation in monthly precipitation is a difference of 0.551″ (1.400 cm). If these thresholds are too low and it takes bigger deviations from normal to be noticeable or longer periods of unusual weather to generate belief, then, as we have shown, length of time to reach majority belief increases. A difference of 1.127° (0.626°C) in average annual temperature, the weather characteristic that most quickly produced majority belief in our climate change models, may be particularly difficult to notice or care about even with the assistance of weather services, public media, utility bills, and vegetation changes. By contrast, the warmest day of the year and occurrences of rain and snow are more memorable, but annual high temperature and annual variation in monthly precipitation take longer to produce three consecutive years of unusual weather.

Second, we have assumed that present weather is compared with historic averages. However, if people take as “normal” the weather they experience when they were young, then the percentage of a locality’s population recalling weather at a certain date in the past will continually decline. Even more problematic is the possibility that people take as normal the weather they experienced in the past 20 or so years. Furthermore, people’s notions of normal weather may be reset when they move to another locality if they assume whatever weather they experience in the first few years after their relocation is normal. All of these situations would mean what is perceived as normal would already incorporate some effect of past climate change, thereby lessening the likelihood of present weather being seen as unusual.

Third, by stopping our simulations when 51% of the localities are in a state of belief, we have implied that things happen when a majority of the localities simultaneously believe in climate change—things like regional, national, or international policy discussions that could lead to changes in regulations, legislation, and programs of action. While things such as these are themselves time-consuming, even the initiation of such events may require more than a simple majority recognizing climate change. Furthermore, the larger the majority of localities needing to be in a state of belief before supralocal responses to climate change occur, the longer it takes to reach that point.⁸

Even with these favorable assumptions built into our models, the results of our simulations demonstrate that waiting on local weather to make the case for global climate change is slow and leaves much to chance. Of course, it was never our intent to suggest that the case for global climate change be left to changes in local weather. Rather, we wanted to see how long it might take if local weather were our only indicator. Having seen how long under the best of reasonable assumptions it might take, we believe our findings give support to Bostrom and Lashof’s (2006) call for opinion leaders to dispel the “local weather equals global climate” model and to replace it with a more realistic mode of thought.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biographies

Robert F. Szafran received his PhD in sociology from the University of Wisconsin at Madison. His research interests include research design, demography, and computer simulations.

Contact: rszafran@sfasu.edu.

Jerry L. Williams received his PhD in sociology from Kansas State University. His research interests include environmental sociology, community development, and phenomenology.

Contact: jerry@williamspage.net.

Jeffery E. Roth received his PhD in geography from the University of Oklahoma. His research interests include historical and environmental geography.

Contact: rothjeffe@sfasu.edu.

References

Abler

(Ed.). (2003). Global change and local places: Estimating, understanding, and reducing greenhouse gases. Cambridge, UK: Cambridge University Press.

Bleda

Shackley

(2005). The formation of belief in climate change in business organisations: A dynamic simulation model (Tyndall Centre Working Paper 68). Norwich, England: Tyndall Centre for Climate Change Research at The University of East Anglia.

Bostrom

Lashof

(2006). Weather it’s climate change? In Moser

S. C.

Dilling

(Eds.), Creating a climate for change: Communicating climate change and facilitating social change (pp. 31-43). Cambridge, UK: Cambridge University Press.

Bray

Shackley

(2004). The social simulation of the public perceptions of weather events and their effect upon the development of belief in anthropogenic climate change. (Tyndall Centre Working Paper 58). Norwich, England: Tyndall Centre for Climate Change Research at The University of East Anglia.

Bader

D. C.

Covey

Gutowski

W. J.

Jr. Held

I. M.

Kunkel

K. E.

Miller

R. L.

. . . Zhang

M. H.

(2008). Climate models: An assessment of strengths and limitations (A report by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research). Washington, DC: Department of Energy, Office of Biological and Environmental Research.

Egan

P. J.

Mullin

(2009, April). How citizens integrate information without ideological cues: Local weather and Americans’ beliefs about global warming. Paper presented at the Midwest Political Science Association meetings, Chicago, IL.

Gilbert

(1997). A simulation of the structure of academic science. Sociological Research Online, 2. Retrieved from http://cress.soc.surrey.ac.uk/~scs1ng/ngpub/paper124_NG.pdf

Griffiths

I. D.

McIntyre

D. A.

(1974). Sensitivity to temporal variations in thermal conditions. Ergonomics, 17, 499-507.

Peiris

D. R.

Crawford

J. W.

NcNicol

J. W.

Marshall

Jefferies

R. A.

(1994, March). An application of belief networks to future crop production. Proceedings of the Tenth Conference on Artificial Intelligence for Applications, San Antonio, TX.

10.

Hansen

J. E.

(2010). Testimony delivered at the Ratcliffe Coal Trial. Retrieved from http://www.columbia.edu/~jeh1/mailings/2011/20110126_RatcliffeTestimony.pdf

11.

Intergovernmental Panel on Climate Change. (2001). Climate Change 2001: Synthesis Report, Third Assessment Report of the Intergovernmental Panel on Climate Change. New York, NY: Cambridge University Press.

12.

Kempton

(1997). How the public views climate change. Environment, 39, 12-21.

13.

Knapp

A. K.

Fay

P. A.

Blair

J. M.

Collins

S. L.

Smith

M. D.

Carlisle

J. D.

McCarron

J. K.

(2002). Rainfall variability, carbon cycling, and plant species diversity in a mesic grassland. Science, 298, 2202-2205.

14.

Leiserowitz

(2006). Climate change risk perception and policy preferences: The role of affect, imagery, and values. Climatic Change, 77, 45-72.

15.

Johnson

E. J.

Zaval

(2011). Local warming: Daily temperature change influences belief in global warming. Psychological Science, 22, 454-459.

16.

London

(2009, March 16). Global warming weather: All weather is local. Huffington Post. Retrieved from http://www.huffingtonpost.com/gabriel-london/global-warming-warning-al_b_175565.html

17.

Madsen

Figdor

(2007). When it rains, it pours: Global warming and the rising frequency of extreme precipitation in the United States. Retrieved from http://cdn.publicinterestnetwork.org/assets/oywshWAwZy-EXPsabQKd4A/When-It-Rains-It-Pours----US---WEB.pdf

18.

Marx

S. M.

Weber

E. U.

Orlove

B. S.

Leiserowitz

Krantz

D. H.

Roncoli

Phillips

(2007). Communication and mental processes: Experiential and analytic processing of uncertain climate information. Global Environmental Change, 17, 47-58.

19.

National Aeronautics and Space Administration. (2005). What’s the difference between weather and climate? NASA’s Earth-Sun Science News Team. Retrieved from http://www.nasa.gov/mission_pages/noaa-n/climate/climate_weather.html

20.

National Oceanic and Atmospheric Administration. (2009). State of the climate: Global analysis—Annual 2008. National Climatic Data Center. Retrieved from http://www.ncdc.noaa.gov/sotc/index.php?report=global&;year=2008&month=ann

21.

Rabe

Borick

(2008). Report of the Virginia climate change survey. Miller Center of Public Affairs. Retrieved from http://www.closup.umich.edu/research/funded/prg/climate-change-opinion-report1.pdf

22.

Schellen

Lichtenbelt

W. M.

Witt

Loomans

Frijns

Toftum

(2008). Thermal comfort, physiological responses and performance during exposure to a moderate temperature drift. Indoor Air, 2008, 17-22.

23.

Spears

P. B.

(1935). Deserts on the march. Norman: Oklahoma University Press.

24.

Turner

(2011, July 27). Memo to climate change pundits: Don’t mention the weather. Mother Nature Network. Retrieved from http://www.mnn.com/earth-matters/climate-weather/blogs/memo-to-climate-change-pundits-dont-mention-the-weather#

25.

United States Historical Climatology Network. (1996). U.S. historical climatology network description. Retrieved from http://cdiac.ornl.gov/r3d/ushcn/ushcn.html#TEXT

26.

United States Historical Climatology Network. (2009). United States historical climatology network (USHCN) Version 2 Serial Monthly Dataset. Retrieved from http://cdiac.ornl.gov/ftp/ushcn_v2_monthly/

27.

Vedwan

Rhoades

(2001). Climate change in the Western Himalayas of India: A study of local perception and response. Climate Research, 9, 109-117.

28.

Ward

R. D.

(1908). Climate. New York, NY: Putnam.

29.

Wilber

C. D.

(1881). The great valleys and prairies of Nebraska and the Northwest. Omaha, NE: Daily Republican Print.

30.

Williams

J. L.

(2001). The rise and decline of public interest in global warming: Toward a pragmatic conception of environmental problems. Huntington, NY: Nova Science.

31.

Zahran

Brody

S. D.

Grover

Vedlitz

(2006). Climate change vulnerability and policy support. Society & Natural Resources, 19, 771-789.