GLOBAL CHANGE | Climate Record

GLOBAL CHANGE | Climate Record

GLOBAL CHANGE Contents Climate Record: Surface Temperature Trends Sea Level Change Upper Atmospheric Change Biospheric Impacts and Feedbacks Climate...

1MB Sizes 0 Downloads 23 Views

GLOBAL CHANGE

Contents Climate Record: Surface Temperature Trends Sea Level Change Upper Atmospheric Change Biospheric Impacts and Feedbacks

Climate Record: Surface Temperature Trends PD Jones, Climatic Research Unit, University of East Anglia, Norwich, UK Ó 2015 Elsevier Ltd. All rights reserved.

Synopsis Surface temperatures have risen by about 0.6  C during the twentieth century. This article addresses the quality of the basic temperature data over the terrestrial and marine domains. The warming during the century has not occurred in a linear fashion but in two periods, from about 1920 to 1945 and since about 1975. Spatial patterns of the change over the twentieth century indicate many regions showing statistically significant warming, but not all. Changes in temperature are also assessed in greater detail: showing that the warming is occurring more by reductions in cold extremes compared to increases in warm extremes and occurring more at night than during the daytime. The twentieth century warming is finally placed in a longer context by considering millennial-length paleoclimatic information from many diverse proxies. The latest evidence shows that the twentieth century has been both the warmest of the millennium and the warming rate during it has been unprecedented. The 1990s (1991–2000) was the warmest decade of the twentieth century and 1998 the warmest year. The first decade of the twenty-first century (2001–10) was warmer again, 0.20  C above the 1990s, with 2010 almost as warm as 1998. The 10 warmest years are all the years from 2001 to 2010, but with 2008 replaced by 1998.

Quality of Temperature Data Any assessment of trends or changes in temperature requires that all the observations have been taken in a consistent manner. Climatologists refer to this property as homogeneity. Time series of temperature are homogeneous if the variations exhibited are due solely to the vagaries of the weather and climate. Numerous nonclimatic factors influence the basic data. Without some form of adjustment, erroneous conclusions can be drawn regarding the course of ‘true’ temperature change. The factors vary depending upon the data source and are briefly considered in the next two subsections for the terrestrial and marine components of the Earth’s surface.

Land It is extremely rare if observational protocols and the environment around an observing location have remained exactly the same during the station’s history. Changes are likely to have occurred with the instruments, their exposure and measurement techniques, in the location of stations and the height of the instruments, in the number and times of observations per day, and the methods used to calculate daily and monthly

110

averages. The two most important factors with respect to longterm consistency are changes brought about by the introduction of Stevenson screens and changes in the environment around some of the stations across the world. Both factors are difficult to deal with using conventional approaches of station homogeneity assessment, as neighboring sites are all likely to be similarly affected. The commonly used louvred screen developed by Stevenson in the 1870s is now the standard around the world, although different countries use variants of a similar design. Prior to this, most thermometers were positioned on polewardfacing walls (i.e., out of direct sunlight), but this poses problems in high-latitude regions in the summer. The issue of the different exposures before screens has recently begun to be addressed by rebuilding the early exposures (using nineteenth century information) and taking modern parallel observations. Comparison of these measurements confirms expectations and indicates that prescreen temperatures are about 0.5  C too warm during the summer months from May to September. Winter temperatures are barely affected by the change. Only the longest series are affected, but countries or regions that introduced screens later (e.g., Australia in the 1910s) have yet to fully allow for the effects in the prescreen parts of their series.

Encyclopedia of Atmospheric Sciences 2nd Edition, Volume 3

http://dx.doi.org/10.1016/B978-0-12-382225-3.00005-0

Global Change j Climate Record: Surface Temperature Trends The second factor is that the location may have been a small town in the nineteenth century, but now could be a city of several million. Development around the site (urbanization) leads to relative warming of city sites compared to, still rural, neighbors. On certain days, particularly with anticyclonic weather, cities can be warmer than rural surroundings by up to 10  C. For monthly averages, this reduces to up to 2  C, more so for inland continental, compared to coastal, locations. Cities that have grown rapidly over the twentieth century tend to be more affected, compared to European locations where development has taken place over many centuries. Assessment of the urbanization influence suggests that the overall influence (at hemispheric scales) is small (up to 0.02  C per decade). Additional factors influencing homogeneity are that most stations have moved at least once during their lifetime. Also, of importance is the time observations that are made each day. Even today, there is no accepted standard, and countries are allowed to choose whatever times suit them. English-speaking countries have tended to use the average of the daily maximum and minimum readings each day to measure daily and monthly averages. Some countries have switched to this method mainly because of its ease, while others retain their national standards (averages of measurements made at fixed hours, between 3 and 24 times per day). Changes to sites or to the methods used to calculate monthly averages influence the time series, often in an abrupt manner (temperatures changing to a new level by up to 2  C in extreme cases). Ideally, when new sites or observation protocols are adopted, parallel measurements are recommended, enabling corrections to be calculated. These corrections are referred to as homogeneity adjustments. Sadly, although clearly recognized as being necessary, few countries carry out sufficient overlapping measurements. The most common problems relate to location moves, particularly to airports in the 1940s and 1950s. Recently, many countries have switched from mercury-inglass thermometers to electrical resistance thermistors, to reduce manpower, automating measurements. The sum total of all these problems can be disentangled, if adequate station history information is available, but it is generally a tedious process locating all the necessary information. In some countries, it is just not available in sufficient detail. Site moves and changes to observation protocols are less important than the widespread changes to screens and changes in the environment around a station as the effects can be of both signs and occur at irregular points in time. The overall effect on large-scale averages is only important if many sites are affected by changes occurring at the same time (such as the introduction of screens and urbanization issues). Homogeneity adjustments are necessary and vital for local scales, but are relatively unimportant at the hemispheric and global scales. Several groups in the UK, USA, Russia, and Japan have extensively analyzed the basic surface temperature data (between two and seven thousand stations), adjusting the data for the abrupt changes and removing urban-affected stations, and have reached similar conclusions about the course of temperature change over the instrumental period since 1850. It is highly unlikely that every problem has been corrected for, but the different techniques used, give confidence that largescale changes over the last 160 years are both real and welldocumented. The agreement on large-scale trends with marine

111

and lower tropospheric temperature estimates (see later discussion of satellite records) confirms this confidence.

Marine Terrestrial parts of the world constitute only 30% of the Earth’s surface, so for a global picture it is vital to include the oceans. Historical temperature data over marine regions are largely derived from in situ measurements of sea surface temperature (SST) and marine air temperature (MAT) taken by ships and buoys. To be of use, each measurement must be associated with a location. Up to 15% of marine data are thought to be mislocated (ships located on the land!) and these values must be discarded. It is obviously harder to reject data still located over the ocean, but all analyses of the raw data also attempt to remove or correct these problems. Plotting ship tracks has helped to considerably reduce the numbers of mislocated reports. Marine data are also beset with homogeneity problems, but they are distinctly different from the terrestrial realm. For MAT data, the average height of ships’ decks above the ocean has increased during the twentieth century, but more importantly, daytime measurements are influenced by the solar heating of the ship, restricting use, at present, to only the nighttime MAT (NMAT). For SST data, the changes in sampling method from uninsulated canvas buckets (generally prior to the early 1940s) to engine intake measurements (early 1940s onward) cause an artificial rise in SST values of 0.3–0.7  C. Recently, the greater preponderance of buoy (as opposed to ship) SST data has led to a further need for adjustments. Data from buoys were rapidly used when they became available, as coverage was dramatically improved. Now with over 20 years of overlap in measurements it is becoming apparent that the absolute temperatures from buoys are slightly cooler (0.1–0.2  C) than those taken by ships. Estimates of SST from satellites are also beginning to be widely used, but these can be offset from in situ measurements by several degrees. In the combination of marine data with land-based surface temperatures, SST data are preferred to NMAT, because they are generally more reliable, principally, as there are at least twice as many observations, daytime MAT values having been contaminated by the ships’ infrastructure. Additionally, the much stronger day-to-day correlation of SST compared to MAT means that averages of a few SST values are much more reliable than comparable averages of MAT data. Absolute values of SST and land air temperatures may differ by up to 10  C near some coastlines, so the two cannot be directly combined. Instead, anomalies are used (departures or differences from average) assuming that anomalies of SST and MAT agree on climatological (monthly and greater) timescales. Correction of the SST data for the change from canvas buckets is achieved using a physical–empirical model to estimate the degree of seawater cooling that occurs in buckets of varying designs. The cooling depends on the ambient weather conditions, but this can be approximated by climatological averages. Corrections are greatest in regions with the largest air–sea temperature differences (i.e., winters compared to summers) and the technique minimizes residual seasonal cycles in pre-World War II (pre-WWII) SST values compared to post-1945 values.

112

Global Change j Climate Record: Surface Temperature Trends

Since both the marine and the land components are independent, the two records can be used to assess each other after they have been separately corrected. The components have been shown to agree with several groups on both hemispheric scales using island and coastal data.

Aggregation of the Basic Data Both the land and the marine data are irregularly located over the Earth’s surface. To overcome the greater density of data in some regions, it is necessary to interpolate the data generally to some form of regular latitude/longitude grid.

Land Differing station elevations and national practices with regard to the calculation of monthly mean temperatures means that interpolation to a regular grid is much more easily achieved by converting all the monthly data to anomalies from a common reference period (often referred to as a climatology). The period with best available data is 1961–90. The simplest interpolation scheme is the average of all stations that are located within each 5  5 grid box. More complex interpolation methods yield essentially the same results on all spatial scales. A potential drawback of gridding schemes is that the variance of grid box time series is affected by changing numbers of stations within each grid box through time, although it is possible to allow for this.

Marine For SST, the aggregation is approached in a somewhat different manner. The changing location of each observation means that it is necessary, by interpolation, to derive the 1961–90 climatology for each 1  1 square of the world’s oceans for each 5-day period (pentad). SST anomaly values with respect to this climatology are then averaged together for each month for each 5  5 grid box, the same as used for the land component.

Combination into One Dataset Combination of the two components occurs in the simplest manner. Anomaly values are taken from each component. They are combined using weights determined by the errors of estimates of the land and the marine part. Around coastal areas this gives greater weight to the marine part.

Hemispherical Global Time Series With the basic data now in 5 latitude/longitude grid boxes, calculation of large-scale averages is relatively simple but must take into account the different sizes of grid boxes in tropical, compared to polar, latitudes. This is simply achieved by weighting each grid box by the cosine of its central latitude value. Figure 1 shows annual hemispheric and global time series for the 1850–2010 period using the HadCRUT3 dataset. The series derived by the other groups are very similar. Table 1

gives monthly linear trend values, estimated by least squares, for the three domains calculated over the 161-year period and for some other subperiods (1901–2010, 1920–44, and 1975– 2010). For the 1901–2010 period, global average surface temperature has risen by 0.83  C, a value that is statistically significant at the 99.9% level. All the monthly values also exhibit a significant warming. The choice of periods such as 1925–44 and since 1975 is determined by looking at Figure 1, so an element of subjectivity could have influenced the choice. Analysis of all possible periods longer than 10 years has been considered in one study. Since 1945, all periods longer than 22 years indicate warming but are only statistically significant for periods ending after about 1990. There is also a segment of significant warming for periods ending in the mid-1940s. All periods longer than 82 years all produce positive trends. While both hemispheres show similar degrees of warming, it is also apparent that many warm and cool years, relative to the underlying trend, are in common. Many anomalous warm years are coincident because they relate to El Niño years in the eastern equatorial Pacific. El Niño events cause somewhat predictable patterns of temperature and precipitation patterns over the world, with more regions experiencing warmer than cooler conditions. The opposite phase of an El Niño is termed a La Niña event. Anomalous patterns of temperature and precipitation also occur here, which to a first order are opposite to those during an El Niño event. Polar regions and much of northern Eurasia are largely unaffected by such an influence. A commonly used measure of the El Niño or La Niña state of the atmosphere is the normalized pressure difference between Tahiti and Darwin, Australia (referred to as the Southern Oscillation index (SOI)). Figure 2 is a scatter plot of the residual annual global temperature averages (the difference between each annual value and the smoothed curve) and the SOI for the 12-month average from July of the previous year through to June of the present year. The SOI explains about 30% of the variance of these residual temperatures. This analysis indicates that the next warm year will likely occur when the next El Niño occurs. A few cool years can be related to the climatic effects of explosive volcanic eruptions, which are large enough to put considerable amounts of dust into the stratosphere. Once there, the dust forms a veil over the Earth, reducing solar radiation and cooling the surface, particularly land areas. Surface cooling of about 0.2–0.3  C followed the eruption of Mt Pinatubo in the Philippines in June 1991, mainly in the northern summer months of 1992 and 1993. Volcanic eruptions, which only affect the troposphere (e.g., Mt St Helens in 1980), have little climatic effect as their ejecta are quickly dispersed by rainmaking processes. The next major explosive volcanic eruption in the tropics is likely to cool global temperatures for 2–3 years following the eruption.

Accuracy of the Hemispheric and Global Series The series in Figure 1 are subject to three sources of error: reductions in coverage earlier in the record; errors associated with the necessary bias adjustments in the basic data (discussed earlier for marine and land data, principally the switch to engine intake measurements in the early 1940s over the oceans

Global Change j Climate Record: Surface Temperature Trends

113

Figure 1 Hemispheric and global temperature averages on the annual timescale (1850–2010 relative to 1961–90). The smooth curves highlight variations on 20-year timescales.

Table 1

Temperature change ( C) explained by the linear trend over four periods: 1850–2010, 1901–2010, 1920–44, and 1975–2010 1850–2010

1901–2010

1920–44

1975–2010

NH

SH

Globe

NH

SH

Globe

NH

SH

Globe

NH

SH

Globe

Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec.

0.77 0.88 0.97 0.75 0.64 0.50 0.50 0.58 0.64 0.78 0.95 0.84

0.69 0.66 0.71 0.72 0.77 0.82 0.73 0.71 0.71 0.71 0.60 0.65

0.73 0.77 0.84 0.74 0.71 0.66 0.61 0.65 0.67 0.74 0.78 0.74

0.81 0.99 1.03 0.93 0.84 0.82 0.77 0.78 0.71 0.75 0.81 0.91

0.80 0.80 0.84 0.80 0.85 0.84 0.86 0.83 0.81 0.83 0.80 0.78

0.80 0.89 0.94 0.87 0.84 0.83 0.82 0.80 0.76 0.79 0.80 0.84

0.26 0.63 0.25 0.51 0.40 0.38 0.47 0.49 0.52 0.62 0.45 0.50

0.55 0.35 0.35 0.36 0.44 0.51 0.62 0.48 0.36 0.44 0.25 0.44

0.40 0.49 0.30 0.43 0.42 0.44 0.55 0.48 0.44 0.53 0.35 0.47

0.65 0.79 0.82 0.75 0.73 0.80 0.88 0.91 0.79 0.91 0.85 0.67

0.44 0.45 0.50 0.51 0.42 0.47 0.45 0.46 0.41 0.44 0.35 0.35

0.55 0.62 0.66 0.63 0.58 0.64 0.66 0.69 0.60 0.68 0.60 0.51

Year

0.73

0.71

0.72

0.84

0.82

0.83

0.46

0.43

0.44

0.80

0.44

0.62

114

Global Change j Climate Record: Surface Temperature Trends

Figure 2 The inverse relationship between the SOI and the residual global temperature (annual value minus the smoother curve seen in Figure 1). The SOI is the normalized sea level pressure difference between Tahiti and Darwin. Residual temperatures are for the calendar year, while the SOI is the average of July of the previous year to June of the current year. Also shown is the regression line (based on annual values for the years 1867–2010) and the correlation between the two series. The regression line slope of w0.8 means that a change of one SOI unit affects global temperature by w0.08  C. Very strong El Niño events can be up to 0.16  C warmer and very strong La Niña years up to 0.16  C cooler.

and the introduction of screens in the nineteenth century and effects of urbanization around some sites for the land); and homogeneity adjustments to land records for site and procedural changes. These are all taken into account in the error estimates and annual hemispheric averages are accurate to within 0.05  C (one standard error). Errors in the midnineteenth century were roughly twice modern values.

Analyses of the Temperature Record The surface record has been extensively analyzed, principally over the past 35 years. The series in Figure 1 has become one of the foremost series in major international reviews of the climate change issue, most recently by the Intergovernmental Panel on Climate Change (IPCC). Here several diverse aspects of the record are analyzed: l

trends in areas affected by monthly extremes; trends in maximum and minimum temperatures; l daily extremes of temperature in two long European series; and l the last 150 years in the context of the last 1000 years. l

Figure 1 clearly shows recent warming since the late 1970s. Five different groups monitor surface temperatures and all show similar courses of change over the last 160 years and similar rates of change to those given in Table 1. Completely independent estimates of temperature change have been developed for the lower part of the troposphere, first from weather balloons (called radiosondes) from the 1940s and more recently by satellite estimates from microwave sounding units aboard polar-orbiting satellites (since the late 1970s). Figure 3 shows a comparison at the global scale, of the two principal groups who collate the satellite data, for the period from 1979 to 2010. The surface temperature indicates a warming of 0.16  C per decade, in exact agreement with one of the satellite records. The difference between the two satellite records relates to the adjustments to their records that must be applied to derive a consistent record over the 32 years. Figure 3 also illustrates the influence of the SOI on global temperatures with the major El Niño event of 1997/1998 and the slightly lesser ones during the 1980s and during 2009/2010. Variability from month to month is markedly greater for the lower troposphere than at the surface, particularly during the larger El Niño and La Niña events.

Global Change j Climate Record: Surface Temperature Trends

115

Figure 3 Comparison of monthly surface and lower tropospheric temperatures for the period 1979–2010. All series have been re-zeroed to the period 1979–2009. The trends given are based on all the years shown from 1979 to 2010.

Trends in Areas Affected by Monthly Extremes The various groups that produce global and hemispheric temperature series also produce spatial temperature maps showing regions that were cooler or warmer than the 1961–90 base period. Similarly, it is possible to produce maps of temperature trends over specified periods. Interpreting these maps is often difficult as they tend to be dominated by the local level of variability. For example, the largest anomalies for a given month or year tend to occur in polar regions. Warming in Siberia may be large, but it is barely significant because of the large year-to-year variability. Trends in some tropical oceanic areas indicate highly statistically significant warming, but it may only be about 0.3–0.7  C. To enable easier intercomparison of trends and extremes, from a local impact point of view, removing the effects of year-to-year variability highlights where significant changes are occurring. An appropriate transformation is the gamma distribution. Each grid box time series, on a monthly basis, is transformed from one in anomalies with respect to 1961–90 to percentiles based on the same period. Percentiles can be easily related to return periods (e.g., the 5th/ 95th percentiles are equivalent to the one in the 20-year return period). Using a normal distribution (i.e., simply dividing the grid box anomaly series by the standard deviation calculated over the 1961–90 period) works almost as well as the gamma distribution, but the latter is better in many regions of the world as monthly temperatures are often significantly negatively skewed. Figure 4 compares the anomaly and percentile method for displaying annual temperatures for 2010. The zero anomaly and the 50th percentile contour are essentially the same in both plots. The percentile map, however, indicates extremely warm annual temperatures over many tropical and oceanic regions that might not warrant a second glance in anomaly form. 2010 shows 34% of the world’s surface with data above the 90th percentile and 3% below the 10th percentile. How unusual is this, compared to other years? Figure 5 shows the percentage of the world’s surface with data with temperatures greater than the 90th percentile (in red) and less than the 10th percentile (in blue) since 1900. An increase in the percentage of the analyzed area with warm extremes is evident (the largest area being 35%

in the warmest year (1998)), but by far the greatest change is a reduction in the percentage of the analyzed area with cold extremes. Some caution should be exercised while interpreting these results because of the large changes in coverage, particularly before 1951. The implicit assumption being made is that the average of the unsampled regions is the same as the average of the sampled regions. Coverage changes since 1951 are minimal, though, and even analyzing only those regions with data for the 1900–20 period produces similar series to those seen in Figure 5. The implications of these series are that before the mid-1970s most of the warming in this century was more apparent through less cold annual averages than excessively warm ones. Over the last 35 years, regions experiencing very warm annual anomalies have begun to increase dramatically.

Trends in Maximum and Minimum Temperatures Up to now, all the surface temperature analyses have been based on monthly mean temperatures. This situation has arisen due to the widespread availability of this variable. As mentioned earlier, English-speaking countries have tended to measure daily and monthly means using maximum and minimum temperatures. Recently, extensive datasets of monthly mean maximum and minimum temperatures have become available since the 1950s. These enable recent warming patterns to be assessed for both day (maximum) and night (minimum) temperatures. The difference between day and night (the diurnal temperature range (DTR)) should prove a useful variable when considering what the causes of changes might be due to. Homogeneity of the series poses more severe problems than for mean temperatures, as the various factors discussed earlier generally cause differential effects in the maximum and minimum series and station history information is even more important to decide upon adjustments. Analyses are restricted to the period 1950–2004 because of data availability issues in many regions of the world. Combining all available land regions, ‘global’ minimum averages warmed by 1.12  C over the 55 years, while maximums warmed by only 0.78  C. The DTR decreased by 0.34  C. Most of the differences in warming rates occurred over the period from 1950 to about 1980. Over the last 25 years, trends have been similar in the two series.

116

Global Change j Climate Record: Surface Temperature Trends

(a)

(b)

Figure 4 Surface temperatures for 2010, relative to the 1961–90 average (a) as anomalies and (b) as percentiles. The percentiles were defined by fitting gamma distributions to the 1961–90 annual deviations relative to the 1961–90 base period for all 5  5 grid boxes with at least 21 years of annual data in this period.

Urbanization influences have been shown to have the same signature (warmer nights compared to days), so these studies have restricted analyses to nonurban stations. In most regions, however, these differential trends can be clearly related to increases in cloudiness which will raise nighttime, compared to daytime, temperatures. Longer records, back to the turn of the twentieth century, are available in a few limited regions. Analyses over the USA and southern Canada, for example,

show little change over the first half of the twentieth century, so the drop in DTR over the period 1950–80 is the main feature of the record.

Daily Temperature Extremes in Long European Series The last two sections have considered extremes on a monthly basis, but public perception of climate change is often

Global Change j Climate Record: Surface Temperature Trends

117

Figure 5 Percentage of the monitored area of the globe having monthly surface temperatures above the 90th percentile (in red) and below the 10th percentile (in blue). Annual averages from these separate monthly analyses were averaged to get the annual values. The percentiles were defined by fitting gamma distributions to each month based on the 1961–90 period, using all 5  5 boxes with at least 21 years of data. The smooth curves highlight variations on decadal timescales.

influenced by daily extremes or runs of warm/cold days (e.g., heat waves or cold surges). In the context of the global warming issue, daily data are relatively unimportant as detection and attribution of human influences are primarily concerned with underlying trends on decadal timescales. In the public and political worlds though, changing frequencies of daily extremes are that how much of the global warming debate is perceived. Daily temperature series present even greater problems for climatologists with respect to homogeneity than monthly data. Site and observation time changes are particularly important and in some cases it may not be possible to fully correct for all the problems. Few long daily temperature series, therefore, are totally homogeneous. Furthermore, the availability of long series in some parts of the world is often restricted to the last 50 years because earlier data have not been digitized (particularly in some developing countries of the world). Changes in the frequency of extremes may be occurring, but without long series it is difficult to judge whether recent changes are really unprecedented. In Europe, however, several 200þ year series have recently been developed which will be ideal for analysis. The public perception of extremes is clearly cold winter and hot summer days, but in different regions it is necessary to define somewhat arbitrarily what is meant by cold and hot.

A cold day threshold of 0  C clearly has important consequences but what is hot in northern Europe clearly differs from what would be regarded as hot in southern Europe. Also, considering only absolute extremes ignores changes that might be taking place in the transition seasons. A better and universally applicable means of defining extremes is to let the data define the thresholds and to allow these to change from place to place and during the year. A number of groups over the recent decade have developed different sets of indices of extremes. These have recently been combined into a large set by the Expert Team on Climate Change, Detection and Indices (ETCCDI, http://www.clivar.org/organization/etccdi/etccdi.php). Here, a relatively simple example is presented, which is applied to three long European series. The first step is an analysis to define the annual cycle of temperature on a daily basis, based on a common period such as 1961–90. Some smoothing of this cycle is necessary as 30 years is a relatively short period for definition. 1961–90 is chosen for compatibility with the other analyses in this section. Variability of a single day’s temperatures from the annual cycle shows greater variability in Europe during winter compared to summer. Also most station data series throughout the year, but particularly in winter, tend to be negatively skewed, so

Global Change j Climate Record: Surface Temperature Trends

a normal distribution would be inappropriate as this would give a bias to the cold day count. Instead, it is necessary to fit a gamma distribution to the daily anomalies for each day of the year, again using the thirty 1961–90 days for each day. Now it is a simple matter to count the number of days above the 90th/95th (warm/very warm) and below the 10th/5th (cold/ very cold) percentiles in a calendar year or in a season. Figure 6 shows counts of warm/cold days for two of the long European series (Central England and Stockholm). Although there are differences between the stations in the timings of change, the overall picture is of an increase in warm days in the second half of the twentieth century, but the largest trend is a reduction in the number of cold days. Recent increases in warm days at the sites with longer records have only just exceeded similar counts in some decades of the eighteenth century. Cold day counts, in contrast, are clearly lower than at any period in the long records. The analysis method is insensitive to the choice of base period, another choice producing similar trends but centered around a different base.

All the last three sections consider extremes in different ways, but all show similar conclusions. Until the recent 35 years, the warming of the twentieth century is mostly manifest, not by increases in warm extremes, but by reduction in cold extremes. Cold extremes often pass by unnoticed by the majority, except in sectors and seasons where they have important effects.

The Last 150 Years in the Context of the Last 1000 Years Global average surface temperature has clearly risen over the last 160 years (Figure 1), but what significance does this have when compared to changes over longer periods of time? The last millennium is the period for which most is known about the preinstrumental past, particularly spatially, but it must be remembered that such knowledge is considerably poorer than since 1850. The millennium, particularly the last 500 years, is also the most important when considering attribution of

140

140 CET

120

120

100

100

80

80

60

60

40

40

20

20

0

0

140

140

No. of warm days (>90th percentile)

Stockholm

Stockholm

120

120

100

100

80

80

60

60

40

40

20

20

0 1750

No. of cold days (< 10th percentile)

No. of warm days (>90th percentile)

CET

No. of cold days (< 10th percentile)

118

0 1800

1850

1900 Year

1950

2000 1750

1800

1850

1900 Year

1950

2000

Figure 6 Numbers of cold days (<10th percentile) and warm days (>90th percentile) for four European locations with long daily records (Stockholm and Central England). The smooth curves highlight variations on decadal timescales (red for warm days and blue for cold days).

Global Change j Climate Record: Surface Temperature Trends recent changes to human influences. Earlier millennia are also important, but they are known to have experienced longer timescale changes in solar irradiance caused by orbital changes (the Milankovitch effect), giving, for example, higher irradiance in summer to northern high latitudes around 9000 years ago. Such differences in insolation mean that comparisons to today are not fair. Summers 9000 years ago in the higher latitudes of the Northern Hemisphere (NH) were warmer, but they experienced an 8% greater amount of solar insolation compared to today. Information about the past millennium comes from a variety of high-frequency and low-frequency proxy sources. High-frequency sources, giving information on the annual timescale include early instrumental records (back to the late seventeenth century in Europe), written historical documents (mainly Europe and the Far East), tree ring densities and widths (mid-to-high latitudes of both hemispheres), ice cores (both polar ice caps and also high-elevation tropical and smaller polar latitude ice caps), corals (tropical), and some highly

119

resolved lake and marine sediments. Low-frequency (decadal to century timescale change) evidence comes from boreholes, glacial advances/retreats and peat, lake, and marine cores. Uncertainties in all proxy information are considerable, both because evidence is restricted to where these written and natural archives survive, and more importantly, all proxy records are only imperfect records of past temperature changes. The last two decades have seen a dramatic improvement in both the availability of past evidence and also in information from diverse regions and sources. Figure 7 compares several different reconstructions of NH temperature change for most of the last millennium. The reconstructions are of different seasons, so based on the instrumental record they would be expected to differ somewhat. None of the series are strictly independent of each other, as they contain some common sources, but each has made different assumptions in their averaging. The most striking feature of the multiproxy averages is the warming over the twentieth century, both for its magnitude and duration. Agreement with the instrumental record

Figure 7 Reconstructions of NH temperatures from several different combinations of proxy data (multiproxy averages). All the series have been smoothed with a 40-year Gaussian filter and all are plotted as departures from the 1961–90 average. The different reconstructions are shown by the colored lines, the black being the instrumental record for April–September for the NH. All the series have been assembled recently and represent cutting edge research in paleoclimatology. Reproduced from Esper, J., Cook, E.R., Schweingruber, F.H., 2002. Low-frequency signals in long tree-ring chronologies for reconstructing past temperature variability. Science 295: 2250–2253. Reproduced from Rutherford, S., et al., 2005. Proxy-based Northern Hemisphere surface temperature reconstructions: Sensitivity to method, predictor network, target season, and target domain. Journal of Climate 18: 2308–2329. Reproduced from Moberg, A., et al., 2005. Highly variable Northern Hemisphere temperatures reconstructed from low- and high-resolution proxy data. Nature 433: 613–617. Reproduced from Pollack, H.N., Smerdon, J.E., 2004. Borehole climate reconstructions: Spatial structure and hemispheric averages. Journal of Geophysics Research 109: D11106, doi: 10.1029/2003JD004163. Reproduced from D’Arrigo, R., Wilson, R., Jacoby, G., 2006. On the long-term context for late twentieth century warming. Journal of Geophysics Research 111: 12, doi: 10.1029/2005JD006352. Reproduced from Hegerl, G.C., Crowley, T.J., Hyde, W.T., Frame, D.J., 2006. Climate sensitivity constrained by temperature reconstructions over the past seven centuries. Nature 440: 1029–1032. Reproduced from Oerlemans, J., 2005. Extracting a climate signal from 169 glacier records. Science 308: 675–677. Reproduced from Mann, M.E., et al., 2008. Proxy-based reconstructions of hemispheric and global surface temperature variations over the past two millennia. Proceedings of the National Academy of Sciences of the United States of America 105, doi: 10.1073/pnas.0805721105. Reproduced from Ljungqvist, F.C., 2010. A new reconstruction of temperature variability in the extra-tropical Northern Hemisphere during the last two millennia. Geografiska Annaler: Series A, Physical Geography 92: 339-351, doi: 10.1111/j.1468-0459.2010.00399.x. Reproduced from Ammann, C.M., Wahl, E.R., 2007. The importance of the geophysical context in statistical evaluations of climate reconstruction procedures. Climatic Change 85: 71–88, doi: 10.1007/s10584-007-9276-x. Reproduced from Juckes, M.N., et al., 2007. Millennial temperature reconstruction intercomparison and evaluation. Climate of the Past 3: 591–609.

120

Global Change j Climate Record: Surface Temperature Trends

should be assumed, as all the components of the series have, to some extent, been calibrated against instrumental data, either locally or as a whole. The twentieth century was the warmest of the millennium and the warming rate during it has been unprecedented. The number of series averaged together in each reconstruction depends on the particular study. Some use as many as possible, while others are restricted to specific proxy types (e.g., trees) or limited so that all proxies extend back a thousand years. Even the most spatially extensive studies only incorporate in number less than 5% of the number of instrumental sites. At the temporal scale plotted in Figure 7, however, the number of contributing proxy series is adequate to reconstruct large-scale temperatures, much more so for the NH compared to the Southern Hemisphere (SH), so Figure 7 only considers the NH. If standard errors are assigned to these series, as was the case for the instrumental period, errors would be considerably larger even for the 50-year timescale plotted. Earlier studies, using considerably fewer proxy datasets, have considered the past millennium and two periods, the Little Ice Age (variously defined as AD 1450–1850) and the Medieval Warm Epoch (less well-defined in the literature, but AD 900–1200 encompasses most earlier works) are often discussed. To some extent, these two periods have become accepted wisdom but the various curves in Figure 7 indicate only partial support. Spatial analysis of the proxy data shows that no century-scale periods in the millennium were universally colder or warmer everywhere, with considerable variability being present. The latter is to be expected even by studying the instrumental period since 1850. Just as the early 1940s were warm in many parts of the world Europe was cold, the early seventeenth century was cool in many regions, but was relatively mild in Iceland. In many respects, therefore, paleoclimatology is in the process of reassessing the evidence for these past periods and further changes are in prospect as more evidence becomes available. The various series in Figure 7 differ in some respects with regard to the coldest and warmest periods of the millennium, but they have all analyzed orders of magnitude more data than available in the early 1970s. The cooler centuries of the millennium were the sixteenth to the nineteenth, the seventeenth being the coldest in Europe, and the nineteenth coldest in North America. These regions are still the best studied and it will be vital in future to extend the knowledge to other areas, particularly in the SH. At present, for every one long SH proxy reconstruction there are at least 10 in the NH. Just as with the instrumental record it is important to gain as much evidence from as many regions as possible, if how global and hemispheric temperatures have varied over this long time are to be fully understood. Contrasts in the timing of changes between regions and particularly between the hemispheres must be recognized if the causes of the changes are to be fully

understood. A more complete understanding of the causes of the changes will allow to determine how much climate can change naturally, enabling to better distinguish the degree of human influence on surface temperature during the twentieth century.

See also: Climate and Climate Change: Climate Variability: Decadal to Centennial Variability; Climate Variability: Nonlinear and Random Effects; Climate Variability: North Atlantic and Arctic Oscillation; Climate Variability: Seasonal and Interannual Variability. Global Change: Biospheric Impacts and Feedbacks; Sea Level Change; Upper Atmospheric Change. Ozone Depletion and Related Topics: Long-Term Ozone Changes. Paleoclimatology: Ice Cores; Varves. Statistical Methods: Data Analysis: Empirical Orthogonal Functions and Singular Vectors; Data Analysis: Time Series Analysis. Tropical Meteorology and Climate: El Niño and the Southern Oscillation: Observation.

Further Reading Böhm, R., Jones, P.D., Hiebl, J., Frank, D., Brunetti, M., Maugeri, M., 2010. The early instrumental warm-bias: A solution for long Central European temperature series, 17602007. Climatic Change 101, 41–67. Hansen, J., Ruedy, R., Sato, M., Lo, K., 2010. Global surface temperature change. Reviews of Geophysics 48, RG4004. doi:10.1029/2010RG000345. Jones, P.D., New, M., Parker, D.E., Martin, S., Rigor, I.G., 1999. Surface air temperature and its changes over the past 150 years. Reviews of Geophysics 37, 173–199. Jones, P.D., Wigley, T.M.L., 2010. Estimation of global temperature trends: What’s important and what isn’t. Climatic Change 100, 59–69. Jones, P.D., Briffa, K.R., Osborn, T.J., et al., 2009. High-resolution paleoclimatology of the last millennium: A review of current status and future prospects. The Holocene 19, 3–49. Karl, T.R., Hassol, S.J., Miller, C.D., Murray, W.L. (Eds.), 2006. Temperature Trends in the Lower Atmosphere: Steps for Understanding and Reconciling Differences. A Report by the U.S. Climate Change Science Program and the Subcommittee on Global Change Research. National Oceanic and Atmospheric Administration, National Climatic Data Center, Asheville, NC, p. 164. Liebmann, B., Dole, R.M., Jones, C., Bladé, I., Allured, D., 2010. Influence of choice of time period on global surface temperature trend estimates. Bulletin of the American Meteorological Society 91, 1485–1491. Mann, M.E., Zhang, Z., Hughes, M.K., et al., 2008. Proxy-based reconstructions of hemispheric and global surface temperature variations over the past two millennia. Proceedings of the National Academy of Sciences of the United States of America 105, 13252–13257. Parker, D.E., 2010. Urban heat island effects on estimates of observed climate change. Wiley Interdisciplinary Reviews: Climate Change 1, 123–133. Thompson, D.W.J., Kennedy, J.J., Wallace, J.M., Jones, P.D., 2008. A large discontinuity in the mid-twentieth century in observed global-mean surface temperature. Nature 453, 646–649. Trewin, B., 2010. Exposure, instrumentation, and observing practice effects on land temperature measurements. WIREs Climate Change, 490–506. DOI 10.1002/ wcc.46. Vose, R.S., Easterling, D.R., Gleason, B., 2005. Maximum and minimum temperature trends for the globe: An update through 2004. Geophysical Research Letters 32, L23822. doi:10.1029/2005GL024379.