This Section focuses on changes and variations in the modern climate record, variations in palaeo-climate are also described. Analyses of the climate record can provide important information about natural climate variations and variability. A major difficulty in using observed records to make deductions about changes resulting from recent increases in greenhouse gases is the existence of natural climatic forcing factors that may add to, or subtract from, such changes. Unforced internal variability of the climate system will also occur further obscuring any signal induced by greenhouse gases. Observing the weather, and converting weather data to information about climate and climate change is a very complex endeavor. Virtually all our information about modern climate has been derived from measurements which were designed to monitor weather rather than climate change. Even greater difficulties arise with the proxy data (natural records of climate sensitive phenomena, mainly pollen remains, lake verves and ocean sediments, insect and animal remains, glacier termini) which must be used to deduce the characteristics of climate before the modern instrumental period began. So special attention is given to a critical discussion of the quality of the data on climate change and variability and our confidence in making deductions from these data. Note that we have not made much use of several kinds of proxy data, for example tree ring data, that can provide information on climate change over the last millennium. We recognize that these data have an increasing potential however their indications are not yet sufficiently easy to assess nor sufficiently integrated with indications from other data to be used in this report.
A brief discussion of the basic concepts of climate change, climate trends etc. together with references to material containing more precise definitions of terms, is found in the Introduction at the beginning of this report.
2.2 Palaeo-Climatic Variations and Change
2.2.1 Climate Of The Past 5,000,000 Years
Climate varies naturally on all time scales from hundreds of millions of years to a few years. Prominent in recent earth’s history have been the 100,000 year Pleistocene glacial-interglacial cycles when climate was mostly cooler than at present (Imbrie and Imbrie 1979) This period began about 2,000,000 years before the present time (BP) and was preceded by a warmer epoch having only limited glaciation, mainly over Antarctica called the Pliocene. Global surface temperatures have typically varied by 5-7oC through the Pleistocene ice age cycles with large changes in ice volume and sea level, and temperature variations as great as 10-15°C in some middle and high latitude regions of the Northern Hemisphere. Since the beginning of the current interglacial epoch about 10,000 BP global temperatures have fluctuated within a much smaller range. Some fluctuations have nevertheless lasted several centuries, including the Little Ice Age which ended in the nineteenth century and which was global in extent.
Proxy data clearly indicate that the Earth emerged from the last ice age 10,000 to 15,000 BP. During this glacial period continental size ice sheets covered much of North America and Scandinavia and world sea level was about 120m below present values. An important cause of the recurring glaciations is believed to be variations in seasonal radiation receipts in the Northern Hemisphere. These variations are due to small changes in the distance of the earth from the sun in given seasons and slow changes in the angle of the tilt of the earths axis which affects the amplitude of the seasonal insolation. These Milankovitch orbital effects (Berger, 1980) appear to be correlated with the glacial-interglacial cycle since glacial arise when solar radiation is least in the extratropical Northern Hemisphere summer.
Variations in carbon dioxide and methane in ice age cycles are also very important factors, they served to modify and perhaps amplify the other forcing effects However, there is evidence that rapid changes in climate have occurred on time scales of about a century which cannot be directly related to orbital forcing or to changes in atmospheric composition. The most dramatic of these events was the Younger Dryas cold episode which involved an abrupt reversal of the general warming trend in progress around 10,500 BP as the last episode of continental glaciation came to a close The Younger Dryas was an event of global significance, it was clearly observed in New Zealand (Salingei 1989) though its influence may not have extended to all parts of the globe (Rind et al 1986). There is as yet no consensus on the reasons for this climatic reversal which lasted about 500 years and ended very suddenly. However, because the signal was strongest around the North Atlantic Ocean, suggestions have been made that the climatic reversal had its physical origin in large changes in the sea surface temperature (SST) of the North Atlantic Ocean. One possibility is that the cooling may have resulted from reduced deep water production in the North Atlantic following large scale melting of the Lauientide Ice sheet and the resulting influx of huge amounts of low density freshwater into the northern North Atlantic Ocean (Broecker et al 1985). Consequential changes in the global oceanic circulation may have occurred (Street Perrott and Perrott 1990) which may have involved variations in the strength of the thermohaline in the Atlantic This closed oceanic emulation involves northward flow of water near the ocean surface sinking in the sub-Arctic and a return flow at depth The relevance of the Younger Dryas to today s conditions is that it is possible that changes in the thermohaline circulation of a qualitatively similar character might occur quite quickly during a warming of the climate induced by greenhouse gases. A possible trigger might be an increase of precipitation over the extratropical North Atlantic (Broecker, 1987), though the changes in ocean circulation are most likely to be considerably smaller than in the Younger Dryas Section. The period since the end of the last glaciation has been characterized by small changes in global average temperature with a range of probably less than 2°C, though it is still not clear whether all the fluctuations indicated were truly global. However, large regional changes in hydrological conditions have occurred, particularly in the tropics. Wetter conditions in the Sahara from 12,000 to 4,000 years BP enabled cultural groups to survive by hunting and fishing in what are today almost the most and regions on earth. During this time, Lake Chad expanded to become as large as the Caspian Sea is today (several hundred thousand km2, Grove and Warren, 1968). Drier conditions became established after 4,000 BP and many former lake basins became completely dry (Street-Perrot and Harnson 1985). Pollen sequences from lake beds of northwest India suggest that periods with subdued monsoon activity existed during the recent glacial maximum (Singh et al 1974), but the epoch 8,000 to 2,500 BP experienced a humid climate with frequent floods.
There is growing evidence that worldwide temperatures were higher than at present during the mid-Holocene (especially 5,000-6,000 BP), at least in summer, though carbon dioxide levels appear to have been quite similar to those of the pre-industrial era at this time. Thus parts of western Europe, China, Japan, the eastern USA were a few degrees warmer in July during the mid- Holocene than in recent decades (Yoshino and Urushibara, 1978, Webb et al 1987, Huntley and Prentice, 1988, Zhang and Wang 1990) Parts of Australasia and Chile were also warmer. The late tenth to early thirteenth centuries (about AD 950-1250) appear to have been exceptionally warm in western Europe, Iceland and Greenland (Alexandre 1987, Lamb, 1988) This period is known as the Medieval Climatic Optimum. China was, however, cold at this time (mainly in winter) but South Japan was warm (Yoshino, 1978) This period of widespread
warmth is notable in that there is no evidence that it was accompanied by an increase of greenhouse gases. Cooler episodes have been associated with glacial advances in alpine regions of the world, such neo-glacial’ episodes have been increasingly common in the last few thousand years. Of particular interest is the most recent cold event, the Little Ice Age, which resulted in extensive glacial advances in almost all alpine regions of the world between 150 and 450 years ago (Grove, 1988), so that glaciers were more extensive 100-200 years ago than now nearly everywhere. Although not a period of continuously cold climate, the Little Ice Age was probably the coolest and most globally extensive cool period since the Younger Dryas. In a few regions, alpine glaciers advanced down-valley even further than during the last glaciation (for example, Miller, 1976). Some have argued that an increase in explosive volcanism was responsible for the coolness (for example, Hammer, 1977, Porter, 1986), others claim a connection between glacier advances and reductions in solar activity (Wigley and Kelly, 1989), such as the Maunder and Sporer solar activity minima (Eddy, 1976), but see also Pittock (1983). At present, agreed explanation for these recurrent cooler episodes The Little Ice Age came to an end only in the nineteenth century. Thus, some of the global warming since 1850 could be a recovery from the Little Ice Age rather than a direct result of human activities. So it is important to recognise that natural variations of climate are appreciable and will modulate any future changes induced by man.
2.2.2 Palaeo-Climate Analogues from the Three Warm Epochs
Three periods from the past have been suggested by Budyko and Izrael (1987) as analogues of a future warm climate For the second and third periods listed below, however, it can be argued that the changed seasonal distribution of incoming solar radiation existing at those times may not necessarily have produced the same climate as would result from a globally-averaged increase in greenhouse gases. • The climate optimum of the Pliocene (about 3,300,000 to 4,300,000 years BP) • The Eemian interglacial optimum (125,000 to 130,000 years BP), • The mid-Holocene (5,000 to 6,000 years BP)
Note that the word “optimum” is used here for convenience and is taken to imply a warm climate However such a climate may not be “optimal” in all senses.
126.96.36.199 Pliocene Climatic Optimum (about 3,300,000 to 4,300,000 BP)
Reconstructions of summer and winter mean temperatures and total annual precipitation have been made for this period by scientists in the USSR. Many types of proxy were used to develop temperature and precipitation patterns over the land masses of the Northern Hemisphere (Budyko and Izrael, 1987). Over the oceans, the main sources of information were cores drilled in the bed of the deep ocean by the American Deep sea Ocean Core Drilling Project.
188.8.131.52 Eemian Interglacial Optimum (125,000-130,000 years BP)
Palaco-botanic oxygen-isotope and other geological data show that the climates of the warmest parts of some of the Pleistocene interglacial were considerably warmer (1 to 2°C) than the modern climate. They have been considered as analogues of future climate (Budyko and Izrael, 1987, Zubakov and Borzenkova, 1990). Atmospheric carbon dioxide reached about 300 ppm during the Eemian optimum but a more important cause of the warmth may have been that the eccentricity of the earth’s orbit around the sun was about twice the modern value, giving markedly more radiation in the Northern Hemisphere summer. The last interglacial optimum (125,000 – 130,000 years BP), has sufficient information (Vehchko et al 1982, 1983, 1984, and CLIMAP, 1984), to allow quantitative reconstructions to be made of annual and seasonal air temperature and annual precipitation for part of the Northern Hemisphere. For the Northern Hemisphere as a whole, mean annual surface air temperature was about 2°C above its immediately pre-industrial value.
184.108.40.206 Climate of the Holocene Optimum (5,000 – 6 000 years BP)
The Early and Middle Holocene was characterized by a relatively warm climate with summer temperatures in high northern latitudes about 3-4°C above modern values Between 9,000 and 5,000 years BP, there were several short lived epochs, the last of which, the mid Holocene optimum from 6,200 to 5,300 years BP (Varushchenko et al., 1980). Each warm epoch was accompanied by increased precipitation and higher lake levels in subtropical and high latitudes
(Singh et al., 1974; Swain et al., 1983). However, the level of such mid- latitude lakes as the Caspian Sea, Lake Geneva and the Great Basin lakes in the USA was lowered ( COHMAP, 1988; Borzenkova and Zubakov, 1984).
2.3 The Modern Instrumental Record
The clearest signal of an enhanced greenhouse effect in the climate system, as indicated by atmosphere/ocean general circulation models, would be a widespread, substantial increase of near-surface temperatures. This section gives special attention to variations and changes of land surface air temperatures (typically measured at about two metres above the ground surface) and sea surface temperatures (SSTs) since the mid-nineteenth century. Although earlier temperature, precipitation, and surface pressure data are available (Lamb, 1977), spatial coverage is very poor. We focus on changes over the globe and over the individual hemispheres but considerable detail on regional space scales is also given.
2.4 Surface Temperature Variations and Change
2.4.1 Hemispheric and Global
Three research groups (Jones et al 1986a, b, Jones, 1988, Hansen and Lebedeff, 1987, 1988, and Vinnikov et al, 1987, 1990), have produced similar analyses of hemispheric land surface air temperature variations from broadly the same data. All three analyses indicate that during the last decade, globally-averaged land temperatures have been higher than in any decade in the past 100 to 140 years. Northern Hemisphere temperatures prior to the climatic discontinuity in the 1920s could be interpreted as varying about a stationary mean climate as shown by the smoothed curve. The nearest approach to a monotonic trend in the Northern Hemisphere time series is the decrease of temperature from the late 1930s to the mid-1960s of about 0.2°C. The most recent warming has been dominated by a relatively sudden increase of nearly 0.3°C over less than ten years before 1982. Of course, it is possible to fit a monotonic trend line to the entire time series such a trend fitted to the current version of the Jones (1988) data gives a rate of warming of 0.53°C/100 years when the trend is calculated from 1881 to 1989 or the reduced if less reliable value of 0.45oC/I00 years if it is calculated from 1861. Clearly, this is a gross oversimplification of the observed variations even though the computed linear trends are highly significant in a statistical sense.
The data for the Southern Hemisphere include the Antarctic land mass since 1957, except for the data of Vinnikov et al (1987 1990). Like the Northern Hemisphere, the climate appears stationary throughout the latter half of the nineteenth century and into the early part of the twentieth century. Subsequently, there is an upward trend in the data until the late 1930s but in the next three decades the mean temperature remains essentially stationary again. A fairly steady increase of temperature resumes before 1970 though it may have slowed recently. Linear trends for the Southern Hemisphere are 0.52oC/100 years from 1881 to 1989, but somewhat less and less reliable, at 0.45°C/100 years for the period 1861 1989. Prior to 1957, data for Antarctica are absent while some other parts of the global land mass lack data as late as the 1920s, for example, many parts of Africa, parts of China, the Russian and Canadian Arctic, and the tropics of South America. In the 1860s, coverage is sparsest, thus, Africa has little or no data and much of North America is not covered. The effect of this drastically changing spatial coverage on hemispheric temperature variations has been tested by Jones et al (1986a, b), who find that sparse spatial coverage exaggerates the variability of the annual averages. The reduction in variability of the Northern Hemisphere annual time series after about 1880 (see Section 220.127.116.11 for a detailed discussion for the combined land and ocean data), suggests that changes of station density since 1900 have had relatively little impact on estimates of hemispheric land temperature anomalies. However, prior to 1900, the decadal uncertainty could be up to 00.1°.CThis is quite small relative to the overall change. Thus, varying data coverage does not seem to have had a serious impact on the magnitude of the perceived warming over land over the last 125 years.
Another potential bias arises from changes in observation schedules. Even today, there is no international standard for the calculation of mean daily
temperature. Thus, each country calculates mean daily temperature by a method of its choice such as the average of the maximum and the minimum, or some combination of hourly leadings weighted according to a fixed formula. As long as each country continues the same practice the shape of the temperature record is unaffected. Unfortunately, few countries have maintained the same practice over the past century, biases have therefore been introduced into the climate record some of which have been corrected for in existing global data sets, but some have not. These biases can be significant; in the USA, a systematic change in observing times has led to a nominal 0.2oC decrease of temperature in the climate record since the 1930s (Karl et al., 1986). The effects of changing observation time have only been partly allowed for in the USA temperature data used in the analyses presented here. So, an artificial component of cooling off rather less than 0.2oC may exist in the USA, part of the temperature analyses for this reason, offsetting the warming effects of temperature of either sign may exist in other parts of the world due to changes observation time but have not been investigated.
Substantial systematic changes in the exposure of thermometers have occurred because thermometers can be affected by the direct rays of the sun, reflected solar radiation, extraneous heat sources and precipitation, there has been a continuous effort to improve their exposures over the last 150 years. Additional biases must accompany these changes in the thermometric record. Since many of the changes in exposure took place during the nineteenth and early twentieth centuries, that part of the record is most likely to be affected. Recently, Parker(1990), has reviewed the earlier thermometer exposures, and how they evolved in many different countries. The effects of exposure changes vary regionally (by country), and seasonally. Thus tropical temperature prior to the late 1920s appear to be too high because of the placement of thermometers in cages situated in open sheds. There is also evidence that for the mid – latitudes prior to about 1880 summer temperatures may be too high and winter temperature too low due to the use of poorly screened exposures. This includes the widespread practice of exposing thermometers on the north walls of the building. These effects have not yet been accounted for in existing analyses ( see Section 18.104.22.168).
Changes in station environment can seriously affect temperature records (Salinger, 1981). Over the years, stations often have minor (usually under 10 km) relocations and some stations have been moved from rooftop to ground level. Even today, international practice allows for a variation of thermometer heights above ground from 1.25 to 2 metres because large vertical temperature gradients exist near the ground, such changes could seriously affect thermometer records. When relocations occur in a random manner, they do not have a serious impact on hemispheric or global temperature anomalies, though they impair our ability to develop information about much smaller scale temperature variations. A bias on the large scale can emerge when the character of the changes is not random. An example is the systematic relocations of some observing stations from inside cities in many countries to more rural airport locations that occurred several decades ago. Because of the heat island effect within cities, such moves tend to introduce artificial cooling into the climate record (Jones et al 1986a, b), attempt in some detail to adjust for station relocations when these appear to have introduced a significant bias in the data but (Hansen and Lebedeff 1988) do not, believing that such station moves cancel out over large time and space averages (Vinnikov et al 1990) do adjust for some of these moves. There are several possible correction procedures that have been or could be, applied to the (Jones 1988) data set (Bradley and Jones, 1985, Karl and Williams 1987). All depend on denser networks of stations than are usually available except in the USA, Europe, the Western Soviet Union and a few other areas.
Of the above problems, increasing urbanisation around fixed stations is the most serious source of systematic error for hemispheric land temperature time series that has so far been identified. A number of researchers have tried to ascertain the impact of urbanisation on the temperature record (Hansen and Lebedeff 1987) found that when they removed all stations having a population in 1970 of greater than 100,000, the trend of temperature was reduced by 0.1°C over 100 years. They speculated that perhaps an additional 0.1°C of bias might remain due to increases in urbanisation around stations in smaller cities and towns. Jones et al (1989) estimate that the effect of urbanisation in then quality- controlled data is no more than 0.1°C over the past 100 years. This conclusion is based on a comparison of then data with a dense network of mostly rural stations over the USA. Groisman and Koknaeva (1990), compare the data from Vinnikov et al (1990), with the rural American data set and with rural stations in the Soviet Union and find very small warm relative biases of less than 0.05°C per 100 years in the USA. Kail et al (1988), find that increases due to urbanisation can be significant (0.1°C), even when urban areas have populations as low as 10, 000. Other areas of the globe are now being studied. Preliminary results indicate that the effects of urbanisation are highly regional and time dependent Changes in urban warming in China (Wang et al , 1990) appear to be quite large over the past decade, but in Australia they are rather less than is observed in the USA (Coughlan et al , 1990). Recently Jones and co workers (paper in preparation) have compared trends derived from their quality-controlled data and those of Vinnikov et al (1990), with specially selected data from more rural stations in the USSR, eastern China, and Australia. When compared with trends from the more rural stations, only small (positive) differences of temperature trend exist in the data used in Jones (1988) and Vinnikov et al (1990) in Australia and the USSR (of magnitude less than 0.05oC/100years). In eastern China, the data used by Vinnikov et al (1990) and Jones (1988) give smaller warming trends than those derived from the more rural stations. This is an unexpected result. It suggests that either
(1) The more rural set is sometimes affected by urbanisation
(2) Other changes in station characteristics over compensate for urban warming bias.
Thus, it is known that the effects of biases due to increased urbanisation in the Hansen and Lebedeff (1987) and the Vinnikov et al (1990) data sets are partly offset by the artificial cooling introduced by the movement of stations from city centres to more rural airport locations during the 1940s to 1960s (Karl and Jones 1990). Despite this, some of these new rural airport locations may have suffered recently from increasing urbanisation. In light of this evidence, the estimate provided by Jones et al (1989) of a maximum overall warming bias in all three land data sets due to urbanisation of 0.1oC/100 years or less is plausible but not conclusive.
The oceans comprise about 61% of the Northern Hemisphere and 81% of the Southern Hemisphere. Obviously a compilation of global temperature variations
must include ocean temperatures. Farmer et al (1989) and Bottomley et al (1990) have each created historical analyses of global ocean SSTs which are derived mostly from observations taken by commercial ships. These data are supplemented by weather ship data and in recent years by an increasing number of drifting and moored buoys. The Farmer et al (1989) analyses are derived from a collection of about 80 million observations assembled in the Comprehensive Ocean-Atmosphere Data Set (COADS) in the USA (Woodruff et al , 1987). The data set used by Bottomley et al (1990) is based on a slightly smaller collection of over 60 million observations assembled by the United Kingdom Meteorological Office. Most, but not all of the observation in the latter are contained in the COADS data set.
Long-term variations of SSTs over the two hemispheres, have been, in general, similar to their land counterparts. The increase in temperature has not been continuous. There is evidence for a fairly rapid cooling in SST of about 0.1oC to 0.2°C at the beginning of the twentieth century in the Northern Hemisphere. This is believed to be real because night marine air temperatures show a slightly larger cooling . The cooling strongly affected the North Atlantic, especially after 1903 and is discussed at length by Helland-Hansen and Nansen (1920). The cool period was terminated by a rapid rise in temperature starting after 1920. This resembled the sudden warming of land temperatures, but lagged it by several years. Subsequent cooling from the late 1950s to about 1975 lagged that over land by about five years, and was followed by renewed warming with almost no lag compared with land data Overall warming of the Northern Hemisphere oceans since the late nineteenth century appears to have been slightly smaller than that of the land and may not have exceeded 0.3°C. In the Southern Hemisphere ocean (remembering that the Southern Ocean has always been poorly measured), there appear to have been two distinct stable climatic periods the first lasting until the late 1920s, the second lasting from the mid 1940s until the early 1960s. Since the mid 1970s, SSTs in the Southern Hemisphere have continued to rise to their highest levels of record. Overall, warming has certainly exceeded 0.3°C since the nineteenth century, but has probably been less than 0.5°C and has been slightly less than the warming of the land. However, if the increases of temperature are measured from the time of their minimum values around 1910, the warming of the oceans has been slightly larger than that of the land. Despite data gaps over the Southern Ocean, the global mean ocean temperature variations tend to take on the characteristics of the Southern Hemisphere because a larger area of ocean is often sampled in the Southern Hemisphere than the Northern Hemisphere. Overall warming in the global oceans between the late nineteenth century and the latter hall of the twentieth century appears to have been about 0.4°C.
Several different types of bucket have been used for sampling made for example, of wood, canvas, rubber or metal but the largest bias arose from an apparently rather sudden transition from various un insulated buckets to ship engine intake tubes in World War II. A complex correction procedure developed by Folland and Parker (1989) and Bottomley et al (1990) which creates geographically varying corrections, has also been used in the same form by Farmer et al. Differences in the two data sets remain, however, primarily because of different assumptions about the mix of wooden versus canvas buckets used during the nineteenth century. Despite recommendations by Maury (1858) to use wooden buckets with the thermometer inserted for four to five minutes, such buckets may have been much less used in practice (Toyabee, 1874; correspondence with the Danish Meteorological Service, 1989) possibly because of damage iron – banded wooden buckets could inflict upon the hulls of ships. Some differences also result, even as recently as the 1970s, because the data are not always derived from identical sources (Woodruff, 1990).
No corrections have been applied to the SST data from 1942 to date. Despite published discussions about the difference between “bucket” and engine intake SST data in this period ( for example James and Fox, 1972), there are several reasons why it is believed that no further corrections, with one reservation noted below, are needed. Firstly, the anomalies are calculated from the mean conditions in 1951 – 1980. So only relative changes in the mix of data since 1942 are important. Secondly, many of the modern “buckets” are insulated (Folland and Parker, 1990) so that they cool much less than canvas buckets. A comparison of about two million bucket and four million engine intake data for 1951- 1981 (Bottomley et al.,1990) reveals a global mean difference of only 0.08oC, the engine intake data being the warmer. This conclusion is strongly supported by the great similarly between time series of globally averaged anomalies of collocated SST and night marine air temperature data 1955 to data.
Less perfect agreement between 1946 and the early 1950s ( SST colder) suggest that uninsulated bucket data may have been more numerous then than in 1951 – 1980, yielding an overall cold bias of up to 0.1oC on a global average.
Marine air temperatures are a valuable test of the accuracy of SST after the early 1890s. Biases of day-time marine air temperatures have been used. The biases arise during the day because overheating of the thermometers and screens by solar insolation has changed as ships have changed their physical characteristics (Folland et al., 1984). On the other hand appreciable biases of night time data currently believed to be confined to nineteenth century and much of the Second World War. Night marine air temperatures have been found to be much too high relative to SST, or to modern values, in certain regions and seasons before 1894, (Bottomley et al., 1990). These values were corrected using SSTs, but subsequently (except in 1941 – 1945), night marine air temperature data constitute independent evidence everywhere, although corrections are also made for the increasing heights of ship decks ( Bottomley et al., 1990). Both SST and night marine air temperature data appear to lay the land data by at least five years during the period of warming from 1920 to the 1940s. However, some of the apparent warmth of the land at this time may e erroneous due to the use of open shed screens in the tropics (Section 22.214.171.124). The above results differ appreciably in the nineteenth century from those published by Ooit et al (1987) who followed the much less detailed correction procedure of Folland et al (1984) to adjust the COADS. SST and all hours marine air temperature data sets Newell et al (1989) also present an analysis quite similar to that of the above authors, based on a UK Meteorological Office data set that was current in early 1988. All these authors obtain higher values of global SST and marine air temperature in the middle to late nineteenth century typically by about 0.1°C and 0.15°C, respectively than aec indicated in this report. It is our best judgement that the more recent analyses represent a real improvement but the discrepancies highlight the uncertainties in the interpretation of early marine temperature records. Yamamoto et al (1990a) have tried to quantify changing biases in the COADS all hours marine air temperature data using a mixture of weather ship air temperature data from the 1940s to 1970s, and selected land an temperature data mainly in three tropical coastal legions to calculate time varying corrections. Based on these corrections, Yamamoto et al (1990b) calculate a global air temperature anomaly curve for 1901-1986 of similar overall character to the night marine air temperature curve, but with typically 0.15°C warmer anomalies in the early pan of the twentieth century, and typically 0.1°C cooler anomalies in the warm period around 1940-1950. Recent data are similar, It could be argued that the collections of Yamamoto et al may be influenced by biases in the land data including warm biases arising from the use of tropical open sheds earlier this century. Warm biases may also exist in some ocean weather ship, day-time air temperature data (Folland 1971). Although, we believe that the night marine air temperature analysis minimises the known sources of error, the work of Yamamoto et al underlines the level of uncertainty that exists in trends derived from marine air temperature data.
126.96.36.199 Land and sea combined
Combined land and sea surface temperatures show a significant increase of temperature from the late nineteenth to the late twentieth century . These data are an average of two data sets, a combination of the Jones land data and the Farmer et al SST data, and a combination of the Jones land data and the UK Meteorological Office SST data. Note that the relative contributions of land and sea to the combined data have varied according to changing data availability over land and ocean. Over the globe the combined data gives an increase of temperature of 0.45°C between the average for the two decades 1881-1900 and the decade 1980-1989. The comparable increase for the Northern Hemisphere is 0.42oC and for the Southern Hemisphere 0.48°C. A similar calculation for the changes of temperature between 1861-1880 and 1980-1989 gives 0.45°C, 0.38°C, and 0.53oC respectively. A linear trend fitted between 1890 and 1989 gives values of 0.50°C/100 years (globe), 0 47°C/100 years (Northern Hemisphere) and 0.53oC/100 years (Southern Hemisphere), a linear trend fitted between 1870 and 1989 gives the reduced values of 0.41°C, 0.19°C, and 0.43oC/100 years respectively. Apparent decadal rates of change of smoothed global combined temperature have varied from an increase of 0.21°C between 1975 and 1985 (largely between 1975 and 1981) to a decrease of 0.19°C between 1898 and 1908 (though data coverage was quite poor around 1900). Surprisingly, the maximum magnitudes of decadal change (warming or cooling) over land and ocean (SST) have been quite similar at about 0 25°C. Smoothed night global marine air temperature showed the largest apparent change around 1900 with a maximum cooling of 0.32°C between 1898 and 1908 though this value is very uncertain.
Combined land and ocean temperature has increased rather differently in the Northern than in the Southern Hemisphere. A rapid increase in Northern Hemisphere temperature during the 1920s and into the 1930s contrasts with a more gradual increase in the Southern Hemisphere. Both hemispheres had relatively stable temperatures attires from the 1940s to the 1970s though with some evidence of cooling in the Northern Hemisphere. Since the 1960s in the Southern Hemisphere, but after 1975 in the Northern Hemisphere temperatures have risen with the overall rise being more pronounced in the Southern Hemisphere. Only a small overall rise was observed between 1982 and 1989.
An important problem concerns the varying spatial coverage of the combined marine and land observations. Ships have followed preferred navigational routes and large areas of the ocean have been inadequately sampled The effect that this may have on global estimates of SSTs has been tested in frozen grid analyses (Bottomley et al., 1990) and in eigenvector analyses (Folland and Colman, 1988). In the frozen grid analyses, global and hemispheric time series were recalculated using data from 5×5° boxes having data nominated in earlier decades, for example, 1861 – 1870. Remarkably, the small coverage of this period appears surprisingly adequate to estimate long term trends probably because the data are distributed widely in both hemispheres throughout the last 125 years. An eigenvector analysis of combined land and ocean data (Colman, personal communication) isolates an under lying signal of century time scale climate change which is surprisingly uniform geographically even though gross regional changes vary because of other factors.
188.8.131.52 Land and Sea
Regional time series suffer from many near-random errors that can nearly cancel in analyses of global and hemispheric temperatures (Kleshchenko et al, 1988). Atlantic and Scandinavian regions around 1940 and 1945 followed by a sharp cooling, and the strong warming in the South Atlantic and much of the Indian Ocean since about 1965. Almost uniformly cooler conditions in the nineteenth century are clearly seen in all zones, extending into the early twentieth century. Warming around 1920-1940 occurs in most zones, except perhaps over the northern part of the Southern Ocean, with a strong warming, exceeding 0.8°C occurring to the north of 60°N over this period. Note that the polar cap (north of 80°N) has insufficient data foe analysis and insufficient data exist to calculate representative zonal means south of 40°S until after 1950. The cooling after 1950 was mainly confined to the Northern Hemisphere, though weak cooling is evident in the Southern Hemisphere tropics between about 1940 and the early 1950s. There was renewed warming in most Southern Hemisphere zones before 1970. This warming continued until the early 1980s but then slowed markedly. However, very little change of temperature is evident over Antarctica (south of 60°S) since records began there around 1957. Renewed warming is seen in the Northern Hemisphere in all zones after the early 1970s, including small rises in high latitudes, a fact hitherto little appreciated probably because of the marked cooling in the Atlantic/ Barents Sea sector in recent decades which is not seen elsewhere in high latitudes.
184.108.40.206 Seasonal Variations and Changes
In the Southern Hemisphere, there is little difference in recent seasonal trends. Of some concern are substantial differences in seasonal trends before 1900 in the Northern Hemisphere. The relative warmth of summer and coolness of winter at that time reflect considerably greater seasonal differences of the same character in the continental interiors of North America and Asia. It is not clear whether a decrease in the seasonal cycle of temperature that commenced around 1880 is real; it could be due to changes in the circulation of the atmosphere or it may reflect large, seasonally dependent biases in some nineteenth century land data. The latter might arise from the progressive changes of thermometer exposure known to have occurred then (Section 220.127.116.11). The Southern Hemisphere (shows a similar decrease in the seasonal cycle of temperature in the last part of the nineteenth century but with less than half the amplitude of that in the Northern Hemisphere.
18.104.22.168 Day-time and Night-time
Because the ocean has a large heat capacity, diurnal temperature variations in the ocean and in the overlying air are considerably muted compared with those over land and from a climatic point of view, are likely to change little. Over land, diurnal variations are much less restricted so the potential for relative variations in maximum and minimum temperature is larger. Such relative changes might result from changes in cloudiness, humidity, atmospheric circulation patterns, windiness or even the amount of moisture in the ground. Unfortunately, it is not yet possible to assess variations of maximum and minimum temperature on a hemisphere or global scale. However, in the regions discussed below, multi decadal trends of day-time and night- time temperatures have been studied and do not always appear to be the same.
2.5 Precipitation and Evaporation Variations and Changes
2.5.1 Precipitation Over Land
Several large-scale analyses of precipitation changes over the Northern and Southern Hemisphere land masses have been carried out (Bradley et al., 1987; Diaz et al., 1989; Vinnikov et al., 1990). These have demonstrated that during the last few decades precipitation has tended to increase in the mid-latitudes, but decrease in the Northern Hemisphere subtropics and generally increase throughout the Southern Hemisphere. However, these large-scale features contain considerable spatial variability. An apparent increase in precipitation has been found over northern Europe (Schönweise and Birrong, 1990) with a suggestion of a decrease in extreme southern Europe, though these data have not yet been corrected for changing instrumental biases. In the tropics, East African rainfall departures from normal show significant decadal variability, but consistent trends are absent. Summer monsoon rainfall in India also reflects multi decadal changes in climate but consistent trends are also absent. The period 1890 – 1920 was characterised by a high frequency of droughts in India, while 1930 – 1964 had a much lower frequency. Since 1965, the frequency of droughts has again been higher relative to 1930-1964 (Gadgil, 1988), mostly in the wet areas of north eastern India (Gregory, 1989). The dramatic drying of sub Saharan Africa in deserves special attention. Various explanations have been proposed reviewed in Duiyan (1989), see also Semazzi et al (1988) and Wolter (1989). The most consistent result of these studies was to show over the last few decades a pattern of anomalously high SSTs in the Atlantic south of about 10°N lower than normal SSTs in the Atlantic to the north of 10°N and higher SSTs in the tropical Indian Ocean. There has been a distinct weakening of some of these patterns recently and a return to near normal rainfall in 1988 and 1989. Such large-scale changes of SST appear to have a major impact on the sub-Saharan atmospheric circulation (Folland et al 1990, Wolter, 1989). Although SST changes appear to be strongly related to the decreased rainfall since the 1950s, they are probably not the only cause (Nicholson, 1989), Folland et al (1990) show however that at least 60% of the variance of Sahel rainfall between 1901 and 1988 on time scales of one decade and longer is explained by worldwide SST variations. Reductions of rainfall occurred at much the same time immediately south of the Sahel and over much of Ethiopia and the Caribbean. It is important to consider the accuracy of the precipitation data sets. Precipitation is more difficult to monitor than temperatures as it varies much more in time and space. A higher spatial density of data is needed to provide an analysis of variations and trends of computable accuracy. High density data often reside within national meteorological centres, but there is no regular international exchange. The number of stations required to sample a regional rainfall climate adequately varies with region and an adequate number may not always be available.
A severe problem for analysing multi decadal variations of precipitation lies in the fact that the efficiency of the collection of precipitation by rain gauges varies with gauge siting construction and climate (Sevruk 1982, Folland 1988, Legates and Willmott 1990). Major influences are the wind speed during rain, the size distribution of precipitation particle sizes, and the exposure of the rain gauge site. Fortunately, appropriate climatological averages of the first two, highly variable quantities can be used to assess usefully their effects over a long enough period (Folland, 1988, Appendix 1). Collection efficiency has tended to increase as operational practices have improved, often in poorly documented ways that may give artificial upward trends in precipitation in some legions. Thus precipitation data are not completely compatible between countries due to the lack of agreed standards. Of particular concern is the measurement of snowfall from conventional gauges where errors of at least 40% in long term collection efficiency can occur. When precipitation errors are expressed as a percentage of the true rainfall, it is not surprising that they tend to be greatest in high latitude windy, climates and least in wet equatorial regions.
Vinnikov et al, (1990) have carried out detailed collections to USSR data for the varying aerodynamic and wetting problems suffered by gauges. These corrections are incorporated in the record though no aerodynamic corrections were thought necessary in summer. In winter the (positive) aerodynamic
corrections can be large and vary from 5% to 40% (the latter for snow). Wetting corrections, which are also positive and tend to be largest in summer, varied typically in the range 4% to 10% and were applied after correction for aerodynamic effects. Despite these large biases, comparisons of data sets over the USSR from Bradley et al (1987), who only partially corrected for biases, and Vinnikov et al who corrected more extensively, show that most of the important long term variations are apparent in both data sets (Bradley and Groisman, 1989). Many of the major variations apparent in precipitation records are evident in hydrological data such as the rise in the levels of the North American Great Lakes, Great Salt Lake and the Caspian Sea during the early 1980s, and the severe desiccation of the Sahel. Nevertheless, the lack of bias corrections in most rainfall data outside the USSR is a severe impediment to quantitative assessments of rainfall trends.
2.5.2 Rainfall Over The Oceans
Quantitative estimates of precipitation over the oceans are limited to the tropics where they are still very approximate. The mean temperature of the upper surfaces of convective clouds deduced from satellite measurements of outgoing long wave thermal radiation (OLR) are used to estimate mean rainfall over periods of days upwards. The colder the clouds the less is OLR and the heavier the rainfall (Section 4 gives references). Nitta and Yamada (1989) found a significant downward trend in OLR averaged over the global equatorial belt 10°N to 10°S between 1974 and 1987, implying an increase of equatorial rainfall over that time. Arkin and Chelhah (1990) have investigated Nitta and Yamada’s results for this Report. They find that inhomogeneity’s in the OLR data arc sufficiently serious to cast doubt on Nitta and Yamada’s conclusions. However the latter’s claim that equatorial SST has risen over this period seems justified (Flohn and Kapala, 1989). This trend is likely to result in increased deep convection and more rainfall there (Gadgil et al 1984, Graham and Barnett, 1987).
Section 2.5 has shown that some regional scale rainfall trends have occurred over land. However much more attention needs to be paid to data quality and to improving data coverage before more comprehensive conclusions can be drawn about precipitation variations over the global land surface. Precipitation cannot yet be measured with sufficient accuracy over the oceans to reliably estimate trends, even though quite modest changes in SST in the tropics could give rise to important changes in the distribution of tropical rainfall (see also Section 2.9.1).
2.5.3 Evaporation from the Ocean Surface
It is difficult to estimate trends in evaporation from the oceans. An increase is however, expected as a result of an increase in greenhouse gases (Section 5). The most important problem concerns the reliability of measurements of wind speed that are an essential component of evaporation estimates. Oceanic wind speeds have apparently increased in recent decades However, Cardone et al (1990) have demonstrated that much of this increase can be explained by changes in the methods of estimating wind speed from the state of the sea surface, and changes in the heights of anemometers used to measure wind speed on ships. Until these problems are substantially reduced, it is considered that estimates of trends in evaporation are unlikely to be reliable.
2.6 Tropospheric Variations and Change
Tropospheric and stratospheric temperatures are central to the problem of greenhouse warming because general circulation models predict that temperature change with enhanced concentrations of greenhouse gases will have a characteristic profile in these layers, with more warming in the mid- troposphere than at the surface over many parts of the globe, and cooling in much of the stratosphere. One of the “fingerprint” techniques for detecting anthropogenic climate change depends in part on an ability to discriminate between tropospheric warming and stratospheric cooling. (Barnett and Schlesinger,1987). Observational studies of variations in recent temperature changes with height have been made by numerous authors, for example Parker (1985), Barnett (1986), Sellers and Liu (1988) and Karoly (1989). Layer mean temperatures from a set of 63 radiosonde stations covering the globe have been derived by Angell (1988). Most stations have operated continuously only since about 1958 (the International Geophysical Year). The network is zonally well distributed, but about 60% of the stations are in the Northern Hemisphere and only 40% in the Southern Hemisphere. Layer mean temperatures from this network have been integrated for the globe.
Temperatures derived from radiosondes are subject to instrumental biases. These biases have not been assessed in the data used by Angell (1988) although there have been many changes in radiosonde instrumentation over the last thirty years. In 1984-85, international radiosonde comparisons were carried out (Nash and Schmidlin, 1987). Systematic differences between various types of radiosonde were determined for a series of flights which penetrated the tropopause. The estimated heights of the 100mb surface generally differed by up to 10-20 geopotentials metres which is equivalent to average differences of 0.25°C in the layer from the surface to 100 mb.
2.6.2 Comparisons of Recent Tropospheric and Surface Temperature Data
A measure of the robustness of the tropospheric data derived by Angell (1988), at least in recent years, can be obtained by comparing his 850-300mb data with ten years of independent satellite measurements analysed by Spencer and Christy (1990) for 1979-1988. Spencer and Christy have used the average of measurements from microwave sounding units (MSU) aboard two USA National Oceanographic and Atmospheric Administration (NOAA) TIROS-N series of satellites to derive global temperatures in the mid troposphere. Although surface and mid-tropospheric data are likely to show rather different changes in their values over individual regions, better but not perfect, coupling is expected when the data are averaged over the globe as a whole The agreement between the three data sets is surprisingly good, despite recent suggestions that it is poor. Thus, the correlations and root mean squared differences between the surface and MSU data are 0.85oC and 0.08°C respectively, while the correlation between the surface and the radiosonde data is 0.91. The correlation between the two tropospheric data sets is, as expected slightly higher at 0.96 with a root mean squared difference 0.02°C. The latter represents excellent agreement given the relatively sparse network of radiosondes. Note that annual values in both tropospheric data sets have nearly twice the variability of the surface values, as measured by their standard deviation. This partly explains why the root mean square difference between the MSU and surface data is appreciably larger than that between the two tropospheric data sets, despite the high correlation This is, arguably, an indication of genuine climatological differences between the inter annual variability of mid-troposphere temperatures and those of the surface. All three data sets show a small positive trend over the period 1979- 1988, varying from 0.04°C/decade for the MSU data to 0.13°C/decade for the surface data These trends are not significantly different over this short period and again reflect surprisingly great agreement. Further discussion of these results is given in Jones and Wigley (1990).
Water vapour is the most abundant greenhouse gas, and its increases are expected to augment the warming due to increases of other greenhouse gases by about 50%. Trenberth et al (1987) estimate that doubling carbon dioxide concentrations would increase the global concentration of water vapour by about 20%, and Hansen et al (1984) estimate a 33% increase. There is evidence that global water vapour has been a few percent greater during the 1980s than during the 1970s (Elliott et al, 1990). Hense et al (1988), and Flohn et al (1990) find a 20% increase in water vapour content in the mid tropospheric over the equatorial Pacific from 1965-1986 with at least a 10% rise between the surface and the 300mb level. Despite great uncertainties in these data, some increase seems to have taken place. Because of numerous changes in radiosondes, a global assessment of variations prior to 1973 is difficult and trends after 1973 have an uncertain accuracy. (See also Section 8).
2.7 Sub-Surface Ocean Temperature and Salinity Variations
The sub-surface ocean data base is now just becoming sufficient for climate change studies in the North Atlantic and North Pacific basins to be carried out. A few, long, local time series of sub-surface measurements exist, sufficient to alert the scientific community to emerging evidence of decadal scale temperature variability in the Atlantic Ocean. Beginning about 1968, a fresh, cold water mass with its origins in the Arctic Ocean appears to have circulated around the sub-Arctic gyre of the North Atlantic Ocean. This event has been described by Dickson et al (1988) as the Great Salinity Anomaly. Some of this cold, fresh water penetrated to the deep waters of the North Atlantic (Brewer et al, 1983). Recently, Levitus (1989a, b, c d) has carried out a major study of changes of sub surface temperature and salinity of the North Atlantic Ocean between 1955-1959 and 1970-1974. 1955-1959 was near the end of a very warm period of North Atlantic surface waters, but by 1970-1974, the subsequent cool period was well developed. Cooler water extended from near the sea surface to 1400m depth in the subtropical gyre (30-50°N). Beneath the tropical gyre, a warming occurred between the two periods North of this gyre there was an increase in the temperature and salinity of the western sub arctic gyre. The density changes associated with these changes in temperature and salinity indicate that the transport of the Gull Stream may have decreased between the two periods. Temperature difference fields along 24.5°N and 36. 5°N presented by Roemmich and Wunsch (1984) based on data gathered during 1981 and the late 1950s, are consistent with these ideas.
Antonov (1990) has earned out a complementary study for the North Atlantic and North Pacific using subsurface temperature data held in the USSR and SST data from the UK Meteorological Office. He finds that zonal averages of temperature changes between 1957 and 1981 show statistically significant cooling in the upper layers and a warming below 600m when averaged over the North Atlantic as a whole. This agrees well with Levitus results for the North Atlantic Basin mean temperature changes (1957 to 1981) for the North Atlantic and North Pacific, as computed by Antonov. The reasons for some of these changes are partially understood For example, the cooling of the upper 1400m of the subtropical was due to an upward displacement of cooler fresher water. Why this displacement occurred is not definitely known but most probably is related to changes in the large scale wind held over the North Atlantic. Of particular importance is the temperature increase of approximately 0.1°C over, on average, a thousand metre thick layer in the deep North Atlantic because it represents a relatively large heat storage Even the upper few metres of the ocean can store as much heat as the entire overlying atmospheric column of air. Scientists have long recognize(Rossby, 1959) that the ocean could act to store large amounts of heat though small temperature changes in its sub-surface layers for hundreds or thousands of years. When this heat returns to the atmosphere/cryosphere system it could also significantly affect climate (Section 6) gives more details.
The magnitude and extent of the observed changes in the temperature and salinity of the deep North Atlantic are thus large enough that they cannot be neglected in future theories of climate change.
2.8 Variations and Changes in the Cryosphere
Snow, ice, and glacial extent are key variables in the global climate system. They can influence the global heat budget through regulation of the exchange of heat moisture, and momentum between the ocean, land, and atmosphere. Accurate information on cryospheric changes is essential for full understanding of the climate system. Cryospheric data are also integrations of the variations of several variables such as temperature, sunshine amount and precipitation, and for sea-ice, changes in wind stress. Therefore caution must be exercised when interpreting a cryosphenc change Variations in the Greenland and Antarctic ice sheets are discussed subsequently.
2.8.1 Snow Cover
Surface-based observations of snow cover are sufficiently, dense for regional climate studies of the low lying areas of the Northern Hemisphere mid-latitudes. Unfortunately, a hemisphere-wide data set of mid-latitude snow cover observations has not yet been assembled (Barry and Armstrong, 1987). In fact sustained high-quality measurements are generally incomplete (Karl et al, 1989). Since 1966 Northern Hemisphere snow cover maps have been produced operationally on a weekly basis using satellite imagery by NOAA. The NOAA data contain snow/no-snow information for 7921 grid boxes covering the globe and were judged by Scialdone and Roebuck (1987) as the best of four data sets which they compared. Deficiencies have been noted by Wiesnct et al (1987) such as until 1975 the charts did not consistently include Himalayan snow cover, there were occasional extensions of the Southern edge of the snow cover beyond observed surface limits; the seasonal variation of sunlight limits polar coverage in the visible wavelengths; and scattered mountain snows are omitted because of the coarse grid resolution. Data are believed to be usable from 1972 with caution, but are better from 1975 onwards. Consistent with the surface and tropospheric temperature measurements is the rapid decrease in snow cover extent around 1980. This decrease is largest during the transition seasons.
Robinson and Dewey (1990) note that the reduction in snow cover extent during the 1980s is largest in Eurasia where they calculate decreases during autumn and spring of about 13% and 9% respectively relative to the 1970s.
2.8.2 Sea-ice Extent and Thickness
There has been considerable interest in the temporal variability of global sea-ice in both the Arctic and Antarctic (for example, Walsh and Sater, 1981; Sturman and Anderson, 1985). This interest has been increased by general circulation model results suggesting that greenhouse warming may be largest at high latitudes in the Northern Hemisphere. It must be recognized, though, that sea- ice is strongly influenced by surface winds and ocean currents so that the consequences of global warming for changes in sea-ice extent and thickness are unlikely to be straightforward. Sea-ice limits have long been observed by ships, and harbour logs often contain reported dates of the appearance and disappearance of harbour and coastal ice. These observations present many problems of interpretation (Barry, 1986), though they are thought to be more reliable after about 1950. Changes and fluctuations in Arctic sea-ice extent have been analysed by Mysak and Manak (1989); they find no long term trends in sea- ice extent between 1953 and 1984 in a number of Arctic ocean regions but substantial decadal time scale variability was evident in the Atlantic sector. These variations were found to be consistent with the development, movement and decay of the “Great Salinity Anomaly” (noted in Section 2.7). Sea-ice conditions are now reported regularly in marine synoptic observations, as well as by special reconnaissance flights, and coastal radar. Especially importantly, satellite observations have been used to map sea-ice extent routinely since the early 1970s. The American Navy Joint Ice Centre has produced weekly charts which have been digitised by NOAA. Since about 1976 the areal extent of sea-ice in the Northern Hemisphere has varied about a constant climatological level but in 1972-1975, sea-ice extent was significantly less. In the Southern Hemisphere since about 1981, sea-ice extent has also varied about a constant level. Between 1973 and 1980 there were periods of several years when Southern Hemisphere sea-ice extent was either appreciably more than or less than that typical in the 1980s. Gloersen and Campbell (1988) have analysed the Scanning Multi-channel (dual polarization) Microwave Radiometer data from the Nimbus 7 satellite from 1978 1987. They find little change in total global ice area but a significant decrease of open water within the ice. Their time series is short, and it is uncertain whether the decrease is real. Sea-ice thickness is an important parameter but it is much more difficult to measure than sea-ice extent. The heat flux from the underlying ocean into the atmosphere depends on sea-ice thickness. Trends in thickness over the Arctic Ocean as a whole could be a sensitive indicator of global warming. The only practical method of making extensive measurements is by upward-looking sonar from submarines. Apart from a very recent deployment of moorings, data gathering has been carried out on voyages by military submarines. In the past, repeated tracks carried out in summer have either found no change in mean thickness (Wadhams, 1989) or variations that can be ascribed to inter annual variability in summer ice limits and ice concentration (McLaren 1989). Recently however, Wadhams (1990) found a 15% or larger decrease in mean sea-ice thickness between October 1976 and May 1987 over a large region north of Greenland. Lack of a continuous set of observations makes it impossible to assess whether the change is part of a long term trend In the Antarctic, no measurements of thickness variability exist and so far only one geographically extensive set of sea-ice thickness data is available (Wadhams et al 1987).
2.8.3 Land Ice (Mountain Glaciers)
Measurements of glacial ice volume and mass balance are more informative about climatic change than those of the extent of glacial ice, but they are considerably scarcer. Ice volume can be determined from transects of bedrock and ice surface elevation using airborne radio-echo sounding measurements. Mass balance studies performed by measuring winter accumulation and summer ablation are slow and approximate, though widely used (Section 9) discusses changes in the Greenland and Antarctic ice-caps so attention is confined here to mountain glaciers.
A substantial, but not continuous, recession of mountain glaciers has taken place almost everywhere since the latter halt of the nineteenth century (Grove, 1988). This conclusion is based on a combination of mass balance analyses and changes in glacial terminus positions, mostly the latter. Evidence for glacial retreat is found in the Alps, Scandinavia Iceland, the Canadian Rockies, Alaska, Central Asia, the Himalayas, on the Equator, in tropical South America, New Guinea, New Zealand, Patagonia, the sub Antarctic islands and the Antarctic Peninsula (Grove 1988). The rate of recession appears to have been generally largest between about 1920 and 1960. Glacial advance and retreat is influenced by temperature precipitation, and cloudiness. For example, at a given latitude glaciers tend to extend to lower altitudes in wetter cloudier, maritime regions with cooler summers than in continental regions. The complex relation between glaciers and climate makes their ubiquitous recession since the nineteenth century remarkable, temperature changes appeal to be the only plausible common factor (Oerlemans, 1988). The response time of a glacier to changes in environmental conditions varies with its size so that the larger the glacier, the slower is the response (Haeberli et al., 1989). In recent decades, glacial recession has slowed in some regions Makarevich and Rototaeva (1986) show that between 1955 and 1980 about 27% of 104 North American glaciers were advancing and 53% were retreating whereas over Asia only about 5% of nearly 350 glaciers were advancing. Wood (1988) found that from 1960 to 1980 the number of retreating glaciers decreased. This may be related to the relatively cool period in the Northern Hemisphere over much of this time. However Patzelt (1989) finds that the proportion of retreating Alpine glaciers has increased sharply since the early 1980s so that retreat has dominated since 1985 in this region. A similar analysis for other mountain regions after 1980 is not yet available
Permafrost may occur where the mean annual air temperatures are less than 1°C and is generally continuous where mean annual temperature is less than 7°C. The vertical profile of temperature measurements in permafrost that is obtained by drilling boreholes can indicate integrated changes of temperature over decades and longer. However, interpretation of the profiles requires knowledge of the ground conditions as well as natural or human induced changes in vegetation cover Lachenbruch and Marshall (1986) provide evidence that a 2 to 4oC warming has taken place in the coastal plain of Alaska at the permafrost surface over the last 75 to 100 years but much of this rise is probably associated with warming prior to the 1910s. Since the 1930s, there is little evidence for sustained warming in the Alaskan Arctic (Michaels, 1990). A fuller understanding of the relationship between permafrost and temperature requires better information on changes in snow cover, seasonal variations of ground temperature, and the impact of the inevitable disturbances associated with the act of drilling the bore holes (Bairy, 1988).
2.9 Variations and Changes in Atmospheric Circulation
The atmospheric circulation is the main control behind regional changes in wind temperature, precipitation soil moisture and other climatic variables. Variations in many of these factors are quite strongly related through large scale features of the atmospheric circulation as well as through interactions involving the land and ocean surfaces One goal of research into regional changes of atmospheric circulation is to show that the changes of temperature rainfall and other climatic variables are consistent with the changes in frequency of various types of weather pattern.
Climates at the same latitude vary considerably around the globe, while variations in regional temperatures that occur on decadal time scales are far from uniform but form distinctive large-scale patterns. The spatial scale of these climatic patterns is partly governed by the regional scales of atmospheric circulation patterns and of their variations. Changes in weather patterns may involve changes in the quasi stationary atmospheric long waves in the extra tropics or in monsoonal circulations (van Loon and Williams, 1976). Both phen omena have a scale of several thousand kilometres. Their large-scale features are related to the fixed spatial patterns of land and sea, topography, sea temperature patterns and the seasonal cycle of solar heating. Persistent large scale atmospheric patterns tend to be wavelike so that regional changes of atmospheric heating if powerful and persistent enough can give use to a sequence of remote atmospheric disturbances. Thus a number of well separated areas of anomalous temperature and precipitation of opposite character may be produced. The best known examples are in part related to the large changes in SSTs that accompany the El Niño Southern Oscillation (ENSO), whereby changes in the atmosphere over the tropical Pacific often associated with the SST, changes there are linked to atmospheric circulation changes in higher latitudes (Wallace and Gutzler, 1981). The 1988 North American drought has been claimed to be partly a response to persistent positive tropical SST anomalies located to the west of Mexico and to the north of the cold La Nina SST anomalies existing at that time (Tienberth et al., 1988) Such localised SST anomalies may themselves have a much larger scale cause (Namias, 1989).
An emerging topic concerns observational evidence that the 11 year solar cycle and the stratospheric quasi biennial oscillation (QBO) of wind direction near the equator are linked to changes in tropospheric circulation in the Northern Hemisphere (van Loon and Labitzke 1988). Coherent variations in tropospheric circulation are claimed to occur over each 11 year solar cycle in certain regions but their character depends crucially on the phase (easterly or westerly) of the QBO. No mechanism has been proposed (of this effect and the data on the QBO cover only about 3.5 solar cycles so that the reality of the effect is very uncertain. However, Bainston and Livesey (1989) in a careful study find evidence for statistically significant influences of these factors on atmospheric circulation patterns in the Northern Hemisphere extra tropics in winter. Many previous largely unsubstantiated claims of links between the 11 year, and other solar cycles, and climate are reviewed in Pittock (1983). Section 2 discusses current thinking about the possible magnitude of the physical forcing of global climate by solar radiation changes in some detail. Several examples are now given of links between changes in atmospheric circulation over the last century and regional-scale variations or trends of temperature.
2.9.1 El Nino-Southern Oscillation (ENSO) Influences
ENSO is the most prominent known source of inter annual variability in weather and climate around the world, though not all areas are affected. The Southern Oscillation (SO) component of ENSO is an atmospheric pattern that extends over most of the global tropics. It principally involves a seesaw in atmospheric mass between regions near Indonesia and a tropical and sub-tropical south cast Pacific Ocean region centred near Easter Island. The influence of ENSO sometimes extends to higher latitudes (see Section 2.9.3) The El Nino component of ENSO is an anomalous warming of the eastern and central tropical Pacific Ocean. In major Warm Events, warming extends over much of the tropical Pacific and becomes clearly linked to the atmospheric SO pattern. An opposite phase of Cold Events with opposite patterns of the SO is sometimes referred to as La Niña. ENSO events occurs every 3 to 10 years and have far leaching climatic and economic influences around the world (Ropelewski and Halpeit, 1987). Places especially affected include the tropical central and East Pacific islands, the coast of north Peru, eastern Australia, New Zealand (Salinger, 1981) Indonesia, India (Parthasarathy and Pant, 1985) and parts of eastern (Ogallo, 1989) and southern Africa (van Heeiden et al., 1988). A fuller description of ENSO can be found in Rasmusson and Carpenter (1982) and Zebiak and Cane (1987). Over India, the occurrence of ENSO and that of many droughts (see Section 2.5.1) is strikingly coincident. Droughts tend to be much more frequent in the first year of an ENSO event though, intriguingly this is often before the ENSO event has fully developed. However, not all Indian droughts are associated with ENSO.
While ENSO is a natural part of the Earth’s climate, a major issue concerns whether the intensity or frequency of ENSO events might change as a result of global warming. Until recently, the models used to examine the climatic consequences of enhanced greenhouse forcing had such simplified oceans that ENSOs could not be simulated. Some models now simulate ENSO like, but not entirely realistic SST variations (Section 4). Unfortunately long term variations in ENSO cannot be studied yet using models The observational record levels that ENSO events have changed in frequency and intensity in the past. The strong SO fluctuations from 1880 to 1920 led to the discovery and naming of the SO (Walker and Bliss, 1932) and strong SO events are clearly evident in recent decades. A much quieter period occurred from the late 1920s to about 1950, with the exception of a very strong multi-year ENSO in 1939-1942 (Trenberth and Shea, 1987; Cooper et al., 1989). Quinn et al (1987) (covering the past 450 years) and Ropelewski and Jones (1987) have documented historical ENSO events as seen on the northwest coast of South America. Therefore, the potential exists for a longer palaeo-record based on river deposits, ice cores, coral growth rings and tree rings.
During ENSO events, the heat stored in the warm tropical western Pacific is transferred directly or indirectly to many other parts of the tropical oceans. There is a greater than normal loss of heat by the tropical oceans, resulting in a short period warming of many, though not all, parts of the global atmosphere (Pan and Oort, 1983). Consequently, warm individual years in the record of global temperatures are often associated with El Niños. Maxima in global
temperatures tend to occur about three to six months after the peak warmth of the El Niño (Pan and Oort, 1983).
From an inter-decadal perspective, ENSO is a substantial source of climatic noise which can dominate the tropical temperature record. It is possible to remove ENSO signals statistically (for example Jones, 1989) to give a smoother global temperature curve; in this way 20 to 30% of decadal and shorter time scale variance is removed. The warming of the globe in the last 15 years than is reduced by about 0.1oC, I.e., by about one half. However, other temperature signals might also be partly removed at the same time, for example, those relating to the effects of volcanic eruptions, though some ability to separate these signals has recently been shown. (Mass and Portman, 1989) (See also Section 2). As it is unclear whether ENSO might and contribute directly to long- term global warming, it seems preferable to ENSO variability as an integral part of the global climate record.
2.9.2 The North Atlantic
The early twentieth century cooling of the Northern Hemisphere occurs by a period of intensified westerlies in the extratropical Northern Hemisphere, especially in the Atlantic sector, that affected most of the year. An extensive discussion is given in Lamb (1977). The global warming which took place in the 1920s and 1930s was largest in the extratropical North Atlantic and in the Artic and coincided with the latter part of the period of intense westerlies. The westerly epoch is regarded as finishing around 1938 ( Makrogiannis et al., 1982). The effects of the enhanced westerlies on surface climate were clearest in winter when there was an absence of very cold outbreaks over Europe and winter were persistently mild. Rogers (1985) noted that the best correlation between temperatures in Europe and wind direction is with the westerly component, largely reflecting whether or not the encroaching air masses have had an oceanic moderating influence imposed on them. Wallen (1986) notes that the period of intense westerly flow also affected summer temperatures in western Europe, making them generally cooler between the late 1890s and about 1920, giving striking decreases in the differences between July and January temperature. This suggests that at least a small part of the decrease in the annual range of northern Hemisphere land surface air temperature seen at this time may be real.
Variations in the westerly index are associated with changes in the depth of the Iceland low pressure centre both near the surface and in the troposphere (van Loon and Rogers, 1978). European temperatures well downstream are quite strongly related to this. Atlantic pressure index is largest, I.e., pressure over Iceland is lowest relative to the Azores. Relative to the level of the westerly pressure index, there is a long-term increase of European winter surface temperature much as noted by Moses et al (1987) for a small set of stations in western Europe. Above normal winter temperatures in Europe tend to go hand- in-hand with below normal temperatures in Greenland and the Canadian Artic where there are increased northerlies as a result of the deeper Iceland Low. Stronger westerlies over the Atlantic, do not, therefore, account for the Artic warming of 1920s and 1930s on their own; in fact, they preceded by 20 years. Iceland (Einarsson, 1977), whereas Greenland and Northern Canada did not warm until the mid – 1920s. (Rogers, 1985). When both Greenland and Northern Europe had above normal temperature, especially in the winter half year, the atmospheric circulation was more zonal around most of the Artic, not just in the Atlantic sector. This was associated with increased cyclonic activity over the whole Artic basin which increased the frequency of zonal flows over the higher latitudes of the continents. Note that the character of the warming experienced in the higher latitude Northern Hemisphere in the 1920s and 1930s differs from that of the mid-1970s to early 1980s when the North Atlantic and Arctic stayed cool, or in parts, cooled further. Inter-decadal changes in the west African monsoon circulation which have particularly affected Sub-Saharan African rainfall (introduced in Section 2.5.1). The main change in atmospheric circulation has been in the convergence of winds into sub Saharan Africa in summer from the north and the south (Newell and Kidson, 1984), less intense convergence gives less rainfall (Folland et al, 1990). Drier years are also often accompanied by a slightly more southerly position of the main wind convergence (rain bearing) zone. The North Atlantic subtropical high pressure belt also tended to extend further southward and eastward during the summer in the dry Sahel decades (Wolter and Hastenrath, 1989).
2.9.3 The North Pacific
Circulation changes in the North Pacific have recently been considerable and have been linked with regional temperature changes.) This index is closely
related to changes in the intensity of the Aleutian low pressure centre. It is also quite strongly linked to a pattern ol atmospheric circulation variability known as the Pacific North American (PNA) pattern (Wallace and Gutzler, 1981) which is mostly confined to the North Pacific and to extratropical North America. All five winter months showed a much deeper Aleutian Low in the period 1977 to 1988 with induced pressure over nearly all the extratropical North Pacific north of about 32 N (Flohn et al 1990) The change in pressure appears to have been unusually abrupt, other examples of such climatic discontinuities have been analysed (Zhang et al., 1983) though discontinuities can sometimes be artefacts of the statistical analysis of irregular time series The stronger Aleutian Low resulted in warmer, moister air being carried into Alaska while much colder air moved south over the North Pacific. These changes account for the large Pacific temperature anomalies tor 1980-1989 which are even clearer for the decade 1977-1986. This decade had a positive anomaly (relative to 1951-1980) of over 1.5°C in Alaska and negative anomaly of more than 0.75°C in the central and western North Pacific. The above changes are likely to have been related to conditions in the equatorial Pacific 1977 – 1987 was a period when much of the tropical Pacific and tropical. Indian Oceans had persistently above normal SSTs (Nitta and Yamada, 1989). Very strong El Nino events and a lack of cold tropical La Niña events in the period 1977-1987 contributed to this situation.
2.9.4 Southern Hemisphere
In Antarctica strong surface temperature inversions form in winter but elsewhere in the Southern Hemisphere maritime influences dominate. The SO has a pronounced influence on precipitation over Australia (Pittock, 1975) and also affects New Zealand temperatures and precipitation (Gordon, 1986). However, the best documented regional circulation-temperature relationship in the Southern Hemisphere is that between an index of the meridional (southerly and northerly) wind (Trenbeith 1976) and New Zealand temperature. The index is calculated by subtracting sea level pressure values measured at Hobail (Tasmania) horn those at Chatham Island (east of New Zealand). A tendency for more northerly mean flow across New Zealand (Hobart pressure relatively low) especially from about 1952 to 1971 has been related to generally warmer conditions in New Zealand after 1950. However a return to more southerly (colder) flow after 1971 is not strongly reflected in New Zealand temperatures so the recent warmth may be related to the general increase in temperature in much of the Southern Hemisphere. This finding indicates that regional temperature changes due to a future greenhouse warming are likely to result from an interplay between large scale warming and changes in local weather patterns.
Clouds modify both the shortwave (solar) and longwave (terrestrial) radiation, the former by reflection and the latter by absorption. Therefore, they may cause a net warming or cooling of global temperature, depending on their type distribution, coverage, and radiative characteristics (Sommerville and Remer, 1984 Cess and Pottei, 1987) Ramanathan et al (1989) show that with todays distribution and composition of clouds, their overall effect is to cool the Earth (Section 3.3.4). Changes in cloudiness are therefore likely to play a significant role in climate change. Furthermore, local and regional climate variations can be strongly influenced by the amount of low, middle and high clouds.
Observations of cloudiness can be made from the Earth’s surface by trained observers from land stations or ocean vessels or by automated systems. Above the Earth’s surface, aircraft or space platforms are used (Rossow 1989, McGuffie and Henderson-Sellers 1989a) Suipnsingly, surface-based observations of cloudiness give closely similar results to space based observations. Careful and detailed inter-comparisons, undertaken as a preliminary part of the International Satellite Cloud Climatology Project (ISCCP) by Sze et al (1986), have demonstrated conclusively that surface and space based observations are highly correlated Space-based observations of cloudiness from major international programs such as ISCCP are not yet available for periods sufficiently long to detect long-term changes.
2.10.7 Cloudiness Over Land
Henderson Sellers (1986 & 1989) and McGuffie and Henderson-Sellers (1989b) have analysed changes in total cloud cover over Europe and North America during the twentieth century. It was found that annual mean cloudiness increased over both continents. Preliminary analyses for Australia and the Indian sub-continent also give increases in cloudiness. The increases are substantial 7% of initial cloudiness/50 years over India 6%/80 years over Europe 8%/80 years for Australia and about 10%/90 years for North America. These changes may partly result from alterations in surface-based cloud observing practices and in the subsequent processing of cloud data. This may be especially true of the large increase in cloudiness apparently observed in many areas in the 1940s and 1950s. At this time (about 1949 or later, depending on the country) the synoptic meteorological code, from which many of these observations are derived generally underwent a major change but not in the USA, USSR and Canada. Observers began recording cloud cover in oktas (eighths) instead of in tenths. When skies were partly cloudy, it is possible that some observers who had been used to making observations in the decimal system converted decimal observations of cloud cover erroneously to the same number of oktas thereby overestimating the cloud cover.
Recently, Karl and Steurer (1990) have compared daytime cloudiness statistics over the USA with data from automated sunshine recorders. They indicate that there was a much larger increase of annual cloud cover during the 1940s than can be accounted for by the small observed decrease in the percentage of possible sunshine. The large increase of cloudiness may be attributed to the inclusion of the obscuring effects of smoke, haze, dust, and fog in cloud cover reports from the 1940s onward (there being no change in the recording practice from tenths to oktas in the USA). The increase in cloudiness after 1950 may be real because an increase is consistent with changes in the temperature and precipitation records in the USA, including the decreased diurnal temperature range.
Observed land based changes in cloudiness are difficult to assess. Nonetheless, total cloud amount appears to have increased in some continental regions, a possibility supported by noticeable reductions in the diurnal range of temperature in some of these regions. Elsewhere the cloudiness record cannot be interpreted reliably.
2.10.2 Cloudiness Over The Oceans
Ocean-based observations of cloud cover since 1930 have been compiled by Warren et al (1988). The data are derived from maritime synoptic weather observations. Then number varies between 100,000 and 2 000,000 each year increasing with time and the geographic coverage also changes. The data
indicate that an increase in marine cloudiness exceeding one percent in total sky covered on a global basis took place from the 1940s to the 1950s. This increase is not reflected in the proportion of observations having a clear sky or a complete overcast. The largest increases were in stratocumulus clouds in Northern Hemisphere mid-latitudes and in cumulonimbus in the tropics. Since 1930, mean cloudiness has increased by 3-4 percent of the total area of sky in the Northern Hemisphere and by about half of this value in the Southern Hemisphere. Fixed ocean weather ships placed after 1945 in the North Atlantic and North Pacific with well trained observers, showed no trends in cloudiness between the 1940s and 1950s when other ship data from nearby locations showed relatively large increases, changes of the same sign as those in available land records (Section 2.10.1). It is clearly not possible to be confident that average global cloudiness has really increased.
2.11 Changes of climate Variability and Climatic Extremes
Aspects of climate variability include those associated with day-to-day changes inter-seasonal and inter annual variations and the spatial variability associated with horizontal gradients. A pervasive problem for assessment of changes in the temporal variability of climate is the establishment of a reference level about which to calculate that variability. For example, the results of an investigation to determine whether the variability of monthly mean values was changing could give different results depending on how the average, or baseline, climate was calculated. If the climate was changing, calculations of variability would depend on whether a fixed baseline was used or whether it was allowed to change smoothly or discontinuously. Additionally, attempts to identify whether may be critically dependent on the definition of the threshold value above which an extreme is defined. An attempt was made for this Report to estimate changes in the number of extremes of monthly average temperature values for a globally distributed set of bout 150 stations, all having at least 60 years of record. The results were found to be very sensitive to the threshold chosen to define the extreme, and to the baseline climatology selected, so in extremes or variability were possible.
Interest in climatic variability often includes that on a daily time-scales: a common question concerns whether daily temperatures above or below given
threshold values, for example, frosts are changing in their frequency. Although considerable local climatic information exists on daily time scales in any national meteorological data centre, such information covering the globe as a whole is not readily available at any one centre. Thus, many of the data needed for a comprehensive assessment of changes in variability still and need to be assembled and a scientifically sound method of analyses needs to be developed. Nevertheless, a few comments can be made about variability which are discussed below.
Several researchers have assessed relationships between anomalously cold or warm seasons and daily temperature variability in the USA. Brinkmann (1983) and Agee (1982) both find reduced day-to-day variability in anomalously warm winters, though Brinkmann (1983) provides evidence of enhanced day-to-day variability during anomalously warm summers. It is likely, however, that these relationships are highly sensitive to the choice of region. There appears to be little relation between inter annual variability and the relative warmth or coldness of decadal averages. Although, Daiz and Quayle (1980) found a tendency for increased variability in the USA during the relatively warm years of the mid-twentieth century (1921-1955), Karl (1988) found evidence for sustained episodes (decades) of very high and low inter annual variability with little change in baseline climate. Furthermore, Balling et al (1990) found no relationship between mean and extreme values of temperature in the desert Southwest of the USA.
2.11.2 Droughts and Floods
An important question concerns variations in areas affected by severely wet (‘ flood ) or drought conditions. However, drought and moisture indices calculated for Australia, parts of the Soviet Union, India, the USA, and China do not show systematic long-term changes. Although this does not represent anything like a global picture, it would be difficult to envisage a worldwide systematic change in variability without any of these diverse regions participating. It is noteworthy that the extended period of drought in the Sahel between 1968 and 1987 exhibited a decreased inter annual variability of rainfall compared with the previous 40 years even though the number of stations used remained nearly the same.
2.11.3 Tropical Cyclones
Tropical cyclones derive their energy mainly from the latent heat contained in the water vapour evaporated from the oceans. As a general rule, for tropical cyclones to be sustained, SSTs must be at or above 26°C to 27°C at the present time. Such values are confined to the tropics, as well as some subtropical regions in summer and autumn. The high temperatures must extend through a sufficient depth of ocean that the wind and wave action of the storm itself does not prematurely dissipate its energy source. For a tropical cyclone to develop, its parent disturbance must be about 7° of latitude or more from the equator. Many other influences on tropical cyclones exist which are only partly understood. Thus, ENSO modulates the frequency of tropical storms in some regions, for example, over the north-west Pacific, mainly south of Japan (Li, 1985, Yoshino, 1990), East China (Fu and Ye, 1988) and in the central and southwest Pacific (Revell and Gaulter, 1986). The reader is referred to Nicholls (1984), Gray (1984), Emanuel (1987) and Raper (1990) for more details.
Have tropical cyclone frequencies or their intensities increased as the globe has warmed over this century. Current evidence, does not support this idea perhaps because the warming is not yet large enough to make its impact felt In the North Indian Ocean, the frequency of tropical storms has noticeably) decreased since 1970 while SSTs have risen here since 1970, probably more than in any other region ( See also Raper, 1990). There is little trend in the Atlantic, though pronounced decadal variability is evident over the last century. There have been increases in the recorded frequency of tropical cyclones in the eastern North Pacific, the southwest Indian Ocean, and the Australian region since the late 1950s. However, these increases are thought to be predominantly artificial and to result from the introduction of better monitoring procedures. Relatively good records of wind speed available from the North Atlantic and western Pacific oceans do not suggest that there has been a change toward more intense storms either.
2.11.4 Temporales of Central America
Temporales are cyclonic tropical weather systems that affect the Pacific side of Central America and originate in the Pacific Inter-tropical Convergence Zone. Very heavy rainfall totals over several days occur with these systems, but unlike hurricanes their winds are usually weak. Their atmospheric structure is also
quite different from that of hurricanes as they possess a cold mid-tropospheric core. Temporales typically last several days, are slow moving and cause damaging floods and landslides in the mountainous regions of Central America. Records of temporales are available since the 1950s. They were markedly more frequent in the earlier than in the later part of this period. Thus there was an average of 2-4 temporales per year in 1952-1961 (Hastenrath, 1988), only 1959 had no temporales. In 1964-1983, the average reduced to 0.45 temporales per year and in 12 of these years there were none. When the evidence in (Section 2.11) is taken together, we conclude that there is no evidence of an increasing incidence of extreme events over the last few decades. Indeed, some of the evidence points to recent decreases, for example in cyclones over the North Indian Ocean and temporales over Central America
The most important finding is a warming of the globe since the late nineteenth century of 0.45 ± 0.15°C, supported by a worldwide recession of mountain glaciers. A quite similar warming has occurred over both land and oceans. This conclusion is based on an analysis of new evidence since previous assessments (Scope 29, 1986) and represents a small reduction in previous best estimates of global temperature change. The most important diagnosis that could not be made concerns temperature variations over the Southern Ocean. Recent transient model results indicate that this region may be resistant to long term temperature change. A data set of blended satellite and ship SST data is now becoming available and may soon provide an initial estimate of recent Southern Ocean temperature changes. Precipitation changes have occurred over some large land regions in the past century but the data sets are so poor that only changes of large size can be monitored with any confidence.
Some substantial regional atmospheric circulation variations have occurred over the last century notably over the Atlantic and Europe. Regional variations in temperature trends have also been quite substantial. This indicates that, in future regional climatic changes may sometimes be quite diverse. Natural climate variations have occurred since the end of the last glaciation. The Little Ice Age in particular invoked global climate changes of comparable magnitude to the warming of the last century. It is possible that some of the warming since the nineteenth century may reflect the cessation of Little Ice Age conditions. The rather rapid changes in global temperature seen around 1920-1940 are very likely to have had a mainly natural origin. Thus, a better understanding of past variations is essential if we are to estimate reliably the extent to which the warming over the last century, and future warming, is the result of an increase of greenhouse gases.