11.1.2 Introduction to Regional Projections

Assessments of climate change projections are provided here on a region-by-region basis. The discussion is organised according to the same continental-scale regions used by Working Group II (WGII) in the Fourth Assessment Report (AR4) and in earlier assessments: Africa, Europe and Mediterranean, Asia, North America, Central and South America, Australia-New Zealand, Polar Regions and Small Islands. While the topics covered vary somewhat from region to region, each section includes a discussion of key processes of importance for climate change in that region, relevant aspects of model skill in simulating current climate, and projections of future regional climate change based on global models and downscaling techniques.

Each of these continental-scale regions encompasses a broad range of climates and is too large to be used as a basis for conveying quantitative regional climate change information. Therefore, each is subdivided into a number of sub-continental or oceanic regions. The sub-continental regions as defined in Table 11.1 are the framework for developing specific regional or sub-continental robust statements of projected change.

Area-averaged temperature and precipitation changes are presented from the coordinated set of climate model simulations archived at the Program for Climate Model Diagnosis and Intercomparison (PCMDI; subsequently called the multi-model data set or MMD). The regions are very close to those initially devised by Giorgi and Francesco (2000) with some minor modifications similar to those of Ruosteenoja et al. (2003). They have simple shapes and are no smaller than the horizontal scales on which current AOGCMs are useful for climate simulations (typically judged to be roughly 1,000 km).

These regional averages have some deficiencies for discussion of the AOGCM projections. In several instances, the simple definition of these boxes results in spatial averaging over regions in which precipitation is projected to increase and decrease. There are also sub-regions where the case can be made for a robust and physically plausible hydrological response, information about which is lost in the regional averages. Partially to help in discussing these features, this chapter also uses maps of temperature and precipitation responses, interpolated to a grid with 128 longitudes by 64 latitudes which is typical of many of the lower-resolution atmospheric models in the MMD.

In the regional discussion to follow, the starting points are temperature and precipitation. Changes in temperature are introduced in each continental section by plotting for each of the regions the evolution of the range of projected decadal mean change for the A1B scenario through the 21st century (simulations hereafter referred to as MMD-A1B). These are put into the context of observed changes in the 20th century by plotting the observed changes and how well the models reproduce these. This summary information is displayed for continental regions in Box 11.1, which also contains details of how the figures were constructed. The equivalent figures for the individual regions of each continental-scale region are displayed in the following sections. These are constructed in the same way as Box 11.1, Figure 1. The 20th-century parts of these figures are also displayed in Section 9.4, where more details on their construction are provided. The discussion on precipitation provides a limited view of hydrological changes. Supplementary Material Figure S11.1 expands on this issue by comparing the annual mean responses in precipitation and in precipitation minus evaporation over the 21st century in the MMD-A1B projections. Over North America and Europe, for example, the region of drying in the sense of precipitation minus evaporation is shifted poleward compared to the region of reduced precipitation. A summary of the more significant hydrological cycle changes from the regional discussions is presented in Box 11.1.

Table 11.1 provides detailed information for each region generated from the MMD-A1B models focusing on the change in climate between the 1980 to 1999 period in the 20th-century integrations and the 2080 to 2099 period. The distribution of the annual and seasonal mean surface air temperature response and percentage change in precipitation are described by the median, the 25 and 75% values (half of the models lie between these two values) and the maximum and minimum values in the model ensemble. Information on model biases in these regional averages for the 1980 to 1999 simulations is provided in Supplementary Material Table S11.1 in a similar format. Maps of biases are referred to in some of the following and are included in the Supplementary Material as well. Data sources used in these comparisons are listed in the table and figure captions where these biases are displayed.

Most of the discussion focuses on the A1B scenario. The global mean near-surface temperature responses (between the period 1980 to 1999 of the 20th-century integrations and the period 2080 to 2099) in the ensemble mean of the MMD models are in the ratio 0.69:1:1.17 for the B1:A1B:A2 scenarios. The local temperature responses in nearly all regions closely follow the same ratio, as discussed in Chapter 10 and as illustrated in Supplementary Material Figures S11.2 to S11.4. Therefore, little is gained by repeating the discussion of the A1B scenario for the other scenarios. The ensemble mean local precipitation responses also approximately scale with the global mean temperature response, although not as precisely as the temperature itself. Given the substantial uncertainties in hydrological responses, the generally smaller signal/noise ratio and the similarities in the basic structure of the AOGCM precipitation responses in the different scenarios, a focus on A1B seems justified for the precipitation as well. The overall regional assessments, however, do rely on all available scenario information.

Given the dominantly linear response of the models, the 2080 to 2099 period allows the greatest clarity of the background climate change underlying the interannual and decadal variability. In the ensemble mean AOGCM projections there is no indication of abrupt climate change, nor does the literature on individual models provide any strong suggestions of robust nonlinearities. Some local temporal nonlinearities are to be expected, for example as the sea ice boundary retreats from a particular location in the Arctic. While the possibility exists that changes of more abrupt character could happen, such as major ocean circulation or land surface/vegetation change, there is little basis to judge the plausibility of these factors (see Chapter 10). Therefore, this discussion is based on this linear picture.

Table 11.1 also provides some simple estimates of the signal-to-noise ratio. The signal is the change in 20-year means of seasonal or annual mean temperature or precipitation. The noise is an estimate of the internal variability of 20-year means of seasonal or annual mean temperature or precipitation, as generated by the models. The signal-to-noise ratio is converted into the time interval that is required before the signal is clearly discernible, assuming that the signal grows linearly over the century at the average rate in the ensemble mean A1B projection. ‘Clearly discernible’ is defined in this context as distinguishable with 95% confidence. As an example, the annual mean precipitation increase in northern Europe (NEU) (Table 11.1) is clearly discernible in these models after 45 years, meaning that the 20-year average from 2025 to 2044 will be greater than the 20-year mean over 1980 to 1999 with 95% confidence, accounting only for the internal variability in the models and no other sources of uncertainty. In contrast, the annual temperature response in Southeast Asia (SEA) rises above the noise by this measure after only 10 years, implying that the average temperature over the period 1990 to 2009 is clearly discernible in the models from the average over the control period 1980 to 1999. This measure is likely an overestimate of the time of emergence of the signal as compared to that obtained with more refined detection strategies (of the kind discussed in Chapter 9). This noise estimate is solely based on the models and must be treated with caution, but it would be wrong to assume that models always underestimate this internal variability. Some models overestimate and some underestimate the amplitude of the El Niño-Southern Oscillation (ENSO), for example, thereby over- or underestimating the most important source of interannual variability in the tropics. On the other hand, few models capture the range of decadal variability of rainfall in West Africa, for example (Hoerling, et al., 2006; Section 8.4).

Also included in Table 11.1 is an estimate of the probability of extremely warm, extremely wet and extremely dry seasons, for the A1B scenario and for the time period 2080 to 2099. An ‘extremely warm’ summer is defined as follows. Examining all of the summers simulated in a particular realisation of a model in the 1980 to 1999 control period, the warmest of these 20 summers can be computed as an estimate of the temperature of the warmest 5% of all summers in the control climate. The period 2080 to 2099 is then examined, and the fraction of the summers exceeding this warmth determined. This is referred to as the probability of extremely warm summers. The results are tabulated after averaging over models, and similarly for both extremely low and extremely high seasonal precipitation amounts. Values smaller (larger) than 5% indicate a decrease (increase) in the frequency of extremes. This follows the approach in Weisheimer and Palmer (2005) except that this chapter compares each model’s future with its own 20th century to help avoid distortions due to differing biases in the different models. The results are shown in Table 11.1 only when 14 out of the 21 models agree as to the sign of the change in frequency of extremes. For example, in Central North America (CNA), 15% of the summers in 2080 to 2099 in the A1B scenario are projected to be extremely dry, corresponding to a factor of three increase in the frequency of these events. In contrast, in many regions and seasons, the frequency of extreme warmth is 100%, implying that all seasons in 2080 to 2099 are warmer than the warmest season in 1980 to 1999, according to every model in this ensemble.

In each continental section, a figure is provided summarising the temperature and precipitation responses in the MMD-A1B projection for the last two decades of the 21st century. These figures portray a multi-model mean comprising individual models or model ensemble means where ensembles exist. Also shown is the simple statistic of the number of these models that show agreement in the sign of the precipitation change. The annual mean temperature and precipitation responses in each of the 21 separate AOGCMs are provided in Supplementary Material Figures S11.5 to 11.12 and S11.13 to 11.20, respectively.

Recent explorations of multi-model ensemble projections seek to develop probabilistic estimates of uncertainties and are provided in the Supplementary Material Table S11.2. This information is based on the approach of Tebaldi et al. (2004a,b; see also section 11.10.2).

Box 11.1: Summary of Regional Responses

As an introduction to the more detailed regional analysis presented in this chapter, Box 11.1, Figure 1 illustrates how continental-scale warming is projected to evolve in the 21st century using the MMD models. This warming is also put into the context of the observed warming during the 20th century by comparing results from that subset of the models incorporating a representation of all known forcings with the observed evolution (see Section 9.4 for more details). Thus for the six continental regions, the figure displays: 1) the observed time series of the evolution of decadally averaged surface air temperature from 1906 to 2005 as an anomaly from the 1901 to 1950 average; 2) the range of the equivalent anomalies derived from 20th-century simulations by the MMD models that contain a full set of historical forcings; 3) the evolution of the range of this anomaly in MMD-A1B projections between 2000 and 2100; and 4) the range of the projected anomaly for the last decade of the 21st century for the B1, A1B, and A2 scenarios. For the observed part of these graphs, the decadal averages are centred on the decade boundaries (i.e., the last point is for 1996 to 2005), whereas for the future period they are centred on the decade mid-points (i.e., the first point is for 2001 to 2010). The width of the shading and the bars represents the 5 to 95% range of the model results. To construct the ranges, all simulations from the set of models involved were considered independent realisations of the possible evolution of the climate given the forcings applied. This involved 58 simulations from 14 models for the observed period and 47 simulations from 18 models for the future. Important in this representation is that the models’ estimate of natural climate variability is included and thus the ranges include both the potential mitigating and amplifying effects of variability on the underlying signal. In contrast, the bars representing the range of projected change at the end of the century are constructed from ensemble mean changes from the models and thus provide a measure of the forced response. These bars were constructed from decadal mean anomalies from 21 models using A1B scenario forcings, from the 20 of these models that used the B1 forcings and the 17 that used the A2 forcings. The bars for the B1 and A2 scenarios were scaled to approximate ranges for the full set of models. The scaling factor for B1 was derived from the ratio between its range and the A1B range of the corresponding 20 models. The same procedure was used to obtain the A2 scaling factor. Only 18 models were used to display the ranges of projected temperature evolution as the control simulations for the other 3 had a drift of >0.2°C per century, which precludes clearly defining the decadal anomalies from these models. However, anomalies from all 21 models were included in calculating the bars in order to provide the fullest possible representation of projected changes in the MMD. Comparison of these different representations shows that the main messages from the MMD about projected continental temperature change are insensitive to the choices made. Finally, results are not shown here for Antarctica because the observational record is not long enough to provide the relevant information for the first part of the 20th century. Results of a similar nature to those shown here using the observations that are available are presented in Section 11.8.

Box 11.1, Figure 1. Temperature anomalies with respect to 1901 to 1950 for six continental-scale regions for 1906 to 2005 (black line) and as simulated (red envelope) by MMD models incorporating known forcings; and as projected for 2001 to 2100 by MMD models for the A1B scenario (orange envelope). The bars at the end of the orange envelope represent the range of projected changes for 2091 to 2100 for the B1 scenario (blue), the A1B scenario (orange) and the A2 scenario (red). The black line is dashed where observations are present for less than 50% of the area in the decade concerned. More details on the construction of these figures are given in Section 11.1.2.

Box 11.1, Figure 2 serves to illustrate some of the more significant hydrological changes, with the two panels corresponding to DJF and JJA. The backdrop to these figures is the fraction of the AOGCMs (out of the 21 considered for this purpose) that predict an increase in mean precipitation in that grid cell (using the A1B scenario and comparing the period 2080 to 2099 with the control period 1980 to 1999). Aspects of this pattern are examined more closely in the separate regional discussions.

Box 11.1, Figure 2. Robust findings on regional climate change for mean and extreme precipitation, drought, and snow. This regional assessment is based upon AOGCM based studies, Regional Climate Models, statistical downscaling and process understanding. More detail on these findings may be found in the notes below, and their full description, including sources is given in the text. The background map indicates the degree of consistency between AR4 AOGCM simulations (21 simulations used) in the direction of simulated precipitation change.

Robust findings on regional climate change for mean and extreme precipitation, drought and snow are highlighted in the figure with further detail in the accompanying notes.

(1) Very likely annual mean increase in most of northern Europe and the Arctic (largest in cold season), Canada, and the North-East USA; and winter (DJF) mean increase in Northern Asia and the Tibetan Plateau.

(2) Very likely annual mean decrease in most of the Mediterranean area, and winter (JJA) decrease in southwestern Australia.

(3) Likely annual mean increase in tropical and East Africa, Northern Pacific, the northern Indian Ocean, the South Pacific (slight, mainly equatorial regions), the west of the South Island of New Zealand, Antarctica and winter (JJA) increase in Tierra del Fuego.

(4) Likely annual mean decrease in and along the southern Andes, summer (DJF) decrease in eastern French Polynesia, winter (JJA) decrease for Southern Africa and in the vicinity of Mauritius, and winter and spring decrease in southern Australia.

(5) Likely annual mean decrease in North Africa, northern Sahara, Central America (and in the vicinity of the Greater Antilles in JJA) and in South-West USA.

(6) Likely summer (JJA) mean increase in Northern Asia, East Asia, South Asia and most of Southeast Asia, and likely winter (DJF) increase in East Asia.

(7) Likely summer (DJF) mean increase in southern Southeast Asia and southeastern South America

(8) Likely summer (JJA) mean decrease in Central Asia, Central Europe and Southern Canada.

(9) Likely winter (DJF) mean increase in central Europe, and southern Canada

(10) Likely increase in extremes of daily precipitation in northern Europe, South Asia, East Asia, Australia and New Zealand.

(11) Likely increase in risk of drought in Australia and eastern New Zealand; the Mediterranean, central Europe (summer drought); in Central America (boreal spring and dry periods of the annual cycle).

(12) Very likely decrease in snow season length and likely to very likely decrease in snow depth in most of Europe and North America.