A Comparison of Simulations of Current Climate from
Two Coupled Atmosphere-Ocean Global Climate Models
Against Observations and Evaluation of Their Future Climates

Report to the National Institute for Global
Environmental Change (NIGEC)

In support of the US National Assessment

Ruth Doherty and Linda O. Mearns

Revised 27 April 1999

Note About Revision

Abstract

A qualitative assessment of climate simulations from two state-of the-art coupled atmosphere and ocean general circulation models (AOGCMs) over North America is presented. The models investigated are the Canadian Climate Centre CGCMI and the British Hadley Centre HADCM2. The focus of this work is to examine seasonal-mean averages of key surface and atmospheric variables and how these compare to the observational data sets of Legates and Wilmott and NCEP/NCAR reanalyses. Comparison of historical control simulations against the Legates and Wilmott surface temperature climatology reveals both models to have a warm bias over Canada and northern US in autumn and winter, a cold bias in the West, and a warm bias in summer and autumn. Both models also display a wet bias over the Rockies, and a wet bias in the Northeast and Canada in the spring and summer months when compared to the Legates and Wilmott precipitation climatology. These biases are generally greater in the CGCMI simulations. Surface pressure and 500 mb geopotential height patterns are also compared between the two models and NCEP/NCAR reanalyses. Both models simulate the low-pressure systems in winter to be too deep. In general, the HADCM2 simulations underestimate and the CGCMI simulations overestimate the strength of the high-pressure systems. Geopotential height patterns are more closely represented in the CGCMI simulations; however, both models exhibit a cold tropospheric bias. Future temperature and precipitation estimates of both these models are also examined for three time slices, 2030, 2060, and 2090. The CGCMI simulations display much more extensive warming than the HADCM2 simulations. Precipitation changes (mainly positive) in these future periods are somewhat similar in both models, but the CGCMI simulates more decreases.

1. Introduction

The Canadian Climate Centre and UK Hadley Centre coupled Atmosphere-Ocean Global Climate Models (AOGCMs) are the main models being used for impact assessments of the United States for the current US National Assessment. This report provides basic information on model performance for use by the impacts community. The goal of this work is to provide some insight into how the climate control runs vary between the AOGCMs, how these compare to seasonal real-world observations, and to compare the future climate scenarios. The climate variables considered are screen-height temperature, precipitation, sea-level pressure, and 500 mb geopotential heights. Of all surface climate variables, temperature and precipitation are generally the focus of climate impact studies. To examine circulation characteristics, which are fundamental in determining global surface patterns, surface sea- level pressure and 500 mb geopotential heights were examined.

2. Models and Data Used

2.1. Coupled atmosphere-ocean simulations

The coupled Canadian Climate Centre AOGCM experiments are referred to as the CGCMI experiments. It is comprised of an atmospheric spectral T32 truncation model (corresponding to 48 latitude by 96 longitude grid points or ~3.75° by 3.75° on a Gaussian grid) with 10 vertical levels (see Table 1). The ocean component has a resolution of ~1.8° by 1.8° and 29 vertical levels. Details of the CGCMI and its configuration are described by Flato et al. (1997), and details of the coupled model control run in Boer et al. (1999a). An ensemble comprised of three 200- year transient simulations has been performed. In these simulations, atmospheric concentrations of greenhouse gases correspond to historical concentrations from 1900 to present, and increase at a rate of 1% thereafter. The direct radiative effect of sulphate aerosols was included by increasing the surface albedo (Reader and Boer, 1997). The indirect effect of sulphate aerosols on the optical properties and lifetime of clouds was not included, since estimates of this forcing effect are highly uncertain (IPCC, 1996). An ensemble of simulations are runs carried out with identical forcing but with different initial conditions. Generally the runs of each ensemble are initialized at certain intervals along the control run. The ensemble from these three simulations of monthly averages of climatic variables from 1900-2100 is available from the Canadian Climate Center web site: www.cccma.bc.ec.gc.ca. The data used in this work was taken directly from the Gaussian 3.75° by 3.75° grid.

The UK Hadley Centre AOGCM experiments are referred to as the HADCM2 experiments. The atmospheric model is a spherical model on a 2.5 ° latitude by 3.75 ° longitude degree rectangular grid. It has 19 vertical levels (Table 1). The ocean component has the same horizontal resolution and 20 vertical levels. Model and experimental design details are provided by Johns et al. (1997). Decadal monthly averages from 1860-2100 are available from the IPCC DDC web-site: ipcc-ddc.cru.uea.ac.uk/dkrz/dkrz_index.html. The simulations presented in this study use historical concentrations of greenhouse gases from 1860-1989, which increase at a rate of 1% per year thereafter and include the effects of sulphate aerosols in the same manner as described for the CGCMI. The first run of an ensemble of four simulations was used here for convenience purposes.

2.2 Observational data

Legates and Wilmott (1990a,b) have produced a monthly average temperature and precipitation climatology for the period 1921-1980, on a 0.5° by 0.5 ° grid. A positive bias in the temperature climatology is present in mountainous regions (e.g., the Rockies), because of a lack of observation stations in higher elevations. Over mountainous areas the temperature climatology is estimated to have a warm bias of 1° C (C. Wilmott 1998 pers. comm.). The precipitation climatology (Legates and Wilmott 1990b) has been corrected for rain-gauge undercatch biases; however, the high spatial variability of precipitation implies that any precipitation climatology will also include biases due to the location of stations. A description of these climatologies including measurement errors, biases and intepolation methods can be found in Legates and Wilmott (1990a, 1990b), Legates and DeLiberty (1993), Legates (1997) and Wilmott and Matsuura (1995).

NCEP/NCAR reanalyses data (see Kalnay et al. 1996 for details) at 2.5 by 2.5 resolution were provided by Dennis Joseph at NCAR. Monthly time-series data for surface sea-level pressure for the period 1958-1989, as well as monthly time series data for 500 mb geopotential height for the period 1968-1989 were used in this study.

2.3 Comparison performed

All data series were regridded onto a common 2 ° by 2 ° grid using a Gaussian interpolation method to facilitate comparisons amongst AOGCMs and against observations. This resolution falls between that of the original observations and the AOGCMs, and thus is a reasonable comparison between the disparate resolutions. Comparisons were performed seasonally for North America and adjacent oceans encompassing the region 2 to 68° N and 169° W to 35 ° W. The spatial and seasonal characteristics of surface-air temperature, precipitation, sea-level pressure and 500 mb geopotential heights under present-day conditions, and the simulated seasonal changes in surface-temperature and precipitation under perturbed climate conditions, were examined at a basic level in this report. A comparison of spatial patterns of seasonal means for climatic subregions in the United States (Northwest US, Basin and Range, Southwest US, and lower Mississippi basin, the Central Plains, the Midwest, Northeast, and the Southeast), and Canada (West and Central Northern Canada, Prairie States, Labrador) and Mexico and Central America, was undertaken.

3. Control Climate Results

Model performance is examined in this next section in terms of how surface temperature, precipitation, mean sea-level pressure and 500 mb geopotential heights from the historical CGCMI and HADCM2 simulations compare with observations. The years 1921-1980 are used from the model control runs, since these years correspond to those years making up the Legates and Wilmott observed data sets.

3.1. Surface temperature

The observed Legates and Wilmott surface-air temperature climatology (1990a) is displayed in Figure 1a on its original 0.5° by 0.5° grid, and interpolated to a 2 ° by 2° grid in Figure 1b. The map-scale is drawn at 5 degree intervals between -25° C and +25° C, which represents a rather coarse scale in relation to climate change impacts studies and scenarios. Since temperature patterns are fairly uniform longitudinally over land and oceans, the interpolation to a coarser grid retains almost all of the spatial detail to be seen at the finer 0.5° by 0.5 ° scale (Figure 1a). Over the Rockies in the Northwest, Basin and Range and Southwest regions, patches of orographically-induced colder temperatures can be seen on the 0.5° by 0.5 ° grid that are not resolved on the coarser 2° by 2° grid. All subsequent comparisons are performed using the surface-air temperature as depicted on the coarse resolution 2 ° by 2 ° grid. The surface-air temperature spatial and seasonal patterns for the same period, predicted by CGCMI and HADCM2 control simulations are shown on their original course grids in Figure 2a (CGCMI-3.75° by 3.75° ) and Figure 3a (HADCM2-2.5 ° by 3.75 ° ). These same fields interpolated onto the same 2 ° by 2 ° grid are depicted in Figure 2b and Figure 3b respectively. Theses seasonal interpolated fields can be seen to preserve the characteristics of the original fields. The overall patterns of surface temperature over the continental US are fairly well reproduced in all seasons by both models.

Both AOGCMs simulate a latitudinal gradient of temperatures: below -20° C over northern Canada and Greenland, -5° C to +10° C over the continental US and 0 ° to 15 ° C in Mexico in winter. In summer, both models estimate temperatures ranging between +5° to +20° C in the northern latitudes of Canada and Alaska, 10 ° to 20° C over the Northwest US, Central Plains and Basin and Range regions, and above 20° C over the rest of the US and Mexico. Ocean temperature patterns also seem similar in the two models ranging from -5 ° to +10° C in the North Atlantic in winter to 0 ° to +15 ° C in summer. In the Pacific off the West coast of the US, temperatures range from +5° to +20° C in winter to temperatures of +10° to +25 ° C in summer.

We created difference plots to quantitatively examine spatial differences between AOGCM estimates and observed data. The most striking differences are to be seen over northern Canada and the Rockies region. Figure 4 and Figure 5 depict spatial and seasonal differences between CGCMI and Legates and Wilmott, and between HADCM2 and Legates and Wilmott respectively.

Both Figure 4 and Figure 5 clearly depict higher winter temperatures of 3-9 ° C simulated by the models, as compared to observations over most of Canada and the northern US, and by more than 12° C over Hudson Bay. Note that temperatures in Hudson Bay are comparable to surrounding land temperatures in the Legates and Wilmott climatology (Figure 1b). The temperature network used in this observed climatology is very sparse over northern Canada and no measurements appear to have been taken in Hudson Bay itself (see Figures 1 and 2, Legates and Wilmott, 1990a). In contrast, the two models define this area as ocean in their land-ocean masks, hence temperatures in Hudson Bay reflect those of the northern Atlantic. As a consequence, the models show winter Hudson Bay temperatures to be substantially warmer than observed and the converse in summer. In spring the HADCM2 simulations (Figure 5) still simulate substantially warmer temperatures in this entire region while in the CGCMI show better agreement with observations (Figure 4). In summer the CGCMI simulations in Figure 4 show a tongue of warmer temperatures (1-3° C ) over northwestern Canada while the HADCM2 simulations warmer temperatures are confined to the extreme North. In contrast to conditions in spring, in autumn the CGCMI simulations overestimate temperatures across Canada to a greater extent than the HADCM2 simulations.

Discrepancies between model simulations and observed data, due to the representation of the orography of the Rockies and the Andes in the AOGCMs, are highly noticeable in Figure 4 and Figure 5. As mentioned in Section 2.2, in mountainous regions the observations are expected to have a warm bias of about 1 ° C. In spring and summer, cooler temperatures generally between -1° to -3 ° C and reaching -6 ° C are estimated in the CGCMI simulations over Alaska, the Northwest US, and Basin and Range. In winter and autumn these differences are mostly confined to the Basin and Range region (Figure 4). The HADCM2 simulations exhibit discrepancies due to orography over the Basin and Range area, Southwest US in winter, and over the Basin and Range and Northwest US and Canada during the other seasons. Large areas of underestimations in the HADCM2 simulations in the Rockies (more than 6° C) are depicted in summer in Figure 5.

Another major area of disagreement over the US occurs over the central plains and the Midwest reaching up to the Canadian prairies in the summer, where both AOGCMs estimate temperatures of 1-3 ° C (reaching maximum differences of 6° C) higher than observations. The CGCMI simulates this region of higher temperatures to extend eastwards over the Northeast US in summer but the warm bias is much smaller in autumn. The HADCM2 simulation exhibits a similar region of warmer temperatures in autumn between 1-3 ° C.

Over large areas of the Pacific and Atlantic, the CGCMI simulation shows a consistent warm bias throughout the year of 1° to 3° C compared to the data of Legates and Wilmott. Over the North Atlantic this warm bias reaches 3° to 6° C in autumn and winter. In contrast, the HADCM2 simulation of ocean temperatures exhibits a greater tendency towards a cold bias generally between 0 ° to -3 ° C over both the Atlantic and the Pacific and in all seasons. Note that the temperature scales in Figure 2 and Figure 3 are too coarse to reveal such biases in oceanic temperatures.

To summarize, similar patterns of temperature errors are to be found in both AOGCMs. A warm bias occurs in winter and autumn over Canada and northern US and North Atlantic, while a cold bias occurs over the West of the US in all seasons and into western Canada in spring and summer. The central US also experiences a warm bias in summer and autumn.

3.2 Precipitation

The observed Legates and Wilmott (1990b) climatology can be seen at its original 0.5° by 0.5° resolution in Figure 6a. Much of the structure is retained when the data is aggregated to the coarser 2° by 2 ° resolution grid. However, the effect of employing a coarser resolution with spatially variable data is seen especially over the Rockies where the finer structural patterns are not resolved in Figure 6b. The map scale depicts precipitation values in the range 0 to 25 mm/day. The CGCMI and HADCM2 simulations are displayed in Figure 7 and Figure 8 at the 2 ° by 2 ° resolution. Figure 7 and Figure 8 clearly show the broad-scale features of centers of precipitation minima (less than 1 mm/day) off the Caribbean and over the West coast of the US. The latter in summer is associated with the Hawaiian higher pressure center. The seasonal northward migration from winter to summer of the center of the West coast precipitation minimum is clearly shown, as is the spring and summer expanse of aridity or low-precipitation cells off the Caribbean. Precipitation maxima of 4-8 mm/day associated with the Aleutian low pressure system (see section 3.1.3) can be seen off the coast of Alaska and the Northwest, particularly in autumn and winter when the high pressure system does not act to block the westerly flow of the jet stream, as is the case in summer. The Northeast, Southeast US and Labrador coasts generally receive precipitation amounts between 2-4 mm/day in autumn and winter, which increase to 4-8 mm/day along parts of the East coast in spring. Precipitation amounts exceeding 10 mm/day, most noticeable in the HADCM2 simulation (Figure 8), can be seen on and off the coast of Central America in summer when the ITCZ has migrated to its northernmost position.

Difference plots between CGCMI and Legates and Wilmott, and between HADCM2 and Legates and Wilmott, are displayed in Figure 7 and Figure 8 Figure 9a and 10a respectively. Again, major areas of disagreement over land are associated with the orography of the Rockies. In Figure 9a (CGCMI-Legates), precipitation amounts over Northwest Canada, Northwest US, and the Basin and Range region are overestimated throughout the year generally by 1-3 mm/day with maximum differences of ~6 mm/day in parts of the Rockies. The HADCM2 simulations (Figure 10a) similarly overestimate rainfall in this region. The CGCMI simulations also show large areas where precipitation amounts are overestimated over Mexico. Rainfall in the Northeast US and Labrador region is overestimated by 1-3 mm/day in the CGCMI simulations in spring and summer (Figure 9a). The HADCM2 simulation shows this overestimation more strongly in spring than summer (Figure 10a). Both models depict the Southeast US and the lower Mississipi Basin to be too dry by up to 3 mm/day, most notably in winter and summer. Over the oceans, neither of the models predict the ITCZ to extend as far North as in the Legates and Wilmott observations in Figure 6b; hence differences greater than 3 mm/day can be observed across the equatorial oceans in both Figure 9a and Figure 10a. Note that the areas of extreme disagreement are more extensive between the CGCMI simulations and the observations (Figure 9a). The CGCMI underestimates precipitation by 1 to 3 mm/day over large areas of the Atlantic and Pacific oceans in autumn and winter (Figure 9a). The HADCM2 simulations show the same biases but confined to a smaller area (Figure 10a). Oceanic areas where both AOGCMs overestimate precipitation by up to 3 mm/day occur off both the East and West coasts of Canada and the US predominantly in spring and summer. The HADCM2 simulations overestimate precipitation in the latter regions more uniformly throughout the year.

Difference plots are expressed as fractional change in Figure 9b and Figure 10b for the CGCMI and HADCM2 simulations. The major fractional differences in precipitation that occur over the Rockies extending from Canada to the Northwest and Basin and Range represent a fractional overestimation in precipitation of 100% to 500%, predominately in winter and spring in the CGCMI simulation. The relative errors are more intense and extensive in the CGCMI simulations than in that of the HADCM2. In the case of the latter, overestimations in precipitation above 200% are mainly confined to Alaska in spring. The drier areas in the Southeast US and lower Mississippi Basin represent an underestimation in precipitation of around 20% to 50% in both models with a small confined area which reaches 100% in winter in the CGCMI simulations.

To summarize, again both models show similar biases. Both models have a wet bias over mountainous regions in the West throughout the year and a wet bias over the northeastern Canada and the US in spring and summer. A dry bias is to be found over the Southeast and lower Mississippi Basin mainly in winter and summer. In general, the HADCM2 simulations show all these regions of biases to be smaller in extent than do the CGCMI simulations.

3.3 Sea-level pressure

Observations of seasonal mean sea-level pressure for the 1980-1989 decade obtained from NCEP/NCAR reanalyses are depicted in Figure 11. Comparisons were also performed for the 1960-69, and 1970-79 decades. The results were similar to those for 80-89, hence only this decade’s results are discussed below. CGCMI and HADCM2 control simulations for the same decade are displayed in Figure 12 and Figure 13. Both AOGCMs reproduce well the locations of the two main low pressure systems over this study region, the Aleutian and Icelandic lows, in autumn and winter. The subtropical high-pressure systems, the Hawaiian and Bermuda highs, are clearly depicted throughout the year. The propagation from winter to summer of these two high pressure systems to their maximum extent and locations northwestward and westward, respectively, is also well represented. In winter over the West coast the Hawaiian high extends furthest westward inland over the Northwest and Southwest of the US. The NCEP/NCAR observations show two centers of high pressure – one over land and the other over the ocean in this region during winter, while both models depict one west coast high pressure center. The CGCMI predictions show this center over land, and the HADCM2 simulations show this center further westward over ocean. In summer when this system has retreated to its maximum westerly location, both AOGCMs depict a low pressure system to evolve over the North and Southwest US extending into the Central Plains (Figure 12 and Figure 13).

Main discrepancies between model results and observations concern the intensity and extent of these main low-pressure and high-pressure centers. Both models estimate the Aleutian and Icelandic lows in autumn and winter to be too deep (c.f. 1000 and 1004 mb contours) relative to the NCEP/NCAR Reanalyses observations. The CGCMI simulations depict a low-pressure band (1008 mb) to stretch over most of Canada between these two low-pressure centers in winter, while the HADCM2 simulations and observations show these centers to be discrete and estimate higher pressures of 1012-1016 mb over this area. This is also the case in autumn. In spring and in autumn the HADCM2 simulations of the Hawaiian high are in better agreement with observations (Figure 11 and Figure 13) than the CGCMI simulations, in terms of their spatial extent and maxima. However, in winter and summer the HADCM2 simulation underestimates the maxima in this high pressure center. The Bermuda high is well replicated in the HADCM2 simulations. In contrast, the CGCMI simulations have a general high bias all year round in their estimations of the magnitude and spatial extent of this system. Both models have a general tendency to overestimate the extent of the 1012 mb contour band, which prevails across the southern latitudes throughout the year. This discrepancy is most noticeable in the HADCM2 simulation (Figure 13).

To summarize, both models simulate the Aleutian and Icelandic low-pressure systems to be too deep. The HADCM2 simulations underestimate the intensity of the Hawaiian high in winter and summer. The CGCMI simulations overestimate the intensity of the Bermuda high throughout the year.

3.4 500 mb geopotential heights

NCEP/NCAR reanalyses of seasonal-mean geopotential heights for the period 1980-89 are displayed in Figure 14. The decade 1970-79 was also analysed and again results were similar, hence the discussion below is restricted to the 1980-89 decade. Both AOGCMs reproduce the wave patterns seen in the NCEP/NCAR reanalyses with crest and trough-like patterns over the West and East coasts of America respectively (Figure 15 and Figure 16). However, model wave patterns seem to be generally flatter compared to observations over land. Since the geopotential heights are a strong function of tropospheric temperatures, the height contours are seen to shift northward with warmer temperatures in summer.

In winter, spring and autumn, over the West coastline of America the CGCMI simulations show a southward displacement of the 5300-5600 m contour level gradient with respect to the NCEP/NCAR observations (Figure 14 and Figure 15). The CGCMI simulations in turn show a gradual displacement in contours with respect to HADCM2, such that the 5400 to 5700 m contours are displaced even further southward in the HADCM2 simulations (by ~100 m, Figure 15 and Figure 16). In winter, the NCEP/NCAR data predicts heights of 5800 m through northern Mexico. The HADCM2 simulations show heights of 5800 m much further South over southern Mexico, while the CGCMI simulations predict heights greater than 5700 m in winter only over the Carribean. Throughout the year, over the East coast of the US and the Labrador coast the CGCMI predictions of heights between 5400-5700 m are also displaced further north (by 100 m) with respect to the HADCM2 simulations. Again, over the East coast these CGCMI height patterns are in better agreement with the NCEP/NCAR reanalayses than the HADCM2 simulations.

To summarize, the general trend is that in all seasons the locations of the contour levels below 5800 m at the West-East coastlines of America and Canada are more closely reproduced in the CGCMI simulations. However, the CGCMI simulations underpredict the extent of regions where geopotential heights reach or exceed 5800 m implying a cold tropospheric temperature bias in the CGCMI AOGCM. The HADCM2 simulations also underestimate the northward extent of the 5800 m geopotential heights suggesting a weaker cold bias.

4. Future Climate Scenarios from CGCMI and HADCM2

Three time slices incorporating ten-year periods centered around 2030, 2060, and 2090 were examined to gain some insight into the range of future predictions of temperature and precipitation estimates from the AOGCMs. As discussed in section 2.5, these scenarios are based on a 1% increase per year in CO2 from 1990 onwards and the inclusion of the direct effects of sulphate aerosols. See Boer et al. (1999b) and Mitchell and Johns (1997) for further details on the Canadian and British model climate change scenarios, respectively.

4.1 Temperatures

A summary table of estimates of increases in surface temperature over the US for these three time slices is presented in Table 2.

2030

Figure 17 and Figure 18 show increases in temperature predicted by the CGCMI and HADCM2 simulations respectively for 2030, relative to the control period of 1921-1980 (Figure 2 and Figure 3 discussed in Section 3.1.1). Over land, a strong winter warming over the northern latitudes is the major feature to be observed. This warming is considerably greater in the CGCMI simulations. The CGCMI simulations depict increases in temperature of 2° to 7 ° C in the 2030 scenario across Canada and the US southwards to ~35 N (Figure 17). Southwards of this, the western half of the US including the Southwest, Basin and Range, Central Plains and Midwest is also estimated to warm by 2° to 4 ° C, while the East coast of the US and Mexico is estimated to be warmer by 1° to 2 ° C. A strong patch of winter warming with simulated increases in temperatures of 4° to 7 ° C can be seen over Hudson Bay for the 2030 scenario. In contrast, the HADCM2 simulations depict smaller increases between 2030 and the control over most of Canada and the US. In the case of the latter, increases above 4° C only occur over Alaska. (Figure 18).

The CGCMI simulations also show a strong spring warming signal (temperature increases above 3 ° C) over westernmost Alaska and over a region encompassing the southern Canadian prairies and Great Plains of the US. More modest warming of 1° to 3 ° C is simulated across the rest of Canada and the US. In spring the HADCM2 simulations generally show an increase in temperatures of 0 ° to 3 ° C over central and West US as well as central Canada, with higher increases of 3° to 4 ° C in Alaska (Figure 18). Over the East US increases of 0-1° C are estimated.

In summer and autumn more uniform warming of 0° to 3 ° C can generally be seen across the US and Canada in both simulations. The CGCMI simulations generally display greater warming than the HADCM2 simulations, which depict extensive regions of small (0 ° to 1 ° C) temperature increases in summer over the Central Plains and east US.

Over the oceans in the CGCMI simulations, a cooling of 1 ° to 5 ° C can be seen over a small region off the Labrador coast in all seasons, most prominent in summer. This cooling is weaker (0° to 1 ° C) but more extensive over the Atlantic in the HADCM2 simulations. The Pacific oceans off the West coast of the US display warming between 0° to 2 ° C in the CGCMI simulations in all seasons, while the HADCM2 simulations display similar patterns but also show patches of cooler temperatures (0 ° to -1° C) in winter and spring.

2060

In the 2060 scenario, over land the CGCMI simulations (Figure 19) again show much greater increases in temperature than the HADCM2 simulations (Figure 20). The CGCMI simulations show intense warming of 3-7 ° C across the whole North American continent in winter and spring. Increases in temperature over Hudson Bay reach above 7° C in the CGCMI simulations, but are no greater than 2-3° C in the HADCM2 simulations. Moreover, the HADCM2 simulations display winter warming to be less extensive than in the CGCMI simulations. Winter temperatures greater than 4° C are confined to Canada, with increases above this to be found in parts of Alaska (Figure 20).

Temperatures increases in the CGCMI simulations in summer are generally around 2-4 ° C over the US and Canada (Figure 19). In autumn, the largest region of increases in temperature of 4 C and above is over the Soutwest, Basin and Range and southern central Plains. Spring-autumn predictions from the HADCM2 simulations also show more modest increases in temperature compared to the corresponding CGCMI predictions. Over most of the continental US and Canada increases in the range of 1-3 ° C are predicted, with the west and central US experiencing greater increases (2-3 ° C) than the east US (1-2 ° C) in autumn (Figure 20).

Over the oceans, the CGCMI simulations show more extensive cooling off the Labrador coast compared to the 2030 CGCMI simulations. This stronger cooling signal at 2060 is not so apparent in the HADCM2 simulations. In the 2060 HADCM2 simulations the Pacific oceans no longer display areas of slight cooling as in the 2030 simulations.

2090

In the 2090 simulations the CGCMI simulations in Figure 21 now show increases in temperature above 4 ° C over most of the North American continent in all seasons. Indeed, much of Canada and the northern US show warming above 9° C in winter. Again, the HADCM2 increases are much more modest in intensity and extent. Increases in winter in Canada reach up to 9° C, and range from 2° to 7 ° C over the US in all seasons (Figure 22). Warming is more uniform from East to West in the CGCMI simulations in contrast to the HADCM2 simulations that show greater warming in the West. In general, the CGCMI simulations show substantially greater and more uniform warming than the HADCM2 simulations in all three future periods.

4.2 Precipitation

Tables 3a-c show estimates of increases in surface precipitation over the US for 2030, 2060 and 2090 respectively. Again, each of these periods are discussed separately below.

2030

Figure 23 and Figure 24 display differences in precipitation between the 2030 period and the 1921-1980 control, for the CGCMI and HADCM2 simulations respectively. The greatest differences in precipitation are located mainly over the oceans off the West coast of the US and over the South and Southeast latitudes, for both models. Over land, the CGCMI predictions for 2030 show mainly decreases in precipitation of 0 to 0.5 mm/day over most of North America in winter (Figure 23). In winter, a band of increased precipitation between 0.5 to 3 mm/day is seen off the West coast of the US and extends into the Northwest of the US in winter in the CGCMI simulations. Another band of increased precipitation can be seen around the Caribbean coast extending northeastward. Over Mexico, Central America and the Southeast US coastline decreases over 0.5 mm/day (reaching a maxima of 2 mm/day) are predicted for 2030 (Figure 23). The HADCM2 simulations display fairly similar patterns (Figure 24). Decreases in precipitation over the Southeast are displaced relative to the CGCMI simulations such that precipitation decreases of similar magnitude are mainly found over the oceans and the Caribbean.

In spring, summer and autumn these same major changes can be observed in the CGCMI simulations with the drying in the Southeast extending down over South America. The maximum of the increased precipitation band off the West coast of the US moves southwestward across the oceans and is greatest in extent in summer (increases of 3 to 5 mm/day). Larger regions of increased precipitation on the order of 0-0.5 mm/day can be observed in these seasons (Figure23). In general, in contrast to the CGCMI simulations, the HADCM2 simulations for 2030 appear to show more extensive regions where modest increases in precipitation of 0 to 0.5 mm/day occur across the North American continent in all seasons.

2060

Comparisons of the 2060 minus control differences reveal similar patterns relative to the 2030- control differences for both models in Figure 25 and Figure 26. The predicted areas of increased precipitation greater than 3 mm/day off the West coast of the US extending into the Northwest US are similar (or slightly more extensive) in 2060 in the CGCMI simulations compared to the HADCM2 simulations (Figure 25 and Figure 26) in winter. Greater drying over the Southeast US can be seen in spring and summer in the CGCMI simulations. In summer there are more areas where decreases in precipitation of more than 0.5 mm/day are predicted in the CGCMI 2060 simulations, as compared to the 2030 simulations. The HADCM2 simulations for 2060 also show greater drying relative to 2030 but only over the oceans in the Southeast and the Southwest (Figure 26). Otherwise the patterns appear to be fairly similar to those in Figure 24 for 2030.

2090

Similar trends appear in the 2090 minus control differences in Figure 27 and Figure 28. The CGCMI simulations show increases in precipitation over the West coast of 5 to 7 mm/day in winter (Figure 27). The HADCM2 simulations show similar estimates of increases in precipitation over the West coast and Northwest in 2090 (Figure 28). Greater and more extensive drying in the CGCMI simulations, up to 3 to 5 mm/day is also estimated in some parts of the Great Plains, Southeast US, and Mexico (Figure 27) in winter, spring and summer compared to the 2060 estimates (Figure 25). The 2090 HADCM2 simulations also reveal further drying in the Southern oceans across the ITCZ relative to the HADCM2 2060 patterns in Figure 26.

In summary, both models are similar in their predictions of increases in precipitation over the West coast. The HADCM2 simulations show greater and more extensive drying in the Southern latitudes, while the CGCMI simulations show drying further northwards in the Southeast US and Mexico.

5. Discussion and Conclusion

Our largely qualitative comparisons of the climate model control runs preclude making any generalization about which model performs better in terms of its ability to simulate present-day climate. Indeed, given the mixed results of the models, it is likely that a more rigorous statistical analysis would also fail to produce a decidedly “better” model. Nevertheless, a few general comparative points can be noted.

The similar spatial resolutions of the two models result in orography problems for both that result in cold and wet biases over the mountainous regions of North America. Moreover, the relatively coarse spatial resolution of both models render detailed regional results for the Western US highly questionable under perturbed climate conditions. In both models the complex topography of the Western US is depicted as one large hill rising from the West coast and descending onto the Great Plains.

The CGCMI shows a much more extensive warm bias which covers much of Canada and the Northern US in winter than does the HADCM2. Cold biases over the Rockies are similar in the models, while wet biases tend to be slightly greater over the Rockies in the CGCMI simulations. The main pressure systems across the US are well represented, although the CGCMI has a tendency to overestimate the pressure of the Bermuda high all year round, while both models underestimate the main low pressure systems. Both models also have a general cold tropospheric bias.

The extent of the errors in these two AOGCMs are within the range of errors found in other state-of-the-art climate model simulations (Kittel et al., 1998; Boville and Gent, 1998). It also must be noted that a comparison of the observed data sets used here against other compatible observational data sets (if available) would give some insight into realistic observational errors. This would be particularly important in the case of precipitation.

However, this is not to say that the errors in the simulations of the current climate are relatively unimportant. Errors in the control run affect the way the models respond to perturbed greenhouse gas and sulfate forcing. Moreover, errors in other parts of the global domain not discussed in this report could affect the model response over the North American continent. Thus, model biases should be taken into account when evaluating future estimates of changes in temperature and precipitation as predicted in the CGCMI and HADCM2 simulations.

Our goal in this report has been to provide researchers who plan to use output from these models for impacts analysis with an overview of the model control runs and perturbed climates over North America on a seasonal and regional basis. It should be noted that the future climates simulated by these models are in no way to be considered predictions or forecasts of the future. They are scenarios of the future and thus inherently uncertain. This uncertainty increases as the spatial scale of focus decreases, i.e., going from continental to regional scales. Researchers should exercise extreme caution in the conclusions they draw from impacts analysis using the output from these climate models, given the uncertainty of the model results, especially on a regional scale.

6. References

Boer, G.J., Flato, G.M., and Ramsden, D., 1999a. A transient climate change simulation with greenhouse gas and aerosol forcing: Experimental design and comparison with the instrumental record for the 20th Century. Submitted to Climate Dynamics.

Boer, G.J., Flato, G.M., and Ramsden, D., 1999b. A transient climate change simulation with greenhouse gas and aerosol forcing: Projected climate for the 21st Century. Submitted to Climate Dynamics.

Boville, G.A. and Gent, P.R., 1998. The NCAR climate system model, version one. J. Climate, 11, 1115-1130.

Flato, G.M., Boer, G.J., Lee, W.G., McFarlane, N.A., Ramsden, D., Reader, M.C., Weaver, A.J. 1997. "The Canadian Centre for Climate Modelling and Analysis Global Coupled Model and its Climate", Canadian Centre for Climate Modelling and Analysis, Atmospheric Environment Service, University of Victoria, B.C.

IPCC, 1996. Climate Change 1995: The Science of Climate Change. Contribution of the Working group I to the Second Assessment Report of the IPCC,. Houghton, J.T., L.G. Meira Filho, B.A. Callander, N. Harris, A. Kattenburg and K. Maskell (eds.). Cambridge University Press, Cambridge, UK, 572 pp.

Johns, T.E., Carnell, R.E., Crossley, J.F., Gregory, J.M., Mitchell, J.F.B., Senior, C.A., Tett, S.F.B., and Wood, R.A. 1997. The second Hadley Centre coupled ocean-atmosphere GCM: model description, spinup and validation. Climate Dynamics, 13, 103-134.

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., Iredell, M., Saha, S.,White, G., Woollen, J., Zhu, Y., Chelliah, M., Ebisuzaki, W., Higgins, W., Janowiak, J., Mo, K.C., Ropelewski, C., Wang, J., Leetmaa, A., Reynolds,R., Jenne, R., and Joseph, D., 1996. The NCEP/NCAR 40-year Reanalysis project 1996. Bull. Am. Meteor. Soc., 77, 437-471.

Kittel, T.G.F., Georgi, F., and Meehl, G.A., 1998. Intercomparison of regional biases and doubled CO2 sensitivity of coupled atmosphere-ocean general circulation model experiments. Climate Dynamics, 14, 1-15.

Legates, D.R. and Wilmott, C.J. 1999a. Mean seasonal and spatial variability in global surface air temperature. Theoretical and Applied Climatology, 41, 11-21.

Legates, D.R. and Wilmott, C.J. 1990b. Mean seasonal and spatial variability in gauge- corrected global precipitation. Int. J. Climatology, 10, 111-127.

Legates, D.R and DeLiberty, T.L. 1993. Precipitation measurement biases in the United States. Water Resources Bulletin, 29, 855-861.

Legates, D.R. 1997. Global and terrestrial precipitation: A comparative assessment of existing climatologies: A reply. Int. J. Climatology, 17, 779-783.

Mitchell, J.F.B. and Johns, T.C., 1997. On modification of global warming by sulfate aerosols. J. Climate, 10, 245-267.

Reader, C. and G.J. Boer, 1997. The modification of greenhouse gas warming by the direct effect of sulphate aerosols. Climate Dynamics, 14, 593-608.

Wilmott, C.J.and Matsuura, K.. 1995. Smart interpolation of annually averaged air temperature in the United States. Journal of Appl. Meteorology, 34, 2577-2586.


The figures captions are linked to Adobe Acrobat Portable Document Format (PDF) files. To view these figures, download the free Adobe Acrobat Reader software. Note: size of the PDF files are noted in MB

Figure Captions

Figure 1a: Legates and Wilmott seasonal-mean surface temperature fields over the period 1921-1980 at original 0.5° by 0.5 ° resolution. DJF = winter; MAM = spring; JJA = summer; SON = fall (and on all subsequent seasonal plots).(1.2MB)

Figure 1b: Legates and Wilmott seasonal-mean surface temperature fields over the 1921-1980 period aggregated to 2° by 2° resolution.(0.2MB)

Figure 2a: CGCMI seasonal-mean surface temperature fields for the period 1921-1980 at the original model 3.75 ° by 3.75 ° resolution.(0.2MB)

Figure 2b: CGCMI seasonal-mean surface temperature fields for 1921-1980 interpolated to 2 ° by 2° resolution.(0.2MB)

Figure 3a: HADCM2 seasonal-mean surface temperature fields for the period 1921- 1980 at the original model 2.5° by 3.75 ° resolution.(0.2MB)

Figure 3b: HADCM2 seasonal-mean model surface temperature fields for 1921-1980 interpolated to 2 ° by 2 ° resolution.(0.2MB)

Figure 4: Seasonal-mean surface temperature difference fields in Centigrade degrees (CGCMI - Legates and Wilmott) for 1921-1980.(0.2MB)

Figure 5: Seasonal-mean surface temperature difference fields in Centigrade degrees (HADCM2 - Legates and Wilmott) for 1921-1980.(0.2MB)

Figure 6a: Legates and Wilmott seasonal-mean precipitation fields over the period 1921- 1980 at original 0.5 ° by 0.5 ° resolution.(1.1MB)

Figure 6b: Legates and Wilmott seasonal-mean precipitation fields over the 1921-1980 period aggregated to 2° by 2 ° resolution.(0.2MB)

Figure 7: CGCMI seasonal-mean precipitation fields for 1921-1980 interpolated to 2° by 2 ° resolution.(0.2MB)

Figure 8: HADCM2 seasonal-mean model precipitation fields for 1921-1980 interpolated to 2 ° by 2 ° resolution.(0.2MB)

Figure 9a: Seasonal-mean precipitation difference fields in mm/day (CGCMI - Legates and Wilmott) for 1921-1980.(0.2MB)

Figure 9b: Seasonal-mean fractional precipitation difference fields (CGCMI - observed / observed) for 1921-1980. Observed is Legates and Wilmott.(0.2MB)

Figure 10a: Seasonal-mean precipitation difference fields in mm/day (HADCM2 - Legates and Wilmott) for 1921-1980. (0.2MB)

Figure 10b: Seasonal-mean fractional precipitation difference fields (HADCM2 - observed/ observed) for 1921-1980. Observed is Legates and Wilmott.(0.2MB)

Figure 11: NCEP/NCAR reanalyses seasonal-mean sea-level pressure fields for 1980-1989 interpolated to 2° by 2 ° resolution.(0.2MB)

Figure 12: CGCMI seasonal-mean sea-level pressure fields for 1980-1989 interpolated to 2 ° by 2° degree resolution.(0.2MB)

Figure 13: HADCM2 seasonal-mean sea-level pressure fields for 1980-1989 interpolated to 2° by 2° resolution.(0.2MB)

Figure 14: NCEP/NCAR reanalyses seasonal-mean 500 mb geopotential height fields for 1980-1989 interpolated to 2° by 2 ° resolution.(0.2MB)

Figure 15: CGCMI seasonal-mean 500 mb geopotential height fields for 1980-1989 interpolated to 2 ° by 2 ° resolution.(0.2MB)

Figure 16: HADCM2 seasonal-mean 500 mb geopotential height fields for 1980-1989 interpolated to 2° by 2° resolution.(0.2MB)

Figure 17: CGCMI seasonal-mean temperature difference fields (2025-2035–1921-1980)(0.2MB)

Figure 18: HADCM2 seasonal-mean temperature difference fields (2025-2035 –1921-1980)(0.2MB)

Figure 19: CGCMI seasonal-mean temperature difference fields (2055-2065–1921-1980)(0.2MB)

Figure 20: HADCM2 seasonal-mean temperature difference fields (2055-2065–1921-1980)(0.2MB)

Figure 21: CGCMI seasonal-mean temperature difference fields (2085-2095–1921-1980)(0.2MB)

Figure 22: HADCM2 seasonal-mean temperature difference fields (2085-2095–1921-1980)(0.2MB)

Figure 23: CGCMI seasonal-mean precipitation difference fields (2025-2035–1921-1980)(0.2MB)

Figure 24: HADCM2 seasonal-mean precipitation difference fields (2025-2035–1921-1980)(0.2MB)

Figure 25: CGCMI seasonal-mean precipitation difference fields (2055-2065–1921-1980)(0.2MB)

Figure 26: HADCM2 seasonal-mean precipitation difference fields (2055-2065–1921-1980)(0.2MB)

Figure 27: CGCMI seasonal-mean precipitation difference fields (2085-2095–1921-1980)(0.2MB)

Figure 28: HADCM2 seasonal-mean precipitation difference fields (2085-2095–1921-1980)(0.2MB)


Tables

Table 1: CGM configurations

Atmosphere Oceans
CGCMI
Spectral T32 Truncation
3.75° x 3.75° horizontal resolution
10 vertical levels
1.8° x 1.8° resolution

29 levels

HADCM2
Spherical
2.5° x 3.75° horizontal resolution
19 vertical levels
2.5° x 3.75° resolution

11 levels

Table 2: Increases in temperature over the US

2030
Winter
2030
Summer
2060
Winter
2060
Summer
2090
Winter
2090
Summer
CGCMI
2 - 4°C
west & central US,
0 - 2°C east US
1 - 3°C
over US
2 - 5°C
west US,
4 - 7°C central US,
1 - 5°C east US
4 - 4°C
over US
3 - 7°C
west US,
4 - 12°C central US,
3 - 7°C east US
3 - 7°C
over US
HADCM2
1 - 4°C
west & central US,
1 - 3°C east US
1 - 2°C
west US,
0 - 3°C central US,
0 - 1°C east US
1 - 3°C
west US
1 - 6°C central US,
2 - 6°C east US
2 - 3°C
west & east US
2 - 4°C central US
2 - 7°C
west & central US,
1 - 5°C east US
3 - 5°C
west US,
0 - 5°C central US,
1 - 3°C east US

Table 3a: Changes in precipitation over western US

2030
Winter
2030
Summer
2060
Winter
2060
Summer
2090
Winter
2090
Summer
CGCMI
0.0 to +3.0
mm/day
-0.5 to +0.5
mm/day
0.0 to +3.0
mm/day
-1.0 to +0.5
mm/day
+0.5 to +10.0
mm/day
-0.5 to +0.5
mm/day
HADCM2
0.0 to +2.0
mm/day
-0.5 to +1.0
mm/day
0.0 to +3.0
mm/day
-0.5 to +0.5
mm/day
0.0 to +10.0
mm/day
-0.5 to +0.5
mm/day

Table 3b: Changes in precipitation over central US

2030
Winter
2030
Summer
2060
Winter
2060
Summer
2090
Winter
2090
Summer
CGCMI
-2.0 to 0.0
mm/day
-1.0 to +0.5
mm/day
-1.0 to +0.5
mm/day
-2.0 to +0.5
mm/day
-2.0 to +0.5
mm/day
-3.0 to +0.5
mm/day
HADCM2
-0.5 to +1.0
mm/day
-0.5 to +1.0
mm/day
0.0 to +1.0
mm/day
-0.5 to +0.5
mm/day
-1.0 to +1.0
mm/day
-0.5 to +2.0
mm/day

Table 3c: Changes in precipitation over eastern US

2030
Winter
2030
Summer
2060
Winter
2060
Summer
2090
Winter
2090
Summer
CGCMI
-2.0 to 0.0
mm/day
-1.0 to +0.5
mm/day
-1.0 to+ 0.5
mm/day
-2.0 to+ 0.5
mm/day
-2.0 to +0.5
mm/day
-2.0 to +0.5
mm/day
HADCM2
0.0 to +1.0
mm/day
-0.5 to +1.0
mm/day
0.0 to +1.0
mm/day
-0.5 to +0.5
mm/day
0.0 to +2.0
mm/day
-0.5 to +2.0
mm/day


Top of page | Back to Main Page | Back to ESIG | Contact Us | About Linda O. Mearns

Environmental and Societal Impacts Group
National Center for Atmospheris Research (NCAR)
PO Box 3000, Boulder, CO 80307 USA
Tel (303) 497-8117; Fax (303) 497-8125

NCAR is sponsered by the National Science Foundation