Sensitivity Experiments Using General Circulation Models

Models can be used in sensitivity experiments to examine the effects of changing a boundary condition (e.g., solar radiation; Syktus et al., 1994) or a process (e.g., sea-ice formation; Vavrus, 1995). By comparing climate associated with the specified change with the "no change" (control) simulation the significance of the process or parameter that was altered can be assessed. Such experiments help in understanding how complex feedbacks (amplifying or minimizing the effect of specified changes) operate and interact within the climate system. Generally a control experiment is run over a number of simulated years "to equilibrium" and a period of years at the end of this run will then be averaged (for a specified month or season) to obtain a reference climate state. This is then "validated" by comparison with modern climate data sets, providing some level of confidence in the model. After the conditions to be investigated have been changed, the simulation is once again run to a new "equilibrium" and a new set of years is averaged. The differences between the control run and the new conditions are considered to be a consequence of those factors that were altered in the sensitivity experiment.

How should the results of sensitivity experiments be evaluated? We know from the outset that each model may have non-trivial limitations in its ability to simulate modern climate. For example, compared to observed climate, one version of the Geophysical Fluid Dynamics Lab (GFDL) atmospheric GCM used in many paleoclimate experiments by Manabe and other researchers (Manabe and Broccoli, 1985a) produces conditions that are too cold in the northern hemisphere, and it generates too much precipitation at high latitudes. It underestimates sea ice in the southern hemisphere and places the Inter-Tropical Convergence Zone (ITCZ) 5°-10° (latitude) too far south. However, the important issue for this and other sensitivity experiments is what differences were induced in the model's climate system by the imposed changes. Certainly, model limitations must be kept in mind to ensure the observed differences are not simply artifacts of the model, but in general, sensitivity experiments can provide valuable insights into the role of boundary conditions and other changes. Other models can also be employed to determine which conclusions are robust, regardless of the model used. On the other hand, the simulations may point to discrepancies between modeled paleoclimates and reconstructions based on proxy data, and these may justify a reconsideration of the paleodata and/or calibration method used. Such considerations have led to an international effort (PMIP — Paleoclimate Model Intercom-parison Project) to compare paleoclimatic reconstructions for specified times, with all models using exactly the same boundary conditions. This will enable the common climatic characteristics to be identified and problems in particular models to be addressed (Joussaume and Taylor, 1995). In a parallel effort (PMAP — Paleoenvironmental Multiproxy Analysis and Mapping Project) paleodata are being compiled and mapped for selected "time slices" to enable direct comparisons to be made between model simulations and paleoclimatic data (see Section 12.4).

In the next section, we first consider changes in orbital forcing. By examining the range of extremes over the late Quaternary, models can address to what extent orbital changes may have been important in initiating glaciation (at one extreme) and in bringing about post-glacial changes in monsoon climates at the other. Then we examine a set of experiments designed to understand how changes in surface boundary conditions during the last glacial maximum (LGM: 18 ka B.P.)35 contributed to climatic conditions prevailing at the time. At 18 ka B.P. orbital conditions differed very little from today and so the dramatically different climate state at that time must have been driven largely by orbital changes at some earlier time, with conditions at 18 ka B.P. responding to (but not necessarily maintaining) changes in the boundary conditions. Finally, the significance of changes in the seasonality of incoming radiation is examined, in a series of experiments at intervals from 18 ka B.P to the present.

12.3.1 Orbital Forcing and the Initiation of Continental Glaciation in the Northern Hemisphere

A number of GCM experiments have been carried out to investigate the role of orbital forcing in the initiation of continental glaciation. Paleoclimatic data from marine sediments (8lsO in benthic forams) and raised corals (recording former sea-level position) indicate that sea level fell by > 50 m from -115-105 ka B.P. Because of a combination of greater eccentricity, lower obliquity, and perihelion close to the (northern) winter solstice, incoming solar radiation at 115 ka was reduced by -7% (40 W m"2) during the northern hemisphere summer (and late spring/early summer

35 The dates of all model experiments are given here in calendar years; 18 ka B.P. was originally thought to be the same in both calendar and radiocarbon years. Later experiments recognized that 18 ka B.P. in radiocarbon years corresponds to -21 ka in calendar years, which alters the relationship between orbital forcing and l4C-dated SSTs, ice-volume estimates, etc. (Kutzbach et al., 1993a). Some experiments have therefore adjusted the orbital parameters accordingly, allowing the large 18 ka B.P. CLIMAP data set (dated in radiocarbon years) to be compared with 21 ka B.P. model simulations (Kutzbach et al1997).

in the southern hemisphere) but was higher at other times (Fig. 12.5). In fact, summer insolation in the northern hemisphere was lower at 115 ka B.P. than at any time in the last 200 ka (see Fig. 2.18). This provides an excellent opportunity to investigate whether orbital forcing alone could have brought about changes in the climate system, sufficient to permit the growth of ice in those areas thought to have been critical in the initiation of glaciation (Ives et al., 1975; Clark et al., 1993).

The earliest attempt to address this problem used a coarse-resolution GCM with modern SSTs and C02 levels prescribed (Royer et al., 1984). This indicated that temperatures in northeastern North America were reduced and precipitation was increased (more than evaporation), suggesting that conditions at 115 ka B.P. were indeed more favorable for ice-sheet growth, but the model was simply too crude to determine if an ice sheet could be sustained. Rind et al. (1989) carried out several experiments using an 8° (latitude) x 10° (longitude) GCM, initially with just modern SSTs and changed orbital parameters, but then with lower C02 levels and SSTs as well. They found that with orbital changes alone ice growth could not be achieved in the model, a conclusion also reached by Phillips and Held (1994) using the GFDL GCM coupled to a mixed layer ocean. Lowering C02 levels by 70 ppm (to 230 ppm) made little difference. Only when a 10-m thick ice sheet was

FIGURE 12.5 Difference in solar insolation (in W m"2) at the top of the atmosphere at pared to present. Negative anomalies (lower radiation at I 15 ka B.P.) are shown as dashed Valdes, 1995).

FIGURE 12.5 Difference in solar insolation (in W m"2) at the top of the atmosphere at pared to present. Negative anomalies (lower radiation at I 15 ka B.P.) are shown as dashed Valdes, 1995).

115 ka B.P. corn-lines (Dong and specified (to reduce surface albedo), C02 levels were lowered to 230 ppm and SSTs were reduced to those associated with full glacial conditions (as determined by CLIMAP 1981) could an ice sheet be maintained (Rind et al., 1989). This may be because the model is too coarse and so does not adequately represent topography and upland surfaces where ice sheets probably first formed. It may also not simulate very well the critically important feedbacks between snow, ice, and low clouds. In similar experiments, Gallimore and Kutzbach (1995), using a low-resolution GCM with a mixed layer ocean, concluded that snow cover could persist in northwestern and northeastern Canada, for -11 mo per yr without a reduction in C02. If C02 levels had been set lower and if cloud cover had not been prescribed, it seems likely that conditions in these areas would have been quite favorable for ice sheet development.

Recently, a higher-resolution GCM has been employed to examine the same problem (Dong and Valdes, 1995). The UGAMP (U.K. Universities Global Atmospheric Modeling Programme) GCM is based on the very successful forecast model of the European Center for Medium Range Weather Forecasts (ECMWF). It has a grid spacing of -2.8° and includes 19 vertical levels, 5 of which are in the lower atmospheric boundary layer (>850 mb). It includes a 50-m thick mixed layer ocean and an interactive surface hydrology. Topographic resolution is quite good, which may be of particular importance in experiments dealing with the initial stages of glaciation. Several studies have pointed out the critical significance of upland plateaus in Labrador, Baffin Island, and Keewatin in the development of a permanent snow cover (Ives et al., 1975). In the NASA-GISS (Goddard Institute for Space Studies) GCM used by Rind et al. (1989), topography is poorly resolved; for example, the Baffin area is represented as <200 m in elevation, whereas in the UGAMP GCM, the value is close to the actual elevation of -550 m.

The UGAMP GCM simulates modern conditions extremely well. Comparisons between modern conditions and 115 ka B.P. (with modern SSTs prescribed) shows pronounced cooling over all continents except Antarctica. Over North America, surface temperatures were 4-6 °C lower, accompanied by increased cloudiness (+10-15%) and increased soil moisture (+25-30%). However, these changes were not sufficient to maintain snow on the ground throughout the year, except in the Canadian High Arctic Islands and in a small area of Tibet. The C02 levels were then reduced slightly (to 280 ppm) and the model was coupled to a shallow mixed layer ocean, allowing SSTs to be determined by the model. This run showed cooling of 8-10 °C in summer over high latitudes, enabling snow to survive all year in many areas (Fig. 12.6). Lower temperatures were brought about by increased sea ice, cooler SSTs in summer, and the presence of permanent snow cover on land, which plays an important role in reinforcing the lower summer insolation effect by sharply increasing surface albedo. This result is quite intriguing because although the model simulated ice growth in northeastern North America and Fennoscandia, where ice sheets are known to have developed, it also generated ice in Siberia and Tibet (Oglesby, 1990). In fact, there has been much debate over whether there really were ice sheets over central Siberia and Tibet during the last glaciation (Velichko et al., 1984; Rutter, 1995; Kuhle, 1991). Although the simulation pertains to an earlier period, and certainly

for 115 ka B.P. with computed SSTs. Shaded areas show snow depth >0.25 m. Many high-latitude regions, and parts of Tibet and Patagonia appear to be "sensitive regions" for the initiation of permanent snow cover (Dong and Valdes, 1995).

does not prove that these areas were glaciated during the LGM (or at any other time), it does point to the sensitivity of these regions in the maintenance of a permanent snow cover (and this is also seen in other models) given certain changes in insolation and lower SSTs at high latitudes. On the other hand, the model also indicates that snow would have built up over much of Alaska and there is no evidence that this has occurred for at least several hundred thousand years; more recent glaciations in Alaska have generally been small and confined to mountainous coastal ranges in the south and the Brooks Range in the north. These experiments therefore raise many interesting questions and provide fertile ground for further field studies, as well as additional model simulations and data-model comparisons.

In both of the models already discussed here, lower levels of C02 (< 280 ppm) were found to contribute to the cooling needed to maintain a permanent snow cover at high latitudes. Syktus et al. (1994), using a GCM coupled to a dynamic upper ocean circulation, set the orbital configuration to 116 ka B.P. and then examined the sensitivity of climate to C02 levels from 260-460 ppm. They found that snow cover does not change very much with orbital forcing until C02 levels drop to some critical level, in the range of 350-450 ppm. However, with lower levels of C02 the climate system is much more sensitive to insolation variations and snow cover can expand rapidly under orbital configurations with low summer insolation. This threshold sensitivity may have played a role in the long-term evolution of glacial-interglacial cycles; if C02 levels were at present-day levels (-350 ppm) or higher (as they may have been in the Pliocene), orbital variations may not have produced the same climate response, leading to the growth of large continental ice sheets, as occurred in the late Quaternary (Li et al., 1998).

Schlesinger and Verbitsky (1996) also found that C02 levels were critical in generating ice sheets at high latitudes of the northern hemisphere. They used a GCM coupled to a mixed-layer ocean to derive climatic conditions resulting from orbital conditions at 115 ka B.P., with C02 levels varying from 326 to 246 ppm. The 246 ppm experiment was carried out to address the combined effects of lower C02 and CH4. The climate so generated was then used as input to a global ice sheet/asthenosphere model. As with the other experiments already discussed here, they found that orbital changes alone were insufficient to produce a significant change in permanent snow cover. However, with an "equivalent C02 level" of 246 ppm, ice sheets built up in North America, most of Siberia, and northeastern Europe (but east of northern Fennoscandia). Over a 10,000-yr period (nominally 115-105 ka B.P.) continental ice volume increased, but not as rapidly as the SPECMAP benthic 8lsO record appears to show. However, there were two important limitations in these simulations. First, the model did not allow any precipitation falling as rain on the newly formed ice sheets to accumulate. When this condition was completely changed (all precipitation falling on the newly formed ice sheets was assumed to be snow) the ice sheets grew much larger (volumetrically and geographically) over a 10,000-yr interval, accounting for 86% of the sea-level equivalent of the SPECMAP record (though this record itself may not only represent sea-level change; see Section 6.3.4). The other limitation is that the model climate calculated for 115 ka B.P. was maintained as constant over the ensuing 10,000-yr period during which the ice sheets were allowed to build up. This means that potentially important positive feedbacks, related to snow-albedo changes, were not included, and these would no doubt have played a key role in the process of ice sheet growth.

One additional factor none of the previous simulations of 115 ka B.P. conditions explicitly considered is the role of snow-albedo feedback resulting from changes in vegetation cover. This issue was examined by Gallimore and Kutzbach (1996) using the National Center for Atmospheric Research (NCAR) Community Climate Model Version 1 (CCM1) coupled to a mixed layer ocean with interactive sea ice. With orbital changes alone, tundra vegetation increased in area by 25% relative to today (effectively expanding by 5° latitude to the south, according to a biome model driven by output from the GCM; Harrison et al., 1995). When the combined effects of insolation change, reduced C02 levels (to 267 ppm) and increased albedo (due to the increased area of tundra) were subsequently considered, summer temperatures over land areas at 60-90° N fell by 8-9 °C, accompanied by more extensive and thicker sea ice. However, other experiments showed that only a slight additional expansion of tundra regions (i.e., > 5° latitudinal expansion) precipitated a dramatic change, with summer temperatures falling by a further 10-15 °C, leading to year-round snow cover at high latitudes. This result points to the potential importance of nonlinear vegetation/snow albedo feedbacks in the initiation of glaciation. Other studies have examined vegetation/albedo feedbacks in different situations (see the next section) and reached similar conclusions: vegetation changes appear to play an important role in the climate system and must be explicitly considered in order to simulate paleoclimatic conditions correctly (i.e., to match the "observed" paleorecord) (Foley et al., 1994; Kutzbach et al., 1996; Crowley and Baum, 1997).

12.3.2 Orbital Forcing and Monsoon Climate Variability

Just as the earth's orbital configuration led to anomalously low solar radiation in the northern hemisphere summer at 115 ka B.P., so the same orbital shifts led to increased seasonality and unusually high summer solar radiation receipts in the early Holocene. At 9 ka B.P., the northern hemisphere received -7% more radiation than today in July (29-76 Wm 2), mainly as a result of perihelion occurring in July (compared to January 3 today) together with greater eccentricity and an increase in axial tilt (24.23° vs 23.45° today). Several model experiments have examined the effect of the orbital changes on the earth's climate (Kutzbach and Otto-Bliesner, 1982; Kutzbach and Guetter, 1986; Kutzbach and Gallimore, 1988; Mitchell et al, 1988; Hewitt and Mitchell, 1998). The main features of all these studies are similar; here we consider the results of Mitchell et al. (1988). They used an 11-layer United Kingdom Meteorological Office GCM (-5° x 7.5° resolution) coupled to a static mixed layer ocean and a simple sea-ice energy balance model. Thus SSTs and sea ice were allowed to change, though the (seasonally varying) heat flux was prescribed and invariant in both the control and 9 ka B.P. experiments. Figure 12.7 (a, b) shows the radiation differences between the two model runs, for incoming solar radiation at the top of the atmosphere and at the surface (after reflection by clouds, scattering and absorption in the atmosphere, etc.) Clearly the main effect is for an increase in the seasonal amplitude of radiation, with higher levels in the northern hemisphere summer, and lower levels in winter (see Fig. 12.5 — the effects are almost the opposite). The excess summer radiation resulted in higher temperatures over the northern hemisphere continents leading to lower pressure and an increase in airflow from the

the atmosphere and (b) at the surface. Negative anomalies (lower radiation at 9 ka B.R) are shaded (Mitchell et al., 1988).

oceans to the land. Thus, monsoon circulations were enhanced, leading to more precipitation in many parts of the Tropics. Figure 12.8 (a, b) summarizes the temperature changes in terms of zonal averages for land and ocean areas, respectively. The pattern of temperature change clearly reflects the radiation anomaly pattern seen in Fig. 12.7b, taking into account a thermal lag associated with surface heating and various feedbacks operating in the climate system. Cooler temperatures in the northern subtropics in late summer are a consequence of increased cloudiness resulting from enhanced monsoonal airflow. A reduction in sea-ice thickness and prolongation of the ice-free season causes large temperature anomalies over high-latitude ocean areas (Fig. 12.8b), which then maintain somewhat warmer conditions over northern continental regions throughout the winter months in spite of lower radiation amounts (bearing in mind that high-latitude winter radiation totals are small anyway, so that advection dominates temperatures at this time of year). Oceanic temperatures stay warm throughout the year over much of the globe because of the large heat capacity of the ocean and the fact that cloud cover over the oceans was reduced (related to increased subsidence over the oceans, compensating for the enhanced uplift over land areas). Thus, solar radiation increased over oceanic regions relative to the control simulation, enhancing the effect of the radiation anomaly.

Considering the geographical distribution of anomalies, summer temperatures were more than 4 °C warmer over most of Eurasia and 2-4 °C warmer over North America (Fig. 12.9). Sea-level pressure in northern continental interiors was lower by up to 6 mb, leading to enhanced monsoonal airflow and to heavier precipitation in a wide swath from northeastern Africa to southeast Asia, and over much of

FIGURE 12.8 Modeled zonal mean differences in air temperature: (9 ka B.P.- control) (a) over land areas, (b) over ocean areas (sea and sea ice). Negative values are shaded (Mitchell et at., 1988).

ISO 120 W 60 0 00 120 E 180

ISO 120 W 60 0 00 120 E 180

FIGURE I 2.9 Modeled temperature difference (°C) between 9 ka B.P. and control run for summer (JuneAugust). Warming was extensive across most of the northern hemisphere at 9 ka B.P. (Mitchell et al., 1988).

northern South America. Increased cloudiness in these areas limited the temperature change so that, on balance, soil moisture levels increased significantly. However, over land areas at higher latitudes, increased levels of evaporation resulting from the large change in temperature were not compensated for by higher rainfall amounts, so soil moisture levels fell over most of North America and northern Eurasia (Fig. 12.10).

These experiments were carried out without taking into account the presence of a substantial ice cap over northeastern North America at 9 ka B.P. When another simulation was run with an ice sheet inserted, temperatures were lowered over the ice sheet and areas downstream (North Atlantic and northwestern Europe), but the principal changes in monsoonal circulation remained, albeit somewhat reduced in intensity.

In many low-latitude regions, orbital conditions at 9 ka B.P. were clearly more favorable than today for the growth of vegetation and the development of lakes in basins of inland drainage. Paleoclimatic evidence supports this scenario (Street-Perrott et al., 1990); the ensuing few thousand years (when radiation anomalies were similar) saw dramatic changes in the environment of arid and semiarid areas from tropical west Africa to India (see Sections 7.6.3 and 9.9.4). At 125 ka B.P. the same orbital configuration led to even greater radiation anomalies, and GCM experiments show that the effects on monsoon climate regimes and high-latitude continental interiors must have been even more intense (Kutzbach et al., 1991; Harrison et al., 1991, 1995). Although limited, paleoclimatic data also point to higher levels of lakes at low latitudes during the last interglacial and drastically reduced sea ice around northern Eurasia at that time (Petit-Maire et al., 1991; Frenzel et al., 1992a).

120 W

120 E

120 W

120 E

90 N

90 N

120 W

120 E

FIGURE 12.10 Modeled soil moisture difference (cm) between 9 ka B.P. and control run for summer (|une-August). Shaded areas indicate drier conditions at 9 ka B.P. (Mitchell et al., 1988).

A careful comparison of the paleoclimatic evidence and the orbitally forced model run shows that soil moisture did not change enough in the 9 ka (or other 6 ka) B.P. experiments to account for the dramatic environmental changes recorded in sediments from now dry Saharan lakes that were much more extensive in the early- to mid-Holocene. Kutzbach et al. (1996) examined this question by comparing results from the NCAR CCM2 model with 6 ka B.P. orbital changes alone, to a model run with the same orbital configuration, plus a change in the land surface of North Africa, from primarily desert (90-100% bare ground) to mainly grassland (80% grass cover). A further experiment also replaced the desert soil with a more loamy organic-rich soil, more typical of a grassland. These changes in surface boundary conditions brought about an additional increase in precipitation, from 12% more than the control run (at 15°-22° N) due to orbital forcing alone, to a 28% increase with changes in radiation, vegetation, and soil type (Fig. 12.11). Although this experiment involved an unrealistically large change in vegetation cover (extending completely across the Sahara from 15°-30° N) it does at least indicate the potential significance of vegetation/soil feedbacks on the climate system and points to the importance of ensuring that such feedbacks are explicitly considered in studies of future climatic changes. Similar conclusions were reached by Foley et al. (1994) and TEMPO36 (1996) using more sophisticated atmosphere/mixed layer ocean models. They focused on the northern treeline at 6 ka B.P. and compared experiments with orbital forcing alone to experiments in which the area of boreal forest was also greatly expanded at the expense of tundra. The lower albedo of forest

36 TEMPO (Testing Earth System Models with Paleo-Observations) is the acronym of an interdisciplinary project led by Kutzbach.

July precipitation (mm per day)

July precipitation (mm per day)

15 20 Latitude

FIGURE 12.1 I July rainfall (mm day"') as a function of latitude, averaged over 0°-50° E in North Africa for modern observations and several GCM simulations. R = orbital configuration of 6 ka B.R; RVS = the same orbital configuration, but with desert vegetation replaced by grassland and desert soil converted to a loamy grassland soil. Control run = modern boundary conditions. Positive vegetation feedbacks enhance the orbital effect of higher rainfall in this region (Kutzbach et al., 1996).

15 20 Latitude

FIGURE 12.1 I July rainfall (mm day"') as a function of latitude, averaged over 0°-50° E in North Africa for modern observations and several GCM simulations. R = orbital configuration of 6 ka B.R; RVS = the same orbital configuration, but with desert vegetation replaced by grassland and desert soil converted to a loamy grassland soil. Control run = modern boundary conditions. Positive vegetation feedbacks enhance the orbital effect of higher rainfall in this region (Kutzbach et al., 1996).

and associated changes in snow cover resulted in additional warming, especially in Spring, again indicating the role of vegetation feedbacks in long-term climate changes.

These modeling studies clearly demonstrate the significance of orbital forcing in generating dramatic changes in the earth's climate. Orbital changes are a necessary, but perhaps not entirely sufficient, condition for the growth and decay of ice sheets at high latitudes and the waxing and waning of the monsoons at lower latitudes. However, many other factors are known to have changed and these must certainly have played important additional roles: changes in trace gases (C02, CH4, N20), continental and volcanic aerosols, vegetation cover and surface albedo, topographic effects of the ice sheets, changes in sea level, sea-ice extent, and the oceanic (thermohaline) circulation regime. Many of these factors can be viewed as additional feedbacks induced by orbital forcing, but there may have been many non-linear internal interactions that became the dominant forcing once some critical threshold was crossed (see Section 6.12). Additional sensitivity experiments can be designed to address these forcings explicitly, alone or in combination, so that the complexity of the climate system can eventually be deciphered.

12.3.3 The Influence of Continental Ice Sheets

The most dramatic change in boundary conditions during the last glacial maximum (LGM) was the presence of large ice sheets over North America and Fennoscandia. One can surmise that the very presence of these ice sheets must have had a significant effect on albedo, airflow, and energy exchange not only over the ice sheets themselves but perhaps regionally as well. A general circulation model simulation with topographically realistic ice sheets37 compared to a similar simulation with only the modern distribution of continental ice enables a quantitative assessment of the effect of large ice sheets to be made, and the role of feedbacks with other parts of the climate system to be examined explicitly. To this end, Manabe and Broccoli (1985a, b) used an atmospheric general circulation model coupled to a mixed layer ocean to investigate climate system conditions with large and extensive (though probably unrealistic!) LGM ice sheets specified according to the "maximum" reconstruction of Hughes et al. (1981) (Fig. 12.12).

Comparison of the ice sheet simulation with the control run indicates that very low air temperatures occur downstream (east) of the Laurentide ice sheet (up to 32 °C colder in winter and 8 °C lower in summer, south of Greenland). Temperatures are also considerably depressed in northeastern Asia. By contrast, little or no

180 150 120 90 60 30W 0 30E 60 90 120 150 180 FIGURE I 2.12 Continental outlines, topography (km a.s.l.), and distribution of continental ice sheet extent and height prescribed for an experiment to assess the role of a thick ice sheet on the atmospheric circulation. Top = present (control run); bottom = LGM experiment (Manabe and Broccoli, 1985a).

180 150 120 90 60 30W 0 30E 60 90 120 150 180 FIGURE I 2.12 Continental outlines, topography (km a.s.l.), and distribution of continental ice sheet extent and height prescribed for an experiment to assess the role of a thick ice sheet on the atmospheric circulation. Top = present (control run); bottom = LGM experiment (Manabe and Broccoli, 1985a).

37 Recent estimates by Peltier (1994) suggest that the height of the ice sheets, especially over North America, may have been overestimated by CLIMAP (1981) for the LGM.

change is observed in the southern hemisphere, which contrasts with marine sediment-based paleoclimatic reconstructions of SSTs, which have SSTs, lower by up to 6 °C in parts of the southern ocean. This result suggests that ice sheets in the northern hemisphere have little impact on southern hemisphere conditions. Although less energy enters the system (because of the higher albedo of the ice sheet) this is compensated for by a reduction in outgoing longwave radiation, so the net effect is small, requiring minimal interhemispheric heat transfer. If the presence of large ice sheets in the northern hemisphere did not cause the southern hemisphere to cool, some other factor (or factors) must be invoked to explain the observed SSTs in this region. To investigate this further, Manabe and Broccoli (1985b) re-ran the LGM ice sheet simulation with C02 levels lowered by 100 ppm (to 200 ppm). This change was sufficient to lower SSTs in the southern hemisphere, bringing the model more in line with paleoclimatic reconstructions (Figure 12.13).

The Laurentide ice sheet (~3 km in height at its center in this simulation) has a significant effect on tropospheric circulation, causing a split airflow around it and

FIGURE 12.13 Latitudinal difference in annually averaged zonal mean SSTs (°C) between several different GCM experiments. EI used modern continental land and sea-ice distribution, modern land albedo for snow-free areas, and a C02 level of 300 ppm. E2 changed only the continental ice and sea-ice distribution to that of the LGM. E3 was the same as E2, but also changed the land albedo for snow-free regions to glacial values. E4 was like E3, but in addition changed C02 levels to 200 ppm. E2-EI thus shows the overall effect of the ice sheets on zonal mean SSTs, indicating little effect on the southern hemisphere. Changing the albedo of unglacierized areas (E3-E2) only further lowers the temperatures slightly. Dropping C02 levels to glacial values (E4-E3) produces a more symmetrical response in both hemispheres but has the major effect on temperatures in the southern hemisphere.The combined effects (E4-EI) show broad similarities to differences between CLIMAP LGM SST reconstructions and present-day conditions (shown as black dots) but the model results indicate colder conditions in low latitudes and much warmer temperatures at high latitudes due to reduced sea-ice formation (Manabe and Broccoli, 1985b).

FIGURE 12.13 Latitudinal difference in annually averaged zonal mean SSTs (°C) between several different GCM experiments. EI used modern continental land and sea-ice distribution, modern land albedo for snow-free areas, and a C02 level of 300 ppm. E2 changed only the continental ice and sea-ice distribution to that of the LGM. E3 was the same as E2, but also changed the land albedo for snow-free regions to glacial values. E4 was like E3, but in addition changed C02 levels to 200 ppm. E2-EI thus shows the overall effect of the ice sheets on zonal mean SSTs, indicating little effect on the southern hemisphere. Changing the albedo of unglacierized areas (E3-E2) only further lowers the temperatures slightly. Dropping C02 levels to glacial values (E4-E3) produces a more symmetrical response in both hemispheres but has the major effect on temperatures in the southern hemisphere.The combined effects (E4-EI) show broad similarities to differences between CLIMAP LGM SST reconstructions and present-day conditions (shown as black dots) but the model results indicate colder conditions in low latitudes and much warmer temperatures at high latitudes due to reduced sea-ice formation (Manabe and Broccoli, 1985b).

the development of a strong jet stream on its southeastern margin (Shinn and Barron, 1989). Cold air flow around the ice sheet caused extensive, thick sea ice to develop in the Labrador Sea and North Atlantic Ocean. Over continental interiors, air temperature differences between the two model simulations are smaller than in most paleoclimatic reconstructions, perhaps because surface albedos (specified in both models) were too low. Finally, maps of soil moisture change show that large areas of the continents south of the ice sheets were significantly drier during the LGM compared to today (due mainly to lower precipitation amounts) and these regions broadly correspond to known loess regions.

Rind (1987) carried out four experiments with a NASA-GISS GCM with 8° x 10° resolution. Starting with a simulation of modern conditions (the control), he added, sequentially, ice-age SSTs (as determined by CLIMAP 1981), a 10-m thick ice sheet over the area occupied by continental ice sheets at the last glacial maximum, and full ice sheets with appropriate elevations. In this way, he was able to investigate the importance of each of these factors to the atmospheric circulation. As in Manabe and Broccoli's experiment, Rind's simulations showed that thick ice sheets have both thermal and topographic effects; because of their elevation and high albedo, surface temperatures are low and atmospheric water vapor above the ice is limited. Consequently infra red (IR) radiation losses are high due to the (effectively) reduced optical thickness of the atmosphere above the ice sheets. Over North America the jet stream bifurcates around the topographic barrier created by the Laurentide ice sheet (in Europe the ice sheet is thinner so this effect is less significant). High pressure at the surface over the ice sheets sets up strong disturbances to zonal (west-east) flow, creating a pronounced meridional circulation regime. However, Rind found that stationary waves established in the northern hemisphere were accompanied by similar changes in the southern hemisphere, implying some sort of dynamic link between the two hemispheres, a feature not seen in Manabe and Broccoli's simulations. Of particular interest in Rind's experiment is the fact that the mass balance of the ice sheet was strongly negative under the prescribed conditions and, had the "ice sheet" been 10-m thick, it would have melted away completely in just a few years. This suggests that the large continental ice sheets at 18 ka B.P., though clearly having a major influence on atmospheric circulation, were not in balance with prevailing climatic conditions. Furthermore, land temperatures and precipitation amounts at low latitudes were higher than indicated by paleoclimatic data. Rind re-ran his simulation with C02 levels 70 ppm lower but this only cooled the atmosphere slightly, not enough to alter the ice mass balance significantly or to reduce low-latitude temperatures and precipitation amounts enough. Only when CLIMAP SSTs were reduced by at least 2 °C at low latitudes were the model simulations brought into alignment with the observational data (Rind and Peteet, 1985). In fact, given the CLIMAP SST boundary conditions specified, the model simulation is not in radiative equilibrium; however, if SSTs are reduced by ~2 °C it brings about an approximate radiative balance.

A further perspective on this problem is provided by Hall et al. (1996). They used the high-resolution UGAMP GCM to simulate LGM climate; ice-sheet elevation, land surface albedo, and CLIMAP SSTs were specified, C02 levels were reduced to 190 ppm, and the orbital configuration was set to that of 21ka B.P. Temperature differences between the control and LGM runs are shown in Fig. 12.14. Winters were dramatically colder over much of the northern hemisphere extra-tropics (up to 50 °C colder over Iceland as a result of the advection of cold Arctic air). In summer, maximum cooling was associated with the Laurentide (-39 °C) and Fennoscandian (-30 °C) ice sheets. Unlike earlier studies, bifurcation of the jet stream around the Laurentide ice sheet was not observed. Storm tracks were concentrated along the southern margins of the ice sheets and sea-ice boundaries, leading to heavier precipitation (much of it as snow) from coastal Alaska and northern British Columbia, across northeastern North America to central Europe (Fig. 12.15). Snowfall was especially heavy over northwestern North America in winter. By considering both accumulation of snow and ablation, Hall et al. estimated that the Laurentide and Fennoscandian ice sheets would have had an overall positive balance, with net accumulation over much of the ice sheets and net ablation confined to a narrow strip along the southern margins. However, they did not use these results as input to an ice sheet model, with ice sheet dynamics, so what the actual mass balance of the ice sheets would have been is not apparent from these results.

based on the UGAMP GCM with CLIMAP-based ice-sheet elevations, SSTs, and sea-ice extent prescribed (Hall etai, 1996).
>182 mm/yr snowfall anomaly >365 mm/yr snowfall anomaly >730 mm/yr snowfall anomaly

FIGURE 12.15 Mean annual snowfall anomalies at 21 ka B.R (relative to the present-day control run) according to the UGAMP GCM, with CLIMAP-based ice-sheet elevations, SSTs, and sea-ice extent prescribed (Hall et ai, 1996).

12.4 MODEL SIMULATIONS: 18 KA B.P. TO THE PRESENT

A series of experiments designed to examine the climatic response to orbital changes and surface boundary conditions has been carried out by Kutzbach and associates, using the NCAR Community Climate Model (CCM) (Kutzbach and Guetter, 1986; COHMAP 1988; Kutzbach et ai, 1993b). The GCM simulations were carried out at 3000-yr intervals, from 18 ka B.P. to the present, for January and July conditions. Radiation anomalies due to orbital changes over this interval are shown in Fig. 12.16. Potentially important changes in boundary conditions are summarized in Fig. 12.17 (Kutzbach and Ruddiman, 1993). As noted earlier, radiation conditions at 18 ka B.P. are similar to those of today (i.e., the control experiment) so the 18 ka simulation is principally a test of altered boundary conditions, like those of Rind (1987) and Manabe and Broccoli (1985b). Simulations for the later periods, especially from 9-3 ka B.P., mainly examine the way in which the atmosphere responds to changes in seasonality, because boundary conditions are similar to those of today (except for a small ice sheet over North America at 9 ka B.P.) but there are pronounced differences in radiation over the annual cycle (Section 12.3.2).

At 18 ka B.P. global temperatures were lower by -3.4 °C (average of January and July means), assuming prescribed CLIMAP SSTs are correct. Other estimates for temperature changes at the LGM are given in Table 12.1. As in other LGM simulations, the hydrological cycle was less intense (i.e., both precipitation and evaporation were lower) in spite of generally higher wind speeds. The NCAR CCM

0 0

Post a comment