ACIA Chapter 5: Cryospheric and Hydrologic Variability

Transcription

1 ACIA Chapter 5: Cryospheric and Hydrologic Variability Contents Summary Page Introduction Page Precipitation and evapotranspiration (J. Walsh, M. Serreze) Page A Background 5.1.B Recent and ongoing changes 5.1.C Projected changes 5.1.D Impacts of projected changes 5.1.E Critical research needs 5.2 Sea ice (J. Walsh, T. Jakobsson) Page A Background 5.2.B Recent and ongoing changes 5.2.C Projected changes 5.2.D Impacts of projected changes 5.2.E Critical research needs 5.3 Snow cover (J. Walsh) Page A Background 5.3.B Recent and ongoing changes 5.3.C Projected changes 5.3.D Impacts of projected changes 5.3.E Critical research needs 5.4 Glaciers and ice sheets (J. Oerlemans, J. O. M. Hagen) Page A Background 5.4.B Recent and ongoing changes 5.4.C Projected changes 5.4.D Impacts of projected changes 5.4.E Critical research needs 5.5 Permafrost 5.5(1) Terrestrial permafrost (V. Romanovsky, O. Anisimov) Page (1).A Background 5.5(1).B Recent and ongoing changes 5.5(1).C Projected changes 5.5(1).D Impacts of projected changes 5.5(1).E Critical research needs External Review January DO NOT CITE OR CIRCULATE

2 5.5(2) Coastal and subsea permafrost (S. Solomon) Page (2).A Background 5.5(2).B Recent and ongoing changes 5.5(2).C Projected changes 5.5(2).D Impacts of projected changes 5.5(2).E Critical research needs 5.6 River and lake ice (T. Prowse) Page A Background 5.6.B Recent and ongoing changes 5.6.C Projected changes 5.6.D Impacts of projected changes 5.6.E Critical research needs 5.7 Freshwater discharge (A. Shiklomanov, N. Savelieva, I. Shiklomanov) Page A Background 5.7.B Recent and ongoing changes 5.7.C Projected changes 5.7.D Impacts of projected changes 5.7.E Critical research needs 5.8 Sea level rise and coastal stability (S. Solomon) Page A Background 5.8.B Recent and ongoing changes 5.8.C Projected changes 5.8.D Impacts of projected changes 5.8.E Critical research needs External Review January DO NOT CITE OR CIRCULATE

3 Summary Recent observational data present a generally consistent picture of cryospheric variations that are shaped by patterns of recent warming and variations of the atmospheric circulation. Sea ice coverage has decreased by 5-10% during the past few decades. The decrease is greater in the summer, when new period-of-record minima have been reached several times in the most recent decade. The coverage of multiyear ice has also decreased, as has the thickness of sea ice in the central Arctic. Snow-covered area has diminished since the early 1970s by several percent over both North America and Eurasia. River discharge over much of the Arctic has increased during the past several decades, and the springtime discharge pulse is occurring earlier on many rivers. The increase of discharge is consistent with an irregular increase of precipitation over northern land areas. Permafrost temperatures over most of the subarctic land areas have increased during the past few decades by several tenths C to as much as 2-3 C. Glaciers throughout much of the Northern Hemisphere have lost mass over the past several decades, as have coastal regions of the Greenland ice sheet. The glacier retreat has been especially large in Alaska since the mid-1990s. During the past decade, glacier melt has resulted in an estimated sea level increase of mm/year. Earlier break-up and later-freeze-up have combined to lengthen the ice-free season of rivers and lakes by 1-3 weeks over the past century in much of the Arctic. The lengthening of the ice- free season has been greatest in the western and central portions of the northern continents. While the various cryospheric and atmospheric changes are consistent in an aggregate sense and are quite large in some cases, it is likely that low-frequency variations in the atmosphere and ocean have played at least some role in forcing the cryospheric and hydrologic trends of the past few decades. Model projections of greenhouse-driven warming indicate a continuation of the recent trends through the next century, although the rates of the projected changes vary widely among the models. For example, Arctic river discharge may increase by an additional 5-25% by the late 21 st century. Trends toward earlier break-up and later freeze-up of Arctic rivers and lakes are also likely if the projected warming occurs. The wastage of Arctic glaciers and the Greenland ice sheet is projected to contribute several cm to global sea level rise by Superimposed on the glacial contributions to sea level change are the effects of thermal expansion and isostatic rebound, which combine to produce a spatially variable pattern of sea level rises of several tens of centimeters in some areas (the Beaufort Sea and much of the Siberian coast) and sea level decreases in other areas (e.g., Hudson Bay, Novaya Zemlya). Models project a 21 st -century decrease of sea ice by more than 50% in the summer season, with a corresponding lengthening by 2-4 months of the navigation season in the Northern Sea Route. Snow cover is projected to continue to decrease, with the largest decreases projected for spring and autumn. Over the next century, permafrost degradation may occur over 10-20% of the present permafrost area, and the outer limit of permafrost may move northward by several hundred km. An acceleration of Arctic coastal erosion and the degradation of coastal permafrost is likely to occur in the next century in response to a combination of Arctic warming, the increase of sea level and the retreat of sea ice External Review January DO NOT CITE OR CIRCULATE

4 5.0 INTRODUCTION The term cryosphere is defined (NRC Canada, 1988) as: That part of the earth s crust and atmosphere subject to temperatures below 0 C for at least part of each year. For purposes of monitoring, diagnosis, projection and impact assessment, it is convenient to distinguish the following components of the cryosphere: sea ice, seasonal snow cover, glaciers and ice sheets, permafrost, and river and lake ice. Each of these variables is addressed in a separate section (Sections ) of this chapter. In addition, we include a section (5.1) addressing precipitation and evapotranspiration, which together represent the net input of moisture from the atmosphere to the cryosphere. Section 5.7 addresses the surface flows that are primary hydrologic linkages between the terrestrial cryosphere and other parts of the Arctic system. Surface flows will be critical to the impacts of cryospheric changes on the terrestrial and marine ecosystems of the Arctic, as well as on Arctic and perhaps global climate. Finally, we conclude with Section 5.8 on sea level variations that may result from changes in the cryosphere and Arctic hydrology. The different components of the cryosphere have widely varying timescales, and some of the components are not in equilibrium with today s climate. In the following sections, we examine each cryospheric component in terms of its recent and ongoing changes, as well as projected changes over the next century. The discussions of change are preceded by summaries of the present distributions of each variable. Each section also includes brief summaries of the impacts of the projected changes, although these summaries rely heavily on references to later chapters in which many of the impacts are covered in more detail. We conclude each section with a brief description of the key research needs that must be met in order to reduce uncertainties in the diagnoses and projections described here. REFERENCE NRC Canada, 1988: Glossary of permafrost and related ground-ice terms Permafrost Subcommittee, National Research Council of Canada 1988, Tech. Memor. 142, 156 pp. 5.1 PRECIPITATION AND EVAPOTRANSPIRATION (J. Walsh, M. Serreze) 5.1.A Background The Arctic s cryosphere and hydrologic system will respond not only to changes in the thermal state of the Arctic, but also to available moisture. For example, higher temperatures will alter the phase of precipitation, the length of the melt season, the distribution of permafrost and the depth of the active layer, with consequent impacts on river discharge, subsurface storages and glacier mass balance. But these systems also depend critically on the balance between precipitation (P) and evapotranspiration/sublimation (collectively denoted as E). The following assessments of P and E help to lay the groundwork for subsequent discussions in this chapter. The distribution of precipitation and evapotranspiration in the Arctic has been a subject of accelerating interest in recent years. Two factors account for this surge of External Review January DO NOT CITE OR CIRCULATE

5 interest. The first is the realization that variations in hydrologic processes in the Arctic have major implications not only for Arctic terrestrial and marine ecosystems, but for the cryosphere and the global ocean. The second arises from the large uncertainties in the pan-arctic distributions of P and E. Uncertainties of even the present-day distributions of P and E are sufficiently large that evaluations of recent variations and trends are problematic. These large uncertainties are due to a combination of the sparse network of in situ measurements of P (several hundred stations, with very poor coverage over Northern Canada and the Arctic Ocean), and the virtual absence of such measurements of E (mostly from fields programs of short duration), in the Arctic; the difficulty of obtaining accurate measurements of solid precipitation, even at manned weather situations, in cold windy environments; the compounding effects of elevation on P and E in topographically complex regions of the Arctic, where the distribution of observing stations is biased toward low elevations and coastal regions; slow progress in the exploitation of remote sensing techniques for the measurement of high-latitude precipitation and evapotranspiration. This relates to the heterogeneous emissivity of snow and ice covered surfaces, difficulties in cloud/snow discrimination and the near-absence of coverage by ground-based radar. Progress in mapping the spatial and seasonal distributions of Arctic precipitation has resulted from the use of information on gauge bias adjustment procedures, e.g., from the WMO s Solid Precipitation Measurement Intercomparison (Goodison et al., 1998). Summaries of precipitation over the Arctic Ocean, where only coastal stations and drifting ice station measurements are available, have recently been completed by Colony et al. (1998), Yang (1999) and Bogdanova et al. (2002). In the latter study, which accounts for all the major systematic errors of precipitation, the annual mean biascorrected precipitation for the central Arctic Ocean was found to be 16.9 cm, which is 32% higher than the measured value. The spatial pattern shows an increase from minimum values of less than 10 cm/yr over Greenland and cm over much of the Arctic Ocean, to more than 50 cm/yr over parts of the North Atlantic subpolar seas. Estimates of evaporation over the Arctic Ocean are scarce. Probably the best estimates are from the one-year SHEBA project in the Beaufort Sea collected during 1997 and These observations show evaporation as nearly zero from October through April peaking in July at about 7 mm (Persson et al., 2002). Estimates of P and E for the major terrestrial watersheds of the Arctic have recently been compiled by Serreze et al. (2003). Basin-averaged values of the annual mean P, P- E, runoff (R) and E (computed in two ways) are shown in Table In this study, P was derived from objectively analyzed fields of gauge-adjusted station measurements; P- E from the atmospheric moisture flux convergences in the NCEP/ NCAR reanalysis; R from river gauges near the mouths of the major rivers; and E from computations of two External Review January DO NOT CITE OR CIRCULATE

6 types of differences: E1, the difference between the independently derived P and P-E; and E2, the difference between basin-averaged P and R. Table Annual mean water budget components in four major drainage basins. See text for definition of symbols. P (mm ) P-E (mm) E1, E2 (mm) R (mm) R/P Ob , Yenisey , Lena , Mackenzie , Note that the two estimates of E differ by as much as 20%, providing a measure of the uncertainty in the basin-scale means of the hydrologic quantities. At least some, and probably most, of the uncertainty arises from biases in P. All basins show summer maxima in P and E, and summer minima in P-E (Fig ). Fig Mean monthly precipitation (P), precipitation minus evapotranspiration (P-E) and evapotranspiration (E) for the four major Arctic watersheds, based on data from 1960 through Units are mm. E was calculated as a residual from P and P-E. Seasonal cycles correspond to water year. [From Serreze et al., 2003]. P-E is essentially zero during July-August in the Mackenzie Basin, and negative during June-July in the Ob Basin, illustrating the importance of evapotranspiration in the hydrologic budget of Arctic terrestrial regions. In addition, about 25% of the July precipitation in the large Eurasian basins is associated with the recycling of moisture from evapotranspiration (Serreze et al., 2003). The relatively low ratios of runoff to precipitation in the Ob Basin are indicative of the general absence of permafrost (19% coverage in this basin, see Table 5.7.2), while the relatively high ratios (and smaller evapotranspiration values) in the Lena and Mackenzie are consistent with the larger proportions of permafrost, which reduces infiltration and enhances runoff. Additional estimates, using data for earlier years, of the freshwater budget components of Arctic and worldwide rivers are provided by Oki et al. (1995; see also A more detailed analysis of the Mackenzie Basin water cycle has recently been provided by Rouse et al. (2003). The present-day hydrologic regimes of the various Arctic subregions are discussed further in Sections 5.3 (Snow cover) and 5.7 (Surface flows). 5.1.B Recent and ongoing changes Given the uncertainties in the climatologies of Arctic P and E, it is not surprising that information on recent variations and trends of these quantities is quite limited. Time series obtained from reanalyses are subject to inhomogeneities resulting from changes in External Review January DO NOT CITE OR CIRCULATE

7 the input data over multidecadal timescales, while trends computed using station data are complicated by measurement errors as noted earlier. The use of in situ measurements for trend determination is further complicated by the changes in the rain/snow ratio during periods of warming or cooling at high-latitude sites (Forland and Hanssen-Bauer, 2000). The IPCC (1996, 2001) has consistently indicated 20 th -century increases of precipitation in northern high latitudes (55-85ºN), as shown in Fig. 2.7 of IPCC (1996). The increase is similar to that in Karl s (1998) Arctic region, which includes the area poleward of 65ºN but excludes the waters surrounding southern Greenland. In both cases, the greatest increase appears to have occurred during the first half of the 20 th century. However, the time series are based on data from the synoptic station network, which is not only unevenly distributed but which has undergone changes over time. Nevertheless, the increase in the early 20 th century is reproduced by some model simulations of 20 th - century climate (Paeth et al., 2002; Kattsov and Walsh, 2002). Groisman and Easterling (1994) present data showing an increase of precipitation over northern Canada (poleward of 55ºN) since For the period since 1960, the gauge-adjusted and basin-averaged precipitation data of Serreze et al. (2003) show little discernible indication of trends in the annual means over the Ob, Yenisey, Lena and Mackenzie basins. However, summer precipitation over the Yenisey Basin shows a decrease of 5-10% over the four decades since The variations of P in these regions are associated with variations the atmospheric circulation. Although they are subject to the caveats that accompany trends of derived quantities in a reanalysis, trends of annual E (determined primarily by summer E) in the NCEP reanalysis are negative in the Ob Basin and positive in the Yenisey and Mackenzie Basin. Serreze et al. (2003) suggest that recent increases in winter discharge from the Yenisey may be associated with a thawing of permafrost in that basin. Additional discussion of recent trends in variables associated with precipitation are presented in Sections 5.3.C (Snow cover) and 5.7.C (Surface flows). 5.1.C Projected changes ACIA s five designated global models (see Chapter 1) were used as a basis for projections of 21 st -century changes of P, E and P-E. The models are denoted here by CGC (Canadian Climate Center), CSM (NCAR s Climate System Model), ECH (ECMWF/Hamburg), GFD (Geophysical Fluid Dynamics Laboratory) and HAD (Hadley Centre). The output of these five models, run with the B2 forcing scenario, are presented here as averages over the Arctic Ocean and the five largest Arctic river basins: the Ob, Yenisey, Pechora, Lena and Mackenzie. For each of the three variables, the projected changes by are generally smaller than the across-model differences in the present-day simulations of these variables by the same models. For example, the ranges of P and E simulated for the period by these models shown in Table External Review January DO NOT CITE OR CIRCULATE

8 Table Across-model ranges of annual means (mm) of P and E for P E Arctic Ocean 220 (ECH) (CGC) 39 (CGC) - 92 (CSM) Ob 708 (CSM) (HAD) 302 (ECH) (HAD) Yenisey 604 (CSM) (CGC) 224 (ECH) (CGC) Lena 552 (CSM) (CGC) 200 (CSM) -312 (HAD) Pechora 493 (CSM) (CGC) 144 (CSM) (HAD) Mackenzie 670 (CSM) (HAD) 330 (CSM) (ECH) In general, the models project modest increases of precipitation by the end of the 21 st century. The changes projected for ACIA s third 21 st -century timeslice, , are shown in Figure as percentages of the present-day ( ) amounts simulated by the same models. The values of P, E and P-E projected for the earlier time slices are Fig Percentage changes of P (upper panel), E middle panel) and P-E (lower panel)projected for by the ACIA-designated climate models. Projected changes are shown for the Arctic Ocean and five major Arctic river basins. Solid circles are five-model means; vertical line segments denote the ranges of the five model projections. generally between the models present-day values and those for the time slice, although sampling variations result in some instances of non-monotonicity, especially when the changes are small. As indicated in Figure 5.1.2, the projected changes of E vary by greater percentages across the models than do the projected changes of P. There is even considerable disagreement among the models concerning the sign of the changes of E; in every region, at least one model projects a decrease of E, although the greater number of the changes of E are positive. However, the base values of E from which the changes occur are much smaller than the corresponding base values of P (Table 5.1.2), so the actual changes of E are generally smaller than the actual changes of P. Of the quantities considered here, the one with the greatest relevance to other parts of the Arctic system is P-E, which represents the net moisture input to the surface from the atmosphere. With one exception (the CSM, which shows by far the weakest greenhouse warming of the five models), the changes of P-E are positive (Figure 5.1.2). The largest increase, 14% (averaged across all models), occurs over the Arctic Ocean, where even the CSM projects an increase of P-E. Over the terrestrial watersheds, the increases range from 6% to 12% (averaged across all models). These changes are considerably smaller than the departures from the means occurring during individual years and even during multiyear periods. Since the changes of P and P-E are generally positive, it is likely that the most consequential changes of these variables will be increases of the frequency and/or duration of wet periods. However, there is a potentially important seasonality that is obscured by the annual averaging of the quantities in Figure The changes of P-E are generally smaller, and occasionally negative, over the major river basins during the warm season. This relative decrease of P-E during summer is the result of two factors: (1) an increase of E in association with the warming, and (2) a longer season with a snow-free surface and above-freezing temperatures in the upper soil layers, resulting in External Review January DO NOT CITE OR CIRCULATE

9 greater evapotranspiration. Consequently, the model projections point to the distinct possibility that increased river flow rates during winter and spring will be accompanied by decreased flow rates during the warm season. The latter is consistent with the results of the Mackenzie Basin Impact Study (Cohen et al., 1997). 5.1.D Impacts of projected changes (1) On other parts of the physical system: The increases of P, and more importantly of P-E, imply an increase of water availability for soil infiltration and runoff. The increases of P-E projected to occur by over the major terrestrial watersheds imply that the annual mean discharge to the Arctic Ocean will increase by 6-12%. Since the annual mean P-E over the Arctic Ocean is projected to increase by 14% over this same time period, a substantial increase of the freshwater supplied to the Arctic Ocean is projected to occur by the later decades of the present century. An increase of the supply of freshwater has potentially important implications for the Arctic Ocean s stratification, for its sea ice regime, and for its freshwater export to the North Atlantic. In addition, increased aquatic transport and associated heat fluxes across the coastal zone may accelerate the degradation of coastal permafrost in some areas. The increases of P and P-E imply generally wetter soils when soils are not frozen, increased surface flows above frozen soils, wetter active layers in the summer, and greater ice content of the upper soil layer during winter. To the extent that the increase of P is manifest as an increase of snowfall during the cold season (Section 5.3), the Arctic Ocean and its terrestrial watersheds will experience increases of snow depth and snow water equivalent, although the seasonal duration may well be shorter if warming accompanies the increase of P. Moreover, the increase of P-E in the annual mean obscure important seasonality. Recent trends of increasing E in the Yeinsey and Mackenzie Basins (Section 5.1.B) raise the possibility that P-E may actually decrease during the summer when E exceeds P, resulting in a drying of soils during the warm season. (2) On ecosystems: The increase of P-E over the terrestrial watersheds will increase the moisture availability in the upper soil layers, favoring plant growth in regions that are otherwise moisture-limited. The caveat noted above is that increases of E during the summer may lead to seasonal (warm-season) drying and reduced summertime river levels. A complicating factor is this scenario is the thawing of permafrost, which could increase the subsurface contribution to streamflow, possibly mitigating the effect of increased E during summer. The increase of river discharge will likely result in enhanced fluxes of nutrients and sediments to the Arctic Ocean, with corresponding impacts on coastal marine ecosystems (Chapter 8). Higher rates of river- and streamflow will likely have large impacts on riparian regions and flood plains in the Arctic if increases of P are accompanied by an increase of flood events, which are likely if P increases during winter and break-up is accelerated. Wetland ecosystems may actually expand in a climate regime of enhanced P-E, with corresponding changes in the fluxes of trace gases (e.g., CO 2 and CH 4 ) across the surface-atmosphere interface. External Review January DO NOT CITE OR CIRCULATE

10 (3) On people: Increases of P and P-E will result in generally greater availability of surface moisture for Arctic residents. In permafrost-free areas, water tables will be closer to the surface, and moisture availability for agriculture will increase. During the springtime period when enhanced P and P-E are likely to increase river levels, the risk of flooding will increase. Lower water levels during the summer would affect river navigation, increase the threat from forest fires, and affect hydropower generation. 5.1.E Critical research needs It is apparent from Tables and that models differ widely in their simulations of P and E, not only in greenhouse simulations but also in their simulations of present-day Arctic climate. The result is a very large range of uncertainty in future rates of moisture supply to the Arctic surface. There is an urgent need to narrow the range of this uncertainty by determining the reasons for the large across-model variances of P and E, and by bringing the models present-day simulations of P and E into closer agreement with observational data. The fact that the observational data are themselves somewhat uncertain points to a need for collaboration between the observational and modeling communities, including the remote sensing community, in reconciling models and data. The most problematic variable of those considered here is evapotranspiration. Despite its direct relevance to the surface moisture budget as well as to terrestrial ecosystems, very few observational data are available for assessing model simulations of E. The 21 st -century simulations summarized here show that the models do not agree even on the sign of the changes of E in the Arctic when a greenhouse scenario of forcing is prescribed. Improved model parameterizations of evapotranspiration will need to address factors such as the effects of vegetation change and simulation of transpiration rates using more realistic vegetation parameters, such as leaf area index instead of a single crop factor. Datasets for the validation and calibration of model-simulated evapotranspiration (which include better use of satellite data) most be considered one of most urgent needs for the development of scenarios of Arctic hydrology. REFERENCES Bogdanova, E. G., B. M. Ilyin and I. V. Dragomilova, 2002: Application of an improved bias correction model to precipitation measured at Russian North Pole drifting stations. J. Hydrometeor., 3, Colony, R., V. F. Radionov and F. J. Tanis, 1998: Measurements of precipitation and snow pack at Russian North Pole drifting stations. Polar Record, 34, Forland, E. J., and I. Hanssen-Bauer, 2000: Increased precipitation in the Norwegian Arctic: True or false? Climatic Change, 46, Goodison, B. E., P. Y. T. Louie and D. Yang, 1998: WMO Solid Precipitation Measurement Eds.), Intergovernmental Panel on Climate Change, Cambridge Univ. Press, 572 pp. Groisman, P. Y., and D. R. Easterling, 1994: Variability and trends of total precipitation and snowfall over the United States and Canada. Journal of Climate, 7, External Review January DO NOT CITE OR CIRCULATE

11 IPCC, 2001: Climate Change 2001: The Scientific Basis (J. T. Houghton, Y. Ding, D. J. Griggs, P. J. van der Linden, X. Dai, K. Maskell and C. A. Johnson, Eds.), Cambridge University Press, 881 pp. IPCC, 1996: Climate Change 1995: The Science of Climate Change (J. T. Houghton et al., Cohen, S. J., 1997: Mackenzie Basin Impact Study (MBIS): Final Report. Canadian Government Publishing, 372 pp., ISBN: Karl, T., 1998: Regional trends and variations of temperature and precipitation. The Regional Impacts of Climate Change: An Assessment of Vulnerability (R. T. Watson, M. C. Zinyowera, R. H. Moss and D. J. Dokken, Eds.), Intergovernmental Panel on Climate Change, Cambridge Univ. Press, Kattsov, V. M., and J. E. Walsh, 2000: Twentieth-century trends of Arctic precipitation from observational data and a climate model simulation. J. Climate, 13, Oki, T., K. Musiake, H. Matsuyami and K. Masuda, 1995: Global atmospheric water balance and runoff from large river basins. Hydrological Processes, 9, Paeth, H., A. Hense and R. Hagenbrock, 2002: Comments on Twentieth-century trends of Arctic precipitation from observational data and a climate model simulation. J. Climate, 15, Persson, P. Ola G., C. W. Fairall, E. L. Andreas, P. S. Guest and D. K. Perovich, 2002: Measurements near the atmospheric surface flux group tower at SHEBA: Near surface conditions and surface energy budget. J. Geophys. Res., 107(C10), doi: /2002JC Rouse, W. R., E. M. Blyth and 14 others, 2003: Energy and water cycles in a highlatitude north-flowing river system. Bull. Amer. Meteor. Soc., 84, Serreze, M. C., D. H. Bromwich, M. C. Clark, A. J. Etringer, T. Zhang and R. Lammers, 2003: The large-scale hydro-climatology of the terrestrial Arctic drainage system. J. Geophys. Res., 108, doi: /2001jd Yang, D., 1999: An improved precipitation climatology for the Arctic Ocean. Geophys. Res. Lett., 26, SEA ICE (J. Walsh, T. Jakobsson) 5.2.A Background Sea ice has long been regarded as a key potential indicator and agent of climate change. In recent years, sea ice has received extensive attention in the news media as well as in the scientific literature because of the apparent reduction of coverage and thickness in the Arctic. Since the potential impacts of these changes on climate, ecosystems and infrastructure are large, sea ice is a highly important variable in an assessment of Arctic change. Because of the routine availability of satellite passive microwave imagery, sea ice coverage has been well monitored since the 1970s. Figure shows the mean concentrations (percent areal coverage) for the calendar months of the climatological maximum (March) and minimum (September) sea ice coverage for the period , as derived from the passive microwave sensors onboard the SMMR and SSMI satellites. External Review January DO NOT CITE OR CIRCULATE

12 The accuracy if passive-microwave-derived ice concentrations vary from approximately 6% during winter to more than 10% during summer. The sea ice variable most Fig March and September mean ice concentration maps from SSMI, [From C. Parkinson, NASA Goddard Space Flight Center]. compatible with pre-satellite information based largely on ship reports is sea ice extent, which is defined as the area of ocean with ice concentration of at least 15%. Arctic sea ice extent, including all subpolar seas except the Baltic, ranges from about 7 million km 2 at its September minimum to about 15 million km 2 at its March maximum. The areal coverage of sea ice (excluding open water poleward of the ice edge) ranges from 5-6 million km 2 in late summer to about 14 million km 2 in the late winter (Parkinson et al., 1999). Interannual variability of the sea ice edge is typically 1-5 degrees of latitude for a particular geographical sector and calendar month. The departures from normal at a particular time vary regionally in magnitude and in sign. While ice extent and areal coverage have historically been used to monitor sea ice, ice thickness is an equally important consideration in the context of the sea ice mass budget. Unfortunately, ice thickness measurements are far less routine, consisting largely of upward-looking sonar measurements from occasional and irregular submarine cruises and, in recent years, from moored sonars on or near the continental shelves. In addition, direct measurements of fast ice thickness have been made for several decades in some coastal regions, and occasional direct measurements have also been made in the central Arctic at manned ice camps. The general pattern of ice thickness has been determined, but it is subject to variations and uncertainties that have not been well quantified. The thickness generally increases from the Siberian side of the Arctic to the Canadian Archipelago, largely in response to the mean pattern of ice drift and convergence (although air temperatures are also generally lower on the Canadian side of the Arctic Ocean). In areas of perennial ice, the seasonal cycle of melt and ablation generally has an amplitude of m. The passive microwave satellites provide information on ice type (e.g., first-year and. multiyear), although only for the portion of the year in which the ice surface is not subject to wetting. Several studies of pertaining to changes of particular types of ice are cited in the following section. The albedo of sea ice is of critical importance for the surface energy budget and for the ice-albedo feedback, which can accelerate ice variations on timescales ranging from the seasonal to the decade-to-century scale of interest in the greenhouse context. While the albedo of sea ice and snow-covered sea ice has been measured through the annual cycle locally, e.g., at ice stations such as SHEBA (Surface Heat Budget of the Arctic Ocean), the albedo of sea ice over scales of ( km) 2 is strongly dependent on the surface state (snow covered vs. bare ice, melt pond distribution, and especially the fraction of open water, i.e., leads and polynyas). A depiction of the climatological and several years of interannual variations of surface albedo in the central Arctic Ocean has been prepared by Robinson et al. (1992), but there have been no such compilations depicting decadal or longer-scale variations, or variations outside the Arctic Ocean, despite the potential value of such datasets for assessments of the ice-albedo-temperature feedback. External Review January DO NOT CITE OR CIRCULATE

13 5.2.B Recent and ongoing changes An apparent reduction of sea ice has occurred over the past several decades, although this reduction varies by region, by season, and by the measure of sea ice. Figure (from Cavalieri et al., 2003) shows the time series of Northern Hemisphere sea ice coverage, in terms of both the seasonal cycle and interannual variations (departures from climatological daily means) over the period , for which passive microwave imagery was available almost continuously. The Arctic ice extent decreased by x 10 6 km 2 /10 yr from 1972 through 2002, but by x 10 6 km 2 /10yr from 1979 through 2002, indicating an acceleration of 20% in the rate of decrease (Cavalieri et al., 2003). The trend in summer (September) is x 10 6 km 2 /10yr, whereas in winter (March) the trend is x 10 6 km 2 /10 yr for the 31-year period. For the more recent 24-year period from , the corresponding summer and winter trends are x 10 6 km 2 /10 yr and x 10 6 km 2 /10 yr, respectively (Cavalieri et al., 2003). These trends contrast with those of sea ice in the Southern Hemisphere, where the trend is either close to zero or slightly positive, depending on the period of analysis. Fig Daily Arctic sea ice extents (upper) and their anomalies (lower) from 1972 through A linear trend line is indicated on the daily extents and a 365-day running mean is included on the daily anomalies [from Cavalieri et al., 2003]. The recent decrease of ice extent over the past few decades is consistent with reports of indigenous peoples in various coastal communities of the Arctic. In particular, the themes of a shortened ice season and a deteriorating ice cover have emerged from studies that drew upon the experiences of residents of Sachs Harbour, Canada and Barrow, Alaska, as well as communities on St. Lawrence Island in the Bering Sea (Krupnik and Jolly, 2002). If the record is extended back several decades to the 1950s, the trends are comparable to those of the satellite period and are statistically significant. Vinnikov et al. (1999) compared the observed trends of the past several decades with estimates of natural variability based on a the GFDL climate model and showed that the decrease of Arctic ice extent is highly unlikely to have occurred by natural variability alone. However, this conclusion is based on the assumption that the natural variability of sea ice can be reliably inferred from a climate model. On longer timescales, the unavailability of sea ice data limits estimates of hemispheric-scale trends. On a regional basis, portions of the North Atlantic subarctic have sufficient data, based largely on historical ship reports and coastal observations, to permit trend assessments over periods exceeding 100 years. Perhaps the best known record is the Icelandic sea ice index, compiled by Thoroddsen (1917) and Koch (1945), with subsequent extensions (e.g., Ogilvie and Jonsson, 2001). The index combines information on the annual duration of sea ice along Iceland s coastline and the length of coastline affected by sea ice. Figure shows that several periods of severe ice years have occurred, especially during the late 1800s and early 1900s, followed by a long interval (from about 1920 to the early 1960s) in which sea ice was virtually absent from the Icelandic waters. However, an abrupt change to severe ice conditions in the late 1960s serves as a reminder that decadal variability is a characteristic of sea ice. In the External Review January DO NOT CITE OR CIRCULATE

14 years since the early 1970s, ice conditions in the vicinity of Iceland have been relatively mild. Fig Annual values of the Koch Index of sea ice along the coasts of Iceland. In an analysis that drew upon ship reports from the ocean waters east of Iceland, Vinje (2001) found that the extent of ice in the Nordic Seas during April has decreased by about 33% since the 1860s (Figure 5.2.4). However, this dataset and longer versions spanning the past several centuries indicate large variations of trends over multidecadal periods. Some multidecadal periods show trends comparable to those of the past several decades. Fig Historical record of April ice extent (2-year running means) in the Nordic Seas (NS) and in their eastern (E) and western (W) subregions. [From T. Vinje (2001), J. Climate, 14, p. 256]. The recent trend of decreasing sea ice has also been identified in the coverage of multiyear sea ice and in ice thicknesses in the central Arctic Ocean. Johannessen et al. s (1999) analysis of passive microwave-derived coverage of multiyear sea ice in the Arctic showed a 14% decrease of wintertime multiyear sea ice between 1978 and More recently, Comiso (2002) has extended to the trend analysis of end-of-summer minimum ice cover. Figure contrasts the ice concentrations at the time of ice minima during the first and second halves of the study period. The decrease is especially large offshore of northern coasts of Russia and Alaska. Comiso s computed rate of decrease of perennial ice, -9% per decade, is consistent with the findings of Johannessen et al. (1999) s trend of multiyear sea ice coverage in the Arctic, and is even slightly greater than the rate of decrease of total ice-covered area over in recent decades (Cavalieri et al., 2003). Fig Averages ice concentrations (color-coded) at time of summer ice minimum for two 11-year periods: (a) 1979 to 1989 and (b) 1990 to (c) is difference field, (b) minus (a), depicting loss of ice from to [From Comiso, 2002, Fig. 2]. A widely cited study of Rothrock et al. (1999), based on a comparison of upwardlooking sonar data from submarine cruises during and , found a decrease of about 40% (1.3 meters) in the sea ice draft (proportional to thickness) of the central Arctic Ocean from the earlier to the later period. Wadhams and Davis (2000) provided further submarine-derived evidence of the thinning of sea ice in the Arctic Ocean. While the findings concerning ice draft and multiyear sea ice coverage are compatible, it should be noted that the trends of ice draft have been evaluated from a relatively small fraction of the past 45 years; Anisimov et al. (2003) show that a one-year shift of Rothrock et al. s (1999) sample of years results in a much weaker trend of ice draft. There are also indications that at least some of the decrease of ice thickness is a consequence of variations of the wind-driven advection of sea ice and that increases of ice thickness in unsampled regions (e.g., offshore of the Canadian Archipelago) may partially offset the decreases in the central Arctic Ocean detected in the 1990s (Holloway and Sou, 2002). Specifically, the ice drafts in the Beaufort Sea (or western Arctic Ocean) appear to have decreased by about 1.5 m between the mid-1980 s and early 1990 s External Review January DO NOT CITE OR CIRCULATE

15 (Tucker et al., 2001). Proshutinsky and Johnson (1997) show that the pattern of Arctic ice drift has historically varied between two regimes, characterized by relatively strong and weak Beaufort anticyclones. The association between the Arctic (or North Atlantic) Oscillation and Arctic sea ice is being used increasingly to explain variations in Arctic sea ice over the past several decades (e.g., Parkinson, 2000; Kwok, 2000; Rigor et al., 2002). The wind forcing associated with this atmospheric mode has been shown to be related to sea ice export from the Arctic Ocean through Fram Strait to the North Atlantic Ocean (Kwok and Rothrock, 1999) and to ice conditions along the northwestern coastline of the Canadian Archipelago (Agnew et al., 2003). However, studies of longer time periods suggest that such associations with Fram Strait ice export may not be temporally robust because of relatively subtle shifts in the centers of action of the North Atlantic Oscillation (NAO) (Hilmer and Jung, 2000). More recent work by Cavalieri (2002) reveals a consistent relationship over decadal time scales between Fram Strait sea ice export and the phase of atmospheric sea level pressure wave 1 at high latitudes. The phase of this wave appears to be a more sensitive indicator of Barents Sea low pressure systems that drive sea ice through Fram Strait than is the NAO index. In general, the role of ice motion in diagnoses of historical changes and projections of future changes is largely unexplored. 5.2.C Projected changes The changes of sea ice projected for the 21 st century by the five global climate models used by ACIA are summarized in this section. In the case of the CGC model, an ensemble of three different 21 st -century simulations was available. These models all predict decreases of sea ice during the 21 st century, although the time series of simulated sea ice contain sufficient variability that increases can be found over occasional intervals of 1-10 years, especially when coverage in specific sectors of the Arctic is examined. Quantitative comparisons of the projected changes of sea ice are hampered by two factors. First, the sea ice variables archived by the various modeling centers vary from model to model, ranging from presence of ice (binary 1/0) to concentration, thickness, and grid cell mass. Since all these variables permit evaluations of ice extent (defined as the area poleward of the ice edge), we have used ice extent in the comparisons among the various models. Second, the sea ice simulated by these models for the present ( control ) climate is generally not in agreement with observed coverage (e.g., Figure 5.2.1), especially when coverage in specific regions is considered. These biases in the control climate will confound interpretations of the model-derived coverage for a future time (e.g., the ACIA time-slices centered on 2020, 2050, 2080), since changes from a biased initial state are unlikely to result in a projected state that is free of biases. In an attempt to optimize the informational content of the projections of sea ice, we have made a very crude adjustment the future (projected) sea ice states of each model by adding to each projection the control-climate bias of sea ice for the particular model, calendar month, longitude. This type of adjustment is an ad hoc procedure for which the need will be eliminated as coupled atmosphere-ocean-ice model simulations become more realistic. In the following synthesis of projections, we include examples of both the raw (unadjusted) projections and the adjusted projections. External Review January DO NOT CITE OR CIRCULATE

16 Figure shows the projected st -century 21 time series of total Northern Hemisphere ice extent for March and September from the five models. The upper panels show the raw (unadjusted) time series, while the lower panels show the adjusted time series. (While the trends and variations are the same in both panels for a particular model, the starting points in 2000 are generally not, owing to the biases in the control climates). Much of the across-model differences among the models unadjusted time series is due to the differences in the initial state (ice extent of the present climate). For example, the unadjusted March ice extents range from approximately 9 to 16 million km 2 for , while the corresponding observational value, averaged over the entire month of March for the past decade, is about 14.5 million km 2. The models raw values show an even greater range in September, varying from about 2 to 11 million km 2, compared to the observational value of approximately 8 million km 2. The CSM model is consistently the model with the largest ice extent, while the CGC model is consistently the model with the least ice. The CGC model s raw values indicate an ice-free Arctic during September by the mid-21 st century, although it should be noted that this model had less than half of the observed ice September coverage at the start of the 21 st century. (There is very little difference among the three ensemble members from the CGC model, indicating that the initial conditions make far less difference than the choice of the model). None of the other models project ice-free summers in the Arctic by 2100, although the ice extent in the HAD and ECH models decreases to about one-third of initial (year-2000) and observed September values by Fig Time series of 21 st -century total N. Hemisphere ice extent projected by five global climate models for March (left panels) and September (right panels). Upper panels show raw model output; lower panels are adjusted for biases in simulated present-day sea ice. During March, the projected decreases of ice extent by 2100 vary among the models from about 2 to 4 million km 2. Unlike September, none of the models are close to icefree in March, although the coverage in the CGC model is only about 10 million km 2, which is about two-thirds of the present-day March extent. A large portion of the differences among the March extents in 2100 is attributable to the differences in the initial (2000) states of the models. Table summarizes the 21 st -century changes of annual mean ice extent projected by the models. The changes in the annual mean ice extent are expressed in km 2 and as percentages of the models original ice extent in Table Changes of annual mean Northern Hemisphere ice extent from 2000 to 2100 Model Change, 10 6 km2 % change Change, 10 6 km2 % change (unadjusted) (unadjusted) (adjusted) (adjusted) CGC from 9.7 to % from 12.3 to % CSM from 16.5 to % from 12.3 to % ECH from 11.9 to % from 12.3 to % GFD from 11.9 to % from 12.3 to % HAD from 12.8 to % from 12.3 to % External Review January DO NOT CITE OR CIRCULATE

17 As Table shows, the greatest reductions of ice cover, both as actual areas and as percentages, occur in the model with the least ice initially (year 2000), while the smallest losses occur in the model with the most ice initially (year 2000). Insofar as ice extent and mean ice thickness are positively correlated, this relationship is not surprising, i.e., the models with the greatest ice extent have the most thick ice, which is more difficult to lose in a climate warming scenario. However, the association found here between the amount of initial ice and the rate of ice retreat does not seem to be present in the CMIP (Coupled Model Intercomparison Project) suite of coupled global models (C. Bitz, pers. Comm.). This lack of association is illustrated by Flato (2003). When regions are defined on the basis of the four 90-degree Arctic sectors adopted for the ACIA framework (Chapter 1), some geographical variations in the modelprojected ice retreat are apparent. However, it should be noted that the regional differences are generally small, and are considerably less than the differences among the models. Wintertime ice retreat, as measured by the changes in March ice extent, is largest in the Bering sector, extending from the Sea of Okhotsk (150ºE) eastward to 120ºW, in three of the models (GFD, HAD, CSM). The March retreat is greatest in the Atlantic sector (30ºW-60ºE) in the CGC model. In the summer, the models show more regional variation in their indication of the greatest retreat. The GFDL model loses the most summer ice in the Bering sector, which becomes ice-free in September by the end of the 21 st century. The Atlantic sector shows the most rapid retreat in the HAD and CGC models, both of which become ice-free by 2100 in the unadjusted results. The CSM model loses little ice in any sector during the summer. Best estimates of the sea ice distributions in the ACIA time slices (2020, 2050, 2080) can be obtained by compositing the adjusted fields of sea ice from the five models. Figure and show these fields for September and March, respectively, expressed in terms of the number of models (out of 5) in which ice is present at least 50% of the time. Comparisons with Figure provide measures of the changes from the present. The distributions in Figures and illustrate the tendency for the reduction of sea ice to be greater, especially as a percentage of the present extent, in September than in March. The values are less than the maximum of 5 over much of the Arctic Ocean, which is now largely ice-covered. The results imply a substantial enhancement of the summertime navigability of the Arctic Ocean, as discussed in Chapter 15. Fig Five-model composite maps of sea ice coverage for September of (a) 2020, (b) 2050, (c) Fig Same as Fig , but for March. The reduction of sea ice in wintertime (March) is somewhat less than in summer, especially when expressed as a percentage of the present coverage, as shown in Figure Most of the Arctic Ocean remains ice-covered in March, although the March ice edge retreats substantially in the subpolar seas. It should be noted, however, that the ice becomes thinner in the central Arctic through the 21 st century, at least in those models for which the ice thickness or mass per grid cell was available. External Review January DO NOT CITE OR CIRCULATE