Arctic change

Data

First, there is a consilience of indicators of climate change in the Arctic, including increased temperatures, diminished sea ice, degraded permafrost, enlarged melt area on Greenland, increased water vapor, decreased snow extent, increasing number of forest fires, increased river discharge, and resulting ecosystem impacts. Certainly one of the most dramatic indicators is the reduction of summer sea ice extent, 40% below the average extent of the 1980s and over 20% below the previous record minimum of 2005. This 1-year drop in sea ice extent is three standard deviations beyond that of the year-to-year variability in the historical sea ice record beginning in 1979.

Global climate models generally project that the temperature increase in the Arctic will be larger than at more southerly latitudes and that the increase due to anthropogenic forcing will have an Arctic-wide character (Chapman and Walsh 2007). Figure 3A shows the annual temperature anomalies in 2001-2007 for the globe and Fig. 3B shows the climate projections for 2020-2029; both signal a rather uniform polar amplification. If one looks more regionally, temperature anomalies for spring 2000-2005 had a minimum increase of +1°C throughout the Arctic, with a hot spot near eastern Siberia of over +3°C, relative to the 19581998 mean (Fig. 4A). Since that time the background temperatures have remained positive but the location of the hot spot has shifted to the Atlantic side of the Arctic as the regional wind pattern has changed. Autumn temperature anomalies for 2005-2007 over much of the central Arctic basin were greater than +6°C in response to thin or no sea ice at the end of summer (Fig. 4B Left); this is also a prediction of climate models. For comparison we show the winter surface temperature anomalies for the AO positive years of 1989-1995 (Fig. 4B Right). Here the increased temperatures are not Arctic-wide and can be explained by the natural variability of the wind pattern associated with the AO with the warm anomalies over Eurasia.

Fig. 3A. Global-annual temperature anomalies for 2001-2007 relative to a 1951-1980 baseline. (From the GISS web site.)

Fig. 3B. IPCC model forecast temperatures for 2020-2029 using a midrange emission scenario (A1B). (From the IPCC Report.)

Fig. 4A. 2000-2005 Spring (Mar-May) near surface air temperature anomalies.

Fig. 4B. Left - 2005-2007 Fall (Oct-Nov) air temperature anomalies. Right - 1989-1995 winter (Dec-Mar) air temperature anomalies when the AO was strongly positive. (From the CDC web site.)
Fig. 5. Spitzbergen and Barrow, AK winter (Nov-Mar) temperature anomalies based on a 19121926 base period.

Likewise, if we look at the temperatures of the 1930s, the Spitzbergen station had temperatures greater than 5°C above its mean before 1930 (Fig. 5). With reference to the 20th century Arctic temperature record (Fig. 2), most of the "Arctic-wide average" 1°C anomaly in the 1930s was contributed by this one station. In fact, Barrow, Alaska showed no positive temperature anomaly throughout the 1930s. If we look year by year in the 20th century and back to the first IPY in 1883 (Wood and Overland 2006), we see only regional positive temperature anomalies, rather than background Arctic-wide anomalies as in the last decade. We take the recent Arctic-wide temperature increases as evidence of a global warming signal in the Arctic.

Models

Reproducing decadal and longer variability in coupled General Circulation Models (GCMs) is a critical test for understanding processes in the Arctic climate system and increasing the confidence in the Intergovernmental Panel on Climate Change (IPCC) model projections. Twentieth century simulations, control runs without external forcing, and 21st century projections are available for 20 coupled GCMs for the IPCC 4th Assessment (Wang et al. 2007). Warm anomalies in the Arctic during the last decade are reproduced by all models. In contrast, only eight models have variance comparable with the variance in the 20th century observations. Since we are interested in when an anthropogenic signal may exceed natural variability, we concentrate on those models which pass an observational selection constraint based on having sufficient variance.

1880 1900 1920 1940 1960 19SO 2000

CCSM3 CSIRO

ECHAM5 IMM

GFDL-CM20

GFDL-CM21

ECHO-G

20C3M Simulation With 5-yr running mean

20C3M Simulation With 5-yr running mean

CCSM3 CSIRO

ECHAM5 IMM

GFDL-CM20

GFDL-CM21

ECHO-G

1880 1900 1920 1940 1960 19SO 2000

Fig. 6. Observed arctic temperatures (heavy black line-similar to Fig. 2) and IPCC model simulations of arctic temperatures for the 20th century. (From Wang et al. 2007.)

Fig. 6. Observed arctic temperatures (heavy black line-similar to Fig. 2) and IPCC model simulations of arctic temperatures for the 20th century. (From Wang et al. 2007.)

Figure 6 shows the observed Arctic temperatures for the 20th century (heavy line) and simulations for different ensemble members from separate models. Ensembles for a given model are different simulations which start with slightly different initial conditions. The range of simulations, which show different timing of maximum and minimum temperature events, shows the influence of the chaotic nature of the climate system. The random timing of mid-century warm anomalies over the different ensemble members suggests that the observed mid-century warm periods are consistent with intrinsic climate variability. The positive departure of all ensemble members at the end of the century supports the conclusion of an external anthropogenic forcing, and thus a global warming signal for the Arctic.

2007 sea ice loss - the fast track of Arctic change

The loss of 40% of sea ice extent on the Pacific side of the Arctic in 2007 relative to climatology resulted from an unusually persistent high surface pressure/ southerly wind pattern from June through August that transported heat and altered cloud distributions. The winds also advected sea ice across the central Arctic toward the Atlantic sector (Gascard et al. 2008). A similar pressure pattern also occurred in 1987 and 1977 with no remarkable effect on sea ice extent. The 2007 event, however, followed the steady preconditioning of the ice pack by two decades of thinning and area reduction (Rigor and Wallace 2004; Nghiem et al. 2007; Overland et al. 2008).

Although it is difficult to attribute a single event to anthropogenic climate change, there are several lines of evidence that support this conclusion. The 2007 ice loss greatly exceeded that in any other year in the observational record. Control runs of global climate models (with no anthropogenic forcing) do not exhibit similar sea ice loss, but large year-on-year decreases are simulated in some model ensemble members with anthropogenic forcing (Holland et al. 2006). While we would not claim that the chain of events in the National Center for Atmospheric Research (NCAR) model in Fig. 7 is identical to those leading up to the 2007 sea ice minimum, several features are similar. The large drop in the model projection of sea ice extent near 2013 in one ensemble member (black line), along with the range of ensemble members (other colors), implicate long-term anthropogenic forcing combined with large intrinsic atmospheric variability and sea-ice related feedbacks. The modeled minimum sea ice cover rebounds in subsequent years from its low value of 4 million square kilometers - comparable to the observed 2007 minimum extent - but it never recovers to 1980-1990 values.

Based on observations and similarity to model projections, the dramatic Arctic sea ice reduction in 2007 was likely caused by a combination of increased temperatures in response to greenhouse gas increases, fortuitous timing in the natural variability of the atmospheric general circulation, and positive feedbacks associated with a reduction in sea ice. Sea ice models without anthropogenic

1980 2000 2020 2040 2060 2080 2100

Fig. 7. Different Arctic Sea ice area projections from the NCAR CCSM3 model. (Modified from Holland et al. 2006.)

1980 2000 2020 2040 2060 2080 2100

Year

Fig. 7. Different Arctic Sea ice area projections from the NCAR CCSM3 model. (Modified from Holland et al. 2006.)

forcing (control runs) do not generate such sea ice minimums, nor does simple extrapolation of temperatures from greenhouse gas warming initiate the observed open water areas. Thus it is not a question of whether global warming or natural variability is the primary cause of the 2007 sea ice event; both were required.

IPCC projections of sea ice loss from the better subset of models show sea ice minimums beyond 2050 (Stroeve et al. 2007; Overland et al. 2008). But these are projections which average over multiple ensembles to give an expected value. It appears that the real world is on a faster trajectory of sea ice loss than the expected value projected by IPCC models. C. Bitz (2008, personal communication) notes that this fast behavior, while it exists in models, occurs in less than 5% of the simulated future years in the NCAR model. Thus, it is important to understand that while summary IPCC projections were based on averages of many model runs, reality is but a single realization. This does not mean that all the IPCC model forecasts are wrong; it just means that there is a difference in the average of the projections over all ensembles and our single realization. Thus the sea ice event of 2007 points to the simultaneous impact of natural variability and global warming in the Arctic, to give us a much earlier timing for sea ice loss than anticipated. The new fast track is consistent with an ice-free summer Arctic before 2030, as suggested by Stroeve et al. (2008).

0 0

Post a comment