University of Nebraska - Lincoln DigitalCommons@University of Nebraska - Lincoln

Papers in Natural Resources Natural Resources, School of

2-2011

Evaluating observed and projected future changes for the Arctic using the Köppen-Trewartha

Song Feng University of Nebraska - Lincoln, [email protected]

Chang-Hoi Ho Seoul National University, Seoul, Korea

Qi Hu University of Nebraska, Lincoln, [email protected]

Robert Oglesby University of Nebraska - Lincoln, [email protected]

Su-Jong Jeong Seoul National University, Seoul, Korea

See next page for additional authors

Follow this and additional works at: https://digitalcommons.unl.edu/natrespapers

Part of the Natural Resources and Conservation Commons

Feng, Song; Ho, Chang-Hoi; Hu, Qi; Oglesby, Robert; Jeong, Su-Jong; and Kim, Baek-Min, "Evaluating observed and projected future climate changes for the Arctic using the Köppen-Trewartha climate classification" (2011). Papers in Natural Resources. 291. https://digitalcommons.unl.edu/natrespapers/291

This Article is brought to you for free and open access by the Natural Resources, School of at DigitalCommons@University of Nebraska - Lincoln. It has been accepted for inclusion in Papers in Natural Resources by an authorized administrator of DigitalCommons@University of Nebraska - Lincoln. Authors Song Feng, Chang-Hoi Ho, Qi Hu, Robert Oglesby, Su-Jong Jeong, and Baek-Min Kim

This article is available at DigitalCommons@University of Nebraska - Lincoln: https://digitalcommons.unl.edu/ natrespapers/291 Published in Climate Dynamics (2011); doi: 10.1007/s00382-011-1020-6 Copyright © 2011 Springer-Verlag. Used by permission.

Submitted October 25, 2010; accepted January 31, 2011; published online February 11, 2011.

Evaluating observed and projected future climate changes for the Arctic using the Köppen-Trewartha climate classification

Song Feng,1 Chang-Hoi Ho,2 Qi Hu,1, 3 Robert J. Oglesby,1, 3 Su-Jong Jeong,2 and Baek-Min Kim 4

1. School of Natural Resources, University of Nebraska-Lincoln, 702 Hardin Hall, Lincoln, NE 68583-0987, USA 2. School of Earth and Environmental Sciences, Seoul National University, Seoul, Korea 3. Department of Earth and Atmospheric Sciences, University of Nebraska-Lincoln, Lincoln, NE, USA 4. Korean Polar Research Institute, Incheon, Korea Corresponding author — Song Feng, email: [email protected]

Abstract eal coverage under the B1, A1b and A2 scenarios, respec- The ecosystems in the Arctic region are known to be very tively. The types (Dc and Do) advance sensitive to climate changes. The accelerated warming for and take over the area previously covered by Ec. The area the past several decades has profoundly influenced the covered by Dc climate expands by 4.61 × 106 km2 (84.6%) lives of the native populations and ecosystems in the Arc- in B1, 6.88 × 106 km2 (126.4%) in A1b, and 8.16 × 106 km2 tic. Given that the Köppen-Trewartha (K-T) climate classifi- (149.6%) in A2 scenarios. The projected redistributions of cation is based on reliable variations of land-surface types K-T climate types also differ regionally. In northern (especially ), this study used the K-T scheme to and , the warming may cause more rapid expansion evaluate climate changes and their impact on vegetation of temperate climate types. Overall, the climate types in 25, for the Arctic (north of 50°N) by analyzing observations 39.1, and 45% of the entire Arctic region are projected to as well as model simulations for the period 1900–2099. change by the end of this century under the B1, A1b, and The models include 16 fully coupled global climate mod- A2 scenarios, respectively. Because the K-T climate classi- els from the Intergovernmental Panel on Climate Change fication was constructed on the basis of vegetation types, Fourth Assessment. By the end of this century, the annual- and each K-T climate type is closely associated with certain mean surface temperature averaged over Arctic land re- prevalent vegetation species, the projected large shift in cli- gions is projected to increase by 3.1, 4.6 and 5.3°C under mate types suggests extensive broad-scale redistribution of the Special Report on Emissions Scenario (SRES) B1, A1b, prevalent ecoregions in the Arctic. and A2 emission scenarios, respectively. Increasing tem- Keywords: Arctic, Köppen-Trewartha climate classifica- perature favors a northward expansion of temperate cli- tion, Fully coupled global climate models, Climate projec- mate (i.e., Dc and Do in the K-T classification) and boreal tion, Vegetation oceanic climate (i.e., Eo) types into areas previously cov- ered by boreal (i.e., Ec) and tundra; and tundra into areas occupied by permanent ice. The tundra region is projected to shrink by −1.86 × 106 km2 (−33.0%) 1. Introduction in B1, −2.4 × 106 km2 (−42.6%) in A1b, and −2.5 × 106 km2 (−44.2%) in A2 scenarios by the end of this century. The Ec The Arctic region is extremely vulnerable to climate climate type retreats at least 5° poleward of its present lo- change and its impacts. Observations show that surface cation, resulting in −18.9, −30.2, and −37.1% declines in ar- air temperatures in the Arctic have warmed at about

1 2 S. Feng et al. in Climate Dynamics (2011) twice the global rate over the past few decades (ACIA One simple, but frequently used method to assess the 2004). This Arctic warming also is expressed through impact of climate change on ecosystems is the Köppen widespread melting of glaciers and sea ice and rising climate classification (Köppen 1936), and its subsequent permafrost temperatures (e.g., Serreze et al. 2007; Zhang modification to the Köppen-Trewartha (K-T) classifica- et al. 2005; Hinzman et al. 2005). Consistent with Arc- tion (Trewartha and Horn 1980). Though environmental tic warming, the amount of rainfall in high latitudes has and historical factors can exert important influences on increased considerably over the past 50 years (Min et natural vegetation at local scales, climate is nonetheless al. 2008), supporting earlier reported increases in Arc- the fundamental factor regulating the broad-scale distri- tic river discharge (ACIA 2005; Peterson et al. 2002; bution of natural vegetation physiognomy and species Hinzman et al. 2005). Additionally, the Arctic warm- composition. This is the main reason that Köppen used ing leads to decreasing sea ice and snow cover as well the natural vegetation of a region as an expression of its as longer snow-free (Chapin et al. 2005; Stone et climate (Köppen 1936). The Köppen and K-T classifica- al. 2002). The shrinking sea ice and snow cover and the tions combine temperature and regimes lengthening of the snow-free in turn reduce sur- and their seasonality into a single metric and thereby face albedo and contribute substantially via a positive classify global climate into several major types. Based feedback to high-latitude warming trends (Chapin et al. on seasonal variations of temperature and precipitation, 2005; McGuire et al. 2006; Jeong et al. 2010a). several sub-climate types in each major climate type A mounting body of evidence indicates that this re- can further be classified. The Köppen classification sys- cent, amplified warming in the Arctic is fueled by hu- tem has been widely used to describe the potential dis- man-induced ‘global warming’ (e.g., Intergovernmen- tribution of natural vegetation based on climatic thresh- tal Panel on Climate Change (IPCC) 2007; Gillett et al. olds thought to drive critical physiological processes 2008). Gillett et al. (2008) examined the mechanisms (Bailey 2009). Indeed, each climate type (major or sub- underlying the observed changes us- climate) is associated with a certain vegetation assem- ing simulations made by multiple climate models in- blage, or ecoregion, under present climate conditions cluded in the IPCC Fourth Assessment (AR4). Their (Bailey 2009; Baker et al. 2010; see also Table 1). There- work demonstrated convincingly that humans have in- fore, by definition the climate types are closely linked deed contributed to recent warming in the Arctic re- to the qualitative features of regional vegetation. Also gion. Additionally, increasing atmospheric concen- of importance, a key advantage of this type of classifica- trations of greenhouse gases are projected to further tion scheme is that it is easy to use with a variety of data contribute to Arctic warming of about 4–7°C over next sets and model outputs. 100 years (ACIA 2004). These results emphasize the ur- A number of previous studies used the Köppen and gent need to understand observed and projected fu- related climate classifications to investigate the poten- ture climate changes and their impact on ecosystems tial impact of past, present, and projected future cli- in the Arctic. mate changes (Fraedrich et al. 2001; Wang and Overland Warming in the Arctic can, and apparently has al- 2004; de Castro et al. 2007; Baker et al. 2010; Gersten- ready caused large shifts in vegetation. Plants in garbe and Werner 2009). Fraedrich et al. (2001) ana- Greenland are flowering at an earlier date; indeed the lyzed changes in climate types over global land regions onset of the growing season occurs earlier (Post et during 1901–1995. They reported that the area covered al. 2009; Matthes et al. 2009). The areal extent of tall by tundra (which primarily appears in the Arctic) sig- shrubs in Alaska’s North Slope tundra region has in- nificantly declined during the 20th century. Wang and creased 1.2% per decade since 1950 (Sturm et al. 2001), Overland (2004) used an updated dataset and reported also supported by indigenous observations in the a rapid decrease in circum-Arctic tundra coverage since same region (Thorpe et al. 2002). Throughout Alaska, 1990. De Castro et al. (2007) analyzed simulations made a majority of the studied sites show a treeline advance by multiple regional climate models and reported that (Lloyd 2005). White (Picea glauca) has expanded the tundra in Northern Fennoscandia may shift to tem- into what was tundra and increased in density in west- perate climate types by 2071–2100 under the Special Re- ern Alaska (Lloyd et al. 2003). During the past 50 years, port on Emissions Scenario (SRES) A2 scenario. Baker 2.3% of the treeless area has been converted from tun- et al. (2010) used a multivariate spatial–temporal clus- dra to forest in Alaska (Chapin et al. 2005). This wide- tering algorithm in conjunction with the K-T classifica- spread expansion of shrubs, and advancing treeline in tion scheme to quantify the impact of temperature and Alaska and other Arctic regions is also supported by precipitation on ecoregions in China. They reported that rapid greening and earlier start of the growing sea- the climate changes projected by the HadCM3 model son, as revealed by satellite-retrieved vegetation indi- under the SRES A1F1 scenario were sufficient to cause ces (Zhou et al. 2001; Tucker et al. 2001; Hinzman et al. shifts in spatial distributions of the majority of ecore- 2005; Jia et al. 2009; Bhatt et al. 2010). gions in China for 2041–2070. However, these previous Evaluating future climate changes for the Arctic using the K-T classification 3

Table 1. The classification criteria, description and the corresponding prevalent vegetation of the Köppen-Trewartha climate classification Climate type Description Prevalent vegetation (Bailey 2009) Classification criteria Do Temperate Oceanic Dense coniferous forests with large trees 4–7 months above 10°C and the coldest month above 0°C Dc Temperate continental Needle leaf and deciduous tall broadleaf forest 4–7 months above 10°C and the coldest month below 0°C Eo Boreal oceanic Needle leaf forest Up to 3 months above 10°C and the temperature of the coldest month above −10°C Ec Boreal continental Tayga (shrub) Up to 3 months above 10°C and the temperature of the coldest month below or equal to −10°C Ft Tundra Tundra The warmest month below 10C but above 0°C Fi Ice cap Permanent ice cover All months <0°C

studies using Köppen and related classification have fo- get region is projected as type C by the end of the 21st cused on climate change during the instrumental pe- century, this climate type is neglected because it only riod, or future changes in specific regions (e.g., China appears in less than 0.1% of the total Arctic area.) Ad- and Europe), and/or for a single SRES scenario. No pre- ditionally, because of the meager precipitation in the vious studies have applied the Köppen and related clas- Arctic, the climate classifications there are merely based sification schemes to a comprehensive examination of on monthly mean air temperature (Table 1). This is at- climate changes throughout the Arctic using both obser- tributed to the fact that the growth of vegetation in cold vations and a suite of projected future climate changes. regions such as the Arctic is mainly regulated by tem- The present study addresses this deficiency and applies perature (Tucker et al. 2001; Zhou et al. 2001). the K-T classification to investigate climate changes in The criteria used to classify the climate types, and the Arctic region, which is defined as north of 50°N. We the prevalent vegetation associated with each type, employ newly updated observations and newly devel- are listed in Table 1. Note that each sub-climate type is oped, statistically-downscaled high-resolution climate dominated by distinct vegetation zones (Köppen 1936; changes projected by 16 fully coupled climate models Trewartha and Horn 1980; Bailey 2009). For example, Ft for various SRES scenarios. is dominated by treeless tundra, while Do is dominated Details of the observed and modeled climate data and by dense coniferous forests with large trees. The shift of the methods used to analyze those data are described in climate types in a specific region due to climate changes Sect. 2. The ability of the climate models to reproduce indicates that the dominant vegetation type in that re- present observed climate types in the Arctic, as well as gion is replaced by other vegetation. projected future changes in climate types, are presented To evaluate climate changes and their impact on in Sect. 3, followed by discussion in Sect. 4 and conclu- vegetation in the Arctic, both observed and projected sion in Sect. 5. future surface air temperatures are examined. Global climate model output, from the World Climate Re- search Programme’s (WCRP’s) Coupled Model Inter- 2. Data and methods comparison Project phase 3 (CMIP3) multi-model da- taset (Meehl et al. 2007), was obtained from http:// The modified K-T climate classification (Trewartha www.engr.scu.edu/~emaurer/global_data/. These and Horn 1980) is used to examine changes in climate data were downscaled as described by Maurer et al. types for the Arctic (north of 50°N). This modified clas- (2009) using the bias-correction/spatial downscaling sification scheme identifies six major climate types using method (Wood et al. 2004) to a 0.5° grid, based on the letters A to F: A-tropical, B-dry climate, C-subtropical, 1950–1999 gridded observations of Adam and Letten- D-temperate, E-boreal, and F-polar. Based on seasonal maier (2003). The statistically-downscaled present-day variations of temperature and precipitation, several sub- control simulations and future climate change projec- climate types are also classified for each major climate tions from 16 fully coupled atmosphere–ocean mod- type. For example, the climate type B is based on mois- els cover the global land surface for the period from ture availability and the other climate types are based 1950 to 2099. The future climate change projections on large-scale thermal zones. Due to the relatively cold include low, median and high greenhouse gas emis- temperatures, only climate types D, E, and F are identi- sion rates, termed SRES B1, A1b, and A2, respectively fied in the Arctic region. (While a small part of the tar- (Nakićenović and Swart 2000). 4 S. Feng et al. in Climate Dynamics (2011)

In addition to the modeled temperature data, two ob- available climate models is often the best determinant served temperature datasets were analyzed. The first is for simulating the mean global climate (e.g., Gleckler et the half-degree resolution temperature dataset devel- al. 2008; Reichler and Kim 2008). This ensemble strategy oped by Adam and Lettenmaier (2003), henceforth AL. can also be valid for regional climate change detection This temperature dataset covers the period 1950–1999 (Pierce et al. 2009). Therefore, the ensemble means of the and was also used to calibrate the modeled tempera- 16 models for each SRES scenario are analyzed in this tures (Maurer et al. 2009). The second is the Terrestrial study. To help evaluate the uncertainties of the model Air Temperature: 1900–2008 Gridded Monthly Time Se- projections, the standard deviations of the model projec- ries (version 2.01) developed by the Center for Climate tions for each SRES scenario are also computed. Addi- Research at University of Delaware, henceforth UD. tionally, to evaluate the temporal variations of the cli- This dataset, obtained from http://www.climate.geog. mate types, a 15-year equal weight smoothing is applied udel.edu/~climate/, merges several updated gauge ob- to the observed and projected temperature data to re- served temperatures that are interpolated to grid points move year to year fluctuations. Fraedrich et al. (2001) covering the global land surface at a 0.5° × 0.5° hori- suggested that a 15-year smoothing is the optimal av- zontal resolution. The spatial interpolations were based eraging interval (window) for the Köppen and related on Willmott et al.’s (1985) spherical implementation of classifications. Shepard’s spatial-interpolation algorithms, which also incorporated digital elevation model-assisted and cli- matological-aided interpolation methods. Compared to 3. Results other existing observation-based land surface air tem- perature data, this dataset can reasonably capture the 3.1. Observed and projected temperature change observed climatology, and departures from the mean state (anomaly fields) both regionally and globally. To understand temporal variations in the Arctic re- Because the two observed temperature datasets used gion as the whole, the areal-weighted average temper- a slightly different number of observations as well as ature anomalies were calculated over the entire domain. different spatial interpolation methods, the long-term Figure 1 shows this domain-averaged temperature for mean of the UD temperature is slightly cooler than the 1900–2099. The observed temperatures show strong in- corresponding AL temperature in most of the Arctic re- terannual variations, which are superimposed on longer gion during the overlapped period, 1950–1999 (figure time-scale multidecadal changes. The temperature in- not shown). The K-T climate classification using the UD creases from 1900 to the middle 1940s, slowly decreases temperatures therefore also shows slightly more area until the middle 1960s, followed by steady increase and covered by colder climate types (e.g., polar climate) amplified warming since the late 1970s. The ensemble of compared to AL temperatures (Table 2). To reduce the modeled temperature and the uncertainties of the model difference between the two observed datasets and for simulations are also depicted in the figure. The large better comparison with the projected climate changes, standard deviations of the 16 model simulations sug- the UD temperature is adjusted so that it has the same gest that some models may do a poor job in simulating monthly climatological mean as the AL temperature. the observed temperature. When the simulations of the Specifically, the monthly anomalies of the UD- temper 16 models are averaged, however, the biases in individ- atures during 1900–2008 were first calculated based on ual models are smoothed out, yielding variations con- the 1950–1999 climatological mean of the UD tempera- sistent with the observations, e.g., cooling trend before ture, then the 1950–1999 monthly climatological mean the 1960s and steady warming trend since the late 1970s. of the AL temperature is added to those monthly anom- These results support the previous finding that multiple alies. The adjusted UD temperatures generate nearly model ensembles usually do a better job in simulating identical climate classifications as the AL temperatures, observed climate changes (Gleckler et al. 2008; Reichler and the IPCC AR4 models, during the overlapped pe- and Kim 2008; Pierce et al. 2009). riod (1950–1999, see Table 2). The agreement between The recent warming is projected to continue under all the observed and modeled data warranting further ex- three SRES scenarios (Figure 1). By the end of this cen- amination the long-term climate change from 1900 to tury, the temperature averaged over the entire the end of the 21st century. Arctic land region is projected to increase by 4.2°C in The K-T climate classification is applied to both ob- B1, 6.1°C in A1b, and 7.1°C in A2 scenarios. The warm- served and projected future temperature changes in the ing in is much weaker than in winter. The sum- Arctic region. Because different models contain differ- mer temperature averaged over the entire Arctic is pro- ent atmospheric and oceanic adjustment processes, the jected to increase by 2.2°C in B1, 3.3°C in A1b, and 3.9°C projected temperature changes by the models differ in A2 scenarios. When all seasons are averaged, the an- somewhat. It has been suggested that the simple aver- nual temperature is projected to increase by 3.1°C in B1, age (or ensemble) of the model outputs made by all the 4.6°C in A1b, and 5.3°C in A2 scenarios. Evaluating future climate changes for the Arctic using the K-T classification 5

0.06 5.3%) 9.4%) 10.0%) 6.2%) 4.9%) 3.8%) 0.08 0.08 0.14 0.15 0.10 ± ± ± ± ± ± ± ± ± − ± ± ±

16.5 25.6 31.9 16.4 15.1 11.4 0.25 0.23 0.17 0.39 0.49 0.25 − − − − − − Fi 1.82 1.53 1.53 1.53 ± 0.002 − ( − ( − ( − ( − ( − (

11.7%) 12.8%) 11.7%) 12.4%) 12.6%) 12.8%) 0.66 0.72 0.66 0.70 0.71 0.72 ± ± ± ± ± ± ± ± ± ± ± ± 32.2 29.6 26.4 42.6 44.2 33.0 1.82 1.67 1.49 2.40 2.50 1.86 − − − − − − Ft 6.00 5.66 5.66 5.64 ± 0.01 − ( − ( − ( − ( − ( − (

4.8%) 5.4%) 9.8%) 11.0%) 6.2%) 6.8%) 1.09 0.84 0.94 1.73 1.95 1.21 ± ± ± ± ± ± ± ± ± ± ± ± 14.8 11.4 30.2 37.1 16.5 18.9 2.91 2.61 2.05 5.32 6.55 3.33 − − − − − − Ec 17.45 17.65 17.65 17.65 ± 0.03 − ( − ( − ( − ( − ( − (

42.2%)

± 23.8%) 31.2%) 23.1%) 46.0%) 29.6%)

0.19 0.25 0.18 0.33 0.36 0.23

± ± ± ± ±

± ± ± ± ± ±

Eo 0.79 0.79 0.79 0.79 ± 0.02 (40.0 (40.1 (31.6 (35.2% (21.8 (38.1 0.31 0.26 0.29 0.17 0.30

42.9%) 47.4%) 28.4%) 27.0%) 23.2%) 32.4%)

± ± 1.55 0.32 1.47 1.26 2.34 2.59 1.77 ±

± ± ±

± ± ± ± ± ±

Dc 5.13 5.43 5.43 5.45 ± 0.03 (77.4 (69.5 (56.9 (126.4 (149.6 (84.6 4.21 3.79 3.01 6.88 8.16 4.61 ) covered by each climate type. The numbers during 2040–2059 and 2080–2099 are the averaged changes and one standard deviation standard one and changes averaged the are 2080–2099 and 2040–2059 during numbers The type. climate each by covered ) 2 km 6

83.7%) 91.4%)

30.9%) 20.7%) 19.1%) 44.1%)

± ± 0.17 0.12 0.11 0.47 0.51 0.25

± ± ± ±

± ± ± ± ± ±

Do 0.51 0.56 0.56 0.56 ± 0.005 0.44 (80.7 0.40 (72.5 0.32 (57.7 0.95 (170.5 1.20 (214.1 0.54 (96.2 ) projected by the 16 models. The numbers in parentheses are the projected changes and one standard deviation in percentage compared to 1950–1999. Bold (Italic) to 1950–1999. Bold deviation in percentage compared changes and one standard The numbers in parentheses are the projected by the 16 models. ) projected 2 km 6 Data sources UD temperature Adjusted UD temperature AL temperature 16 models A1b Observed and projected future changes of climate types in the Arctic region (north 50°N) Table 2. Years 1950–1999 2040–2059 A1b A2 B1 2080–2099 A2 B1 The numbers during 1950–1999 are the total area (in 10 of total area (in 10 numbers indicate expanding (shrinking) area coverage compared to the period 1950–1999 6 S. Feng et al. in Climate Dynamics (2011)

Figure 1. Areal weighted temperature changes in the Arc- tic during 1900–2099. The temperature anomalies are based on the 1950–1999 climatological mean. The black solid line is the temperature anomalies based on the adjusted tempera- ture dataset from University of Delaware, and green dashed line is the temperature anomalies based on the temperature data- set developed by Adam and Lettenmaier (2003). The pink, red and blue lines are the ensembles of the projected temperature changes under SRES B1, A1b, and A2 scenarios, respectively. The yellow shading shows the standard deviations of the tem- perature anomalies from the 16 model projections under A1b scenarios Figure 2. Spatial distributions of the projected temperature changes during 2080–2099 under SRES A1b scenario. The pro- The warming signals, however, are not homoge- jected changes are the ensemble of the 16 fully coupled mod- neously distributed across the Arctic (Figure 2). In win- els. The contour lines show the standard deviation of the pro- ter, warming of 2–10°C is projected by the end of this jected changes among the 16 models century under the A1B scenario. Strongest warming (>7°C) appears along the Arctic coast regions. Moder- ate warming (5–6°C) appears in most of southern Rus- sia and southern . The weakest warming (2–3°C) Russia and the Arctic coastal regions. The weakest occurs in southern Greenland, , and Western Eu- warming (<2.0°C) is projected for southern Greenland, rope. By contrast, the regional warming is much weaker Iceland and . in summer. The strongest summer warming (3–4°C) oc- The spatial distribution of the annual temperature curs in the southern portions of the Arctic, i.e., south- warming is similar to that during winter, except the central Canada and southwestern Siberia. Moderate magnitude of the warming is slightly smaller (about warming (2–3°C) is projected for Alaska, far-eastern 2–8°C warming). Despite the overall strong warm- Evaluating future climate changes for the Arctic using the K-T classification 7 ing throughout the Arctic, the warming in south- ern Greenland, Iceland, and western Europe is rela- tively weak (1.5–2.0°C). The projected weak warming in these regions is likely caused by accelerating melt of snow in Greenland and sea ice in the Arctic Ocean (Dima and Lohmann 2007) as energy is going into melt- ing, rather than warming. Though the spatial distribu- tions of the projected warming are noticeably different between winter and summer, the spatial distributions of the model uncertainties are fairly similar for all sea- sons. These uncertainties among the model projections are evaluated by calculating the standard deviation of the temperature changes projected by the 16 models. As shown in Figure 2, large uncertainties (that is, large standard deviation of the projected changes among the 16 models) of the projected warming occur in northern Greenland and Arctic coastal regions. The uncertain- ties in southern Canada and southern Russia are rela- tively small. These projected temperature warmings are comparable with previous studies for the Arctic regions (ACIA 2004). Our results, however, are based on an en- semble of 16 models included in IPCC AR4. The statisti- cally-downscaled high spatial resolution data also pro- vide more local detail of the projected changes in the Arctic region. Nonetheless, it is reassuring that our re- sults are similar to those obtained previously as it in- creases confidence in their robustness.

3.2. Spatial distribution of the K-T climate types

To examine the impact of these large Arctic warm- ings on vegetation type, the K-T climate classification is calculated using both observed and projected temper- ature datasets. Figure 3a shows the spatial distribution of each climate type during 1950–1999 based on the ad- justed UD temperature dataset. The spatial distribution of the climate types based on the AL temperature data- set and the ensemble of the 16 climate models during 1950–1999 is very similar to that based on the adjusted UD temperature (see Table 2, figures omitted). The tem- perate oceanic climate (Do) is found in Western Europe, and some small regions near the west Canadian coast. The tundra climate (Ft) is mainly found in northern Canada and coastal regions around the Arctic Ocean. Scattered regions of tundra are also found in the moun- Figure 3. Spatial distribution of K-T climate sub-types. a) is tains of southern Alaska, and in far-eastern Russia. deduced from the long-term average (1950–1999) tempera- The spatial distribution of tundra in the Arctic closely ture dataset of the adjust UD temperature, and b) and c) are matches those regions enclosed by mean summer tem- deduced from the ensemble of the projected temperature dur- peratures between −5 and 5°C (figure not shown). Mat- ing 2040–2059 and 2080–2099 under A1b scenarios. The contour thes et al. (2009) analyzed the growing degree days, lines in b) and c) outline regions with 9 or fewer models as- the accumulated temperature for daily mean tempera- signed the same climate types as the ensemble ture warmer than 5°C, in the Arctic region. Their results showed that annual growing degree days in regions oc- (2004) also showed that the vegetation cover in regions cupied by tundra climate are normally less than 600°C, occupied by tundra climate is quite low, with NDVI val- suggesting very little heating energy available for veg- ues ranging from 0.1 to 0.4 in July and August. The cor- etation growth in those regions. Wang and Overland respondence of regions covered by tundra with mean 8 S. Feng et al. in Climate Dynamics (2011) summer temperatures between −5 and 5°C, few grow- made by multiple regional climate models (de Castro et ing degree days, and low vegetation growth all support al. 2007). previous studies (Köppen 1936; Bailey 2009) that sug- While the ensemble of the 16 global models shows gest the K-T climate types can reasonably describe the a systematic redistribution in climate types, there are dominant vegetation throughout the Arctic. some differences among the models. Because it is im- To evaluate the impact of projected climate change practical to display the projected climate types for each on vegetation assemblages, the spatial distributions of model, Figure 3 shows the uncertainties of the 16 mod- the climate types as simulated by the 16 climate mod- els in describing the projected changes in K-T climate els are analyzed. For simplicity, only the spatial distri- types. The uncertainties are evaluated by comparing bution of the climate types projected by A1b (ensem- the K-T climate types projected by the model ensem- ble of the 16 models) in the middle (2040–2059) and end ble with individual models. In particular, for a given (2080–2099) of this century are displayed in Figure 3. grid cell, if more than 10 models projected the same cli- Compared to present-day conditions (Figure 3a), notice- mate type as the ensemble mean, it suggests that major- able shifts in climate types are projected for the middle ity of the models (two thirds) are in agreement for that and end of this century. For example, the areal extent of grid cell. It also implies that model projections for that tundra in Alaska is substantially reduced during 2040– grid cell contain fewer uncertainties. For a majority of 2059, being replaced by forests of the boreal continental Arctic regions, at least 10 or more models projected the climate (Ec). The tundra will be further reduced, being same climate types as the ensemble (Figure 3). Regions mainly restricted to the north coast of Alaska by 2080– with less agreement among the models (shown in con- 2099. In northern Canada, the warming pushes the dis- tour lines) are mainly located in western and middle Si- tribution of tundra poleward to the coast of the Arctic beria. These disagreements become larger during 2080– Ocean and adjacent islands during 2040–2059. The tun- 2099 compared to during 2040–2059, suggesting that the dra will be restricted to the islands in the Arctic Ocean biases in individual model increase with time. These during 2080–2099. On the other hand, the melting of disagreements are consistent with the models contain- snow and ice in Greenland following the warming will ing different atmospheric, oceanic and land surface pro- reduce the permanent ice cover (Fi), giving its territory cesses. These disagreements, however, are relatively up to tundra (Ft). small and mostly located along the boundary of climate Following the northward contraction of tundra and types. They also suggest that, despite the biases in indi- permanent ice, the boreal oceanic (Eo), Do, and Dc types vidual models, the ensemble of the 16 models may well are projected to expand northward. In eastern Europe describe the projected climate changes. and western Siberia (30°E–90°E), the boreal continen- The total areas occupied by each climate type during tal climate (Ec) is found south of 60°N during 1950–1999 2040–2059 and 2080–2099 for all the three SRES scenar- (Figure 3a). With the projected warming, this climate ios are listed in Table 2. As seen in the table, noticeable type is projected to retreat to approximately 62°N dur- changes are projected for these two periods. Overall, ing 2040–2059 and 65°N during 2080–2099 (Figure 3). the areas occupied by polar climate (i.e., Fi and Ft) and On the other hand, the Do and Dc climate types advance boreal continental climate (i.e., Ec) are projected to de- into the areas originally covered by Ec. In middle and cline, while the temperate (i.e., Do and Dc) and boreal eastern North Asia (90°E–150°E), the Ec climate occurs oceanic (i.e., Eo) climate types are expected to expand in from south of 50°N to the Arctic coast during 1950–1999. the Arctic region. Of these 6 climate types, the Ec and This climate type almost disappears, and is replaced Ft show the most decline with the warming. The tun- by the Dc climate during 2080–2099 in regions south of dra cover is expected to shrink by −1.82 × 106 km2 and 55°N in middle and eastern North Asia. In other words, −1.67 × 106 km2 (or −32.2 and −29.6%, respectively) un- with the projected warming under the mid-range ARES der SRES A1b and A2 scenarios by 2040–2059. Addi- A1b scenario, the Ec climate zone in middle and eastern tional 0.9 × 106 km2 to 1.0 × 106 km2 reductions are pro- North Asia shifts from south of 50°N during 1950–1999 jected for the two scenarios, respectively, by the end to north of 55°N by the end of the 21st century. A more of this century. By contrast, the reduction in tundra is than 5° northward retreat of Ec climate is also projected less under the stronger stabilization SRES scenario B1. for during 2080–2099 (Figure 3). The simulated area occupied by Ec will be reduced by In Europe, the Do climate is projected to expand −2.05 × 106 km2 (−11.4%), −2.91 × 106 km2 (−16.5%), and north and east into Western Europe (around 25°E) by −2.61 × 106 km2 (−14.8%) during 2040–2059 under SRES the end of this century. The tundra in Scandinavia dur- B1, A1b, and A2 scenarios, respectively. The reduction ing 1950–1999 will be substantially reduced during increases to −3.33 × 106 km2 (−18.9%), −5.32 × 106 km2 2040–2059, almost totally disappearing by 2080–2099. (−30.2%), and −6.55 × 106 km2 (−37.1%) for the three Scandinavia will then be dominated by the Eo climate. scenarios, respectively, during 2080–2099. On the other Similar changes are also projected using simulations hand, the area occupied by Dc, Do, and Eo types are pro- Evaluating future climate changes for the Arctic using the K-T classification 9 jected to expand with all three SRES scenarios. The most noticeable expansion is projected for Dc, with a greater than 3.0 × 106 km2 increase in Dc projected during 2040– 2059. Coverage then increases by 4.61 × 106 km2 (or 84.6%), 6.88 × 106 km2 (or 126.4%), and 8.16 × 106 km2 (or 149.6%) under B1, A1b, and A2 scenarios, respec- tively, during 2080–2099. As shown in Figure 3, the ex- pansion of Dc is mainly because the area occupied by this climate type in 2080–2099 increases by at least 5° north of its present-day conditions (1950–1999).

3.3. Temporal variations of the K-T climate types

The evolution of the Arctic regions occupied by each climate type has also been analyzed. Figure 4 shows the temporal variations of the total area occupied by each climate type during the entire analysis period 1900– 2099. The observations show a weak trend toward re- ducing tundra cover from the beginning of the 20th century to the 1940s. This decrease in tundra coverage leveled off from middle 1940 to 1970s, followed by an even more abrupt decrease during the recent 40 years. Similar trends have been observed by Wang and Over- land (2004), using a different observed temperature da- taset for the period 1901–2000. The models projected a steady decline in tundra coverage. The projected decline rates in tundra coverage are −0.16 × 106 km2 (−2.7%), −0.22 × 106 km2 (−3.8%), and −0.23 × 106 km2 (−4.0%) per decade for the B1, A1b and A2 scenarios, respectively. The changes of Dc coverage during the instrumental period also show interdecadal variations. The area cov- Figure 4. Time series of the total areas occupied by each cli- ered by Dc steadily expands from the early 1900s to the mate types in the Arctic region. The black solid lines are the late 1940s, but then slowly shrinks until the late 1970s, temporal variations based on the adjusted temperature da- followed by a steady expansion over the last few de- taset from University of Delaware, and green dashed lines are cades (Figure 4). This multidecadal change in Dc cov- based on the temperature dataset from Adam and Lettenma- erage is closely related to the observed temperature ier (2003). The pink, red and blue lines are the ensemble of the changes in the Arctic during the last 100 years (Fig- projected total area changes under SRES B1, A1b, and A2 sce- ure 1). Recent expansions in Dc coverage are projected narios, respectively. The yellow shading shows the standard de- to continue by all the SRES scenarios. The area occupied viations of the total area coverage of the 16 model projections by Dc is projected to increase by 0.43 × 106 km2 (7.8%), under A1b scenarios. A 15-year equal weight smoothing is ap- 6 2 6 2 plied to the observed and projected temperature data before 0.72 × 10 km (13.0%), and 0.81 × 10 km (14.7%) per the K-T climate classification is calculated. decade for the B1, A1b, and A2 scenarios, respectively. In order to evaluate the change of climate types on regional scales, the temporal variations of the areal cov- into those areas formerly occupied by Ft and Ec. On erage of each climate type for (50°N– the other hand, the Do climate type is also projected to 75°N and 12°W–40°E) and Alaska (50°N–75°N and move north and east, expanding into regions previously 130°W–168°W) were also analyzed. In northern Europe, covered by Dc and Eo (Figure 3). The expansion of Do the projected changes of each climate type are more coverage markedly increases after 2040, consistent with complicated as compared to the entire Arctic region, the slow decline in Dc and Eo (Figure 5). suggesting differing regional responses to large scale Alaska is dominated by both Ec and Ft climate types. warming (Figure 5). Persistent expansions in Do cover- The two occupy about 1.95 × 106 km2 (or 90% of the land age are projected by all SRES scenarios, while the cov- region in Alaska) during the instrumental period (Fig- erage of Dc and Eo are projected to slowly expand un- ure 6). Tundra covers about 0.72 × 106 km2 (or 32% of til the 2040s, and then slowly decline. The warming Alaska) during the 1950 and 1960s, then gradually de- allows the Dc and Eo types to move north and eastward, clines until the 1990s. The coverage of tundra then 10 S. Feng et al. in Climate Dynamics (2011)

Figure 5. Same as Figure 4 but for Northern Europe (50–75°N Figure 6. Same as Figure 4 but for Alaska (50–75°N and 130– and 12°W–40°E). The ‘Fi’ climate type is not shown because it 168°W). The ‘Fi’ climate type is not shown because it does not does not exist in Europe. exist in Alaska. slightly increased in recent years. Decreasing area oc- age of Dc and Do combined is less than 0.015 × 106 km2 cupied by tundra is projected for all three SRES scenar- of the total area in Alaska before 2050, but then is pre- ios, so that it just covers about 0.2–0.3 × 106 km2 (or 10– dicted to increase sharply by 2100. These results suggest 15% of Alaska, depending on the SRES scenarios) by the that, though the temperature in the Arctic is projected to end of this century. This reduction in tundra is largely increase steadily under all SRES scenarios, when a tip- replaced by increasing coverage of Ec. The coverage of ping point in temperature is reached, an abrupt shift can Ec and Ft both varied on bi-decadal and longer times- occur in regional climate types and vegetation. Climate cales, but the fluctuations and the trend of their cover- changes are known to have caused large and abrupt age are nearly out of phase during the instrumental pe- shifts in regional vegetation during the Holocene (e.g., riod. This overall out-of-phase relationship between the Claussen et al. 1999; Cole 2010); this study suggests that coverage of the two climate types is projected to con- warming in the future may also trigger significant shifts tinue until the 2050s. Ec is projected to increase until the in Arctic regional ecosystems. middle of this century, followed by an even sharper de- Figures 3, 4, 5, and 6 all suggest that the observed and cline. Similar decline in areal coverage in the second half projected future climate changes are sufficient to cause of this century is also projected for the Eo climate type. large shifts in the spatial distribution of climate types. Those projected declines after the 2050s are consistent For successive 15-year intervals, Figure 7 shows the per- with the northward shift of Do and Dc (Figs. 3 and 6) centage of total area in the Arctic assigned to specific during that same period into a region where they did climate types, as compared to the 1950–1999. From the not occur during the observational period. The cover- 1900 to 1950s only about 3–5% of the total Arctic area Evaluating future climate changes for the Arctic using the K-T classification 11

are dwarfed compared to the projected changes under different SRES scenarios. As shown in Figure 7, the cli- mate types in about 25, 39.1, and 45% of the Arctic are projected to change by the end of this century under B1, A1b and A2 scenarios, respectively. In other words, un- der these emission scenarios, the current dominant veg- etation may be replaced by different vegetation by one- quarter to nearly one-half of the Arctic land area by the end of this century. To better understand the differences between the cur- Figure 7. Time series of the percentage of area in the Arctic as- rent climate classifications and those projected for the signed different climate types compared to the present day condition (1950–1999). A 15-year equal weight smoothing is future, the redistributions of climate types during 2040– applied to the observed and projected temperature data before 2059 and 2080–2089 in the Arctic region are each ana- the K-T climate classification is calculated. Then the percent- lyzed. For simplicity, only the projected changes un- age of area in the Arctic assigned different climate types com- der A1b scenarios are shown in Figure 8. About 26.1% pared to 1950–1999 during each 15 years interval is calculated. (39.1%) of the Arctic regions are assigned to a differ- ent K-T climate type in 2040–2059 (2080–2099) com- pared to the present-day conditions. As shown in the shows different climate types from 1950 to 1999, sug- Figure, major transfers take place from Ft to Ec, and gesting that the climate regimes are fairly stable dur- from Ec to Eo and Dc climate types. Additionally, the ing this period. The most recent 15 year period (1994– changes of climate types all follow the same direction, 2008), however, shows distinct differences compared to e.g., from colder climate types to warmer climate types. the 1950–1999 period, with about 9% of the region hav- The reduced ice covered regions (Fi) are taken over by ing different climate types. This is consistent with the Ft. The large decline in tundra in turn will largely be re- general consensus that the most recent 20 years repre- placed by Ec. The Ec will be mostly replaced by Dc cli- sents a period with accelerated global warming (IPCC mate. As a result, the area occupied by Dc has the larg- 2007). These recent changes in climate types, however, est projected increase, followed by Do and Eo. The

Figure 8. Transfers between different K-T climate types in the Arctic during a 2040–2059 under SRES A1b scenarios. The numbers above/below each climate type indicate the total area (in 106 km2) and percentage of total Arctic area (in parentheses) occupied by each type during 1950–1999 and 2040–2059, respectively. The numbers by the arrows indicate redistribution of area and percent- age of total Arctic area (in parentheses) between climate types. The results shown are the ensemble of the 16 models. b is same as a but for 2080–2099 12 S. Feng et al. in Climate Dynamics (2011)

Polar climate and Ec are expected to be reduced sub- As mentioned in Sect. 2, a key advantage of the K-T stantially with the projected warming in the Arctic re- classification is that it is simple, being only defined by gion. Though the redistributions of the climate types in temperature and precipitation. Further, the results are Arctic during the 20th century are very small (Figure 7), easy to interpret and understand. Because of this rela- the much larger redistributions projected for the future tive simplicity, it is possible to use the method to eval- suggest that the warming will cause large shifts in cli- uate the impact of projected climate changes on veg- mate regimes (Figure 8). etation based on numerous, multiple-model outputs and multiple future scenarios. However, like all simple methods, using the K-T climate classification to evalu- 4. Discussion ate vegetation changes has its limitations. For example, the K-T classification is not able to address the effect

This study used the K-T climate classification to of CO2 fertilization (Piao et al. 2007) and other non-cli- evaluate climate changes in the Arctic, as based on re- mate factors, such as local soil type, nutrient limitation, sults from a number of global climate models. Because human land use changes, permafrost dynamics, com- each climate type is associated with a certain vegeta- petition among plant species, and wild fire (Hobbie et tion assemblage, the redistribution of climate types al. 2002; Goetz et al. 2005; Tchebakova et al. 2009; Soja suggests concomitant changes in Arctic vegetation. An- et al. 2007) on the distribution of local and regional other approach involves use of models that use appro- vegetation. Pests and diseases may also expand their priate biophysics to ‘dynamically’ compute the veg- geographic ranges as climate warms, increasing stress etation for a region for a given climate regime. The on vegetation growth (Soja et al. 2007). Therefore, the response of Arctic vegetation to climate change has relationships between climate and vegetation may been simulated by several such dynamic vegetation not be the same in the future as under current condi- models, including the -- dynamic tions. (These limitations also affect dynamical vege- global vegetation model (LPJ DGVM) (Sitch et al. 2003; tation models.) Moreover, the K-T classification only Callaghan et al. 2005), the 4 model (e.g., Kaplan considers a few climate-vegetation assemblages, which and New 2006; Epstein et al. 2007), the Terrestrial Eco- hardly represent the current range of vegetative diver- system Model (TEM, McGuire et al. 2000; Thompson et sity throughout the Arctic. Importantly, feedbacks of al. 2005; Euskirchen et al. 2009), the BIOME-BGC model vegetation changes back onto the surface climate can- (Engstrom et al. 2006), the Canada climate-vegetation not be explicitly accounted for. Previous studies (e.g., model (CCVM, Lenihan and Neilson 1995), the Alaska Chapin et al. 2005; McGuire et al. 2006; Jeong et al. Frame-based Ecosystem Code (ALFRESCO, Rupp et al. 2010a) suggest that the feedbacks of vegetation on cli- 2000), the ArcVeg (Epstein et al. 2000) and the TreeMig mate can be important and should be considered in fu- model (Lischke et al. 2007). Some of the dynamic veg- ture climate change impact assessment. etation models (e.g., the LPJ DGVM) were incorpo- The K-T classification, though simple and with limita- rated into global climate models to better understand tions, has also yielded results consistent with those from the interactions and feedbacks of vegetation on climate the dynamic vegetation models. The K-T classification (Levis et al. 1999, 2004). These dynamic models vary only identifies a few climate-vegetation assemblages, in the types and detail of the ecological and biophys- while the dynamic vegetation models generally provide ical processes incorporated, the controlling and input much more resolution of vegetation types. Therefore, it model variables, and the representation of vegetation is impractical to quantitatively compare our results with types in the Arctic (Epstein et al. 2007). Comprehen- those of the dynamic vegetation models. Nevertheless, sive reviews of the vegetation models and their mod- the modern distributions of K-T climate types in the eling strategies were given by Woodward and Lomas Arctic resemble the major vegetation types simulated (2004) and Epstein et al. (2007). However, the dynamic by advanced vegetation models (Epstein et al. 2007; Ka- models require many input parameters, whose values plan and New 2006), as well as by what few observa- may not be readily available, especially for future sce- tional studies exist (e.g., Sturm et al. 2001; Thorpe et al. narios. These models are also computer-intensive (Ep- 2002, Lloyd et al. 2003; Lloyd 2005). For example, Sturm stein et al. 2007), making it impractical to use them to et al. (2001) showed widespread decrease in tundra cov- evaluate the impact of climate changes on vegetation erage and a distinct increase in the coverage and density when forced by multiple climate models and multiple of spruce trees along the tree lines by using long-term future scenarios. This is especially important because it ground photographs. Those observed changes in tree has been suggested that, in order to reduce the bias in- lines are consistent with the northward shift of Ec and herent in individual models, an ensemble of multiple Dc climate types in our results. climate model outputs are necessary for robust climate Callaghan et al. (2005) summarized modeled Arc- change impact assessment (e.g., Gleckler et al. 2008; tic vegetation changes resulting from global warming. Reichler and Kim 2008; Pierce et al. 2009). They found that most of the dynamic vegetation mod- Evaluating future climate changes for the Arctic using the K-T classification 13 els projected shrinkage of tundra coverage. Much of 5. Conclusion the tundra (between 11 and 50%, depending on spe- cific region and model) will be replaced by northward This study evaluated the temperature changes in the shift of boreal forest when the atmospheric CO2 con- Arctic region (north of 50°N) using observations and centrations are doubled. Our results yield 33.0–44.2% simulations for the period 1900–2099 made by 16 fully percent shrinkage in tundra coverage by the end of coupled climate models. Our examination shows mul- this century, well within the projected changes in tun- tidecadal variations of temperature during the instru- dra coverage made by the dynamic vegetation mod- mental period, consistent with the temperature re- els. Additionally, the treeline is projected to move cord for the entire . A consistent north in all sectors of Arctic (Figure 3), which is also warming in the Arctic is observed since the late 1970s. consistent with dynamic model projections (Cal- The recent warm trends are projected to continue under laghan et al. 2005; Bonan et al. 1992; Foley et al. 1994; the three SRES scenarios (B1, A1b, and A2). Compared Levis et al. 1999; Jeong et al. 2010a). Recent modeling to present-day conditions, the annual temperatures studies (Kaplan and New 2006; Epstein et al. 2007) us- are projected to increase by 2–8°C in the Arctic by the ing BIOME 4 predict that, with a 2°C global warm- end of this century under the A1b scenario. The warm- ing (which possibly will happen by the middle of this ing signals in the annual mean temperature are not ho- century), the boreal forest will move north, with the mogeneously distributed in the Arctic, with the largest northern limit trees reaching up to 400 km from the warming (>5°C) in coastal regions, and lesser warming present tree line. Figure 3 also indicated that the Dc (3–5°C) in the southern parts of the Arctic (between 50– climate type will displaced northward by about 2–3° 60°N). The weakest warming (2–3°C) occurs in the high- in the middle of this century, consistent with the BI- latitude North Atlantic realm. The spatial distribution OME 4 model. of the warming signals in winter is very similar to the Our results and the dynamic vegetation models all annual mean temperature, except the magnitude of the project large redistributions of vegetation in the Arc- warming is stronger, 2–10°C. The projected warming tic region. The changes in vegetation are broadly con- in summer is much weaker (1.5–4.2°C), with strongest sistent with observed vegetation changes in the Arctic warming in the southern Arctic, and weaker warming region (e.g., Sturm et al. 2001; Thorpe et al. 2002, Lloyd in the Arctic coastal regions. When averaged over the et al. 2003; Lloyd 2005). However, the observed rate of entire Arctic land region, annual mean temperatures in change is smaller than the projections in this study and the Arctic are projected to increase by 3.1, 4.6 and 5.3°C the vegetation models (Callaghan et al. 2005). There under the B1, A1b, and A2 scenarios, respectively, by are several constraints to vegetation changes, a domi- 2080–2099. The winter temperature is projected to in- nant one being the dispersal of seeds, followed by the crease by 4.2, 6.1, and 7.1°C, and the summer temper- germination and establishment of seedlings (Epstein et ature is projected to increase by 2.2, 3.3, and 3.9°C un- al. 2007). The response of vegetation therefore usually der B1, A1b and A2 scenarios, respectively, by the end lags changes in climate. For example, shrub density in of this century. tundra regions has seen a rapid increase on decadal The projected warming leads to large shifts in cli- time scales (Arft et al. 1999), but boreal forest expan- mate regimes in the Arctic regions. The areas occupied sion has seen a much slower response on century time by polar climate types (Ft and Fi) and subarctic conti- scales (ACIA 2004; Epstein et al. 2007). Furthermore, nental climate (Ec) type are projected to steadily decline, increasing drought conditions may help offset any po- while the areas covered by temperate (Dc and Do) and tential benefits of warmer temperatures and reduce the boreal oceanic climate (Eo) types are expected to steady overall vegetation growth (biomass) in the Arctic re- expand. The tundra region is projected to decline by gion (e.g., Barber et al. 2000; Angert et al. 2005; Bunn −1.86 × 106 km2, −2.4 × 106 km2, and −2.5 × 106 km2, or et al. 2007; Jeong et al. 2010b). This suggests that in- −33.0, −42.6, and −44.2% by the end of this century un- creasing temperature, with no comparable increase in der the B1, A1b and A2 scenarios, respectively. The Ec precipitation, may lead to reduced vegetation growth climate type will retreat at least 5° north of its present in the future. Other non-climate factors, e.g., local hu- day location, resulting in −18.9, −30.2, and −37.1% de- man activity, land use change, permafrost thawing, as clines in areal coverage under the B1, A1b and A2 sce- well as pest outbreaks and fire may also locally affect narios, respectively. Following the retreat of tundra and the response of vegetation to temperature warming in Ec climate types, the temperate climate advances into the Arctic. Therefore, the redistributions of vegetation the areas currently covered by Ec. The area covered by suggested by the K-T classifications obtained from this Dc climate is expected to expand by 4.61 × 106 km2 (or study do not mean that the projected changes of vege- 84.6%), 6.88 × 106 km2 (or 126.4%) and 8.16 × 106 km2 (or tation will really happen during this century. More de- 149.6%) under B1, A1b and A2 scenarios, respectively. tailed studies accounting for both climate and non-cli- The redistribution of K-T climate types differ region- mate factors are needed. ally. In Europe, the areal coverage’s of Dc and Eo are 14 S. Feng et al. in Climate Dynamics (2011) projected to slowly expand until 2040s, then slowly de- Callaghan TV et al. (2005) Arctic tundra and polar ecosys- cline. The Do climate, however, is projected to abruptly tems. In: Arctic Climate Impact Assessment (ed) Arctic cli- expand after the 2040s. In Alaska, the regions occupied mate impact assessment: Scientific report. Univ. Press, Cambridge, pp 243–352 by boreal climate types (Eo and Ec) are projected to in- Chapin FS III et al (2005) Role of land-surface changes in Arctic crease until the 2050s, whereas accelerated expansion of summer warming. Science 310:657–660 temperate climate types (i.e., Do and Dc) is projected af- Claussen M, Kubatzki C, Brovkin V, Ganopolski A, Hoelzmann ter the 2050s. The redistribution of each climate type is P, Pachur H (1999) Simulation of an abrupt change in Sharan slightly smaller under the stabilization B1 scenario com- vegetation in the mid-Holocene. Geophys Res Lett 26:2037–2040 Cole KL (2010) Vegetation response to early Holocene warming pared to the business as usual A1b and A2 scenarios. as an analogy for current and future changes. Conserv Biol Temporal variations of the K-T climate types in the 24:29–37 Arctic were also analyzed. The recent 15 year period de Castro M, Gallardo C, Jylha K, Tuomenvirta H (2007) The use (1994–2008) shows distinct differences compared to the of climate-type classification for assessing climate change- ef 1950–1999 period, with about 9% of the Arctic having fects in Europe from an ensemble of nine regional climate different climate types. These recent changes in climate models. Climatic Change 81:329–341 Dima H, Lohmann G (2007) A hemispheric mechanism for the At- types are projected to continue and be amplified under lantic Multidecadal Oscillation. J Climate 20:2706–2719 all three SRES scenarios. The climate types in about 25, Engstrom R, Hope A, Kwon H, Harazono Y, Mano M, Oechel W 39.1, and 45% of the Arctic are projected to change by (2006) Modeling evapotranspiration in Arctic coastal plain the end of this century under the three scenarios. ecosystems using a modified BIOME-BGC model. J Geophys Res 111:G02021. doi:10.1029/2005JG000102 Epstein HE, Walker MD, Chapin FS III, Starfield AM (2000) A transient, nutrient-based model of arctic plant community re- Acknowledgments — We thank the editor, Dr. Edwin K. sponse to climatic warming. Ecol Appl 10:824–841 Schneider, and two anonymous reviewers for Climate Dy- Epstein HE, Kaplan JO, Lischke H, Yu Q (2007) Simulating future namics for their constructive comments, which have led to changes in arctic and sub-arctic vegetation. Comput Sci Eng improvement of this manuscript. This research was funded 9(4):12–23 by the Korea Arctic Multidisciplinary Program under grant Euskirchen ES, McGuire AD, Chapin FS III, Yi S, Thompson CC PP10090, CATER 2006–4204 and by USDA Cooperative Re- (2009) Changes in vegetation in northern Alaska under scenar- search Project NEB-40-040. ios of climate change 2003–2100: Implications for climate feed- backs. Ecol Appl 19:1022–1043 Foley JA, Kutzbach JE, Coe MT, Levis S (1994) Feedbacks between References climate and boreal forests during the Holocene epoch. Nature 371:52–54 Adam JC, Lettenmaier DP (2003) Adjustment of global gridded Fraedrich K, Gerstengarbe F-W, Werner PC (2001) Climate shifts precipitation for systematic bias. J Geophys Res 108:1–14 during the last century. Climatic Change 50:405–417 Angert A, Biraud S, Bonfils C, Henning CC, Buermann W, Pinzon Gerstengarbe F-W, Werner PC (2009) A short update on Koep- pen climate shifts in Europe between 1901 and 2003. Climatic J, Tucker CJ, Fung I (2005) Drier summer cancel out the CO2 uptake enhancement induced by warmer springs. Proc Natl Change 92:99–107 Acad Sci 102:10823–10827 Gillett NP, Stone DA, Stott PA, Nozawa T, Karpechko AY, Hegerl Arctic Climate Impact Assessment (ACIA) (2004) Impacts of a GC, Wehner MF, Jones PD (2008) Attribution of polar warm- warming arctic: Arctic climate impact assessment. Cambridge ing to human influence. Nat Geosci 1:750–754 University Press, Cambridge Gleckler PJ, Taylor KE, Doutriaux C (2008) Performance metrics Arctic Climate Impact Assessment (ACIA) (2005) Arctic climate for climate models. J Geophys Res. doi: 10.1029/2007JD008972 impact assessment: Scientific report. Cambridge Univ. Press, Goetz SJ, Bunn AG, Fiske GJ, Houghton RA (2005) Satellite-ob- Cambridge served photosynthetic trends across boreal North America as- Arft AM et al (1999) Responses of tundra plants to experimental sociated with climate and fire disturbance. Proc Natl Acad Sci warming: Meta-analysis of the international tundra experi- 102:13521–13525 ment. Ecol Monogr 69:491–511 Hinzman LD et al (2005) Evidence and implications of recent cli- Bailey RG (2009) Ecosystem geography: From ecoregions to sites, 2nd mate change in Northern Alaska and other Arctic regions. Cli- edn. Springer, New York, p 251 matic Change 72:251–298 Baker B, Diaz H, Hargrove W, Hoffman F (2010) Use of the Köp- Hobbie SE, Nadelhoffer KJ, Högberg P (2002) A synthesis: The pen-Trewartha climate classification to evaluate climatic refu- role of nutrients as constraints on carbon balances in boreal gia in statistically derived ecoregions for the People’s Republic Arctic regions. Plant Soil 242:163–170 of China. Climatic Change 98:113–131 IPCC (2007) Climate change 2007: the physical science basis. In: Barber VA, Juday GP, Finney BP (2000) Reduced growth of Alas- Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt kan white spruce in the twentieth century from temperature- KB, Tignor M, Miller HL (eds) Contribution of working group I induced drought stress. Nature 405:668–673 to the fourth assessment report of the intergovernmental panel on Bhatt US et al (2010) Circumpolar Arctic tundra vegetation change climate change. Cambridge University Press, Cambridge, p 996 is linked to sea ice decline. Earth Interact 14(8):1–20 Jeong S-J, Ho C-H, Park T-W, Kim J, Levis S (2010a) Impact of veg- Bonan GB, Pollard D, Thompson SL (1992) Effects of boreal forest etation feedback on the temperature and its diurnal range vegetation on global climate. Nature 359:716–718 over the Northern Hemisphere during summer in a 2 × CO2 Bunn AG, Goetz SJ, Kimball JS, Zhang K (2007) Northern high-lat- climate. Clim Dyn. doi: 10.1007/s00382-010-0827-x itude ecosystems respond to climate change. Eos Trans AGU Jeong S-J, Ho C-H, Kim B-M, Feng S, Medvigy D, Kim Y-W, 88:333–335. doi: 10.1029/2007EO340001 Lee H-H (2010b) Conspicuous circumpolar greening in the Evaluating future climate changes for the Arctic using the K-T classification 15

end of growing season over the Arctic region. J Geophys Res Post E et al (2009) Ecological dynamics across the Arctic associ- (submitted) ated with recent climate change. Science 325:1355–1358 Jia GJ, Epstein HE, Walker DA (2009) Vegetation greening in the Reichler T, Kim J (2008) How well do coupled models simulate to- Canadian arctic related to decadal warming. J Environ Monit day’s climate? Bull Am Meteor Soc 89:303–311 11:2231–2238 Rupp TS, Chapin FS III, Starfield AM (2000) Response of subarctic Kaplan JO, New M (2006) Arctic climate change with a 2°C global vegetation to transient climatic change on the Seward Penin- warming: Timing, climate patterns, and vegetation change. sula in North-West Alaska. Global Change Biol 6:541–555 Climatic Change 79:213–241 Serreze MC, Holland MM, Stroeve J (2007) Perspectives on the Köppen W (1936) Das geographisca system der Klimate. In: Köp- Arctic’s shrinking sea-ice cover. Science 315:1533–1536 pen W, Geiger G (eds) Handbuch der Klimatologie. 1. C. Gebr, Sitch S, Smith B, Prentice IC, Arneth A, Bondeau A, Cramer W, Borntraeger, pp 1–44 Kaplan JO, Levis S, Lucht W, Sykes MT, Thonicke K, Venevsky Lenihan JM, Neilson RP (1995) Canadian vegetation sensitivity to S (2003) Evaluation of ecosystem dynamics, plant geography projected climatic change at three organizational levels. Cli- and terrestrial carbon cycling in the LPJ dynamic global vege- matic Change 30:27–56 tation model. Glob Change Biol 9:161–185 Levis S, Foley JA, Pollard D (1999) Potential high-latitude vege- Soja AJ, Tchebakova NM, French NHF, Flannigan MD, Shugart tation feedbacks on CO2-induced climate change. Geophys Res HH, Stocks BJ, Sukhinin AI, Parfenova EI, Chapin III FS (2007) Lett 26:747–750 Climate-induced boreal forest change: Predictions versus cur- Levis S, Bonan GB, Vertenstein M, Oleson KW (2004) The commu- rent observations, Global Planet. Change 56:274–296 nity land model’s dynamic global vegetation model (CLM-DGVM): Stone RS, Dutton EG, Harris JM, Longenecker D (2002) Ear- Technical description and user’s guide. Technical Note NCAR/ lier spring snowmelt in northern Alaska as an indica- TN-459 + IA, National Center for Atmospheric Research, tor of climate change. J Geophys Res 107:D10, 4089. doi: Boulder, Colorado, p 50 10.1029/2000JD000286 Lischke K, Bolliger J, Seppelt R (2007) Dynamic spatio-tempo- Sturm M, Racine C, Tape K (2001) Increasing shrub abundance in ral landscape models. In: Kienast F et al. (eds) A Changing the Arctic. Nature 411:546–547 world: Challenges for landscape research. Springer, Dordrecht, pp Tchebakova NM, Parfenova E, Soja AJ (2009) The effect of cli- 273–296 mate, permafrost and fire on vegetation change -in Si Lloyd AH (2005) Ecological histories, ecological futures: what beria in a changing climate. Environ Res Lett 4. doi: recent changes at treeline reveal about the future. Ecology 10.1088/1748-9326/4/4/045013 86:1687–1695 Thompson CC, McGuire AD, Clein JS, Chapin FS III, Beringer J Lloyd AH, Rupp TS, Fastie CL, Starfield AM (2003) Patterns (2005) Net carbon exchange across the Arctic tundra-boreal and dynamics of treeline advance on the Seward Peninsula, forest transition in Alaska 1981–2000. Mitig Adapt Strat Glob Alaska. J Geophys Res 108:8150. doi:10.1029/2001JD000570 Change 11:805–827 Matthes H, Rinke A, Dethloff K (2009) Variability of observed Thorpe N, Eyegetok N, Hakongak N, Elders K (2002) The earth temperature-derived climate indices in the Arctic. Global is faster now: Indigenous observations of arctic environmental Planet Change 69:214–224 change. In: Krupnik I, Jolly D (eds) Research Consortium of the Maurer EP, Adam JC, Wood AW (2009) Climate model based con- , Fairbanks, AK, pp 201–239 sensus on the hydrologic impacts of climate change to the Trewartha GT, Horn LH (1980) An introduction to climate, 5th edn. Rio Lempa basin of . Hydrol Earth Syst Sci McGraw-Hill, New York, p 437 13:183–194 Tucker C, Slayback D, Pinzon J, Los S, Myneni R, Taylor M (2001) McGuire AD, Clein JS, Melillo JM, Kicklighter DW, Meier RA, Vo- Higher northern latitude NDVI, growing season trends from rosmarty CJ, Serreze MC (2000) Modelling carbon responses of 1982 to 1999. Int J Biometeorol 45:184–190 tundra ecosystems to historical and projected climate: Sensitiv- Wang M, Overland JE (2004) Detecting Arctic climate change us- ity of pan-Arctic carbon storage to temporal and spatial varia- ing Köppen climate classification.Climatic Change 67:43–62 tion in climate. Global Change Biol 6(supplementary):141–159 Willmott CJ, Rowe CM, Philpot WD (1985) Small-scale climate McGuire AD, Chapin FS III, Walsh JE, Wirth C (2006) Integrated maps: A sensitivity analysis of some common assumptions regional changes in arctic climate feedbacks: Implications for associated with grid-point interpolation and contouring. Am the global climate system. Annu Rev Environ Resour 31:61–91 Cartogr 12:5–16 Meehl GA, Covey C, Delworth T, Latif M, McAvaney B, Mitch- Wood AW, Leung LR, Sridhar V, Lettenmaier DP (2004) Hydro- ell JFB, Stouffer RJ, Taylor KE (2007) The WCRP CMIP3 multi- logic implications of dynamical and statistical approaches model dataset: A new era in climate change research. Bull Am to downscaling climate model outputs. Climatic Change Meteorol Soc 88:1383–1394 62:189–216 Min S-K, Zhang X, Zwiers F (2008) Human-induced Arctic moist- Woodward FI, Lomas MR (2004) Vegetation dynamics–simulating ening. Science 320:518–520 responses to climatic change. Biol Rev 79:643–670 Nakićenović N, Swart R (2000) Special report on emission scenarios. Zhang T, Frauenfeld OW, Serreze MC, Etringer AJ, Oelke C, Mc- A special report of working group III of the intergovernmental panel Creight JL, Barry RG, Gilichinsky D, Yang D, Ye H, Feng L, on climate change. Cambridge University Press, Cambridge Chudinova S (2005) Spatial and temporal variability in active Peterson BJ et al (2002) Increasing river discharge to the Arctic layer thickness over the Russian Arctic drainage basin. J Geo- Ocean. Science 298:2171–2173 phys Res 110. doi: 10.1029/2004JD005642 Piao S, Friedlingstein P, Ciais P, Viovy N, Demarty J (2007) Grow- Zhou LM, Tucker CJ, Kaufmann RK, Slayback D, Shabanov NV, ing season extension and its impact on terrestrial carbon cy- Myneni RB (2001) Variations in northern vegetation activity cle in the Northern Hemisphere over the past 2 decades. Global inferred from satellite data of vegetation index during 1981 to Biogeochem Cycles 21:GB3018. doi: 10.1029/2006GB002888 1999. J Geophys Res 106:20,069–20,083 Pierce DW, Barnett TP, Santer BD, Gleckler PJ (2009) Selecting global climate models for regional climate change studies. Proc Natl Acad Sci 106:8441–8446