Climatic Change

, Volume 111, Issue 2, pp 215–239 | Cite as

Vertical structure of recent arctic warming from observed data and reanalysis products

  • Vladimir A. Alexeev
  • Igor Esau
  • Igor V. Polyakov
  • Sarah J. Byam
  • Svetlana Sorokina
Article

Abstract

Spatiotemporal patterns of recent (1979–2008) air temperature trends are evaluated using three reanalysis datasets and radiosonde data. Our analysis demonstrates large discrepancies between the reanalysis datasets, possibly due to differences in the data assimilation procedures as well as sparseness and inhomogeneity of high-latitude observations. We test the robustness of arctic tropospheric warming based on the ERA-40 dataset. ERA-40 Arctic atmosphere temperatures tend to be closer to the observed ones in terms of root mean square error compared to other reanalysis products used in the article. However, changes in the ERA-40 data assimilation procedure produce unphysical jumps in atmospheric temperatures, which may be the likely reason for the elevated tropospheric warming trend in 1979–2002. NCEP/NCAR Reanalysis data show that the near-surface upward temperature trend over the same period is greater than the tropospheric trend, which is consistent with direct radiosonde observations and inconsistent with ERA-40 results. A change of sign in the winter temperature trend from negative to positive in the late 1980s is documented in the upper troposphere/lower stratosphere with a maximum over the Canadian Arctic, based on radiosonde data. This change from cooling to warming tendency is associated with weakening of the stratospheric polar vortex and shift of its center toward the Siberian coast and possibly can be explained by the changes in the dynamics of the Arctic Oscillation. This temporal pattern is consistent with multi-decadal variations of key arctic climate parameters like, for example, surface air temperature and oceanic freshwater content. Elucidating the mechanisms behind these changes will be critical to understanding the complex nature of high-latitude variability and its impact on global climate change.

1 Introduction

The vertical structure of the arctic atmosphere is shaped by a strongly negative surface radiation balance and poleward heat and moisture advection in the troposphere. On average, the Arctic north of 70°N lacks about 100 W/m2 in the radiation heat balance, resulting from the difference between the incoming and outgoing solar radiation at the top of the atmosphere (Nakamura and Oort 1988). The heat balance is maintained through meridional heat transport from lower latitudes, which varies between 85 W/m2 in summer and 111 W/m2 in autumn according to Nakamura and Oort (1988) and between 85 and 121 W/m2 according to Overland and Turet (1994). The vertical structure of this heat advection controls mean atmospheric lapse rate and therefore the vertical heat exchange processes that define the fraction of heat used to warm the near-surface atmospheric layers. In the summertime, the bulk of heat advected to the Arctic is spent on heating the surface. In wintertime, the advected heat can substantially warm the arctic atmosphere as has been suggested by radiative models (e.g. Overland and Guest 1992).

Vertical turbulent exchange in the arctic atmosphere is generally weak. A very common and well-known feature of the arctic atmosphere, especially during the cold season, is the frequent occurrence of near-surface air temperature inversions (Sverdrup 1933). The surface temperature does not correlate well with the tropospheric temperature during strong inversion events because the super-stable boundary layer is decoupled from the tropospheric circulation (Tjernstroem 2005). Surface HEat Budget of the Arctic Ocean (SHEBA) data show more frequent occurrence of cold than warm events in the surface temperature record (Uttal et al. 2002). This skewing of temperature behavior in the boundary layer is not well simulated by general circulation models (Beesley et al. 2000; Rinke et al. 2006; Byrkjedal et al. 2008). Weak sensitivity of boundary layer temperature to tropospheric advection could be one reason for biases in reanalysis data and model simulations during the cold season (Beesley et al. 2000; Tjernstroem et al. 2005; Rinke et al. 2006).

Understanding changes in the atmospheric lapse rate resulting from differential temperature trends at different heights is important for understanding the nature of tendencies in the Arctic environment. Indeed, the recent documented surface changes have been substantial (Serreze et al. 2000; Serreze and Francis 2006; Francis and Hunter 2007; Bekryaev et al. 2010); the recent surface temperature trend in the Arctic is about twice as large as the Northern Hemisphere trend (IPCC 2007). In the 1950s and 1960s, however, the temperature trends were negative and opposite to the global temperature trends (Polyakov et al. 2003; Johannessen et al. 2004). The spatial pattern of the arctic surface air temperature (SAT) trends is also very heterogeneous; moreover, the trends from different data sources are not necessarily coherent (Kuzmina et al. 2008).

The atmospheric temperature in the Arctic exhibits large natural variability on a wide range of time scales from synoptic to multi-decadal (e.g. Overpeck et al. 1997; Polyakov et al. 2003; Bengtsson et al. 2004a). Available instrumental temperature records (particularly from the free atmosphere) are not long enough to enable us to resolve slow processes. This is one reason for poor understanding of the mechanisms behind variability in the Arctic. There are, however, indications that the pattern of Arctic warming is partially controlled by the Arctic Oscillation (AO) and North-Atlantic Oscillation (NAO) (Thompson and Wallace 1998).

It has been demonstrated that the stratospheric circulation in the polar regions is closely linked to surface conditions. For example, the Eurasian snow cover extent in October controls, to some degree, the AO/NAO behavior during the following winter (Cohen and Entekhabi 1999; Cohen and Barlow 2005; Cohen and Fletcher 2007). The polar stratosphere responds to varying surface conditions, as in the example above; in addition, it can also be a driver of surface changes. For example, changes in the stratospheric circulation in the Southern Hemisphere due to declining ozone concentrations (Marshall et al. 2004) could be one explanation for the increase of sea-ice extent near Antarctica (Gillett et al. 2008). Various mechanisms for observed climate changes in the two hemispheres have been discussed in Gillett et al. (2008) and Turner et al. (2007); the important role of the stratospheric circulation is stressed. Understanding the reasons underlying changes in the arctic atmosphere is therefore important as it is a crucial component of the tightly-coupled Arctic climate system.

It has been argued that the surface warming in the Arctic should be preceded by an elevated atmospheric warming induced by lateral heat transport (Flannery 1984; Schneider et al. 1997; Alexeev 2003; Rodgers et al. 2003; Alexeev et al. 2005; Langen and Alexeev 2005, 2007). Using the European Center for Medium Range Weather Forecasts Re-Analysis data (ERA-40), Graversen et al. (2008) found such an elevated warming in the winter and summer temperature trends, which they argue was not strongly linked to poleward atmospheric heat transport. This elevated warming has been questioned (Bitz and Fu 2008; Grant et al. 2008; Thorne 2008) and shown to be most likely a result of changes in the ERA-40 data assimilation system. Detailed analysis of spatiotemporal patterns of arctic warming (including the reported elevated 1979–2002 warming in the arctic troposphere) is one of the purposes of this study. A possible explanation of those long-term changes in the arctic troposphere and stratosphere is another purpose of the article.

The plan of the article is the following. In section 2 we describe the data used for the study. In section 3 we study temperature trends for the 1979–2002 period from different reanalysis products. In sections 4 and 5 we test the robustness of the elevated ERA-40 warming against radiosonde data and other reanalysis products. The sensitivity of trends to changes in the time interval will be investigated. In section 6 we analyze the most recent (post-1990) temperature trends in the Arctic, in an attempt to diagnose possible reasons for those changes and to see if any of the station data support the elevated warming reported in ERA-40. This section is followed by the discussion/conclusions.

2 Data

2.1 Meteorological stations and reanalysis products

There is a variety of long-term data sources available for the arctic atmosphere. A reasonable network of Arctic coast and island stations provides routine land-based meteorological observations. Measurements made from Russian patrol ships (Kuzmina et al. 2008) represent another source of data. A vast array of data from meteorological stations is available at the National Snow and Ice Data Center (NSIDC) web site (http://www.nsidc.org). An increasingly large number of satellite data products for the arctic atmosphere have become available since the beginning of the satellite era in the 1970s.

However, no routine atmospheric observations were made over the Arctic Ocean before the beginning of the Russian North Pole drifting stations program in the mid-1930s (Kahl 1998). One to three of these stations were operating each year in the Arctic since 1950. This program was significantly downscaled in 1991. No radiosondes from the Russian North Pole drifting stations were launched after 1991 until 2007 (A.P.Makshtas, personal communication).

The North American Regional Reanalysis (NARR: Mesinger et al. 2006) represents an important source of data for various diagnostic and validation purposes. NARR uses a limited-area NCEP Eta model and data assimilation system. This model has high spatial and temporal resolution (32 × 45 km in space and 3-hourly output in time). A wide variety of available data of different origin was assimilated, especially over the continental United States. The immediately available data covers the period from 1979 to 2005. However, we chose not to use NARR because the article is dealing with the high Arctic and the locations analyzed here are either outside of the NARR domain or too close to the boundary of this reanalysis product.

The following datasets were used in this study.

IABP/POLES

is the International Arctic Buoy Program (IABP) dataset (Rigor et al. 2000; Chen et al. 2002). The Polar Science Center of the Applied Physics Laboratory, University of Washington, in collaboration with IABP participants, has maintained a network of drifting Argos buoys in the Arctic Ocean since 1979. The dataset used in this study covers 1979–2004. The data were obtained from http://iabp.apl.washington.edu.

IGRA dataset.

For upper air profiles we use data from the International Global Rawinsonde Archive (IGRA) dataset (Gaffen 1996). This dataset contains most existing arctic radiosonde data, including both a daily and monthly mean archive. The data coverage varies in time by region and country. Most observations start at the surface and go as high as 20 mb, especially in recent decades. The quality of radiosonde data is compromised by a variety of problems, including inhomogeneity of observations and processing problems (Gandin 1988; Schwartz and Doswell 1991; Gaffen 1994). In general, quality assurance procedures for sounding data rely on principles of internal consistency, basic physical relationships, and/or statistical methods which are illuminated in Collins (2001) and literature cited therein. All the soundings are processed with quality controls (http://www.ncdc.noaa.gov/oa/climate/igra/index.php). A total of 113 IGRA stations are located north of 60°N. Many stations in the Russian Arctic stopped launching radiosondes by the mid-1990s; therefore number of stations with reasonable continuous coverage over long enough period is limited. A list of stations from IGRA used for this study and criteria for their selection are given below in section 2.3.

ERA-40.

Is the second-generation reanalysis dataset (Uppala et al. 2005). The ERA-40 assimilation procedure was significantly improved starting in about 1979 (Bengtsson et al. 2004a, b; Uppala et al. 2005). We use the 2.5 × 2.5 gridded ERA-40 dataset available from the European Center for Medium-range Weather Forecasts (ECMWF) website, which we find acceptable for the purposes of this study.

NCEP Reanalyses.

We use both the older National Center for Environmental Protection (NCEP) reanalysis product (the so-called NCEP-1, Kalnay et al. (1996)) and the newer NCEP-Department of Energy (DOE) Atmospheric Model Intercomparison Project (AMIP)-II Reanalysis (we will call it NCEP-2, for brevity), described in Kanamitsu et al. (2002). As noted in Kanamitsu et al. (2002), the newer product can be used “…as a supplement to the NCEP-National Center for Atmospheric Research (NCAR) reanalysis especially where the original analysis has problems. The differences between the two analyses also provide a measure of uncertainty in current analyses.” Data from both products are available to the present day, which makes them very useful for comparison with the observations.

JRA-25.

The Japanese 25-year ReAnalysis (JRA-25) was conducted by the Japanese Meteorological Agency (JMA) in collaboration with the Central Research Institute of Electric Power Industry. The available dataset covers period from 1979 to 2004. The data assimilation was done using 3DVar. The global model’s resolution used for the reanalysis was T106 (Onogi et al. 2007).

2.2 Problems with reanalysis data

Major data assimilation systems use a large variety of data in one form or another. However, data coverage is not uniform in space and time. Discontinuities in observational systems can potentially impact the quality of reanalysis data. Bromwich and Wang (2005) argued, for example, that accuracy of reanalysis data may suffer over areas with sparse observations.

Bengtsson et al. (2004a, b) questioned the quality of trends computed from the ERA-40 reanalysis for the period 1958–2001 in the context of changes to the global observing system. The ERA-40 global mean temperature in the lower troposphere has a trend of +0.11 K dec−1 over the period of 1979–2001, which is slightly higher than the microwave sounding unit (MSU) measurements, but within the estimated error limit. For the period 1958–2001, however, the warming trend was larger (0.14 K dec−1), but Bengtsson et al. (2004a, b) found this increase to be an artifact of changes in the observing system. When these corrections are introduced, the warming trend is reduced to 0.10 K per decade.

Simmons et al. (2004) compared monthly-mean anomalies in SAT from the ERA-40 and NCEP/NCAR reanalyses with corresponding values from the Climate Research Unit (CRU) dataset CRUTEM2v (Jones and Moberg 2003). Least-square linear trends were found to be significantly lower for both reanalysis projects, but ERA-40 trends are within 10% of CRU for the whole Northern Hemisphere when computed from 1979 onwards. There is, however, a warm model bias present at middle and high latitudes and a cold bias at low latitudes. The ECMWF model (the basis for ERA-40) produces a cold bias at mid- and high-troposphere in data-sparse regions. This feature may amplify the troposphere temperature trends in ERA-40 because the Arctic data coverage has varied between 1979 and 2002. Trends and variability in ERA-40 throughout the planetary boundary layer (1,000 mb to 850 mb layer) are generally similar to those at the surface from the late 1970s onwards.

Bromwich and Wang (2005) pointed out that some of the ERA-40 tropospheric cold bias was introduced by the satellite instrument High-resolution Infra-Red Sounder (HIRS)-2 on the Television Infra-Red Observation Satellite (TIROS) Operational Vertical Sounder (TOVS). A quote from the ECMWF website (http://www.ecmwf.int/research/era/ERA-40/Data_Services/section3.html): “ … A further problem of concern is cold bias in the lower troposphere (below ~500 hPa) over ice-covered oceans in both the Arctic and the Antarctic. A related problem in Arctic precipitation has also been identified. These polar cold biases arise from the assimilation of HIRS radiances. Changes to the thinning, channel-selection and quality control of the infrared data that were introduced for analyses from 1997 onwards to reduce the tropical precipitation bias have also virtually eliminated the cold polar biases.” TOVS satellite input to ERA-40 started in 1979 and was replaced by ATOVS in 1998. Its replacement corrected the negative tropospheric bias.

Trenberth et al. (2001) showed that while the MSU and NCEP/NCAR reanalysis temperatures show fairly good agreement, large discrepancies with the ERA-40 temperatures indicate that changes in the satellite observing system may have adversely affected the ECMWF reanalyses, especially in the tropics. The temperature discrepancies have a complex vertical structure that is not fully understood. Changes in the observing system limit the applicability of the reanalysis products to some climate studies.

Despite the problems, Bromwich and Wang (2005) and Bromwich et al. (2007) find NCEP/NCAR and ERA-40 reanalysis reliable to some extent in the Arctic during the modern satellite era (post 1979). According to these studies, one of the most significant problems is treatment of clouds and the associated radiation budget, which produces excessively strong short-wave radiation over land and therefore surface temperatures that are too high.

For the discussion throughout this article, it is important to keep in mind that reanalysis products are essentially the result of sophisticated data interpolation procedures with dynamical constraints. In the areas scarcely covered by direct measurements of temperature, satellite information acquires relatively large weight so that any change in satellite instruments can have an implication for the quality of assimilated air temperature. Because of the possible strong decoupling between the surface and the free atmosphere, potential errors introduced at the surface or aloft sometimes cannot be corrected by the assimilation system. Vertical temperature profiles in the Arctic are almost never nearly adiabatic.

2.3 Radiosonde station data used in this study

To test reanalyses data quality we selected ten arctic stations with data coverage, some starting as early as the 1950s. We included stations with relatively good data coverage. Many Russian stations stopped launching radiosondes after the break-up of the Soviet Union in 1991, which explains why we did not include more stations from the Russian Arctic. Data coverage and station locations are listed in Table 1. Geographical locations are plotted in Fig. 1—the stations are spread all over the Arctic in a more or less uniform manner. We decided to keep Jan Mayen station since it is located above 70°N although it is often viewed as a North Atlantic station rather than as an Arctic station.
Table 1

Station list. Shaded areas represent approximate periods of data availability

Fig. 1

Stations used for the analysis

3 Temperature trends from reanalysis datasets

Reanalysis datasets are used here in order to test the existence of the elevated high-latitude warming pattern found in ERA-40 (Graversen et al. 2008). We demonstrate that all three reanalysis datasets, ERA-40, JRA-25 and NCEP-1, show substantially different patterns of high Arctic warming during 1979–2002 for all four seasons (Fig. 2). We used an algorithm described in Wigley (2006) for assessing trends and their significance (direct link to the exact description of the algorithm is found here: http://www.climatescience.gov/Library/sap/sap1-1/finalreport/sap1-1-final-appA.pdf).
Fig. 2

Linear trend of zonal mean air temperature (K/decade) as a function of height (hPa) and latitude calculated over 1979-2001 period for four seasons calculated from ERA-40, JRA-25 and NCEP-1 reanalysis datasets. Only significant trends are shown

The 1,000 mb temperature trend in ERA-40 is lower approximately north of 70–80°N for all seasons except spring compared with that from NCEP-1. There is a pronounced maximum in ERA-40 and less so in JRA-25 temperature trend at 925 mb in the winter and at 700 mb in the summer. Autumn trends are dramatically different as well; ERA-40 and JRA-25 do not show any maximum in trend near the surface, unlike the corresponding trend from NCEP/NCAR reanalysis. In all cases the largest trend maxima in ERA-40 temperatures are found in the area north of 80°N, corresponding to the Arctic Ocean. Note that no regular observations are available for that area.

Spatially averaged (70–90°N) winter temperature difference between ERA-40 and NCEP-1 is shown in Fig. 3a. Temperatures derived from these two datasets differ substantially at the end of the ERA-40 period, starting approximately after 1980 until 1998. Fig. 3b and 3c show the same difference at two different levels—925 mb and 600 mb, for ERA-40 and JRA-25.
Fig. 3

(a) – Winter air temperature (degrees K) averaged over 70-90N, difference between ERA-40 and NCEP-1; (b) – the same difference between ERA-40 and NCEP-1 at 925mb (black open circles), JRA-25 and NCEP-1 (green circles); (c) – same as in (b), except for the values were plotted at 600mb

The horizontal structure of the trends from various datasets at different levels is shown in Figs. 4 and 5. The summer trend in ERA-40 at 700 mb is significantly greater than the surface trend (Fig. 4), while the ERA-40 winter trend has a maximum at 925 mb (Fig. 5). These ERA-40-based elevated patterns of warming are among the major findings of Graversen et al. (2008). However, these two features are not present in trends calculated from other reanalysis products. NCEP-1, NCEP-2, and JRA-25 trends presented in Figs. 4 and 5 do not show a stronger elevated warming in the winter or the summer. There is only one place where a weak elevated warming can be seen, which is in the NCEP-1 summer trend at 925 mb (central panel in Fig. 4, 3rd row). Note a disagreement between the datasets even at the surface; e.g. much stronger winter warming in the Canadian Archipelago is seen in NCEP-1 than in the other datasets (Fig. 5, column 1, row 3). The Beaufort and East Siberian seas are two other areas where reanalysis products disagree at the surface.
Fig. 4

Summer temperature trends (1979-2002), K/decade for ERA-40, NCEP-1, NCEP-2, and JRA-25 at different pressure levels: 1000mb, 925mb, and 700mb

Fig. 5

Winter temperature trends (1979-2002), K/decade for ERA-40, NCEP-1, NCEP-2, and JRA-25 at different pressure levels: 1000mb, 925mb, and 850mb

Geographically, the area of the biggest disagreement between tropospheric trends from ERA-40 and other datasets (JRA-25, NCEP-1 and NCEP-2) is located in the high Arctic. This disagreement could potentially be tested at several high-latitude stations. However, Ny Alesund lacks a sufficiently long radiosonde record. The Franz Joseph Land and Severnaya Zemlya stations stopped launching sondes in the early 1990s. Other stations in the Canadian Archipelago with long radiosonde records (e.g. Eureka, Thule) are located outside the area of interest. The only station close to the area of interest is Alert, where observations started as early as 1963 and have continued until present. We will test the robustness of reanalysis temperature trends against observations at Alert in the next section.

4 Temperature trends from reanalysis and radiosondes at Alert station

Alert station is particularly suitable for assessment of reanalysis products due to its proximity to the high Arctic and its location in an area of persistent sea-ice cover that significantly reduces spatio-temporal inhomogeneities.

Because 1997 is identified as a problem year for the ERA-40 dataset, we test reanalysis products against Alert station data for two time windows. The first period covers 1979–2002, which represents our ‘standard’ ERA-40 period. The second period covers an earlier period of 1976–1997. Note that the two periods do not have the same time length. We chose the second period for two reasons. By ending it in 1997 we avoid problems related to changes in the ERA-40 data assimilation system, while by starting in 1976 we avoid problems associated with the abrupt changes in the Arctic that occurred around that time due to the shift in the Pacific Decadal Oscillation (PDO) index. The PDO index was predominantly negative before 1976. The associated changes in the large-scale circulation have been shown to affect many climate parameters in the Arctic (Mantua et al. 1997; Hartmann and Wendler 2005). In particular, the SAT trend calculated for 1951–2001 in Fairbanks is positive. However, this overall positive trend is “… strongly biased by the sudden shift in 1976 from the cooler regime to a warmer regime. When analyzing the total time period from 1951 to 2001, warming is observed; however, the 25-yr period trend analyses before 1976 (1951–75) and thereafter (1977–2001) both display cooling, with a few exceptions”, according to Hartmann and Wendler (2005).

Winter trends for the two periods derived from the two reanalysis products ERA-40 and NCEP-1 are shown in Fig. 6. For the first period the vertical trend profiles show substantial differences; the NCEP-1 trend is the closest to the trend calculated using radiosonde data, while the ERA-40 trend overestimates the observed trend. A typical value for the standard error of the calculated trends at 700 mb is about 0.3–0.4 K/decade. However, for the second, earlier, period the trend profiles from the two reanalysis products are very similar and, more importantly, they are both much closer to the radiosonde trends. ERA-40 temperatures are colder before 1997 and a big jump occurs in 1998, which is the primary reason for the strong warming trend. This is also true for other seasons (not shown). Our analysis for spring, summer, and autumn showed that ERA-40 systematically overestimates lower- and mid-tropospheric trends for 1979–2002 while trends calculated for 1976–1997 for both ERA-40 and NCEP-1 are very close to IGRA radiosonde data. Unfortunately, NCEP-2 and JRA-25 data are not available before 1979.
Fig. 6

(a) Winter temperature trend at Alert for 1979-2002 (K/decade); (b) trend for 1976-1997; (c) RMS error calculated using monthly means for 1979-2002, all months (K): ERA-40, open circles; NCEP-1, green circles; radiosonde IGRA data, magenta triangles and significance estimate represented by one standard error. (d) Difference between ERA-40 and IGRA (black circles) and difference between NCEP-1 and IGRA (green circles) at 700mb, degrees K

Based on this analysis, we argue that the significantly closer agreement between two reanalysis products for 1976–1997 compared to 1979–2002 suggests that ERA-40 temperature trends for the latter period might not be very robust and could therefore be considered as an artifact of changes in the data assimilation system. Therefore, trends derived from this product (ERA-40) should be treated with caution. However, the root mean square errors calculated using monthly mean temperatures for all months from January 1979 through September 2002 demonstrate that ERA-40 is generally better than NCEP-1 (Fig. 6c) or, actually, all other reanalysis products used here (not shown) in terms of deviation from the observations. This indicates that ERA-40 better captures seasonal-scale variability in the Arctic; it better captures synoptic variability as well, as shown in Graversen et al. (2008).

5 Lower stratospheric temperature trends

This section is devoted to analysis of Arctic upper-tropospheric/lower-stratospheric temperature trends. We use NCEP-1 and radiosonde data wherever available to verify the reanalysis results. It is known (Ramaswamy et al. 2006; IPCC, 2007) that the global stratosphere has been cooling for the last couple of decades. However, Fig. 7 shows that the temperature trend in the lower stratosphere at Alert changes its sign from cooling to warming when we shift the time window from 1976 to 1997 to 1979–2002. Here we present a closer look at this phenomenon, without explaining the physics behind this trend in much detail.
Fig. 7

(Left): Temperature anomalies for different stations at 100mb from radiosonde dataset IGRA; straight lines show approximate trends before and after 1990. Green circles in the lower left panel show the principal component of the first EOF of the sea level pressure (multiplied by 3 to match the scale) calculated using seasonal means (DJF) for 1949-2008. Values near each station’s names indicate correlation of air temperatures with the principal component of the first EOF of sea level pressure. (Right): Temperature trends (black thin lines) as a function of height with error estimates for the same stations calculated for the period before 1990 (grey shaded area) and from 1990 to 2008 (green shaded area)

Shown in Fig. 7 are the stations’ radiosonde winter temperatures at 100 mb as a function of year and temperature trends as a function of height for periods before and after 1990. There is a visible, prolonged minimum in the temperature at 100 mb in the late 1980s-early 1990s. This temperature minimum explains our choice for the break point between the two analysis periods. The right panels show that at most stations the temperature trends at around 100 mb reversed from one period to the next, with strongest change observed at the stations located in the Canadian Arctic: Alert, Eureka, Resolute Bay, and Thule. Figure 1 from Ramaswamy et al. (2006) also shows a similar result.

A possible mechanism for those trends would be a change in the atmospheric circulation. The lower most panel of Fig. 7 shows air temperature variations at 100 mb in Tiksi along with ‘AO-like’ index. We calculated EOFs of seasonal mean sea level pressure fields (December-January-March) and used the first EOF’s principal component (PC) as our ‘AO’ index. The spatial structure of this EOF is very similar to the conventional Thompson and Wallace (1998) picture of AO (not shown here). Indeed, the overall strength of the polar vortex at 100 mb height weakened after 2000 compared to that prior to 1980, and the center of the vortex shifted toward the Siberian Arctic (Fig. 8). This shift resulted in the position of the temperature minimum that moved towards Siberia, and in pronounced warming over the Canadian Arctic. The isolines of the 100 mb geopotential height serve as a good proxy for the streamfunction of the flow. The weakening of vortex strength also resulted in the overall warming of the lower stratosphere almost everywhere throughout the Arctic. These changes of the 100 mb temperature trend from cooling before 1990 to warming after 1990 are also seen in the radiosonde IGRA data (Fig. 8e, f) for various locations in the Arctic. Figure 9a, b shows first EOFs of 100 mb seasonal mean geopotential height and air temperature. The structure and location of changes in the geopotential height and air temperature shown in Fig. 8 are very similar to the EOFs of the corresponding fields. On top of that, their principal components highly correlate to our ‘AO-like’ index (Fig. 9c). For example, the geopotential height- and air temperature’s PCs correlation coefficient is 0.92, the geopotential heights’- and AO PCs correlate at 0.71, the air temperature and AO PCs correlate at 0.56.
Fig. 8

(a) Winter air temperature at 100mb (degrees C) calculated from NCEP-1 for 1980-90 (color contour lines) and 2000-2008 (black lines); (b) Geopotential height at 100mb (km) calculated for 1980-90 (color contour lines) and 2000-2008 (black lines); (c) Difference between winter air temperatures (degrees C) at 100mb calculated for 2000-2008 minus 1980-1990; (d) Same as in (c), except for geopotential height (km); (e) Winter air temperature trend at 100mb (degrees K/decade) before 1990; (f) Same as in (e), except for trend after 1990. Blue and red circles show corresponding temperature trends calculated from IGRA dataset at stations used in the analysis

Fig. 9

(a) – First EOF of NCEP-1 winter (DJF) air temperature at 100 mb over the 1949-2008 period; (b) – same as in (a) except for the 100 mb geopotential height; (c) – principal components of the 1st EOF of the air temperature from (a), black line; principal component of the 1st EOF of geopotential height at 100 mb from (b), green line; and principal component of the 1st EOF of the sea level pressure, red line

Therefore we conclude that most of the variability and long-term changes in atmospheric air temperature can be explained by changes in the dynamics of the Arctic Oscillation.

6 Structure of trends since 1990

6.1 Surface trends

We now estimate robustness of the most recent trends in the Arctic atmosphere from NCEP-1, NCEP-2, and IABP/POLES. As mentioned earlier, there is some disagreement between these datasets. For example, Figure 6 shows that both NCEP and ERA-40 root mean square errors have a big maximum at the surface for Alert station. The near surface (1,000 mb) 1979–2002 winter temperature trends in the Beaufort Sea are unexpectedly negative according to NCEP-2 and positive according to NCEP-1. Also, the reanalysis datasets disagree on the magnitude of the trends. This motivated us to compare NCEP products with IABP/POLES data for the period of 1990–2004. The choice of the shorter time period is explained by the availability of the IABP/POLES data. For the sake of simplicity we chose to omit the JRA-25 SAT trends, because as we will see from the following the datasets demonstrate substantial disagreement so that adding one more product will not change this result.

Figure 10 shows the SAT trend from two reanalysis products, NCEP-1 and NCEP-2, and the IABP/POLES dataset. The three sources of data provide a generally consistent picture for all seasons, although geographically the differences are quite substantial. A pattern of winter warming (Fig. 10, upper row) is captured by all three datasets, with a positive warming trend located on the American side of the Arctic and a slightly negative trend on the Siberian side. The maximum differences are located in the Beaufort, Chukchi, and East Siberian seas. There are disagreements in the Canadian Arctic, not only on the magnitude of the trends, but also on the positions of the maxima. Areas of warming are more widespread in NCEP-1 and IABP/POLES datasets in all seasons compared to NCEP-2. NCEP-2 trends (Figs. 5 and 10) have more areas with negative temperature changes than NCEP-1, ERA-40, and IABP/POLES temperature trends.
Fig. 10

Temperature trends at 2 meters for different datasets and seasons (K/decade) calculated for 1990-2004. Only areas with significant trends are plotted

One of the centers of disagreement is in the Beaufort Sea. Comparison of SAT trends for the Beaufort Sea limited to 170°W–130°W, 72°N–81°N from NCEP and IABP/POLES datasets shows that NCEP-1 overestimates the rate of warming (Fig. 11). NCEP-2 tends to produce warmer SATs for the time period around 1990 and then to converge with IABP/POLES data towards 2004, which results in negative trends for both, contradicting the trend from the IABP/POLES dataset. The IABP/POLES winter trend averaged over the Beaufort Sea is positive, although it is weaker than the NCEP-1 trend; this results from a stronger negative bias in NCEP-1 SAT compared to IABP/POLES data. One interesting observation from Fig. 11 can be made: all datasets show wider spread during 1980–90 compared to the latest decade, which could be an indicator of improving data coverage in the area.
Fig. 11

Winter surface air temperature (degrees C) averaged over the Beaufort Sea (170W-130W, 72N-81N) using NCEP-1 (black line), NCEP-2 (green line), JRA-25 (purple line), and IABP/POLES data (red line with circles)

6.2 Lower stratospheric warming

The reported earlier winter warming trend in the lower stratosphere (Fig. 6) is confirmed by data from most of the stations used in our analysis (Fig. 12). It is most pronounced in the Canadian Arctic (Alert, Eureka, Resolute Bay, Thule) with values reaching as high as 5 K/decade at Alert. The Russian Arctic (Dikson, Tiksi) and the North Atlantic (Bjornoya, Jan Mayen) show significant positive winter temperature trends as well. Fairbanks and Barrow station data show a weak lower tropospheric warming trend in the spring and no warming signal in the winter. The lower stratospheric warming trend extends into the spring at some of the stations (Alert, Eureka, Resolute Bay, Dikson, Tiksi); this could be explained by the same mechanism associated with the change in the lower stratospheric circulation.
Fig. 12

Temperature trends throughout the year at different stations (K/decade) for 1991-2008. NCEP-1: black thin line; radiosonde IGRA data: shaded area having width of two standard errors, centered around the trend (not shown)

The Arctic warming has been accelerating since 1990, therefore it would be natural to look for faster elevated warming in the latest data, assuming that this air is coming from the warmer lower latitudes. However, the only station showing faster elevated warming in the lower troposphere is Tiksi in the winter; Tiksi data also show some hints of faster warming in the summer (Fig. 12). The NCEP-2 results (not shown here) are similar to the NCEP-1-based results. Note that the NCEP-calculated trends are in a reasonable agreement with the station data for all seasons (Fig. 12).

7 Discussion /conclusions

Recent temperature increases in the Arctic are larger than elsewhere. This is a matter of great concern due to the impact the rising temperatures can have on the Arctic and global climate systems. The importance of the consequences brings about hot debates concerning the spatio-temporal structure of the changes in the Arctic and the mechanisms driving these changes. One of the main topics of the debate is whether the Arctic warming is primarily local in nature, or is induced by changes in global circulation patterns.

Robustness of the recently-reported elevated tropospheric warming trend found in ERA-40 in the Arctic (Graversen et al. 2008) was tested using NCEP/NCAR reanalysis and JRA-25 datasets as well as the radiosonde data archive IGRA. This trend has been questioned in a number of studies. Thorne (2008) compared ERA-40 with the zonal mean radiosonde-based HadAT2 (Thorne et al. 2005) dataset and satellite retrievals and concluded that lack of observations could be the primary reason for non-robustness of ERA-40 trends poleward of 80°N. Grant et al. (2008) compared ERA-40 trends with trends calculated from a subset of arctic radiosonde data and came to a similar conclusion. Using the HadCRUtv3 dataset and satellite data, Bitz and Fu (2008) showed that the Arctic mid- and lower troposphere winter warming in ERA-40 has been greatly overestimated. Serreze et al. (2009) came to a similar conclusion about surface-based polar amplification using the NCEP/NCAR reanalysis dataset and JRA-25.

This study tests the robustness of the reported ERA-40 trends based on other reanalysis products (JRA-25, NCEP-1 and NCEP-2) and detailed analysis of radiosonde data from one particular station (Alert). Our analysis suggests that the ERA-40 elevated warming trend calculated for 1979–2002 could be an artifact of changes in the data assimilation system. According to the ECMWF documentation, 1997–98 were problem years for ERA-40 because a change in the satellite input resulted in a cold tropospheric bias before 1998. Our analysis of ERA-40 data indicates that in 1998 the temperature in the lower troposphere experienced an unphysical jump. Faster elevated warming in the atmosphere is not confirmed by other reanalysis products, or by the radiosonde dataset IGRA. The arctic warming accelerated even more after 2002; therefore, we extended our analysis to 2007. However, the most recent trends (for 1990–2007) do not exhibit a pattern of faster elevated warming in the lower or middle troposphere. Instead, we found that significant changes are occurring in the lower stratosphere.

Main results of this study can be formulated as follows:
  1. 1.

    Reanalysis products disagree with each other over trends for 1979–2002. ERA-40 shows faster elevated warming in the central Arctic, unlike any of the NCEP/NCAR products. NCEP/NCAR trends tend to be in better agreement with trends calculated from the radiosonde IGRA dataset than trends calculated from ERA-40, especially in the free atmosphere. ERA-40 shows a consistently warmer trend in the low- and mid-troposphere.

     
  2. 2.

    The extent of disagreement between trends depends on the time period chosen for the analysis. During the earlier period (1976–1997) the disagreement between trends derived from reanalysis products and the IGRA dataset is smaller, which indicates that the last few years of the ERA-40 data period are contaminated by a spurious trend in the temperature, which is not found in other products. This jump in the temperature around 1997 has been discussed in Bromwich and Wang (2005) and documented on the ECMWF website (http://www.ecmwf.int) and has been identified as due to the change in satellite instrument from HIRS to ATOVS.

     
  3. 3.

    The uncertainty in temperature trends is too great to make any conclusive statements about the faster elevated warming in the lower troposphere in the Arctic during the last two decades. The only station showing elevated warming similar to the warming described by Graversen et al. (2008) is Tiksi. All other stations used in our analysis do not show any indication of faster elevated warming in the troposphere in any season.

     
  4. 4.

    Disagreement in temperature trends between the datasets used for the analysis is substantial even at the surface. All the “hotspots” of disagreement are in regions with sparse data coverage. There is a major disagreement between reanalysis products and IABP with regard to the trend at the surface for 1990–2004. The recent winter warming signal over the Beaufort Sea is statistically significant according to IABP/POLES and NCEP-1. However, other reanalysis products disagree substantially over trend magnitude and even sign: for example, NCEP-2 results show a significant negative SAT trend over the Beaufort Sea.

     
  5. 5.

    Our analysis of radiosonde data from the IGRA dataset revealed a change in the temperature trend in the lower stratosphere (200 mb to 70 mb) around 1990 when the trend changed its sign from negative to positive. This signal is robust to a varying degree throughout the array of available stations in the Arctic with sufficiently long temperature records. This lower stratospheric warming signal is most pronounced in the Canadian Arctic.

     
  6. 6.

    This pattern of temporal changes may be associated with multi-decadal fluctuations on time scales of 50–80 years, which are known to be exceptionally strong in the Arctic and North Atlantic. Polyakov et al. (2008) demonstrated a strikingly coherent pattern of long-term variations of the key arctic climate parameters and strong coupling of long-term changes in the Arctic climate system with those at lower latitudes. Remarkably coherent low-frequency variations are expressed by the arctic SAT, Arctic Ocean fresh water content and intermediate Atlantic Water core temperature, fast-ice thickness, and North Atlantic sea surface temperature. For example, associated with this variability, the arctic SAT record shows two warmer periods in the 1930–40s and in recent decades, and two colder periods early in the 20th century and in the 1960–70s. The observed stratospheric air temperature variations are consistent with this pattern. The long-term changes in the upper troposphere/ lower stratosphere seem to occur together with changes at the surface, including the extent of Eurasian snow cover and sea ice. Elucidating the mechanisms behind these relationships will be critical to our understanding of the complex nature of low-frequency variability found in the Arctic and at lower latitudes, and its impact on climate change.

     

Notes

Acknowledgements

The study was supported by the National Science Foundation grants ARC 0909525, ARC 0652838 (VA, IP), National Oceanographic and Atmospheric Administration, Japan Agency for Marine-Earth Science and Technology (VA, IP), and the University of Alaska Fairbanks (SB). IE and SS were supported by the Norwegian Research Council projects PAACSIZ 178908/S30 “Planetary Boundary Layer Feedbacks Affecting the Polar Amplification of Arctic Climate Change in Seasonal Ice Zone”, POCAHONTAS 178345/S30 “Polar Climate and Heat Impact on the Arctic Shelves”, and NORCLIM 178245/S30 “Norwegian Climate Assessment”. ECMWF ERA-40 data used in this project were provided by ECMWF and obtained from the ECMWF data server. NCEP data used in this project were provided by NCEP and obtained from the NCEP data server. The JRA-25 data were obtained from the JRA-25 website: http://jra.kishou.go.jp/JRA-25/index_en.html. The authors thank David Bromwich for useful discussions.

References

  1. Alexeev VA (2003) Sensitivity to CO2 doubling of an atmospheric GCM coupled to an oceanic mixed layer: A linear analysis. Clim Dyn 20:775–787Google Scholar
  2. Alexeev VA, Langen PL, Bates JR (2005) Polar amplification of surface warming on an aquaplanet in “ghost forcing” experiments without sea ice feedbacks. Clim Dyn. doi:10.1007/s00382-005-0018-3
  3. Beesley JA et al (2000) A comparison of cloud and boundary layer variables in the ECMWF forecast model with observations at Surface Heat Budget of the Arctic Ocean (SHEBA) ice camp. JGR-atmospheres 105(10):12337–12349CrossRefGoogle Scholar
  4. Bekryaev RV, Polyakov IV, Alexeev VA (2010) Role of polar amplification in long-term surface air temperature variations and modern arctic warming. J Climate 23(14):3888–3906Google Scholar
  5. Bengtsson L, Hagemann S, Hodges KI (2004a) Can climate trends be computed from reanalysis data? JGR-atmospheres 109. doi:10.2001/2004JD00
  6. Bengtsson L, Hodges KI, Hagemann S (2004b) Sensitivity of the ERA40 reanalysis to the observing system: Determination of the global atmospheric circulation from reduced observations. Tellus A 56:456–471CrossRefGoogle Scholar
  7. Bitz CM, Fu Q (2008) Arctic warming aloft is dataset dependent. Nature 455:E3–E4CrossRefGoogle Scholar
  8. Bromwich DH, Wang S-H (2005) Evaluation of the NCEP–NCAR and ECMWF 15- and 40-Yr reanalyses using rawinsonde data from two independent arctic field experiments. Mon Weather Rev 133:3562–3578CrossRefGoogle Scholar
  9. Bromwich DH, Fogt RL, Hodges KI, Walsh JE (2007) A tropospheric assessment of the ERA-40, NCEP, and JRA-25 global reanalyses in the polar regions. J Geophys Res 112. doi:10.1029/2006JD007859
  10. Byrkjedal O, Esau I, Kvamstoe N-G (2008) Sensitivity of simulated wintertime Arctic atmosphere to vertical resolution in the ARPEGE/IFS model. Clim Dyn 30:687–701CrossRefGoogle Scholar
  11. Chen Y, Francis JA, Miller JR (2002) Surface temperature of the Arctic: Comparison of TOVS satellite retrievals with surface observations. J Climate 15:3698–3708CrossRefGoogle Scholar
  12. Cohen J, Entekhabi D (1999) Eurasian snow cover variability and Northern Hemisphere climate predictability. Geophys Res Lett 26:345–348CrossRefGoogle Scholar
  13. Cohen J, Barlow M (2005) The NAO, the AO, and global warming: How closely related? J Climate 18:4498–4513CrossRefGoogle Scholar
  14. Cohen J, Fletcher C (2007) Improved skill for Northern Hemisphere winter surface temperature predictions based on land-atmosphere fall anomalies. J Climate 20:4118–4132CrossRefGoogle Scholar
  15. Collins WG (2001) The operational complex quality control of radiosonde heights and temperatures at the national centers for environmental prediction. Part I: Description of the method. J Applied Meteor 40:137–151CrossRefGoogle Scholar
  16. Flannery BP (1984) Energy-balance models incorporating transport of thermal and latent energy. J Atm Sci 41:414–421CrossRefGoogle Scholar
  17. Francis JA, Hunter E (2007) Changing fabric of the arctic blanket, Environmental Research Letters, 2, Article 045011Google Scholar
  18. Gaffen DJ (1994) Temporal inhomogeneities in radiosonde temperature records. J Geophys Res 99:3667–3676CrossRefGoogle Scholar
  19. Gaffen DJ (1996) A digitized metadataset of global upper-air station histories, NOAA Technical Memorandum ERL ARL-211Google Scholar
  20. Gandin LS (1988) Complex quality control of meteorological observations. Mon Wea Rev 116:1137–1156CrossRefGoogle Scholar
  21. Gillett N et al (2008) The Arctic and Antarctic: Two faces of climate change. Eos 89:177–178CrossRefGoogle Scholar
  22. Grant AN, Broennimann S, Haimberger L (2008) Recent Arctic warming structure contested. Nature 455:E2–E3CrossRefGoogle Scholar
  23. Graversen RG et al (2008) Vertical structure of recent Arctic warming. Nature 451:53–56CrossRefGoogle Scholar
  24. Hartmann B, Wendler G (2005) On the significance of the 1976 Pacific climate shift in the climatology of Alaska. J Clim 18:4824–4839CrossRefGoogle Scholar
  25. IPCC (Intergovernmental Panel for Climate Change) (2007) Fourth assessment report. The physical sciences basis. Contribution of working group I to the fourth assessment report of the IPCC, Cambridge University Press, ISBN 978-0-521-88009-1, 996pGoogle Scholar
  26. Johannessen OM et al (2004) Arctic climate change: Observed and modelled temperature and sea-ice variability. Tellus A 56:328–341CrossRefGoogle Scholar
  27. Jones PD, Moberg A (2003) Hemispheric and large scale surface air temperature variations: An extensive revision and an update to 2001. J Clim 16:206–223CrossRefGoogle Scholar
  28. Kahl J (1998) Daily Arctic Ocean rawinsonde data from Soviet drifting ice stations. National Snow and Ice Data Center, digital media, Boulder, COGoogle Scholar
  29. Kalnay E et al (1996) The NCEP/NCAR 40-year reanalysis project. BAMS 77:437–471CrossRefGoogle Scholar
  30. Kanamitsu M et al (2002) NCEP–DOE AMIP-II reanalysis (R-2). BAMS 83:1631–1643CrossRefGoogle Scholar
  31. Kuzmina S et al (2008) High northern latitude surface air temperature: Comparison of existing data and creation of a new gridded dataset 1900–2000. Tellus A 60:289–304CrossRefGoogle Scholar
  32. Langen PL, Alexeev VA (2005) Analysis of 2xCO2 sensitivity in an aquaplanet GCM using fluctuation-dissipation theorem, GRL, L23708Google Scholar
  33. Langen PL, Alexeev VA (2007) Polar amplification as a preferred response in an aquaplanet GCM. Clim Dyn. doi:10.1007/s00382-006-0221-x
  34. Mantua NJ et al (1997) A Pacific interdecadal climate oscillation with impacts on salmon production. BAMS 78:1069–1079CrossRefGoogle Scholar
  35. Marshall GJ et al (2004) Causes of exceptional atmospheric circulation changes in the Southern Hemisphere. GRL 31:L14205. doi:10.1029/2004GL019952 CrossRefGoogle Scholar
  36. Mesinger F et al (2006) North American regional reanalysis. BAMS 87. doi:10.1175/BAMS-87-3-343
  37. Nakamura N, Oort AH (1988) Atmospheric heat budgets of the polar regions. J Geophys Res 93:9510–9524CrossRefGoogle Scholar
  38. Onogi K, Tsutsui J, Koide H, Sakamoto M, Kobayashi S, Hatsushika H, Matsumoto T, Yamazaki N, Kamahori H, Takahashi K, Kadokura S, Wada K, Kato K, Oyama R, Ose T, Mannoji N, Taira R (2007) The JRA-25 Reanalysis. J Meteor Soc Japan 85:369–432CrossRefGoogle Scholar
  39. Overland JE, Guest PS (1992) The Arctic snow and air-temperature budget over sea ice during winter. JGR Oceans 96:4651–4662CrossRefGoogle Scholar
  40. Overland JE, Turet P (1994) Variability of the atmospheric energy flux across 70°N computed from the GFDL data set. In: The Polar Oceans and Their Role in Shaping the Global Environment, Nansen Centennial Volume, Geophysical Monograph 85, Johannessen, O., R. Muench, and J. Overland (eds.), AGU pp 313–325Google Scholar
  41. Overpeck J et al (1997) Arctic environmental change of the last four centuries. Science 278:1251–1256CrossRefGoogle Scholar
  42. Polyakov IV et al (2003) Observationally based assessment of polar amplification of global warming. GRL 29:1878. doi:1029/2001GL011111 CrossRefGoogle Scholar
  43. Polyakov IV, Alexeev VA, Belchansky GI et al (2008) Arctic Ocean freshwater changes over the past 100 years and their causes. J of Climate 21:364–384CrossRefGoogle Scholar
  44. Ramaswamy V et al (2006) Anthropogenic and natural influences in the evolution of lower stratospheric cooling. Science 311:1138–1141CrossRefGoogle Scholar
  45. Rigor IG, Colony RL, Martin S (2000) Variations in surface air temperature observations in the Arctic, 1979–97. J Clim 13:896–914CrossRefGoogle Scholar
  46. Rinke A et al (2006) Evaluation of an ensemble of Arctic regional climate models: Spatial patterns and height profiles. Clim Dyn. doi:10.1007/s00382-005-0095-3
  47. Rodgers KB et al (2003) A tropical mechanism for Northern Hemisphere deglaciation. Geochem Geophys Geosys 4(5):1046. doi:10.1029/1003GC000508 CrossRefGoogle Scholar
  48. Schneider EK, Lindzen RS, Kirtman BP (1997) A tropical influence on the global climate. J Atmos Sci 54:1349–1358CrossRefGoogle Scholar
  49. Schwartz BE, Doswell CA III (1991) North American rawinsonde observations: Problems, concerns, and a call to action. Bull Amer Meteor Soc 72:1885–1896CrossRefGoogle Scholar
  50. Serreze MC et al (2000) Observational evidence of recent change in the northern high latitude environment. Clim Chang 46:159–207CrossRefGoogle Scholar
  51. Serreze MC, Francis JA (2006) The Arctic Amplification debate. Clim Change. doi:10.1007/s10584-005-9017-y
  52. Serreze MC et al (2009) The emergence of the surface-based Arctic amplification. The Cryosphere Discussions 2:601–622CrossRefGoogle Scholar
  53. Simmons AJ et al. (2004) Comparison of trends and low-frequency variability in CRU, ERA-40, and NCEP/NCAR analyses of surface air temperature, J. Geophys, Res., 109, D24115, doi:10.1029/2004JD005306
  54. Sverdrup HO (1933) The Norwegian North Polar expedition with the Maud, Vol. II, Meteorology, Geophysical Institute Bergen, pp 331Google Scholar
  55. Thompson DWJ, Wallace JM (1998) The Arctic Oscillation signature in the wintertime geopotential height and temperature fields. Geophys Res Lett 25:1297–1300CrossRefGoogle Scholar
  56. Thorne PW et al (2005) Revisiting radiosonde upper air temperatures from 1958 to 2002. J Geophys Res 110:D18105. doi:10.1029/2004JD005753 CrossRefGoogle Scholar
  57. Thorne PW (2008) Arctic tropospheric warming amplification? Nature 455:E1–E2CrossRefGoogle Scholar
  58. Tjernstroem M (2005) The summer Arctic boundary layer during the Arctic Ocean Experiment 2001 (AOE-2001). Boundary-Layer Meteorology 117:5–36CrossRefGoogle Scholar
  59. Tjernstroem M et al (2005) Modeling the Arctic Boundary Layer: An evaluation of six ARC­MIP regional-scale models with data from the SHEBA project. Boundary-Layer Meteorology 117:337–381CrossRefGoogle Scholar
  60. Trenberth KE et al (2001) Quality of reanalyses in the Tropics. J Clim 14:1499–1510CrossRefGoogle Scholar
  61. Turner J, Overland JE, Walsh JE (2007) An Arctic and Antarctic perspective on recent climate change. Int J of Climatology 27:277–293CrossRefGoogle Scholar
  62. Uppala SM et al (2005) The ERA-40 reanalysis, Quarterly. J of the Royal Met Society 61:3493–3512Google Scholar
  63. Uttal T et al (2002) Surface heat budget of the Arctic Ocean. BAMS 82:255–275CrossRefGoogle Scholar
  64. Wigley TML (2006) Statistical issues regarding trends, Appendix A from the US Climate change science program report “Temperature trends in the lower atmosphere: steps for understanding and reconciling differences” (http://www.climatescience.gov/Library/sap/sap1-1/finalreport/). Direct link to Appendix A from the report: http://www.climatescience.gov/Library/sap/sap1-1/finalreport/sap1-1-final-appA.pdf

Copyright information

© Springer Science+Business Media B.V. 2011

Authors and Affiliations

  • Vladimir A. Alexeev
    • 1
  • Igor Esau
    • 2
    • 3
  • Igor V. Polyakov
    • 1
  • Sarah J. Byam
    • 1
  • Svetlana Sorokina
    • 2
    • 3
  1. 1.International Arctic Research CenterUniversity of Alaska FairbanksFairbanksUSA
  2. 2.Thormoehlensgate 47, Nansen Environmental and Remote Sensing CenterBergenNorway
  3. 3.Allegaten 55, Bjerknes Centre for Climate ResearchBergenNorway

Personalised recommendations