1. Introduction
The Arctic is a region that experiences extreme conditions. The change from polar day to polar night over the year causes large annual amplitude in the amount of energy received from the sun. The change in surface properties and exchange of heat at the surface over the year is also very large. The open ocean has the ability to absorb most of the incoming solar radiation while it also, via turbulent transfer, is effective in releasing both sensible and latent heat. To a large extent the Arctic Ocean is presently covered with sea ice during all seasons, a surface that reflects solar light very efficiently and also exerts control of the near-surface temperature while freezing and melting. During most of the year, the free atmospheric air is warmer than the surface, thus setting up conditions for shallow boundary layers under strong inversions with restricted turbulent exchange as a consequence. The snow-covered surface is well isolated from the underlying surface (sea ice or land) from which it follows that rapid surface temperature changes follow if the net radiation conditions change. The Arctic is also to a very large extent cloud covered, often with low clouds that have a strong influence on the net surface energy conditions at the surface. Most of the year clouds act to heat the surface in the Arctic; only during a short period in summer may clouds have a net negative impact on the surface energy budget (e.g., Intrieri et al. 2002; Shupe and Intrieri 2004).
The Arctic climate change is observed to occur at a higher rate than the global change. It is also expected to continue to do so as projected with global climate models [GCMs; Arctic Climate Impact Assessment (ACIA) 2005; Solomon et al. 2007]. The across-model spread is, however, much larger in this region, which can be attributed both to more natural variability (Sorteberg et al. 2005) but also to large model differences in the process representation (Arctic Climate Impact Assessment 2005). There are a number of explanations for this rapid change presented in the literature (e.g., Arctic Climate Impact Assessment 2005; Manabe and Wetherald 1975; Graversen et al. 2008): the surface–albedo feedback, the more easily heated shallow boundary layers, changes in net meridional transport of heat, and moisture in the atmosphere or ocean, cloud feedbacks etc. To target the integrated effect, and for future projections, numerical models faithfully describing the processes involved are needed.
Several studies of various aspects of the present-day simulations in the Coupled Model Intercomparison Project phase 3 (CMIP3; Meehl et al. 2007) have been performed (e.g., Chapman and Walsh 2007; Eisenman et al. 2007; Gorodetskaya et al. 2008; Sorteberg et al. 2007; Vavrus et al. 2008). Chapman and Walsh (2007) reported a cold bias of 2 K in the annual mean surface air temperature over large parts of the Arctic region for a 14-GCM composite compared to 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40). On the other hand, the ERA-40 documented a warm bias compared with the Surface Heat Budget of the Arctic Ocean (SHEBA) observations of about 1 K near the surface (Tjernström and Graversen 2009). The simulated annual cycle of the Arctic surface energy balance (70°–90°N) is discussed in Sorteberg et al. (2007). They find both the downward and upward longwave radiation to be underestimated wintertime in many models. The across-model variability of longwave radiation is largest during winter. The across-model variability in the simulated sensible and latent heat fluxes is substantial and the annual mean for the region north of 70°N varies between −5–13 and 3–13 W m−2, respectively (positive numbers are defined as upward). The corresponding values for the reanalysis products are 0 and 13 W m−2 (ERA-40) and −10 and 13 W m−2 [National Centers for Environmental Prediction (NCEP)]. These discrepancies may seem large but can partly be explained by across-model differences in the distribution of open water and sea ice where the boundary layer turbulence conditions are vastly different.
The complex problem of the Arctic climate processes may be reduced by only examining the large-scale energy budget as is done by Serreze et al. (2007). They conclude that the net surface flux has first-order impacts on the atmospheric large-scale energy budgets of the Arctic, and that a net difference on the order of 1 W m−2 is important, as a sustained net surface heat flux of this amount over a year equals about 0.1 m of sea ice melt (at its melting point).
In this paper, we explore nine of the models contributing with simulations of the twentieth century to CMIP3. The models were selected based on the availability of the output fields and is the same subset of models that was analyzed in Karlsson and Svensson (2011). The focus of the present paper is on the processes important for the wintertime energy balance at the Arctic surface, its relation to the surface skin temperature, and to the temperature and humidity structure in the lower atmosphere.
In the high Arctic, north of the polar circle, a negligible amount of shortwave radiation reaches the surface during the winter months [December–February (DJF)]. If we further restrict our attention to the Arctic Ocean, most of the area is covered by sea ice, see Fig. 1. Open ocean is mainly found in the Greenland and Barents Seas. The large differences in boundary layer conditions over the different surfaces, discussed above, motivate that the analysis is performed separately over sea ice and open ocean. This was not explicitly done in the previously mentioned studies on the CMIP3 models skill to realistically simulate different aspects of the Arctic climate (Chapman and Walsh 2007; Eisenman et al. 2007; Gorodetskaya et al. 2008; Sorteberg et al. 2007; Vavrus et al. 2008).
Averaged sea ice fraction for DJF as modeled by the nine GCMs (see Table 1), the ERA-Interim, and the satellite observations (Nimbus).
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
In Karlsson and Svensson (2011), the influence of the simulated clouds on the modeled winter surface temperature is discussed when using these temporal and spatial restrictions. The model differences in simulated clouds, in terms of cloud condensate amount and phase, were found to be very large. Also, different models showed very different relationships between cloud condensate and surface cloud forcing. As expected, a somewhat robust relationship was found both in the observations and most of the models such that grid points with high monthly mean total cloud fraction were associated with high surface cloud warming. More surprisingly, the simulated surface skin temperature spread between models could not be explained by relating it to the surface cloud forcing, as anticipated considering the substantial influence that clouds have on the surface energy budget (e.g., Walsh and Chapman 1998). In this paper, we discuss other parameters relevant for the surface energy budget and the surface skin temperature, especially the vertical structure of temperature and humidity. It is fundamental for GCMs that these processes are correctly described to be able to simulate the Arctic temperatures for the present climate as well as in future climate scenarios.
2. Models and data
a. Observations and reanalysis data
There is a lack of long-term in situ observations of the different terms in the surface energy budget in the Arctic, especially over the Arctic Ocean (e.g., Sorteberg et al. 2007). This is particularly true for the turbulent fluxes, which cannot be observed by remote sensing from space. The only available observationally constrained datasets are from reanalysis projects. Here, we use the data from the publicly available 1.5° ECMWF ERA-Interim dataset (Simmons et al. 2007; Uppala et al. 2008) available from 1989 and onward. Improvements since ERA-40 (Uppala et al. 2005) include upgrades to a 12-h four-dimensional variational data assimilation (4DVAR) system, more extensive assimilation of satellite radiances, improved model physics, and improved horizontal and vertical resolution to T255 and 60 vertical layers. Since the available observations for assimilation are limited in the Arctic, one should be cautious to interpret the data as the “truth” in this remote region. However, the abundance of observations at lower latitudes and sea ice extent should at least constrain the properties of the air masses that enter and exit the Arctic.
For radiative fluxes, skin surface temperature, and cloud fraction, the extended Advanced Very High Resolution Radiometer (AVHRR) Polar Pathfinder product (APP-x) version 2.0 (Key 2002; Wang and Key 2003, 2005a,b) are used. It should be mentioned that satellite remote sensing of clouds using passive instruments (e.g., AVHRR) is a challenge in the Arctic. Successful wintertime cloud detection relies on contrasts in the outgoing terrestrial radiation, which is not always apparent considering that surface temperature inversions are commonly occurring phenomena (e.g., Raschke et al. 1992; Eastman and Warren 2010). However, the APP-x retrievals are optimized for high-latitude conditions, and the data have shown reasonable agreement with data from the SHEBA campaign (Maslanik et al. 2001). The APP-x monthly mean images at 0400 and 1400 local time are combined to a monthly mean. The data cover the period January 1982 to December 1999 and are gridded on a 25 × 25 km2 equal-area grid.
To distinguish between open ocean and sea ice for the period of APP-x observations, we have used monthly mean sea ice concentration from Nimbus-7 Scanning Multichannel Microwave Radiometer (SMMR) and Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave Imager (SSM/I) passive microwave data (Cavalieri et al. 1996). The sea ice data, which come on a 25 × 25 km2 stereographically projected grid, are interpolated bilinearly to the APP-x grid.
b. The coupled models
The analyzed model data are taken from the World Climate Research Programme (WCRP)’s CMIP3 multimodel dataset [prepared for the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report (AR4)]. Table 1 lists the nine GCMs included in our study (see http://www-pcmdi.llnl.gov/ for more info). The simulations studied are from the twentieth-century experiment (20C3M), and we are analyzing monthly mean data on the model grid for December, January, and February of the relevant parameters for the period 1980–99. These nine models represent the subgroup of models in the CMIP3 archive, which provide the monthly mean parameters needed for our analysis. The sea ice fraction is bilinearly interpolated to the atmospheric grid in the models when the horizontal resolution differs between the atmospheric and the ocean modules. The vertical profiles of temperature and specific humidity are not available on the model grid but on standard pressure levels; we have used 1000, 925, 850, 700, 600, 500, 400, and 300 hPa.
List of models that are analyzed in the study and their horizontal and vertical resolution.
3. Method
The analysis is performed for the area north of the Polar Circle (>66.6°N), either as averages over all surface types over the entire region or for different categories of surfaces and latitude bands—66.6°–70°N, 70°–80°N, and >80°N. Each ocean model grid point is classified as open ocean, marginal ice zone, or sea ice depending on the monthly mean sea ice fraction thresholds of 10% and 90%. Grid points with a land fraction of 20% or more are considered land points. The individual gridpoint monthly mean surface values and vertical column data are averaged over the surface categories for each of the three months. These monthly averages for each year (1980–99) are treated as individual data points; thus, each model generates 19 × 3 = 57 monthly mean values per surface category and latitude band, from which median and percentiles are calculated. The ERA-Interim and the APP-x data are treated the same way as the model data except that the ERA-Interim covers 18 years (1989–2007) and the APP-x data covers 17 years (1982–99). ERA-Interim also has much higher vertical grid resolution, containing 20 levels below 300 hPa. Although the natural variability is large in the Arctic, the difference in time period for the various datasets does not influence the main findings of the study.
Obviously, the modeled sea ice fraction has an influence on the sizes of the analyzed areas covered by the different categories over the latitude bands. Presented in Table 2 is the percentage of the ocean area covered by the surface categories. The number of grid points in each model differs substantially due to the different horizontal resolution (Table 1). The resolution also affects the land–sea mask; the percentage of ocean north of 66.6°N varies between 48% and 63%. The area categorized as open ocean in the models, varies between 7% and 26% with the coupled model of the Institute of Numerical Mathematics (INM) having the largest value, see Fig. 1. INM is also the model with the largest fraction of marginal ice zone and does not have any solid sea ice in the southernmost latitude band. The satellite observations (Nimbus) and ERA-Interim have about 15% of the ocean north of 66.6°N categorized as marginal sea ice, about 20% as open ocean, and the remainder as sea ice. The majority of the models have as much or larger fraction of sea ice than Nimbus and ERA-Interim. The open ocean areas are found in the Atlantic sector (Fig. 1) in both models and observations, however, the models show very variable fractional coverage. The open ocean category is not discussed for the northernmost area (north of 80°N).
Monthly mean fraction (DJF) of the ocean area north of the Arctic Circle categorized as sea ice, marginal ice zone, and open ocean. Included in the table is also the mean ocean area fraction north of the Arctic Circle.
During DJF the net solar contribution to the surface energy budget is negligible this far north. The APP-x data and the model results show both a maximum monthly mean of about 1 W m−2. Consequently, when evaluating the surface cloud forcing (SCF), we only take the longwave component into account. In the calculations of the net surface energy flux, we also neglect the shortwave component and examine the longwave components of the radiation together with the turbulent sensible and latent heat fluxes, with the sign convention of positive being upward. Unfortunately, the so-called conductive flux, the flux to and from the surface from below, is not generally available in the CMIP3 dataset. Whenever the surface is covered by snow, which is the case for DJF, the low conductivity of the snow should limit this flux. However, since the surface temperature gradient between the water temperature at the bottom of the ice (Tw = −1.8°C) and the surface temperature easily can be very large, this heat flux is not negligible. Estimates using the data taken at the SHEBA site show monthly means of this flux between 4 and 12 W m−2 (Persson et al. 2002).
4. Results
a. The area north of the Polar Circle
The Arctic is losing energy to space at the top of the atmosphere (TOA) during most of the year, with the exception of a short period during summer (Serreze et al. 2007). This loss is balanced by energy transport from lower latitudes both in the ocean and the atmosphere and moderated by the ocean heat content as well as by the growth and melting of sea ice. The net energy loss at the top of the atmosphere during winter (DJF, north of 66.6°N) estimated from the APP-x data is 170 W m−2 while the ERA-Interim gives a value of 174 W m−2 (Fig. 2); these values correspond well to earlier estimates (e.g., Serreze et al. 2007). The 1-K difference in surface skin temperature between APP-x and ERA-Interim, which according to Stefan–Boltzman Law corresponds to a 3.5 W m−2 difference in the radiative energy emission and explains the noted difference in net energy loss. The various GCMs’ net TOA radiation spans between 156 and 174 W m−2 (Fig. 2), which is a considerable spread for climatological values implying a large variation in the meridional transport of energy and/or ocean heat release.
Wintertime (DJF) top of the atmosphere outgoing longwave radiation (W m−2) plotted against the surface skin temperature (K) over sea ice–covered ocean north of 66.6°N for the nine GCMs, the ERA-Interim, and the APP-x data. The figure shows the median values (symbol) as well as the 25 to 75 percentile range (lines).
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
Figure 2 shows a high correlation between the surface skin temperature and the TOA radiative flux. The modeled surface skin temperature range is 11 K that, at these temperatures, would give a net change in the outgoing surface radiation of 36 W m−2, the simulated net radiation at the TOA only spans half of that. This implies the importance of the radiative properties of the simulated atmospheres. The analyzed climate models show wintertime mean cloud fractions between 35% and 95% over the sea ice, with the APP-x and ERA-Interim having values of 68% and 82%, respectively, and no clear relation between the mean cloud fraction and surface skin temperature is found (Karlsson and Svensson 2011). The amount of cloud is, however, not as important as its radiative properties and their total effect can be expressed by the SCF. Figure 3 shows the SCF plotted against the surface skin temperature for the sea ice region. There is no clear relationship between the modeled SCF and the surface skin temperature apparent in the figure. The entire spread of modeled surface skin temperature is covered with a range of 15 to 20 W m−2 in SCF. On the other hand, about the same surface skin temperature can be reached with values of SCF from 17 to almost 40 W m−2. Thus, there must be other factors influencing the surface skin temperature.
Wintertime (DJF) surface cloud forcing (W m−2) plotted against the surface skin temperature (K) north of 66.6°N over sea ice–covered ocean for the nine GCMs, the ERA-Interim, and observations. The figure shows the median values (symbol) as well as the 25 to 75 percentile range (lines).
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
b. The terms in the surface energy budget
First, we will examine the turbulent heat fluxes. Figure 4a shows sensible, latent, and net heat fluxes for all models and the ERA-Interim for the area north of 66.6°N. The median of the monthly mean values as well as the variability (5 and 95 percentiles) are shown in the figure. It is clear that the net turbulent heat flux is small when considering the entire polar cap. The latent heat flux is positive in all GCM with values from 5 to 14 W m−2. The majority of the models also have a positive (upward) sensible heat flux while the two models with the coarsest horizontal resolution have substantially negative values [−7 and −6 W m−2 for Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled General Circulation Model (CGCM) and INM, respectively]. The climate model of the Geophysical Fluid Dynamics Laboratory (GFDL) also has a negative (downward) sensible heat flux but smaller (−2 W m−2). The ERA-Interim gives negligible sensible heat flux contribution and the net turbulent heat flux is 11 W m−2, in agreement with the values reported earlier for ERA-40 (Serreze et al. 2007; Sorteberg et al. 2007). Only one model shows a negative total turbulent heat flux (CGCM) while the majority of the models show an upward flux between 10 and 18 W m−2.
Wintertime (DJF) median (symbols) and 5–95 percentile range (lines) of turbulent heat fluxes (W m−2) for nine GCMs (see Table 1) and the ERA-Interim for (a) all surfaces north of the Polar Circle, (b) over open ocean for the latitude bands (left) 66.6°–70°N and (right) 70°–80°N, and (c) over sea ice for the latitude bands (left) 66.6°–70°N, (middle) 70°–80°N, and (right) 80°–90°N. The (top) sensible, (middle) latent, and (bottom) net turbulent heat fluxes are shown. Positive fluxes are upward. Note that INM does not have any value for the southernmost latitude band in (c) since it does not have any sea ice category south of 70°N.
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
The small net turbulent heat flux for the entire polar cap is made up of strongly different turbulence regimes over the open water and the sea ice. This is obvious from Figs. 4b and 4c, which show the corresponding values over the open ocean and over sea ice, respectively, presented in latitude bands (only the two southernmost for the open ocean category). The net monthly median heat fluxes over the open ocean are in the range 100 to 200 W m−2, with the ERA-Interim in the lower range. The sensible and latent turbulent fluxes contribute about half of this flux each. The intermodel monthly variability over open ocean is large as shown by the 5 and 95 percentiles. When moving northward, the turbulent heat fluxes increase in all models but INM, which is the model with lowest median values and small variability. The increase is expected since the air temperature decrease farther north while the open water temperature is restricted to the freezing point, making it possible for the unstable temperature stratification to support more turbulence and a larger flux.
Restricted by the cool temperatures of the ice resulting in stably stratified air, the heat fluxes are much lower over the sea ice (Fig. 4c). Most of the models and ERA-Interim show a very low median monthly flux changing from positive (upward) to negative (downward) flux when going north. At the southernmost latitude band, the latent heat flux is dominating. The net heat flux is gradually changed to be dominated by the sensible heat flux while going northward.
Next, we turn to the longwave flux. Figure 5a shows its components and the longwave net flux for the area north of 66.6°N. The net radiative energy loss at the surface in the APP-x data is estimated to be about 29 W m−2 while the ERA-Interim gives a value of 40 W m−2, somewhat lower than reported for ERA-40 (~50 W m−2, Serreze et al. 2007; Sorteberg et al. 2007). The various GCMs give net fluxes between 29 and 47 W m−2, a range of almost 20 W m−2, with all models giving equal or higher values than the APP-x. The modeled monthly variability is small (<8 W m−2).
Wintertime (DJF) median (symbols) and 5–95 percentile range (lines) of longwave radiative fluxes (W m−2) for nine GCMs (see Table 1), the ERA-Interim, and the APP-x data for (a) all surfaces north of the Polar Circle, (b) over open ocean for the latitude bands (left) 66.6°–70°N and (right) 70°–80°N, and (c) over sea ice for the latitude bands (left) 66.6°–70°N, (middle) 70°–80°N, and (right) 80°–90°N. The (top) downward, (middle) upward, and (bottom) net longwave radiative fluxes are shown. Note that INM does not have any value for the southernmost latitude band in (c) since it does not have any sea ice category south of 70°N.
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
The overestimation in the net longwave flux at the surface in most cases arises from the underestimation of the downward component (Fig. 5a) compared with the APP-x data. As was the case for the turbulent heat fluxes, the contribution from the different surface categories plays a role, thus presented in Figs. 5b and 5c are the fluxes over open ocean and sea ice for the different latitude bands. The net longwave over open ocean is overestimated in all models and ERA-Interim compared with the APP-x. Most of the models show a too large outgoing component consistent with the fact that they also have higher surface skin temperature than the APP-x data (not shown). This, in combination with a too small incoming radiation, gives large net fluxes and possibly too large ocean heat loss.
The net radiation fluxes over the sea ice also have positive biases for most models when compared with the APP-x (Fig. 5c). For this region, most models show comparable or colder surface skin temperatures than the observations and ERA-Interim, which agree fairly well. Thus, the positive biases do not originate from the outgoing longwave radiation. The underestimation of the incoming longwave radiation is noteworthy; the models underestimate this component by up to 35 W m−2 at the northernmost region. Some models [CGCM, GFDL, the climate configuration of the Met Office Unified Model (HadCM), and the Hadley Centre Global Environmental Model (HadGEM)] show much lower temperatures, at least for the two northernmost regions. The same models also show far too low incoming longwave radiation; these are also the models with the lowest SCF (see Fig. 3). The L’Institut Pierre-Simon Laplace (IPSL) and the Model for Interdisciplinary Research on Climate (MIROC) models also have very low SCF (Fig. 3); however, they do not have the same problem with too low surface skin temperatures.
c. The net flux at the surface
Two categories of models can be identified based on the downward component of the longwave radiation over the sea ice (Fig. 5c), one with reasonably simulated values and small latitudinal variation [Community Climate System Model (CCSM), ECHAM, INM, and MIROC] and one with too low values also associated with too much change (compared with observations) while moving northward (CGCM, GFDL, HadCM, HadGEM, and IPSL). ERA-Interim falls in the first category. There is no clear organization according to the resolution of the atmospheric model (see Table 1). However, the surface skin temperature is not merely determined by the radiative fluxes. In Fig. 6, the median net flux at the surface, that is, the sum of the radiative and turbulent fluxes, is presented for the three latitude bands over sea ice. In the figure, the overall net fluxes and the surface skin temperature seem to decrease while proceeding northward. The two model categories are becoming gradually more visible with increasing latitude, which can possibly be translated to time that the air has spent north of the Arctic Circle. Since the different behavior of the model categories seems to originate from the ability of the atmosphere to radiate back to the surface, and not primarily from the total cloudiness or the SCF, we proceed to examine the vertical profiles of temperature and humidity.
Net energy flux at the surface (W m−2) over sea ice plotted against the surface skin temperature (K) for three latitude bands for the nine GCMs and the ERA-Interim. (bottom) Note that INM is not presented since it does not have any sea ice category south of 70°N.
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
d. Profiles of temperature and humidity
There are substantial Arctic regional differences in the near-surface structure of temperature and humidity. This is shown in Fig. 7 where mean profiles from ERA-Interim, taken over the different surfaces for the southernmost latitude band, are presented. Air found over open water at this latitude band is warm and moist; the air over land or sea ice is already at this latitude much colder and dryer. This difference is present all through the troposphere and is also evident in the mean tropopause height, which is higher over the open ocean (not shown). A surface-based inversion is already present in the sea ice profiles while the open ocean profile indicates near-neutral stratification in the lowest 500 m. The fraction of open ocean is not large and mainly found in the Atlantic sector (see Fig. 1), thus the median for all surfaces is closer to the sea ice profile.
Median vertical profiles of (a) temperature (K), (b) specific humidity (g kg−1), and (c) relative humidity (%) over entire area (solid), open ocean (dashed–dotted), sea ice (dashed), and over land (dotted) from ERA-Interim over the southernmost latitude band 66.6°–70°N.
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
Already in this southernmost latitude band, considerable differences are noted in the simulated net energy flux at the surface (Fig. 6). The net flux is highly dependent on the state of the atmosphere. The magnitude of the turbulent heat fluxes depends on the temperature and moisture stratification of the lower troposphere and the downward longwave radiation depends on the temperature, moisture and cloudiness of the atmosphere. Across-model differences are found in these properties of the air masses that enter and exit the Arctic (Fig. 8, note that the two model categories can be identified by dashed or solid lines). The structure of the median air mass in the southernmost latitude band for all surfaces is presented in Figs. 8a–c). In most models, the inversion is seen in both temperature and specific humidity. ERA-Interim profiles tend to be in between the too cold (solid lines) and the warmer (dashed lines) models.
Median vertical profiles of temperature (K), specific humidity (g kg−1), and relative humidity (with respect to water) over (a)–(c) entire area, (d)–(f) open ocean, and (g)–(i) sea ice from the GCMs and ERA-Interim over the southernmost latitude band 66.6°–70°N. Note that INM is not presented in (g)–(i) since it does not have any sea ice category south of 70°N.
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
Consistent with ERA-Interim, the profiles over the open ocean are well mixed in all models but one (IPSL). However, the median temperatures in the lower-atmospheric layers differ with more than 5 K, with most models being warmer than the ERA-Interim (Fig. 8d). All GCMs have a relative humidity (with respect to water) above 80% at the 925-hPa level most with relative humidity both above and below this level (Fig. 8f).
The median temperature profiles in the southernmost latitude band over sea ice all show the characteristic Arctic inversion (e.g., Tjernström and Graversen 2009) with highest temperatures at 850 hPa (Fig. 8g). At this level, the spread of the model-median temperatures is about 8 K, at the surface the spread exceeds 10 K, which implies differences in the low-level stratification and thus in the turbulent transfer of heat and moisture. The median temperature difference between 1000- and 850-hPa levels varies between 1.3 and 7.5 K. The ERA-Interim gives a value of 3.2 K. However, the range in dry static stability cannot directly be translated to variations in the strength of the turbulent heat flux (Fig. 4b). The actual flux is determined by the intensity of the turbulence, determined by the buoyancy and shear, in combination with the temperature gradient. The variation in specific humidity at 850 hPa is also very large (0.5–1.0 g kg−1, Fig. 8h), with all models but two being more humid than the ERA-Interim (0.6 g kg−1), basically following the temperature bias. The relative humidity is highest at the surface (Fig. 8i), due to the temperature inversion, and all models but one (CGCM) show values higher than 80%.
From Fig. 8, we can conclude that the air masses in the southernmost Arctic have very different properties in the various models. Air masses with such different temperature and humidity radiate considerably differently and cool at different rates. To examine how the airmass properties change with distance from the Arctic circle, we investigate the changes in the modeled temperature and humidity profiles over the sea ice between the southernmost latitude bands and the latitude bands covering 70°–80° and 80°–90°N (Fig. 9). The change in median temperature for ERA-Interim between the two southernmost sectors (Fig. 9a) is less than 1 K for the entire layer below 700 hPa. The inversion strength is increasing somewhat since the cooling and drying at 925 hPa are smaller than at the surface (Figs. 9a,b). All GCMs show larger cooling and drying than the ERA-Interim. Most models show profiles of temperature change (Fig. 9a) with similar shape as the ERA-Interim but with much too large a decrease, magnitudes of 6 K near the surface, and 3 K at 700 hPa. Two models (HadGEM and MIROC), however, have the smallest cooling near the surface indicating a decrease in the inversion strength. CGCM and MIROC are the only two models with similar shape as the ERA-Interim profile for the humidity change (Fig. 9b) although larger decrease. The modeled drying at the surface is in the range between 0.1 and 0.3 g kg−1 while the ERA-Interim have a slightly lower value.
Difference between the median profiles of temperature (K) and specific humidity (g kg−1) over sea ice between the latitude band (a),(b) 70°–80°N and (c),(d) 80°–90°N and the southernmost latitude band 66.6°–70°N from the GCMs and ERA-Interim. Note that INM is not included since it does not have any sea ice category south of 70°N.
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
When comparing the air mass around the North Pole with the one close to the Polar Circle (Figs. 9c,d) the too strong cooling and drying in most GCMs become even more evident. In terms of profile shape, ECHAM and HadGEM have the most similar cooling and drying profiles as compared to ERA-Interim (Figs. 9c,d). ECHAM also has very similar properties of the air mass that enters the area as ERA-Interim (Fig. 8) but too large SCF (Fig. 3). The HadGEM air is far too cold and dry near the Polar Circle (Fig. 8) and has much lower SCF and surface skin temperature (Fig. 3). The temperature change is largest in the GFDL, HadCM, and IPSL models, which shows 5-K cooling or more below 700 hPa. These models also have the most dramatic change in humidity. Note that at the southernmost area the temperature profiles of both GFDL and HadCM show good resemblance with ERA-Interim at the southernmost area while IPSL shows a warm bias as large as 3 K below 700 hPa (Fig. 8d). The specific humidity is too high in all three models (Fig. 8e) and the IPSL has the highest relative humidity (Fig. 8f). These three models, together with CGCM and HadGEM, are the ones that we already have categorized as models with too much change in the net surface skin temperature. Some of these models thus have a too cold atmosphere to start with and some have a too large change in the atmospheric temperature and humidity.
e. Downwelling longwave radiation and vertical structure
Modeled downwelling longwave radiation at the surface for clear- and all-sky conditions plotted against (a),(c) mean atmospheric temperature (°C) and (b),(d) precipitable water (kg m−2) over all surfaces and latitude bands. (top and bottom) Plotted are all monthly mean values for the two groups of models and for ERA-Interim. (1) and (2) from Zhang et al. (2001) are shown as black solid lines as well as fitted functions for modeled clear sky (dashed lines) and all sky (solid lines).
Citation: Journal of Climate 24, 22; 10.1175/2011JCLI4012.1
From Fig. 10 it is clear that the two identified groups of models have different mean atmospheric temperatures. The models with generally colder surface skin temperature (Figs. 10a,b) individually show monthly averaged mean-layer temperatures as low as −40°C (Fig. 10a). Furthermore, it is clear that this group of GCMs is below the ERA-Interim points when considering the precipitable water (Fig. 10b). Since the change in downwelling longwave radiation is inversely proportional to the amount of precipitable water, changes in the radiation become large when drying an already dry atmosphere [as pointed out in Zhang et al. (2001)]. In general, for both groups the clear-sky values are on the low side compared with the ERA-Interim. Every model seems to have about the same functional behavior for their clear-sky values but negatively shifted with a few W m−2, which can play a crucial role for the net heat balance at the surface. The impact of the clouds is clearly seen but nevertheless in these models the total downwellling radiation is underestimated compared with the ERA-Interim.
In Figs. 10c and 10d the model results for the GCMs that are associated with generally warmer surface skin temperatures are shown. The fitted functions for these models diverge substantially with most models (except INM) having a smaller slope for clear- and all-sky conditions. This group of models is also more moist, rarely less than 1.5 kg m−2 (Fig. 10d). This means that these models do not enter the very steep part of the function that leads to a dramatic drop in the ability of the atmosphere to interact with the longwave radiation. From the rather large gap between the clear-sky values and all-sky values (Figs. 10c,d), it is also clear that this category of models are associated with the strongest cloud forcing (Fig. 3). Many of the modeled precipitable water contents are larger than found in ERA-Interim, again indicating that these models have too much humidity (see section 4d).
5. Discussion and conclusions
In this paper, the turbulent and radiative heat fluxes during the Arctic winter night are examined. From the analysis, it is clear that there are wide differences between in Arctic airmass properties in the various GCMs, and these differences both have an effect on the simulated surface skin temperature and are also affected by it. Two model categories emerge and the difference between these groups is most clearly seen when the net heat flux at the surface is correlated with the surface skin temperature in the northernmost region (Fig. 6). The group with approximately correct surface skin temperature (the APP-x data gives an average surface skin temperature of 245 K) also displays slightly better simulated vertical profiles of temperature and humidity in the southernmost region of the Arctic (Figs. 8g–h), compared with the ERA-Interim. All models show too much change in the vertical temperature and humidity profiles in between latitude bands (Fig. 9). This is reflected in the downwelling longwave radiation, which becomes too low for clear-sky conditions (Fig. 10). If clouds are formed, with their positive surface forcing, they can compensate for the too low clear-sky conditions and help hinder the surface temperature to decrease. Models (CCSM and ECHAM) that manage to keep the surface skin temperature realistic have a higher surface cloud forcing (Figs. 3 and 10) likely because they also have high wintertime liquid water content (Karlsson and Svensson 2011).
Although the presence of clouds, especially the ones with high liquid water content, have positive influence on the net radiation at the surface, it is not so easy to comprehend their effect on the vertical structure of the temperature and humidity profiles; especially since the CMIP3 data are only available as monthly means. When a cloud forms, latent heat is released. However, this is likely much less energy (depends on the amount of condensate) than is lost by the enhanced radiative cooling that occurs due to the cloud layer.
Also the longwave radiative cooling for clear air varies between models. Offline radiation calculations, with a radiation model (Fu 1991; Fu and Liou 1992, 1993) using median averaged temperature and humidity profiles for the area north of the polar circle from the GCMs, show that the variation in mean cooling rate for the air below 700 hPa varies between −1.0 and −1.25 K day−1 (−1.05 K day−1 for ERA-Interim). However, there is no clear relation between a particular model’s cooling rate and either the mean-layer temperature or the precipitable water (not shown). For clear-air conditions it would thus take, on average, between 2 days and a week to cool the air as much as the differences we find between 66.6° and 70°N and at 80°–90°N (Fig. 9c). From these radiation calculations it is also clear that it is primarily the humidity profile, not the temperature profile, that determines the clear air downwelling radiation in these very cold and dry conditions; this was also found in Zhang et al. (2001) but for the spring melt season. Thus, we conclude that the atmospheric moisture content is highly important for the surface radiation balance in the Arctic, especially over sea ice. The clouds are also important but their effects are difficult to see directly in the results. They seem instead to play the most important role in the evolution of the temperature and humidity profiles that in turn affect the surface energy budget.
ERA-Interim does not show much difference in either temperature or humidity across the Arctic region (Figs. 9c,d). The GCMs have larger differences, which could be related to radiation issues (as discussed above) or possibly also to the mean time that the air spend in the Arctic. A model with less synoptic activity would lead to more stagnant conditions when air over the polar cap can cool and dry further than it is allowed to if there are more frequent events with warm-air intrusions from lower latitudes (Sorteberg and Walsh 2008). Most GCMs have problems with the sea level pressure in the Arctic (Chapman and Walsh 2007). In winter, the GCM composite shows a large positive pressure bias over the Barent Sea. This is explained by the GCM-simulated storm track that ends in Barents Sea instead of further north in the Kara Sea as the ERA-40 results show. The models in this study that tend to cool too much, and are analyzed in the Chapman and Walsh (2007) study (CGCM, GFDL, HadCM, and IPSL), are also associated with a stronger positive bias in the annual sea level pressure than the GCMs with a smaller temperature bias (CCSM, ECHAM, and INM); CCSM even has a strong annual negative bias in the Arctic Ocean sea level pressure. A consequence is often too much ice in the Barents Sea (Fig. 1) with too low surface temperatures (Chapman and Walsh 2007). The GCMs inability to make the storms enter the Eurasian portion of the Arctic Ocean has an impact on the airmass properties and thus also on the surface temperature.
Another uncertainty that arises in the assessment of the energy fluxes important for the surface skin temperature is the conductive heat flux from the ocean through the sea ice, which may play a crucial role. Unfortunately, this could not be investigated using the CMIP3 dataset since the various models did not report the heat flux through the sea ice. However, it is clear that the flux from the ocean potentially can be a source of the large variations between the different models, since both the sea ice extent and thickness varies substantially between the simulations (Gerdes and Köberle 2007). The thickness of sea ice and snow is important for the strength of the conductive flux of heat upward through the ice (Eisenman et al. 2007).
The turbulent heat fluxes display large variations between the models. These probably originate from several reasons—differences in the boundary layer parameterizations, surface stability functions, and the near-surface profiles of wind, temperature, and humidity. During winter, stably stratified conditions are common over the sea ice (Tjernström and Graversen 2009), and it is well-known that these conditions are problematic to simulate (Dethloff et al. 2001; Cuxart et al. 2006). The vertical resolution also plays an important role for the turbulence under these conditions (Byrkjedal et al. 2008; Svensson and Holtslag 2009). Higher vertical resolution would additionally benefit the radiation calculations, especially in presence of strong vertical gradients as often are found in the Arctic (Edwards 2009).
Over the ocean, on the other hand, the turbulent regime is either nearly neutral or convective. Compared to the sea ice–covered ocean this creates very different magnitudes of the turbulent fluxes and even different sign (Fig. 4). Many of the GCM grid boxes consist of a combination of these two surfaces, specified with fractions of sea ice and open ocean. Normally, the surface energy balance is solved using the atmospheric profile to generate surface temperatures and turbulent fluxes for the fractional surfaces separately. The turbulent fluxes are then combined, according to the surface fractions, and are passed back to the atmosphere as a boundary condition. The underlying assumption is that the blending height is below the first model level, that is, no difference because of the surface conditions would be detectable in the atmosphere at this height, which is commonly violated in this region. Moreover, the surface exchange schemes typically utilize surface layer theory that is only valid in roughly the lowest 10% of the boundary layer, which requires the lowest grid point to be close to the surface; this is typically not the case.
Substantial differences between the GCM results have been found in this study. Examination over different surfaces and regions individually brings some structure to the modeling problems. The emerging picture is that profiles of atmospheric moisture must be well simulated as the air enters the Arctic. This of course implies that the temperature profiles need to be correctly simulated as well since the humidity is strongly related to the temperature. If not, unrealistic cloud cover and/or amount of condensate are needed to compensate for the loss in downwelling longwave radiation since it declines exponentially as the precipitable water decreases. Too low humidity will thus lead to unrealistic cool ice surface temperatures.
Acknowledgments
We acknowledge the modeling groups, the Program for Climate Model Diagnosis and Intercomparison (PCMDI) and the WCRP’s Working Group on Coupled Modelling (WGCM) for their roles in making available the WCRP CMIP3 multimodel dataset. Support for this dataset is provided by the Office of Science, U.S. Department of Energy. ECMWF ERA-INTERIM data used in this study have been obtained from the ECMWF data server. Discussions with Michael Tjernström are greatly appreciated.
REFERENCES
ACIA, 2005: Arctic Climate Impact Assessment. Cambridge University Press, 1042 pp.
Byrkjedal, Ø., I. Esau, and N. G. Kvamstø, 2008: Sensitivity of simulated wintertime Arctic atmosphere to vertical resolution in the ARPEGE/IFS model. Climate Dyn., 30, 687–701.
Cavalieri, D., C. Parkinson, P. Gloersen, and H. J. Zwally, cited 1996: Sea ice concentrations from Nimbus-7 SMMR and DMSP SSM/I passive microwave data, (1982–1999). National Snow and Ice Data Center. [Available online at http://nsidc.org/data/nsidc-0051.html.]
Chapman, W. L., and J. E. Walsh, 2007: Simulations of arctic temperature and pressure by global coupled models. J. Climate, 20, 609–632.
Cuxart, J., and Coauthors, 2006: Single-column model intercomparison for a stably stratified atmospheric boundary layer. Bound.-Layer Meteor., 118, 273–303.
Dethloff, K., C. Abegg, A. Rinke, I. Hebestadt, and V. F. Romanov, 2001: Sensitivity of Arctic climate simulations to different boundarylayer parameterizations in a regional climate model. Tellus, 53A, 1–26, doi:10.1034/j.1600-0870.2001.01073.x.
Eastman, R., and S. G. Warren, 2010: Arctic cloud changes from surface and satellite observations. J. Climate, 23, 4233–4242, doi:10.1175/2010JCLI3544.1.
Edwards, J. M., 2009: Radiative processes in the stable boundary layer: Part I. Radiative aspects. Bound.-Layer Meteor., 132, 349–350, doi:10.1007/s10546-009-9402-6.
Eisenman, I., N. Untersteiner, and J. S. Wettlaufer, 2007: On the reliability of simulated Arctic sea ice in global climate models. Geophys. Res. Lett., 34, L10501, doi:10.1029/2007GL029914.
Fu, Q., 1991: Parameterization of Radiative Processes in Vertically Nonhomogeneous Multiple Scattering Atmospheres. Ph.D. dissertation, University of Utah, 259 pp.
Fu, Q., and K. N. Liou, 1992: On the correlated k-distribution method for radiative transfer in nonhomogeneous atmospheres. J. Atmos. Sci., 49, 2139–2156.
Fu, Q., and K. N. Liou, 1993: Parameterization of the radiative properties of cirrus clouds. J. Atmos. Sci., 50, 2008–2025.
Gerdes, R., and C. Köberle, 2007: Comparison of Arctic sea ice thickness variability in IPCC Climate of the 20th century experiments and in ocean–sea ice hindcasts. J. Geophys. Res., 112, C04S13, doi:10.1029/2006JC003616.
Gorodetskaya, I. V., L.-B. Tremblay, B. Liepert, M. A. Cane, and R. I. Cullather, 2008: The influence of cloud and surface properties on the arctic ocean shortwave radiation budget in coupled models. J. Climate, 21, 866–882.
Graversen, R. G., T. Mauritsen, M. Tjernström, E. Källén, and G. Svensson, 2008: Vertical structure of recent Arctic warming. Nature, 451, 53–56, doi:10.1038/nature06502.
Intrieri, J. M., M. D. Shupe, T. Uttal, and B. J. McCarty, 2002: An annual cycle of Arctic cloud characteristics observed by radar and lidar at SHEBA. J. Geophys. Res., 107, 8030, doi:10.1029/2000JC000423.
Karlsson, J., and G. Svensson, 2011: The simulation of arctic clouds and their influence on the winter surface temperature in present-day climate in the CMIP3 multi-model dataset. Climate Dyn., 36, 623–635.
Key, J., 2002: The Cloud and Surface Parameter Retrieval (CASPR) system for polar AVHRR user’s guide, version 4. Cooperative Institute for Meteorological Satellite Studies, University of Wisconsin—Madison, 68 pp. [Available online at http://stratus.ssec.wisc.edu/caspr/documentation.html.]
Manabe, S., and R. T. Wetherald, 1975: The effects of doubling the CO2 concentration on the climate of a general circulation model. J. Atmos. Sci., 32, 3–15.
Maslanik, J. A., J. Key, C. W. Fowler, T. Nguyen, and X. Wang, 2001: Spatial and temporal variability of satellite-derived cloud and surface characteristics during FIRE-ACE. J. Geophys. Res., 106, 15 233–15 250, doi:10.1029/2000JD900284.
Meehl, G. A., C. Covey, T. Delworth, M. Latif, B. McAvaney, J. F. B. Mitchell, R. J. Stouffer, and K. E. Taylor, 2007: The WCRP CMIP3 multimodel dataset: A new era in climate change research. Bull. Amer. Meteor. Soc., 88, 1383–1394.
Persson, P. O. G., C. W. Fairall, E. L Andreas, P. S. Guest, and D. K. Perovich, 2002: Measurements near the Atmospheric Surface Flux Group tower at SHEBA: Near-surface conditions and surface energy budget. J. Geophys. Res., 107, 8045, doi:10.1029/2000JC000705.
Raschke, E., P. Bauer, and H. J. Lutz, 1992: Remote sensing of clouds and surface radiation budget over polar regions. Int. J. Remote Sens., 13, 13–22.
Serreze, M. C., A. P. Barrett, A. G. Slater, M. Steele, J. Zhang, and K. E. Trenberth, 2007: The large-scale energy budget of the Arctic. J. Geophys. Res., 112, D11122, doi:10.1029/2006JD008230.
Shupe, M., and J. Intrieri, 2004: Cloud radiative forcing of the Arctic surface: The influence of cloud properties, surface albedo, and solar zenith angle. J. Climate, 17, 616–628.
Simmons, A. S., C. Uppala, D. Dee, and S. Kobayashi, 2007: Era-Interim: New ECMWF reanalysis products from 1989 onwards. ECMWF Newsletter, No. 110, ECMWF, Reading, United Kingdom, 25–35.
Solomon, S., D. Qin, M. Manning, M. Marquis, K. Averyt, M. M. B. Tignor, H. L. Miller Jr., and Z. Chen, Eds., 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.
Sorteberg, A., and J. E. Walsh, 2008: Seasonal cyclone variability at 70°N and its impact on moisture transport into the Arctic. Tellus, 60A, 570–586.
Sorteberg, A., T. Furevik, H. Drange, and N. G. Kvamstø, 2005: Effects of simulated natural variability on Arctic temperature projections. Geophys. Res. Lett., 32, L18708, doi:10.1029/2005GL023404.
Sorteberg, A., V. Kattsov, J. E. Walsh, and T. Pavlova, 2007: The Arctic surface energy budget as simulated with the IPCC AR4 AOGCMs. Climate Dyn., 29, 131–156.
Svensson, G., and A. A. M. Holtslag, 2009: Analysis of model results for the turning of the wind and related momentum fluxes in the stable boundary layer. Bound.-Layer Meteor., 132, 261–277.
Tjernström, M., and R. G. Graversen, 2009: The vertical structure of the lower Arctic troposphere analysed from observations and the ERA-40 reanalysis. Quart. J. Roy. Meteor. Soc., 135, 431–443.
Uppala, S. M., and Coauthors, 2005: The ERA-40 Re-Analysis. Quart. J. Roy. Meteor. Soc., 131, 2961–3012.
Uppala, S. M., D. Dee, S. Kobayashi, P. Berrisford, and A. Simmons, 2008: Towards a climate data assimilation system: Status update of ERA-interim. ECMWF Newsletter, No. 115, ECMWF, Reading, United Kingdom, 12–18.
Vavrus, S., D. Waliser, A. Schweiger, and J. Francis, 2008: Simulations of 20th and 21st century Arctic cloud amount in the global climate models assessed in the IPCC AR4. Climate Dyn., 33, 1099–1115, doi:10.1007/s00382-008-0475-6.
Walsh, J. E., and W. L. Chapman, 1998: Arctic cloud–radiation–temperature associations in observational data and atmospheric reanalyses. J. Climate, 11, 3030–3045.
Wang, X., and J. Key, 2003: Recent trends in arctic surface, cloud, and radiation properties from space. Science, 299, 1725–1728.
Wang, X., and J. Key, 2005a: Arctic surface, cloud, and radiation properties based on the AVHRR Polar Pathfinder dataset. Part I: Spatial and temporal characteristics. J. Climate, 18, 2558–2574.
Wang, X., and J. Key, 2005b: Arctic surface, cloud, and radiation properties based on the AVHRR Polar Pathfinder dataset. Part II: Recent trends. J. Climate, 18, 2575–2593.
Zhang, T., K. Stamnes, and S. A. Bowling, 2001: Impact of the atmospheric thickness on the atmospheric downwelling longwave radiation and snowmelt under clear-sky conditions in the arctic and subarctic. J. Climate, 14, 920–939.
