Abstract

The Arctic summer sea ice has diminished fast in recent decades. A strong year-to-year variability on top of this trend indicates that sea ice is sensitive to short-term climate fluctuations. Previous studies show that anomalous atmospheric conditions over the Arctic during spring and summer affect ice melt and the September sea ice extent (SIE). These conditions are characterized by clouds, humidity, and heat anomalies that all affect downwelling shortwave (SWD) and longwave (LWD) radiation to the surface. In general, positive LWD anomalies are associated with cloudy and humid conditions, whereas positive anomalies of SWD appear under clear-sky conditions. Here the effect of realistic anomalies of LWD and SWD on summer sea ice is investigated by performing experiments with the Community Earth System Model. The SWD and LWD anomalies are studied separately and in combination for different seasons. It is found that positive LWD anomalies in spring and early summer have significant impact on the September SIE, whereas winter anomalies show only little effect. Positive anomalies in spring and early summer initiate an earlier melt onset, hereby triggering several feedback mechanisms that amplify melt during the succeeding months. Realistic positive SWD anomalies appear only important if they occur after the melt has started and the albedo is significantly reduced relative to winter conditions. Simulations where both positive LWD and negative SWD anomalies are implemented simultaneously, mimicking cloudy conditions, reveal that clouds during spring have a significant impact on summer sea ice while summer clouds have almost no effect.

1. Introduction

The Arctic sea ice has decreased and thinned dramatically in recent decades (e.g., Serreze et al. 2007; Rothrock et al. 2008). This trend is evident in all seasons, but the most rapid decline in sea ice is observed in September—the Arctic sea ice extent (SIE) has decreased at a rate of about 13% decade−1 between 1979 and 2014 (IPCC 2013; http://nsidc.org/arcticseaicenews/2014/10/). Besides this trend, the Arctic SIE also shows large year-to-year variability (Serreze et al. 2007). For example, in September 2013 the Arctic SIE was observed to be 1.7 million square kilometers higher than in September 2012, while September 2007 had an about 1 million square kilometers lower SIE than September 2006.

The processes behind the year-to-year variability are not yet well understood. One of the key drivers for the large interannual variability is likely the atmospheric circulation (Ogi and Wallace 2007; Maslanik et al. 2007; Kapsch et al. 2013). Anomalies in atmospheric circulation impact sea ice thermodynamics, such as ice melt (e.g., Graversen et al. 2011; Kapsch et al. 2013; Devasthale et al. 2013) and sea ice dynamics (e.g., the ice-volume export out of the Arctic basin; Wang et al. 2009). They also impact the oceanic heat flux, for example, because of wind-driven advection of warm or cold water from adjacent seas, which has a direct effect on sea ice melt (Wang et al. 2009; Wu et al. 2012).

Changes in atmospheric circulation alter the atmospheric heat and moisture transport into the Arctic (e.g., Woods et al. 2013; Screen et al. 2011). Kapsch et al. (2013) showed that in years with anomalously low September sea ice concentration (SIC), more moisture was transported during the preceding spring into the area with the largest ice retreat, leading to positive anomalies of clouds and atmospheric water vapor. As clouds and water vapor absorb and reemit upwelling longwave radiation back toward the surface, more downwelling longwave radiation (LWD) is absorbed by the surface in April and May during these years over the Beaufort and East Siberian Seas. Also, Francis et al. (2005) found a significant correlation between the amount of LWD and summer sea ice variability: LWD anomalies within 80 days prior the ice-edge minimum extent explain up to 40% of the ice-edge variability in most of the Arctic regions.

Some previous studies pointed toward the importance of clouds for the seasonal evolution of the sea ice cover. Liu and Key (2014) found that a decreased wintertime cloud cover in 2013 resulted in a cooling of the surface and allowed for a greater-than-normal ice growth. They argued that this led to the recovery of the ice cover in 2013 after the 2012 record minimum. In contrast, Eastman and Warren (2010) found the largest correlations between spring and fall cloud amounts and September SIE, indicating a specific importance of LWD during these two seasons for the September SIE.

Several studies analyzed the 2007 record ice minimum. By investigating the contribution of clouds and radiation anomalies Kay et al. (2008) suggested, using spaceborne observations, that less clouds during summer, corresponding to less LWD but more downwelling shortwave radiation (SWD), caused the 2007 SIE minimum. Schweiger et al. (2008), using an ice-ocean model, questioned the importance of the anomalous low cloud cover for the sea ice minimum. Graversen et al. (2011) argued that the area of the negative cloud anomaly in Kay et al. (2008) did not correspond to the region of the greatest ice loss. They concluded that the area that experienced the largest ice loss in 2007 coincided with an above-average cloud cover during summer due to an anomalous flow of warm and humid air into that region (Graversen et al. 2011).

The conflicting conclusions from the aforementioned studies indicate that the importance of clouds and water vapor, and thus ultimately of LWD and SWD, for the seasonal sea ice evolution is not well understood. In this study, the impact of realistic anomalies of LWD and SWD on sea ice is investigated. A slab-ocean, global climate model (see section 2) is used. By applying SWD and LWD anomalies both separately and in combination and for different seasons, the relative effect and the seasonal importance of the two types of radiation are examined. We address the question whether positive LWD anomalies in spring and early summer lead to a larger summer sea ice decline than anomalies imposed later in the year. The latter hypothesis relates to the fact that feedback processes lead to a larger sea ice decline over summer when initiated early. A positive LWD anomaly early in the season might initiate an earlier melt onset (Maksimovich and Vihma 2012), hence, a negative ice anomaly early in the season. The earlier the negative ice anomaly occurs the more shortwave radiation is absorbed by the surface throughout the melt season due to the positive ice–albedo feedback. This additional input of heat during the melt season acts to amplify the melt and to warm the upper ocean layers (Perovich et al. 2007). The energy absorbed by the ocean mixed layer shifts the freeze-up toward a later date while it is released in fall, with a variety of consequences (e.g., less snow accumulation, reduction in the ice growth rate; e.g., Bitz et al. 1996).

2. Models and experimental setup

a. Model description

In the present study the Community Earth System Model, version 1 (CESM1), is used (Hurrell et al. 2013). CESM1 is a fully coupled global climate model, consisting of model components for the atmosphere, land, sea ice, and ocean. The Community Atmosphere Model, version 4.0 (CAM4), is used as the atmospheric component. CAM4 is a global model with a finite-volume dynamical core, described in detail by Neale et al. (2013). The land component of CESM1 is the Community Land Model, version 4 (CLM4; Lawrence et al. 2011; Oleson et al. 2010).

For the sea ice, the Community Ice Code, version 4 (CICE4; Hunke and Lipscomb 2008) is used, originally developed at the Los Alamos National Laboratory. CICE4 includes elastic–viscous–plastic sea ice dynamics (Hunke and Dukowicz 1997) and energy conserving thermodynamics (Bitz and Lipscomb 1999). Sea ice is subdivided into five categories (following Thorndike et al. 1975) and processes such as ridging and rafting (Rothrock 1975; Lipscomb et al. 2007) and melt ponds are represented. Further, a new shortwave radiation parameterization (Brieglieb and Light 2007), which allows for absorption and scattering of solar radiation within the individual layers of sea ice, melt ponds, and snow, results in a more realistic representation of the surface albedo (Holland et al. 2012).

The default ocean model component is the Parallel Ocean Program, version 2 (POP2; Danabasoglu et al. 2012), developed also at the Los Alamos National Laboratory. In this study, however, a slab-ocean model (SOM) is used instead of the full-depth ocean model POP2, as the coupling with the deep ocean is not important at the seasonal time scales considered here. The SOM is a mixed layer ocean with prescribed lateral ocean heat transport. Although the SOM is a simplified ocean model, it shows similar Arctic climate sensitivity as an active, full-depth ocean model (Kay et al. 2012). The prescribed ocean heat fluxes were calculated from an equilibrated model coupled with a full-scale ocean, POP2, and run for present-day climate conditions (2000 perpetual conditions; Bitz et al. 2012).

For the present study CESM1 was run on a 1.9° × 2.5° horizontal grid for the atmospheric component and on a 1° × 1° grid for the ocean and sea ice components.

b. Experimental design

1) Longwave forcing experiments

The effect of enhanced LWD in spring is implemented by adding a fixed amount of longwave radiation over the Arctic area (Fig. 1, top). This anomaly is compensated for, by reducing LWD over all remaining grid points in the Northern Hemisphere, whereby no net global climate forcing is induced by the Arctic LWD anomaly. Note that the main conclusions drawn in this study do not change in case this compensation is not implemented. The Arctic LWD anomaly is applied only over ocean grid points within an area north of 70° (Fig. 1). No LWD perturbation is added to the Atlantic Ocean side of the Arctic: Barents, Kara, and Greenland Seas. In this area sea ice conditions are governed by more complex atmosphere–ocean interactions and are strongly influenced by the inflow of warm Atlantic water (Smedsrud et al. 2013).

Fig. 1.

(top) Forcing area for the surface downward radiation. (bottom) Longitude section where the downward shortwave radiation is applied. The blue area in the top panel represents all ocean grid points within the chosen study area (black box), where a positive forcing was added. In all other grid points on the Northern Hemisphere a negative forcing was applied to compensate the positive forcing. For the SWD, an additional constraint is applied: solar insolation (direct, near infrared) at 0015 UTC is shown in the bottom panel. The black box indicates the area where the shortwave perturbation is implemented for a particular time step of the day [see section 2b(2) for details].

Fig. 1.

(top) Forcing area for the surface downward radiation. (bottom) Longitude section where the downward shortwave radiation is applied. The blue area in the top panel represents all ocean grid points within the chosen study area (black box), where a positive forcing was added. In all other grid points on the Northern Hemisphere a negative forcing was applied to compensate the positive forcing. For the SWD, an additional constraint is applied: solar insolation (direct, near infrared) at 0015 UTC is shown in the bottom panel. The black box indicates the area where the shortwave perturbation is implemented for a particular time step of the day [see section 2b(2) for details].

2) Shortwave forcing experiments

The SWD anomalies are implemented similar to the LWD anomalies with the following two exceptions. 1) The SWD perturbation is only applied in grid boxes where sunlight is present. To compare with the LWD anomalies applied on the entire area, the SWD anomalies (forcing as well as compensation) are scaled by the sunlit area. 2) The surface albedo in the perturbation and compensation areas is very different because the forcing area is mostly sea ice covered while the compensation area is mostly ice free. To neutralize the global forcing accurately, we additionally scale the compensating SWD by the amount of the absorbed SWD [SWN = (1 − α)SWD, where SWN is the net shortwave radiation at the surface]. Accordingly, the SWD perturbation (W m−2) can be expressed as

 
formula

and the SWD compensation (W m−2) can be expressed as

 
formula

where α is the surface albedo; the indices P and C indicate perturbation and compensation area, respectively; and is the scaling factor for the sunlit area: is the area composing the whole Northern Hemisphere and represents the area with sunlight (Fig. 1). The latter is defined as a box extending from the equator to the pole. The eastern and western boundaries are given by the longitudes of local sunrise or sunset at the equator. The box is shifted in time following the local noon at the center of the box and is repeatedly calculated in each time step. Note that we require that the magnitude of the SWD in all grid boxes within the box needs to exceed the magnitude of the SWD compensation during the entire time period the perturbation is applied (hereafter referred to as the perturbation period; Fig. 1). As surface properties are temporally changing, we calculate the scaling parameters for each perturbation period separately. Further, as solar radiation is absent between fall and spring in the northernmost latitudes, a SWD perturbation is only applied in months that experience sunlight during the entire perturbation period in all grid cells within the aforementioned longitudinal box.

In CAM4 SWD is split into two wave bands and direct and diffuse radiation. These SWD components are dependent on the spatially and seasonally varying atmospheric and surface properties, such as clouds, water vapor, and other atmospheric constituents as well as the surface albedo. Hence, we distribute the perturbation of 12 W m−2 between the four SWD components (visible and near infrared, and direct and diffuse) according to their relative percentage to the total SWD flux within the respective perturbation and compensation area during the perturbation period.

3) Experiments with shortwave and longwave forcing

Clouds reflect shortwave radiation to space and emit longwave radiation to the surface. Hence, a positive LWD anomaly is accompanied by a negative SWD anomaly at the surface. To account for this we designed two experiments where positive LWD anomalies are compensated for by negative SWD anomalies. The effect of clouds on LWD and SWD is estimated by taking the difference between cloudy and clear-sky fluxes from a 20-yr climatology of the control run. To avoid energy inconsistencies within the model, due to negative SWD anomalies, the LWD and SWD anomalies are implemented by adding a relative fraction to the “raw” LWD and SWD values. The fraction is calculated by dividing the corresponding LWD or SWD anomaly (e.g., 12 W m−2) by the climatological mean LWD or SWD over the area and months the anomaly is applied. This fraction is first multiplied and then added to the corresponding LWD or SWD value in each time step. For example, a LWD perturbation of +12 W m−2 during April and May can be expressed as

 
formula

where is the LWD climatology over the perturbation area and April–May (AM). As for the experiments described in sections 2b(1) and 2b(2) a compensating forcing is applied accordingly. Note, that this method removes the necessity of the aforementioned box for the shortwave radiation.

A caveat of this method is that the magnitude of the anomalies cannot be controlled as accurately as for the previously described method: the anomalies are not consistent throughout the period they are applied, they dependent on the absolute magnitude of LWD and SWD. Hence, feedback mechanisms impact the magnitude of the perturbations. However, comparisons between experiments where LWD and SWD perturbations were applied separately according to this method and the LWD and SWD experiments described in sections 2b(1) and 2b(2) show that the magnitude of the perturbations are almost identical, as are the sea ice responses.

4) Statistical significance

An ensemble technique is used to account for internal model variability and to test the statistical significance of the results. Every ensemble consists of 60 members, initiated with different initial conditions. The initial conditions are taken from consecutive years of a 60-yr equilibrated control run. Because of the large ensemble size, the data are approximately normally distributed (central limit theorem; Wilks 2006) and hence a Student’s t test is applied to test the null hypothesis that the ensemble means of the control and perturbed ensemble are equal. Note that in order to analyze one whole year, the run time for each simulation is adjusted according to the months the perturbation is applied, resulting in a total run time ranging from 1 year to 18 months.

5) Determination of seasonal transition

The onset of melt and freeze-up are calculated to study the effects of the flux anomalies on the seasonal transition of sea ice. Here we follow the definition of Rigor et al. (2000) and define the melt onset as the day when the 14-day running median of the surface air temperature (SAT) exceeds −1°C for the first time during spring. Freeze-up is calculated correspondingly as the day when the SAT drops below −1°C for the first time in fall. The melt onset is calculated for each grid cell that has a SIC larger than 80% during 98% of the days during the preceding January through April (Mortin et al. 2014). The same criteria are used for the freeze onset but the ice-cover criteria are applied for the following winter season (January–April).

c. Discussion of experimental design

A large advantage of model experiments, such as those undertaken here, is that hypothetical worlds can be constructed where processes are implemented separately or omitted so that their effects can be studied in isolation but under controlled settings. These worlds are by nature artificial and may have little in common with the real world. However, experiments such as these provide an important alternative to observational analysis aimed at investigating the dynamics of the climate. When it comes to the impact on sea ice of atmospheric variability such as cloud alterations, surface radiation plays a key role and it has long been debated whether large summer ice anomalies are induced by shortwave or longwave radiation anomalies or a combination of the two. Here we take advantage of a model to study the effect on sea ice separately and in combination for the two types of radiation. To our knowledge such a study has not been undertaken before.

However, there are some caveats associated with the experiments. A realistic LWD or SWD anomaly at the surface is triggered by an increase or decrease of clouds and water vapor within the atmosphere. Clouds not only alter the surface energy fluxes but also have a significant impact on atmospheric processes (e.g., precipitation or the atmospheric stability; Pithan et al. 2014). Therefore, clouds alter boundary layer dynamics and trigger processes that feed back to LWD and SWD at the surface. By implementing the LWD and SWD anomalies directly at the surface, these cloud feedbacks are not represented realistically. An alternative approach would be to manipulate the clouds themselves and thereby the effects that they have on surface radiation. However, this method raises several other issues, such as whether to alter the radiative properties of already preexisting clouds or manipulate the cloud cover. In addition such changes would likely provide unforeseen effects such as direct changes of the atmospheric temperatures that would alter the tropospheric stability. Finally, the cloud changes would be difficult to implement so as to maintain control over the amount of surface forcing, chosen here based on results from previous studies using reanalysis (Kapsch et al. 2013). In fact, the approach used in this study allows for a full control of the LWD and SWD perturbations. Thus, this is a dedicated approach for identifying the importance of LWD and SWD anomalies during different seasons.

3. Results

a. The effect of spring downward longwave radiation on sea ice

We first explore the sensitivity of the modeled sea ice to enhanced LWD in spring. Five ensembles with different LWD perturbations over the Arctic area (Fig. 1) during April and May, ranging from +6 to +18 W m−2, were computed. We chose a perturbation of 6 W m−2 as a lower limit, because a positive spring LWD anomaly of on average 6 W m−2 was evident in years with a low September SIE between 1979 and 2010 (Kapsch et al. 2013). In the following we investigate the ensemble mean of the perturbed ensemble relative to the ensemble mean of the control simulations (hereafter referred to as CTRL).

All ensembles show a decreased ice thickness, SIC and SIE in the September succeeding the spring LWD anomaly. For the 6 W m−2 ensemble, September SIC is decreased by about 3% over the area the perturbation was applied, although the response is not statistically significant due to a relatively large interensemble variability. Kapsch et al. (2013, their supplementary information) find a SIC response of about the same magnitude over a similar area. This illustrates that CESM1 is able to realistically simulate the seasonal sea ice response to a spring LWD anomaly, as compared to Kapsch et al. (2013).

It takes a 9 W m−2 perturbation for a significant decrease of the SIC to persist through September (Figs. 2a,b). The energy provided by the LWD perturbation (Fig. 3b) is absorbed by the surface in April and May, as indicated by a positive surface energy balance anomaly (net shortwave and longwave radiation plus turbulent fluxes; Fig. 3d). Since surface temperatures are below the freezing point in April and May, the energy surplus leads to a warming of the ice surface. If surface temperatures reach the melting point of snow and sea ice, additional energy goes into surface melt instead (Fig. 2f). Hence, melt onset starts significantly earlier in the perturbed ensembles than in CTRL (Fig. 4). The earliest melt onset occurs on the Pacific Ocean side of the Arctic and latest on the Atlantic Ocean side.

Fig. 2.

Sea ice properties and anomalies for the experiments with an enhanced downward longwave radiation (LWD) in spring of different magnitudes: (a) sea ice concentration and (b) its anomaly (%), (c) sea ice thickness and (d) its anomaly (m), (e) melt rate (snow accumulation and top and bottom melt) and (f) its anomaly (mm day−1), and (g) sea ice extent and (h) its anomaly (106 km2). All anomalies are calculated as the average differences between each perturbed ensemble and the CTRL ensemble; the results in (a)–(f) are calculated over the area in the top panel of Fig. 1. Gray shadings represent the CTRL ensemble spread (±1σ). Horizontal lines mark statistically significant differences using a Student’s t test (α = 0.05).

Fig. 2.

Sea ice properties and anomalies for the experiments with an enhanced downward longwave radiation (LWD) in spring of different magnitudes: (a) sea ice concentration and (b) its anomaly (%), (c) sea ice thickness and (d) its anomaly (m), (e) melt rate (snow accumulation and top and bottom melt) and (f) its anomaly (mm day−1), and (g) sea ice extent and (h) its anomaly (106 km2). All anomalies are calculated as the average differences between each perturbed ensemble and the CTRL ensemble; the results in (a)–(f) are calculated over the area in the top panel of Fig. 1. Gray shadings represent the CTRL ensemble spread (±1σ). Horizontal lines mark statistically significant differences using a Student’s t test (α = 0.05).

Fig. 3.

As in Fig. 2, but for surface fluxes: (a),(b) LWD; (c),(d) net radiation plus turbulent fluxes (LWN); (e),(f) SWD; (g),(h) SWN; and (i),(j) the sensible plus latent heat fluxes (SH + LH). All fluxes are in W m−2 and are defined positive if downward.

Fig. 3.

As in Fig. 2, but for surface fluxes: (a),(b) LWD; (c),(d) net radiation plus turbulent fluxes (LWN); (e),(f) SWD; (g),(h) SWN; and (i),(j) the sensible plus latent heat fluxes (SH + LH). All fluxes are in W m−2 and are defined positive if downward.

Fig. 4.

(left) The ensemble mean of the timing of (top) melt onset and (bottom) freeze-up for CTRL. The remaining panels show the difference between perturbed simulations and CTRL with various magnitudes of downward longwave perturbation in spring above each column. Black lines indicate significant differences (Student’s t test; α = 0.05). For the differences, negative values indicate an earlier melt and positive values indicate a later freeze-up in the perturbed simulations.

Fig. 4.

(left) The ensemble mean of the timing of (top) melt onset and (bottom) freeze-up for CTRL. The remaining panels show the difference between perturbed simulations and CTRL with various magnitudes of downward longwave perturbation in spring above each column. Black lines indicate significant differences (Student’s t test; α = 0.05). For the differences, negative values indicate an earlier melt and positive values indicate a later freeze-up in the perturbed simulations.

The early melt onset leads to a significant increase in the melt rate of snow and sea ice in all ensembles during April and May (Fig. 2f) and manifests in a more rapid decline of the SIC (Fig. 2b). In April and May, SICs in the perturbed ensembles are significantly reduced as compared to CTRL; note that the anomaly in SIE does not appear until later (Fig. 2h). The increase of the surface temperatures and the reduction of the ice cover result in a larger fraction of open water and an increased energy flux from ocean to atmosphere in the form of latent and sensible heat (Fig. 3j). Because of the latent heat exchange, more water vapor and clouds are present in the atmosphere during spring (Fig. 5). This enhances the atmospheric greenhouse effect and results in an increased emission of LWD toward the surface. This positive feedback acts alongside the direct effect of the LWD perturbation itself, as indicated by a larger positive LWD anomaly at the surface than what is applied (Fig. 3b). Both these anomalies are important contributors to the increased melt and the accelerated decline of the sea ice cover in April and May. Intuitively, the larger the LWD perturbation in spring is, the larger the sea ice cover declines in spring (Fig. 2). Further, the earlier melt onset results in a decreased SIC and lower surface albedo. Hence, more shortwave radiation is absorbed by the surface from April throughout the succeeding summer months (Fig. 3h). This also contributes to an enhanced melt and the reduction of SIC and SIE in the subsequent months. By September SIC is significantly reduced over the perturbation area and the ice becomes almost 10 cm thinner than in CTRL in the 12 W m−2 ensemble (Fig. 2). The sea ice differences between perturbed and CTRL experiments are largest in the central Arctic and the Laptev Sea (Fig. 6). Large areas of open water allow for a release of the additional heat stored in the upper ocean layers due to a reduced ice area in summer. From September to November upward sensible and latent heat fluxes are significantly increased (Figs. 3b,j). This results in a positive water vapor and cloud feedback (Fig. 5b), positively contributing to the negative ice anomalies. The warmer ocean and the related feedback process are to a large extent responsible for a later freeze-up in the April–May ensembles (Fig. 4). The ice remains significantly thinner over the major part of the Arctic until the end of the year (Figs. 2d and 6). In conclusion enhanced LWD in spring can have a significant effect on ice thickness, SIC, and SIE throughout the whole year, provided that it is large enough.

Fig. 5.

As in Fig. 2, but for (a),(b) the vertically integrated cloud water path (liquid and ice; g m−2) and (c),(d) the vertically integrated precipitable water (specific humidity; 102 g m−2). Note the different magnitudes for the units.

Fig. 5.

As in Fig. 2, but for (a),(b) the vertically integrated cloud water path (liquid and ice; g m−2) and (c),(d) the vertically integrated precipitable water (specific humidity; 102 g m−2). Note the different magnitudes for the units.

Fig. 6.

(top) The ensemble mean sea ice thickness and (middle top) sea ice concentration for CTRL for (left)–(right) May, July, September, and November, respectively. (middle bottom),(bottom) The difference in sea ice thickness and concentration between an ensemble with a 12 W m−2 LWD perturbation in April and May and the CTRL ensemble. Negative values relate to a smaller fraction/thinner ice in the perturbed ensemble. Statistically significant differences, using a two-sided Student’s t test (α = 0.05), are bounded by black contours.

Fig. 6.

(top) The ensemble mean sea ice thickness and (middle top) sea ice concentration for CTRL for (left)–(right) May, July, September, and November, respectively. (middle bottom),(bottom) The difference in sea ice thickness and concentration between an ensemble with a 12 W m−2 LWD perturbation in April and May and the CTRL ensemble. Negative values relate to a smaller fraction/thinner ice in the perturbed ensemble. Statistically significant differences, using a two-sided Student’s t test (α = 0.05), are bounded by black contours.

These results are in line with earlier findings stating that melt onset in the Arctic is primarily triggered by an increase in LWD (Maksimovich and Vihma 2012; Persson 2012). The earlier melt onset evokes cloud–water vapor feedbacks in spring and the ice–albedo feedback over summer (Curry et al. 1996). Perovich et al. (2007) point toward the specific importance of the ice–albedo feedback for the seasonal ice evolution; they estimated that for each day of earlier melt, energy sufficient to melt 3 cm of ice is absorbed between April and August. In CESM1 a LWD perturbation of 12 W m−2 in spring shifts the melt onset to on average nine days earlier and thins the ice by about 10 cm. This is 17 cm less than expected based on the findings by Perovich et al. (2007). A possible explanation for this smaller response in CESM1 could be related to a constant lateral ocean heat flux. In a slab-ocean setup the heat transported between grid cells is always constant, as it is prescribed (see section 2b). Thus, the lateral heat transport in CTRL is the same as the lateral heat transport in the perturbed simulations. With LWD perturbation in April–May (LWDAM) the sea ice melts earlier, resulting in larger areas of open water relative to CTRL. As partially or completely ice-free grid cells absorb more energy than ice-covered grid cells, more energy is stored in the ocean with LWDAM. However, the prescribed lateral heat flux prevents the heat from being transported to adjacent grid cells, for example, the heat transport from ocean grid cells to grid cells covered by sea ice. An increased heat exchange between ocean and sea ice grid cells would likely contribute to enhanced lateral and bottom melt in ice-covered grid cells. This shortcoming possibly results in an underestimation of sea ice melt in the present study.

An LWD perturbation of at least 12 W m−2 is necessary to achieve clearly significant responses in all sea ice parameters (SIC, SIE, and sea ice thickness) in a 60-member CESM1 ensemble (Figs. 2b,d). Hence, a perturbation of 12 W m−2 is used in the following to explore the processes in more detail and to investigate how important an LWD perturbation during other months than April and May is for the seasonal sea ice evolution.

b. The effect of seasonally varying downward longwave radiation on sea ice

In the following a LWD perturbation of 12 W m−2 was induced during different months of the year. Almost all ensembles with a positive LWD perturbation show a significant decrease of the SIC and SIE in September (Fig. 7a). The largest negative SIC anomalies develop in the ensembles with increased LWD in April–May, LWD perturbation in June–July (LWDJJ), and LWD perturbation in August–September (LWDAS).

Fig. 7.

As in Fig. 2, but for time series of the anomalies for SWD or LWD perturbations during different months of (a) sea ice concentration (%), (b) sea ice thickness (m), (c) sea ice extent (106 km2), and (d) melt rate (mm day−1).

Fig. 7.

As in Fig. 2, but for time series of the anomalies for SWD or LWD perturbations during different months of (a) sea ice concentration (%), (b) sea ice thickness (m), (c) sea ice extent (106 km2), and (d) melt rate (mm day−1).

In June and July surface temperatures are already around the melting point—the average melt onset over the perturbation area in CTRL occurs around 7 June (Fig. 4). Thus, LWDJJ (Fig. 8a) results in an almost immediate response of the surface in the form of melt (Fig. 7d), indicated by a positive surface energy balance in June and July (Fig. 8c). Because more of the energy goes directly into melt the ice albedo reduces more rapidly for LWDJJ than for LWDAM. Consequently, more shortwave radiation is absorbed by the surface and the ice–albedo feedback for LWDJJ is stronger (Fig. 8d). The effects of the LWD perturbation, together with the increased ice–albedo feedback, are the main contributors to the accelerated melt rate in June and July for LWDJJ (Fig. 7d). The large difference between the September sea ice response for LWDAM and LWDJJ can largely be attributed to the immediate activation of the ice–albedo feedback for LWDJJ. The SWD is significantly reduced during June and July in the ensemble with LWDJJ, which cannot be attributed to feedbacks associated with water vapor and clouds, as was the case for LWDAM (Figs. 8a,b). Cloud water and atmospheric water vapor are not anomalously large for LWDJJ during these months (Figs. 8f,g). Rather, we speculate that the negative SWD anomaly may be a secondary effect of the surface albedo and clouds. Clouds reflect the shortwave radiation that is emitted directly from the sun but also reflect the upward shortwave radiation that is reflected by the surface or the clouds below; the SWD at the surface is the sum of the directly transmitted and reflected radiation. For LWDJJ the ice cover decreases rapidly, leading to a significantly decreased surface albedo (Fig. 7a). The lower the surface albedo, the less SWD is reflected back to the atmosphere and the less is rereflected back to the surface by the clouds. These processes lead to a reduction of the amount of reflected SWD at the surface during June and July for LWDJJ (Fig. 8b). Still, the amount of absorbed shortwave radiation is significantly larger for LWDJJ than in CTRL (Fig. 8d), emphasizing the importance of the solar radiation and the surface–albedo feedback for the sea ice evolution. As in LWDAM the freeze-up in LWDJJ is significantly delayed by on average seven days (Fig. 9), because of the additional ocean heat uptake and aforementioned water vapor–cloud feedback in fall (Figs. 8e–g). Notably, the ice remains significantly thinner until the end of the year (Fig. 7b).

Fig. 8.

Anomalies in (a) LWD, (b) SWD, (c) surface energy balance, (d) net shortwave radiation, and (e) latent and sensible heat fluxes (W m−2); and in (f) vertically integrated cloud water (ice and liquid; g m−2) and (g) precipitable water (102 g m−2) for ensembles with LWD or SWD perturbations in different months. The anomaly is defined as the difference between the perturbed and control ensembles.

Fig. 8.

Anomalies in (a) LWD, (b) SWD, (c) surface energy balance, (d) net shortwave radiation, and (e) latent and sensible heat fluxes (W m−2); and in (f) vertically integrated cloud water (ice and liquid; g m−2) and (g) precipitable water (102 g m−2) for ensembles with LWD or SWD perturbations in different months. The anomaly is defined as the difference between the perturbed and control ensembles.

Fig. 9.

Differences of (top) melt onset and (bottom) freeze onset for four different sets of (left)–(right) LWD and SWD anomalies applied during different months relative to the CTRL ensemble. Negative values indicate a later melt and positive values a later freeze-up in the perturbed simulations. LWD anomalies are induced in June–July (JJ) and August–September (AS), and SWD anomalies are induced in April–May (AM) and JJ.

Fig. 9.

Differences of (top) melt onset and (bottom) freeze onset for four different sets of (left)–(right) LWD and SWD anomalies applied during different months relative to the CTRL ensemble. Negative values indicate a later melt and positive values a later freeze-up in the perturbed simulations. LWD anomalies are induced in June–July (JJ) and August–September (AS), and SWD anomalies are induced in April–May (AM) and JJ.

The processes that contribute to the ice anomaly in the ensemble with LWDAS are similar to those associated with LWDJJ. The surface temperatures are near the melting point of snow and sea ice; thus, most of the positive LWD anomaly leads almost instantaneously to an increased snow and/or ice melt (Figs. 8c and 7d). The subsequent response in SIC initiates a significant ice–albedo feedback, although not as strong as in LWDJJ (Fig. 8d). The response is likely weaker as solar insolation is already reduced through the seasonal solar cycle. Further, the ice–albedo feedback has a shorter time to act than in LWDJJ, thus, allowing for less absorption of energy and consequently a weaker sea ice response (Figs. 7a and 8d). Nevertheless, the additional ocean heat uptake leads to a significant delay of the fall freeze-up over the entire perturbation region (Fig. 9). As discussed earlier additional ocean heat uptake leads to a reduced ice growth, thus, the ice remains significantly thinner until the end of the year (Fig. 7).

All other ensembles with a LWD perturbation in the preceding winter [October–November (LWDON), December–January (LWDDJ), and February–March (LWDFM)] exhibit only small anomalies in terms of SIC and thickness (Fig. 7b). The ice thickness is reduced by only a few centimeters over the perturbation area during the months succeeding the perturbation (Fig. 7b). A significant decrease of the SIC is only evident in late summer/fall in the LWDFM ensemble, when the ice is at its annual minimum (Figs. 7a,b). One reason for the relatively small response is that the temperatures are far below the freezing point. The energy provided by the LWD perturbation is not sufficient to raise surface temperatures above the freezing point and to trigger melt (Figs. 7d and 8c). However, in the LWDFM ensemble the extra energy likely reduces the ice growth during these months, leading to thinner ice when melt starts later in spring. Consequently, the LWDFM ensemble shows an earlier melt onset (not shown). This is consistent with findings by Liu and Key (2014) who argued that below-average clouds in the winter of 2012 led to a surface cooling, allowing for a greater ice growth and a thickening of the sea ice cover. However, in the present study the surface response to enhanced LWD in winter is not statistically significant and much smaller than the response to a LWD perturbation in other months, indicating a relative small contribution of winter clouds to September ice anomalies. Note, that Liu and Key (2014) only considered a single year. The impact of LWD on the seasonal sea ice evolution has a large variability due to interannual variability of the climate system; there are single ensemble members in our experiment that have a much stronger response. Here we considered a 60-member ensemble in order to filter out such variability.

One reason for the different sea ice response in different months is that the chains of processes, which lead to the respective ice anomalies, differ depending on the timing of the LWD perturbation. A LWD perturbation in April–May results in an early melt onset, which evokes several feedback processes that positively contribute to ice melt; such as cloud, water vapor, and the ice–albedo feedback (section 3a). Extra energy in June–July has a more direct effect on the sea ice, as most of the energy goes directly into melt and immediately sets off a significant ice–albedo feedback that acts during the summer. The same is true for a perturbation in late summer (August–September); however, for the late summer perturbation the ice–albedo feedback has less time to act, hence, the magnitude of the September sea ice anomaly is smaller. The simulations indicate that the ice–albedo effect is important for development of the September SIE, but also that the relative timing of the extra forcing and the feedback processes are important in determining the sequence of events.

c. The role of downward shortwave radiation for the seasonal sea ice evolution

The role of SWD for the sea ice evolution is widely discussed in the literature (e.g., Nussbaumer and Pinker 2012; Choi et al. 2014). Here we explore the effect of SWD anomalies on the seasonal sea ice evolution in a similar way as for the LWD anomalies. The SWD perturbation was only applied in April–May (SWDAM) and June–July (SWDJJ).

An increased SWD in spring, of the same magnitude as that applied with LWD, has no significant impact on the SIC and thickness over the perturbation area (Fig. 7a). In April and May most of the Arctic is covered by snow and sea ice; in CTRL the melt onset does not occur before 7 June (Fig. 4). Because of a high surface albedo, of about 0.8, most of the extra SWD is reflected back toward the atmosphere (Fig. 8d). Less than 20% of the radiation associated with the SWD perturbation is absorbed by the surface, leading to a slightly higher melt rate than in CTRL and an earlier melt onset of about three days (Fig. 9). On an Arctic-wide scale this, however, has no significant impact on the SIC or thickness. Hence, no ice–albedo feedback is triggered and the ice cover does not change significantly from the ice cover in CTRL.

In June, the month of melt onset, the surface albedo in the model is reduced to about 0.6. Thus, a large amount of the energy provided by the SWD perturbation in the ensemble with SWDJJ is absorbed by the ice and ocean surface (Fig. 8b). A large part of this energy goes directly into ice melt, leading to an accelerated melt rate in both June and July (Fig. 7d); thus, this ensemble is, in a sense, similar to LWDJJ. The more melt the larger the ice–albedo feedback, which acts to amplify the melt (Fig. 8d). By September SIC is significantly decreased and the ice is thinner than in CTRL (Fig. 7b). As for the LWDAM and LWDJJ ensembles, the SWDJJ ensemble is also associated with an anomalously low ice cover in fall, which initiates cloud and water vapor feedbacks that lead to a warming of the surface and a later fall freeze-up over the East Siberian Sea (Fig. 9). Note, that the September ice anomaly for SWDJJ shows a similar magnitude as the ice anomalies for LWDAM and LWDAS but is less pronounced than for LWDJJ (Figs. 7a,b). This can be contributed to the high surface albedo in June, which allows only about half of the SWD to be absorbed by the surface.

d. The role of clouds for the seasonal sea ice evolution

To relate these findings to the discussion on clear or cloudy conditions favoring sea ice melt (Kay et al. 2008; Schweiger et al. 2008; Graversen et al. 2011), we performed experiments where both the LWD and SWD anomalies where implemented simultaneously. In April–May the effect of clouds on SWD is 1.5 times larger than on LWD, and in June–July the corresponding fraction is 2.8 [see section 2b(3)]. Thus, two experiments were conducted, one with anomalies of 12 W m−2 of LWD and −18 W m−2 of SWD in April–May and the other with 12 W m−2 of LWD and −34 W m−2 of SWD in June–July (Fig. 10). The results suggest that the cloud LWD effect on sea ice dominates over the SWD effect during spring, hence, a cloudier sky has a negative impact on the sea ice evolution; it leads to a negative September sea ice anomaly. In summer, however, cloudier conditions do not result in a significantly negative SIC anomaly; longwave warming balances shortwave cooling. The findings are in favor of the results by Schweiger et al. (2008), who found that negative cloud anomalies during June and July did not substantially contribute the record-low September sea ice extent in 2007. Rather, on an Arctic-wide scale the SWD and LWD effect of clouds cancel each other during summer.

Fig. 10.

As in Fig. 7, but for ensembles with a 12 W m−2 LWD perturbation in April–May (AM) and June–July (JJ), and a 12 W m−2 LWD perturbation paired with a −18 W m−2 SWD perturbation in April–May (LWD/SWD AM) and a −34 W m−2 SWD perturbation in June–July (LWD/SWD JJ).

Fig. 10.

As in Fig. 7, but for ensembles with a 12 W m−2 LWD perturbation in April–May (AM) and June–July (JJ), and a 12 W m−2 LWD perturbation paired with a −18 W m−2 SWD perturbation in April–May (LWD/SWD AM) and a −34 W m−2 SWD perturbation in June–July (LWD/SWD JJ).

The results are to a large extent consistent with earlier studies investigating the seasonal effect of clouds on the September ice extent. For example Eastman and Warren (2010) found that the cloud cover in spring (March–May) is significantly anticorrelated with the September SIE, which is agreement with the results presented in the present study. Further, they found no significant correlation between the total cloud cover in summer and the September SIE.

4. Conclusions

The role of clouds and water vapor for the seasonal sea ice evolution has widely been discussed (e.g., Kay et al. 2008; Nussbaumer and Pinker 2012; Kapsch et al. 2013; Liu and Key 2014). One of the most important local effect of clouds is the alteration of the downward radiation to the surface. In this study a comprehensive set of model experiments is conducted in order to investigate the impact of surface downward radiation in the Arctic sea ice area. Special attention is drawn to the dependence of the September sea ice extent on the seasonal presence of radiation anomalies. The effect of downward longwave and shortwave radiation on sea ice is studied separately and in combination. Hereby the present study follows up on the discussion by Kay et al. (2008), Schweiger et al. (2008), Graversen et al. (2011), and Kapsch et al. (2013) regarding the role of anomalous clear and cloudy sky for the Arctic September sea ice extent. The results largely support previous findings, indicating that a positive (negative) anomaly in longwave radiation in spring leads to a negative (positive) anomaly in SIE in September (Graversen et al. 2011; Kapsch et al. 2013).

The following key conclusions are drawn:

  1. The timing of the LWD perturbation is important. A significant impact on the September SIC and SIE is only evident when the LWD perturbations are induced during months when surface temperatures are lower than but approach the melting point of snow and ice—for example, in spring (April–May) or in summer (June–September) when melting has already started. An LWD perturbation during winter (October–March) has only a minor effect on the ice cover.

  2. The earlier a LWD perturbation is applied during summer, the more time for feedback processes to interact with the surface and the larger the September SIC anomaly.

  3. A positive LWD perturbation has a larger impact on the September sea ice extent than an equivalent SWD perturbation during the same time. The energy provided by a LWD anomaly goes directly into melt, while a large part of the anomalous SWD is reflected by the surface.

  4. In spring and summer a given cloud anomaly alters the downward shortwave radiation more than the downward longwave radiation. However, for the effects on the sea ice, the response in the surface temperature and the albedo need to be taken into account. A positive LWD anomaly in spring, representing an increase of clouds, causes a decline of the sea ice even if compensated for by the corresponding negative SWD anomaly induced by that cloud increase, which is double the magnitude of the LWD anomaly. A similar experiment during June–July, with an almost tripling of the negative SWD perturbation relative to that of the positive LWD anomaly, reveals that during summer the LWD and SWD effect of clouds on the sea ice cancel each other; within this experimental set-up, summer clouds seem to have little effect on the ice cover.

In this study only the effect of Arctic-wide LWD and SWD anomalies is investigated by applying a simplified forcing in a coupled climate model. Future work should investigate what effect regional cloud cover anomalies have on the sea ice thickness, SIC, and SIE and ice thickness. Future research should also identify the origin of cloud anomalies in different times of the season; are the cloud anomalies of local or remote origin? Kapsch et al. (2013) argued that LWD anomalies in spring are triggered by advection of humid air in spring. Identifying the origin of cloud or water vapor anomalies could be advantageous for seasonal sea ice predictions, considering the importance of clouds for the ice evolution.

Acknowledgments

The CESM project is supported by the National Science Foundation and the Office of Science (BER) of the U.S. Department of Energy. The CESM1.1 simulations were performed on resources provided by the Swedish National Infrastructure for Computing (SNIC) at Linköping, Sweden, under the project SNIC 2014/1-244: “Arctic atmospheric variability and its implication for summer sea-ice melt.” We thank Hamish Struthers at SNIC for his assistance in setting up CESM1.1 on Triolith. Special thanks to Jonas Mortin for helpful discussions around the analysis of melt and freeze onset.

REFERENCES

REFERENCES
Bitz
,
C. M.
, and
W. H.
Lipscomb
,
1999
: An energy-conserving thermodynamic model of sea ice. J. Geophys. Res., 104, 15 669–15 677, doi:.
Bitz
,
C. M.
,
D. S.
Battisti
,
R. E.
Moritz
, and
J. A.
Beesley
,
1996
:
Low-frequency variability in the Arctic atmosphere, sea ice, and upper-ocean climate system
.
J. Climate
,
9
,
394
408
, doi:.
Bitz
,
C. M.
,
K. M.
Shell
,
P. R.
Gent
,
D. A.
Bailey
,
G.
Danabasoglu
,
K. C.
Armour
,
M. M.
Holland
, and
J. T.
Kiehl
,
2012
:
Climate sensitivity of the Community Climate System Model, version 4
.
J. Climate
,
25
,
3053
3070
, doi:.
Briegleb
,
B. P.
, and
B.
Light
,
2007
: A Delta-Eddington multiple scattering parametrization for solar radiation in the sea ice component of the Community Climate System Model. NCAR Tech. Note NCAR/TN-472+STR, NCAR, 100 pp. [Available online at http://nldr.library.ucar.edu/repository/assets/technotes/TECH-NOTE-000-000-000-832.pdf.]
Choi
,
Y.-S.
,
B.-M.
Kim
,
S.-K.
Hur
,
S.-J.
Kim
,
J.-H.
Kim
, and
C.-H.
Ho
,
2014
:
Connecting early summer cloud-controlled sunlight and late summer sea ice in the Arctic
.
J. Geophys. Res. Atmos.
,
119
,
11 087
11 099
, doi:.
Curry
,
J. A.
,
W. B.
Rossow
,
D.
Randall
, and
J. L.
Schramm
,
1996
:
Overview of Arctic cloud and radiation characteristics
.
J. Climate
,
9
,
1731
1764
, doi:.
Danabasoglu
,
G.
,
S. C.
Bates
,
B. P.
Briegleb
,
S. R.
Jayne
,
M.
Jochum
,
W. G.
Large
,
S.
Peacock
, and
S. G.
Yeager
,
2012
:
The CCSM4 ocean component
.
J. Climate
,
25
,
1361
1389
, doi:.
Devasthale
,
A.
,
J.
Sedlar
,
T.
Koenigk
, and
E. J.
Fetzer
,
2013
:
The thermodynamic state of the Arctic atmosphere observed by AIRS: Comparisons during the record minimum sea ice extents of 2007 and 2012
.
Atmos. Chem. Phys.
,
13
,
7441
7450
, doi:.
Eastman
,
R.
, and
S. G.
Warren
,
2010
:
Arctic cloud changes from surface and satellite observations
.
J. Climate
,
23
,
4233
4242
, doi:.
Francis
,
J. A.
,
E.
Hunter
,
J. R.
Key
, and
X.
Wang
,
2005
:
Clues to variability in Arctic minimum sea ice extent
.
Geophys. Res. Lett.
,
32
,
L21501
, doi:.
Graversen
,
R. G.
,
T.
Mauritsen
,
S.
Drijfhout
,
M.
Tjernström
, and
S.
Mårtensson
,
2011
:
Warm winds from the Pacific caused extensive Arctic sea-ice melt in summer 2007
.
Climate Dyn.
,
36
,
2103
2112
, doi:.
Holland
,
M. M.
,
D. A.
Bailey
,
B. P.
Briegleb
,
B.
Light
, and
E.
Hunke
,
2012
:
Improved sea ice shortwave radiation physics in CCSM4: The impact of melt ponds and aerosols on Arctic sea ice
.
J. Climate
,
25
,
1413
1430
, doi:.
Hunke
,
E. C.
, and
J. K.
Dukowicz
,
1997
:
An elastic–viscous–plastic model for sea ice dynamics
.
J. Phys. Oceanogr.
,
27
,
1849
1867
, doi:.
Hunke
,
E. C.
, and
W. H.
Lipscomb
,
2008
: CICE: The Los Alamos sea ice model documentation and software user’s manual version 4.1. Los Alamos National Laboratory Tech. Rep. LA-CC-06-012, 76 pp.
Hurrell
,
J. W.
, and Coauthors
,
2013
:
The Community Earth System Model: A framework for collaborative research
.
Bull. Amer. Meteor. Soc.
,
94
,
1339
1360
, doi:.
IPCC
,
2013
: Climate Change 2013: The Physical Science Basis. Cambridge University Press, Cambridge, 1535 pp, doi:.
Kapsch
,
M.-L.
,
R. G.
Graversen
, and
M.
Tjernström
,
2013
:
Springtime atmospheric energy transport and the control of Arctic summer sea-ice extent
.
Nat. Climate Change
,
3
,
744
748
, doi:.
Kay
,
J. E.
,
T.
L’Ecuyer
,
A.
Gettelman
,
G.
Stephens
, and
C.
O’Dell
,
2008
:
The contribution of cloud and radiation anomalies to the 2007 Arctic sea ice extent minimum
.
Geophys. Res. Lett.
,
35
,
L08503
, doi:.
Kay
,
J. E.
,
M. M.
Holland
,
C. M.
Bitz
,
E.
Blanchard-Wrigglesworth
,
A.
Gettelman
,
A.
Conley
, and
D.
Bailey
,
2012
:
The influence of local feedbacks and northward heat transport on the equilibrium Arctic climate response to increased greenhouse gas forcing
.
J. Climate
,
25
,
5433
5450
, doi:.
Lawrence
,
D. M.
, and Coauthors
,
2011
:
Parameterization improvements and functional and structural advances in version 4 of the Community Land Model
.
J. Adv. Model. Earth Syst.
,
3
,
M03001
, doi:.
Lipscomb
,
W. H.
,
E. C.
Hunke
,
W.
Maslowski
, and
J.
Jakacki
,
2007
:
Ridging, strength, and stability in high-resolution sea ice models
.
J. Geophys. Res.
,
112
,
C03S91
, doi:.
Liu
,
Y.
, and
J. R.
Key
,
2014
:
Less winter cloud aids summer 2013 Arctic sea ice return from 2012 minimum
.
Environ. Res. Lett.
,
9
,
044002
, doi:.
Maksimovich
,
E.
, and
T.
Vihma
,
2012
:
The effect of surface heat fluxes on interannual variability in the spring onset of snow melt in the central Arctic Ocean
.
J. Geophys. Res.
,
117
,
C07012
, doi:.
Maslanik
,
J. A.
,
C.
Fowler
,
J.
Stroeve
,
S.
Drobot
,
J.
Zwally
,
D.
Yi
, and
W.
Emery
,
2007
:
A younger, thinner Arctic ice cover: Increased potential for rapid, extensive sea-ice loss
.
Geophys. Res. Lett.
,
34
,
L24501
, doi:.
Mortin
,
J.
,
S. E. L.
Howell
,
L.
Wang
,
C.
Derksen
,
G.
Svensson
, and
R. G.
Graversen
,
2014
:
Extending the QuikSCAT record of seasonal melt–freeze transitions over Arctic sea ice using ASCAT
.
Remote Sens. Environ.
,
141
,
214
230
, doi:.
Neale
,
R. B.
,
J.
Richter
,
S.
Park
,
P. H.
Lauritzen
,
S. J.
Vavrus
,
P. J.
Rasch
, and
M.
Zhang
,
2013
:
The mean climate of the Community Atmosphere Model (CAM4) in forced SST and fully coupled experiments
.
J. Climate
,
26
,
5150
5168
, doi:.
Nussbaumer
,
E. A.
, and
R. T.
Pinker
,
2012
:
The role of shortwave radiation in the 2007 Arctic sea ice anomaly
.
Geophys. Res. Lett.
,
39
,
L15808
, doi:.
Ogi
,
M.
, and
J. M.
Wallace
,
2007
:
Summer minimum Arctic sea ice extent and the associated summer atmospheric circulation
.
Geophys. Res. Lett.
,
34
,
L12705
, doi:.
Oleson
,
K. W.
, and Coauthors
,
2010
: Technical description of version 4.0 of the Community Land Model (CLM). NCAR Tech. Note NCAR/TN-478+STR, 257 pp. [Available online at http://www.cesm.ucar.edu/models/cesm1.0/clm/CLM4_Tech_Note.pdf.]
Perovich
,
D. K.
,
B.
Light
,
H.
Eicken
,
K. F.
Jones
,
K.
Runciman
, and
S. V.
Nghiem
,
2007
:
Increasing solar heating of the Arctic Ocean and adjacent seas, 1979–2005: Attribution and role in the ice–albedo feedback
.
Geophys. Res. Lett.
,
34
,
L19505
, doi:.
Persson
,
P.
,
2012
:
Onset and end of the summer melt season over sea ice: Thermal structure and surface energy perspective from SHEBA
.
Climate Dyn.
,
39
,
1349
1371
, doi:.
Pithan
,
F.
,
B.
Medeiros
, and
T.
Mauritsen
,
2014
:
Mixed-phase clouds cause climate model biases in Arctic wintertime temperature inversions
.
Climate Dyn.
,
43
,
289
303
, doi:.
Rigor
,
I. G.
,
R. L.
Colony
, and
S.
Martin
,
2000
:
Variations in surface air temperature observations in the Arctic, 1979–97
.
J. Climate
,
13
,
896
914
, doi:.
Rothrock
,
D. A.
,
1975
:
The energetics of the plastic deformation of pack ice by ridging
.
J. Geophys. Res.
,
80
,
4514
4519
, doi:.
Rothrock
,
D. A.
,
D. B.
Percival
, and
M.
Wensnahan
,
2008
:
The decline in arctic sea-ice thickness: Separating the spatial, annual, and interannual variability in a quarter century of submarine data
.
J. Geophys. Res.
,
113
,
C05003
, doi:.
Schweiger
,
A. J.
,
J.
Zhang
,
R. W.
Lindsay
, and
M.
Steele
,
2008
:
Did unusually sunny skies help drive the record sea ice minimum of 2007?
Geophys. Res. Lett.
,
35
,
L10503
, doi:.
Screen
,
J. A.
,
I.
Simmonds
, and
K.
Keay
,
2011
: Dramatic interannual changes of perennial Arctic sea ice linked to abnormal summer storm activity. J. Geophys. Res., 116, D15105, doi:.
Serreze
,
M. C.
,
M. M.
Holland
, and
J.
Stroeve
,
2007
:
Perspectives on the Arctic’s shrinking sea-ice cover
.
Science
,
315
,
1533
1536
, doi:.
Smedsrud
,
L. H.
, and Coauthors
,
2013
:
The role of the Barents Sea in the Arctic climate system
.
Rev. Geophys.
,
51
,
415
449
, doi:.
Thorndike
,
A. S.
,
D. A.
Rothrock
,
G. A.
Maykut
, and
R.
Colony
,
1975
:
The thickness distribution of sea ice
.
J. Geophys. Res.
,
80
,
4501
4513
, doi:.
Wang
,
J.
,
J.
Zhang
,
E.
Watanabe
,
M.
Ikeda
,
K.
Mizobata
,
J. E.
Walsh
,
X.
Bai
, and
B.
Wu
,
2009
:
Is the Dipole Anomaly a major driver to record lows in Arctic summer sea ice extent?
Geophys. Res. Lett.
,
36
,
L05706
, doi:.
Wilks
,
D. S.
,
2006
: Statistical Methods in the Atmospheric Sciences. 2nd ed. Academic Press, 627 pp.
Woods
,
C.
,
R.
Caballero
, and
G.
Svensson
,
2013
:
Large-scale circulation associated with moisture intrusions into the Arctic during winter
.
Geophys. Res. Lett.
,
40
,
4717
4721
, doi:.
Wu
,
B.
,
J. E.
Overland
, and
R.
D’Arrigo
,
2012
:
Anomalous Arctic surface wind patterns and their impacts on September sea ice minima and trend
.
Tellus
,
64A
, 18590, doi:.