1. Introduction
One of the most visible manifestations of recent global climate change has been the decrease of Arctic sea ice cover at the end of the September melt season. Recent years have exhibited record minimum September extents (e.g., 2007 and 2012), relative to all other years in the satellite record, as well as an accelerated trend of decline in the minimum September extent (e.g., Comiso et al. 2008). To help explain the decrease in the September minimum, many prior studies have focused on physical mechanisms that may explain the observed sea ice variability or the decreasing sea ice extent. The physical mechanisms are pathways by which the variability in the Arctic sea ice is connected to another phenomenon or region in the earth’s climate system. These physical mechanisms contain one or more components (or building blocks) that together help build the connection and include direct ocean influences (e.g., Shimada et al. 2006) and thermodynamical and dynamical atmospheric influences.
Important atmospheric thermodynamical influences on Arctic sea ice trend and variability include anomalies in components of the surface energy budget, such as positive near-surface downwelling shortwave (Kay and Gettelman 2009) and longwave (Dong et al. 2014) radiation anomalies. Important atmospheric dynamical influences include anomalous poleward energy transport into the Arctic (Graversen et al. 2011), anomalous transport of sea ice out of the Arctic Basin by persistent anomalous winds (Ogi et al. 2008; J. Zhang et al. 2008; Ogi and Wallace 2012), anomalous regional atmospheric circulation patterns (Wang et al. 2009), abnormal summer storm activity (Screen et al. 2011), and the influence of important low-frequency modes and teleconnection patterns [e.g., the Pacific–North American (PNA) pattern (L’Heureux et al. 2008); the Arctic Oscillation (AO) and North Atlantic Oscillation (NAO) (Yamamoto et al. 2006; Ukita et al. 2007)]. All of the abovementioned components are part of proposed individual mechanisms that help explain sea ice variability on an interannual or decadal scale.
Another change that has been connected to the trend in the September minimum extent is the length of Arctic sea ice melt season, defined as the period from the first melt of the ice in boreal spring to the start of freeze-up in autumn. Several studies have noted that the length of the melt season leading up to the September minimum has become longer, with a trend toward both an earlier initial date of melt and later date of autumn freeze-up (Belchansky et al. 2004; Markus et al. 2009; Stroeve et al. 2014). The decrease in surface albedo during melt increases the amount of incoming solar radiation absorbed at the surface during the melt (Perovich et al. 2007). Additionally, the breakdown of negative feedbacks in energy flux that is unique to a melting sea ice surface helps support large surface net radiative fluxes. Surface turbulent and convective energy fluxes and upwelling longwave radiation fluxes become insensitive to downwelling radiative flux anomalies over melting sea ice (Persson 2012). Thus, the earlier onset of melt potentially contributes to the trend in the September minimum sea ice extent by increasing the cumulative amount of longwave and shortwave radiation absorbed at the surface (e.g., Stroeve et al. 2014). Additionally, the total area of multiyear ice (i.e., sea ice that survives at least one melt season) has decreased and been replaced with first-year ice (Maslanik et al. 2007). First-year ice exhibits a greater albedo decrease compared to multiyear ice when melting first occurs and meltwater appears on the ice surface (Perovich and Polashenski 2012); thus, the recent change in sea ice age has potentially added to the excess amount of extra incoming shortwave radiation absorbed at the surface when an earlier melt occurs.
Both the date of melt onset and freeze-up of Arctic sea ice at point locations have been shown to be affected by atmospheric synoptic eddies in individual cases. In the analysis of the initial melt of sea ice at the Surface Heat Budget of the Arctic Ocean (SHEBA) experiment site, Persson (2012) showed that the onset of melt occurred within the warm sector of an open frontal wave. Synoptic-scale eddies before the melt onset also helped warm the surface ice and snow layer toward the melting point. In terms of the surface energy budget, these synoptic eddy events increased downwelling longwave radiation, thus increasing the net longwave radiation input at the surface. The combination of cloud cover and above-freezing air aloft transported poleward by the individual synoptic eddies contributed to the increase in downwelling radiation that helped warm the surface toward the melting point. Synoptic-scale eddies were shown to be similarly important for the initial melt of landfast ice in the Canadian Archipelago (Else et al. 2014). Other studies have stated that the same transient warm-air advection and cloud cover help drive rapid sea ice melt in individual cases later in the melt season (Tjernström et al. 2012). Thus, understanding how atmospheric eddies interact with sea ice melt is important to understanding sea ice extent variability throughout the entire melt season.
Given the potential importance of the lengthening of the melt season, particularly the trend toward an earlier Arctic-mean melt, we investigate further the role of atmospheric eddies as a trigger to the initial melt of Arctic sea ice. The purpose of this paper is to explore how atmospheric eddies contribute to the interannual variability and trend in the Arctic-mean melt onset date of sea ice. Also, we examine the atmospheric eddy contribution to individual events where the melt onset of Arctic sea ice occurs over a large area. We view the eddy effect on the Arctic-mean and local melt date as a physical mechanism with components that include both the eddy influence on meridional heat and moisture transport from lower latitudes and the eddy influence on the downwelling longwave and shortwave radiation at the surface through changes in cloud cover. Thus, we take an integrative view of the mechanism that is responsible for the variability and trend. To expand on previous studies of the interaction between atmospheric eddies and the initial melt of sea ice, we use gridded surface radiative flux data to look at the spatial distribution of radiative flux anomalies associated with individual cases of early melt onset. We also focus on the influence of different types of atmospheric eddies, classified by their frequencies (i.e., high-, low-, and subseasonal-frequency eddies with periods of 2–7, 10–30, and 30–90 days, respectively), to determine which type is most crucial in triggering the melt in the individual cases . Following this introduction, we outline the data and methods in section 2, and present the major results in section 3. A summary of the results and concluding remarks are presented in section 4.
2. Data and methods
The melt onset date data are taken from version 3 of the melt onset data archived at the National Snow and Ice Data Center (NSIDC; Bliss and Anderson 2014). In this dataset, the melt onset is calculated using the advanced horizontal range algorithm (AHRA), first described in Drobot and Anderson (2001). In the AHRA, the melt onset date is derived from the difference between microwave brightness temperatures at two different frequencies from the Scanning Multichannel Microwave Radiometer (SMMR) and the Special Sensor Microwave Imager (SSM/I)–Special Sensor Microwave Imager/Sounder (SSMIS) satellite instruments. The difference between the brightness temperatures is a function of the presence or absence of liquid water on a frozen surface, where the difference exceeding a certain threshold indicates the presence of meltwater at the surface. The first day that the brightness temperature difference exceeds the threshold at a particular grid point in the data is considered the melt onset date. To account for noise and short transitory melt onset events that may adversely affect the accuracy of the brightness temperature data and result in a spuriously early melt onset date, Bliss and Anderson apply a 10-day time series averaging window to the melt onset data in the creation of their dataset. The data exist on an Equal-Area Scalable Earth (EASE) grid with a spatial resolution of 25 km and covers the period 1979–2012. This dataset contains an area of missing data around the pole north of 84.5°N latitude for the period 1979–87 and north of 87.2°N for the period after 1987. For consistency, when calculating the trend in initial melt date over the period 1979–2012, we consider only the melt onset date data south of 84.5°N for all years. The temporal resolution of the data is daily after 1987 and 2 days prior to 1987. When quantifying the total melt on a given day in a particular region, we count the total number of grid boxes that exhibit melt on that date in the dataset. For dates before 1987, the daily count is calculated as half of the total count in the raw data if data exist on that date and as half of the count recorded on the previous day if data do not. The mean melt onset date at any location is simply the mean of the time series of initial melt dates at that location. The mean melt onset date over a defined area is the mean of the collection of mean melt dates over the defined area.
Two datasets are used for the calculation of atmospheric quantities. For heat transport, temperature, and wind quantities, we use the NASA Modern-Era Retrospective Analysis for Research and Applications (MERRA) reanalysis dataset on a 0.5° latitude × 0.67° longitude grid (Rienecker et al. 2011). The MERRA reanalysis dataset represents the state of the atmosphere over the Arctic very well, and it has been shown to be among the reanalysis datasets that best match independent observations of atmospheric conditions in the Arctic (Lindsay et al. 2014). For longwave and shortwave radiative flux data, we use datasets from the NASA Clouds and the Earth’s Radiant Energy System (CERES) project (Wielicki et al. 1996) that contain all-sky longwave and shortwave surface fluxes. These daily data, the level 3 synoptic 1° (SYN1deg) surface product, covers the period 2000–14. The data are derived from observations by the Terra and Aqua satellites, part of the NASA Earth Observing System (EOS). This product is also the source of cloud fraction data used in the analysis. For the estimate of the bias and error of the CERES longwave and shortwave fluxes, we utilize ground truth flux data and directly comparable CERES flux data provided by the CERES/ARM Validation Experiment (CAVE; Rutan et al. 2001).
When calculating spatial averages, we define the Arctic polar cap as all points north of 70°N. This latitude boundary is also the boundary over which we calculate the meridional heat transport. The Arctic cap mean melt date is the average melt that occurs within the polar cap. When looking at individual sectors across the latitude circle, we longitudinally average over 5° sectors before plotting the data. The melt date mean (i.e., the mean melt date over the sector) and melt count (i.e., the number of grid boxes that exhibit melt for that date) are done over these 5° longitude sectors and are bounded by 70° and 84.5°N, except in the time–longitude plots for individual years. For the melt date mean, we focus on boreal spring and early summer melt. Subsequently, melt dates before day 90 (i.e., before 31 March) are discarded before calculating the mean.
3. Results
a. Trend in mean melt date and eddy heat transport climatology
There is a strong association between the mean melt onset date over the entire Arctic polar cap and the September minimum sea ice extent. Figure 1 shows the time series of the September minimum sea ice extent and the mean melt onset date across the polar cap north of 70°N. A strong positive correlation (r = 0.74, significant at the 99% level) exists between the mean melt date and the September minimum for the entire 1979–2012 period, such that lower September minimum sea ice extent values are associated with earlier polar cap mean initial melt dates. It is plausible that the coincident declining trends in the two time series are physically linked. Stroeve et al. (2014) noted that the trend of an earlier melt onset since 1979 has led to a large increase in additional heat stored in the upper ocean during the sea ice melt season, leading to thinner ice and reduced Arctic sea ice extent at the end of the melt season. It is possible that the declining trend of the two time series presented in Fig. 1 are connected by this mechanism, or that the correlation of the trends is due to an independent physical mechanism, such as the Arctic-wide increase in 925-hPa temperature occurring in all seasons (Stroeve et al. 2014). When comparing the detrended time series, the correlation is reduced and no longer statistically significant (r = 0.12). The lack of correlation between the detrended time series may be due to the separation in time between the spring melt onset and September minimum extent, where in that time gap other independent mechanisms acting during the melt season modify the evolution of sea ice melt and play more critical roles in determining the total amount of sea ice melt in individual seasons.
Plot of Arctic polar cap mean melt date (left y axis; days after 1 Jan) and the September minimum sea ice extent (right y axis; km2) time series from 1979 to 2012. Polar cap is defined as the area above 70°N.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
Two important characteristics in the declining trend of the September minimum extent and Arctic-mean melt onset date appear in Fig. 1. In the September minimum extent time series, the declining trend is not constant but accelerating after year 2000. This is consistent with the acceleration in the September minimum extent decline noted by Comiso et al. (2008). The melt onset time series shows a sudden steplike decrease in 1988–90, followed by a more gradual decrease. Both of these trend patterns will appear again in the subsequent analysis.
A closer look at the longitudinal distribution of the mean melt onset date reveals that the trend in the mean melt onset date is not evenly distributed across all longitudes. In Fig. 2a, the values shown are the mean melt values in a 5° longitude sector, bounded by 70° and 84.5°N. Two interesting features exist in this figure. First, there is a clear steplike decrease in the melt date from 90°E to 130°W, similar to the decline in the Arctic cap mean melt onset date in Fig. 1. The area where the steplike decrease is occurring corresponds to an area that extends eastward across the Laptev, East Siberian, Chukchi, and Beaufort Seas. Since this region is adjacent to the Siberian and Alaskan coasts, we will refer to this sector as the Siberian–Alaskan sector throughout the manuscript. The decrease is most pronounced around and after 1990 and 2000. We speculate that the steplike nature of the decline in mean melt date across the Siberian–Alaskan sector is partly related to important changes in large-scale modes of Arctic large-scale atmospheric variability. Around 1990, the AO was consistently positive, which resulted in increased atmospheric heat transport into the Arctic and the Siberian–Alaskan sector (Rigor et al. 2002; Zhang et al. 2003). Since the year 2000, the appearance of a meridional-oriented dipole atmospheric circulation pattern has also increased atmospheric and ocean heat transport into the Siberian–Alaskan sector [i.e., the Arctic rapid change pattern (X. Zhang et al. 2008) and the Arctic dipole (Wang et al. 2009; Overland and Wang 2010)]. Second, there is large interannual variability, with years that show consistently early or late mean initial melt across large longitude sectors (e.g., 1990, 2000, and 2009) relative to surrounding years.
(a) Mean melt onset date across the polar cap (north of 70°N) for each year from 1979 to 2012. Each value plotted is the average across a 5° longitude sector, starting from 180° longitude (i.e., the value plotted at 100°E is the average in the box bounded by 100° and 105°E from east to west and 70° and 90°N from north to south). (b) As in (a), but for the mode of the melt onset date in each 5° longitude sector.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
These two characteristics, the steplike decrease and the consistent values across large bands of longitude, also appear when plotting the mode (Fig. 2b). The mode shown in Fig. 2b is the most common melt onset date in each 5° longitude sector for a particular year. In other words, the date shown in Fig. 2b is the date at which there is the largest area of melt onset in that longitude sector. Nearly identical mode dates in several adjacent sectors are a signature of a large melt event, which is sufficiently large in spatial extent to result in maximum initial melt in several adjacent sectors for that particular year. A close inspection of Fig. 2b reveals that there are many bands of similar or identical mode dates oriented horizontally across the plot. For example, in the year 1990, a band of nearly identical mode values are located across the East Siberian and Chukchi Seas, from 150°E to 150°W. Thus, in 1990, there was a common initial melt date across that longitude band. This event and other melt events are possibly initiated by a common large-scale atmospheric or oceanic feature. Interestingly, there is also a steplike decreasing trend in the mode of the melt date across the Siberian–Alaskan sector, suggesting that the most common melt onset date has also been occurring earlier in recent years; similar to what is observed for the mean melt onset date.
Climatologically, at the time of the initial melt, the magnitude of the direct eddy heat transport by the different frequency eddies is of similar order and positive. Figure 3a shows the climatological meridional heat transport by high-, low-, and subseasonal-frequency eddies from days 110 to 140 (20 April–20 May). We choose this date range to focus on dates around and immediately preceding the mean melt date across most sectors of the Arctic, to look at a period where anomalous heat transport may trigger or support the melt onset. The subseasonal- and high-frequency eddy heat transport is greater than the low-frequency eddy heat transport across most longitude sectors, with maximum values broadly found across Greenland and the Greenland Sea (300°–30°E). In the Siberian–Alaskan sector, the values of heat transport by all eddy scales are at a broad minimum. The high-frequency heat transport is greater than the low-frequency and subseasonal-frequency transport across most of the sector, suggesting that direct eddy heat transport is primarily by synoptic-scale eddies.
(a) Climatology of lower-tropospheric meridional heat transport across 70°N by high-, low-, and subseasonal-frequency eddies (H, L, and S, respectively), averaged from days 110 to 140 in 15° bins, corresponding to the date range 20 April–20 May. High, low, and seasonal eddies are defined by the 2–10-, 10–30-, and 30–90-day frequency bands, respectively. Heat transport values are mass weighted and vertically averaged from 1000 to 500 mb, and the data are horizontally averaged identically as in Fig. 2. (b) Climatology of significant cross terms of the lower-tropospheric meridional heat transport across 70°N. Date range and averaging is as in (a). Term Tbar is the seasonal mean temperature calculated by applying a 90-day moving average filter to the raw temperature data.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
The climatological magnitude of the cross terms, which represent the interaction of eddies in the three defined frequency bands with each other and the mean wind and temperature fields, is presented in Fig. 3b. The cross term of greatest magnitude is the product of the subseasonal-frequency wind and climatological mean temperature near 0° longitude, denoted here as VS×Tbar. This term represents the interaction of the Arctic wind response to intraseasonal-scale phenomena both within the Arctic and at lower latitudes, such as the Madden–Julian oscillation, and the Arctic-mean temperature field. The bar term here is calculated using the seasonal average (i.e., centered 90-day mean) to capture the mean temperature field for boreal spring. Within the Siberian–Alaskan sector, the VS×Tbar quantity peaks around 180° longitude. The product of the high-frequency wind and seasonal mean temperature (i.e., the VH×Tbar term) is also climatologically positive in some parts of the Siberian–Alaskan sector. As we will later show, these two terms are important and associated with the initialization of significant melt onset episodes that we explore in section 3c, even though the climatological high-frequency cross term is lower in magnitude compared to the other terms. The product of the low-frequency wind and seasonal mean temperature (i.e., the VL×Tbar term) is climatologically positive in some parts of the Siberian–Alaskan sector, and is the second-largest cross term in the climatology in terms of magnitude. However, we will show later that this term plays a more minor role in the initialization of significant initial melt episodes that we explore in section 3c. The remaining cross terms have very small magnitudes (magnitudes of less than 1 K m s−1) relative to the terms presented in Fig. 3b, and thus are not shown in the figure.
The magnitude of the direct eddy heat transport terms in Fig. 3a is an order of magnitude less than the magnitude of the cross terms in Fig. 3b. The direct eddy heat transport is the heat transport generated by both the anomalous wind and temperature fields associated with the eddies. Although these direct eddy terms are small, the values are all positive, confirming the net effect of transient eddies to transport heat poleward (Nakamura and Oort 1988). In cross-term eddy heat transport (e.g., VH×Tbar), the anomalous winds associated with the eddies transport heat contained within the climatological temperature field. The larger-magnitude cross-term transport means that the eddies increase the meridional heat transport into the Arctic polar cap more by meridionally advecting heat contained in the existing temperature field than by meridionally advecting the heat contained in the temperature anomalies generated by the eddies. It is the interaction of the eddies with the mean state that transport more heat from the lower latitudes across 70°N than the direct eddy transport. Additionally, it will be shown subsequently that the cross-term meridional heat transport will be more closely associated with the individual large-area melt onset events than the direct eddy meridional heat transport.
b. Connection between eddy heat transport and mean melt onset date trend
To investigate the relationship between the stepwise change in the melt onset, the surface temperature, and lower-tropospheric heat and moisture transport quantities, we explore the time evolution of each quantity in more detail, focusing on the Siberian–Alaskan sector. In the time evolution of the melt onset date each year in the Siberian–Alaskan sector (Fig. 4a), there is a steplike trend toward an earlier mean melt onset date, with a sharp decrease around 1990 and a steady decrease after the year 2000. The mean melt date decreases from day 167 (16 June) to day 138 (18 May) from 1979 to 2012, a decline of 29 days over 34 years.
(a) Total count of the number of 25 km × 25 km grid boxes exhibiting the melt signal for each day and year in the Siberian–Alaskan sector of interest (90°E–130°W). White line is the mean melt date in the sector of interest for each year. (b) Total lower-tropospheric meridional heat transport (K m s−1) across 70°N in each longitude sector, averaged over days 110–140. (c) Average surface (2 m) temperature in the sector of interest (°C), averaged over 70°–90°N. Red line is identical to the white line in (a). Black dashed line denotes the first date when temperatures in the atmospheric column above the sector reach 0°C across 2% of the grid points in the sector. Blue line denotes when the maximum mean temperature in the atmospheric column above the sector first reaches 0°C. Thick black line is the 0°C isotherm. (d) Total daily lower-tropospheric meridional heat transport (K m s−1) across 70°N for each year in the 1979–2012 period, in a sector containing all longitudes within 5° of the location of maximum-area melt for that particular year. Markers denote the day of the maximum-area melt onset for each year, with larger open markers denoting anomalously early melt onset and filled markers denoting events where the melt onset is the most anomalously early for that year. Thin (thick) black line denotes first date that temperatures greater than 0°C appear in the atmospheric column over the maximum-area melt over greater than 2% (20%) of the area. Red line denotes the first date when the mean maximum temperature in the atmospheric column over the melt area first reaches 0°C. (e) Total daily lower-tropospheric meridional water vapor transport (m s−1) across 70°N for each year in the 2000–12 period, in a sector containing all longitudes within 5° of the location of maximum melt for that particular year. Markers denote the day of the maximum-area melt onset for each year, and are highlighted as in (d). Days when the total cloud fraction in the sector is greater than 0.75 are shaded.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
Unlike the steplike decreasing trend in the mean melt onset date in Fig. 4a, the trend in meridional heat transport into the Arctic (Fig. 4b) over the 34-yr period is not apparent. Although the heat transport is positive in the Siberian–Alaskan sector longitudes, there is no increasing trend in heat transport at any longitude during the 30-day period of days 110–140. A similar lack of trend exists for heat and moisture transport at any longitude or any other 30-day period around the mean melt onset date (not shown). The lack of trend suggests that another heat or moisture transport component, such as zonal transport, or another process that is coincident with, but not directly related to the positive meridional heat transport into the Arctic, is driving the trend in the Arctic-mean melt onset date. It is not necessarily the magnitude of the lower-tropospheric heat and moisture transport from lower latitudes across 70°N that is important to the recent trend in the melt onset date.
The evolution of the surface temperature and maximum temperature in the atmospheric column more closely matches the trend in the melt onset (red line in Fig. 4c). The date on which the mean surface temperature first reaches 0°C steadily declines from day 170 to day 160 over the 1979–2012 period. Both the time series of dates when the maximum temperature in the atmospheric column first reaches 0°C over an appreciable area of the Siberian–Alaskan sector (black line in Fig. 4c) and the dates when the mean of the atmospheric column maximum temperature over the sector first reaches 0°C (blue line in Fig. 4c) exhibit a steady decline over the entire period. Additionally, the atmospheric column maximum temperature time series is significantly and positively correlated with the mean melt onset date after 1990 (correlation coefficient: 0.72). The results shown in Figs. 4b and 4c suggest that the earlier appearance of temperatures above 0°C at the surface and in the lower troposphere, especially in the last 15 years of the 1979–2012 period, have occurred in conjunction with the earlier melt onset and may be contributing to the earlier melt onset. The earlier appearance of above-freezing temperatures is not related to an increase in the meridional transport of heat from lower latitudes, but due to other factors increasing the temperature of the lower troposphere.
Another prominent feature of Fig. 4a is the discrete events where a large area of melt occurs over a few days (e.g., day 138 in 1990, day 133 in 2011, day 140 in 2002, among others). We mark the date of the largest-area melt onset events for each year across the Arctic polar cap in Figs. 4d and 4e. If the event is anomalously early, relative to the 1979–2012 climatology, then we mark the event with a larger open marker. If the event is the most anomalous for that year, based on an area-weighted mean anomaly over the melt area, then the event is highlighted with a filled marker. Many, but not all, maximum-area melt events are associated with large positive values of heat and moisture transport either on the date of the event or immediately preceding it (Figs. 4d and 4e), especially if the maximum-area melt onset event is anomalously early. Net heat convergence in the lower troposphere over the melt area is also found in 25 of the 34 maximum-area melt onset cases (not shown), supporting a local atmospheric column temperature increase over the melt onset area.
Based on the hypothesis presented in Persson (2012), we would expect that the melt onset events would occur when temperatures exceed 0°C and cloud fraction increases within the atmospheric column over the area of melt. The results presented in Fig. 4 fulfill that expectation in two important ways. First, there is evidence that there is a relationship between the melt onset and the appearance of above-freezing temperatures in the atmospheric column above the area of melt onset. The declining trends in the date when the mean maximum temperature in the atmospheric column over the sector first reaches 0°C (blue line in Fig. 4c) and the sector mean melt date (red line in Fig. 4c) closely match. Additionally, many of the individual maximum-area melt onset events occur when the mean maximum temperature in the atmospheric column above the whole melt area first reaches 0°C (red line in Fig. 4d), generally above the surface at the 900–700-mb pressure levels (not shown). Second, cloudy (cloud fraction greater than 0.75) conditions across the melt onset area are also common for the maximum-area melt onset events (Fig. 4e). The melt events occur in a period of increased cloudiness relative to earlier in the year, consistent with the climatology for cloud cover across the high Arctic (poleward of 70°N) that show an sharp increase in total cloud fraction in May relative to previous months of the year (Eastman and Warren 2010). Although the hypothesis presented by Persson (2012) helps in the explanation of the trend and individual melt event relationships described above, the step change in the mean melt date time series in 1988–90 (red line in Fig. 4c) is not related to any corresponding change in the tropospheric temperature.
Even though trends in meridional heat transport across 70°N appear not to be related to the recent trend in an earlier melt onset, Figs. 4d and 4e suggest a potential connection between large positive tropospheric meridional heat and moisture transport and the maximum-area melt onset events. Positive heat and moisture transport is coincident with many of the maximum-area melt onset events. In the next section, we will investigate in detail how heat transport and other factors help trigger these individual melt onset events by investigating two events in 1990 and 2011. These two maximum-area melt onset events are among the earliest events in the 1979–2012 period, occurring on days 138 and 133, respectively (Fig. 4a). The melt on day 138 (133) in 1990 (2011) occurred over an area equal to 12.41% (7.73%) of the total area in the Siberian–Alaskan sector and 6.70% (4.38%) of the total area that experiences melt over the Arctic polar cap, the largest melt event in area in both spatial domains for that year. They are also located wholly within the Siberian–Alaskan sector.
c. Initial melt events in individual years
We first examine the melt event on day 138 in the year 1990 (see Fig. 4a) to determine which meridional heat transport term is most associated with the melt event. In Fig. 5a, we plot the daily evolution of the melt onset count at each longitude. The melt onset event is clearly found centered at day 138 and extending from 150°E to 150°W. The event is also by far the greatest melt event in magnitude in any sector in the initial melt season of 1990. Figure 5b shows the time series of total heat transport and other important eddy heat transport components, with day 0 indicating the maximum-area melt onset date. Immediately preceding the peak of the melt event is a period of large-magnitude positive total heat transport, starting from 4 days before the melt event. Although there is little change in the mean maximum temperature in the atmospheric column above the melt onset area, above-freezing temperatures first appear over the area at day −2. A major component of the total heat transport in this period is the cross term containing the product of the HF wind and the climatological mean temperature, with a nearly equal contribution from the product of the subseasonal-frequency (SF) wind and the climatological mean temperature. The low-frequency cross term is only a minor contributor to the peak in heat transport around day 0. The change in the total heat transport closely matches the change of the HF cross term with time. The heat transport from the SF cross term and climatology are both positive, and thus the total heat transport remains above zero for most of the period highlighted. Additionally, these two terms were identified in Fig. 3b as significant components of the climatological meridional heat transport in this sector. Like in the climatology in Fig. 3, the magnitude of the heat transport directly resulting from the eddies (not shown) is much smaller than the cross terms (solid lines in Fig. 5); thus, the meridional heat transport that results from the interaction of the eddies with the background temperature field may be more crucial to the melt onset in this event than the meridional heat transport generated purely by the eddy temperature and meridional wind fields. In subsequent plots, we omit including the direct eddy transport due to its small magnitude.
(a) Daily count of the total number of 25 km × 25 km grid boxes exhibiting an initial melt signature in the year 1990. (b) Total heat transport and important cross-term meridional heat transport terms across 70°N for the melt event identified in (a) (extending from 150°E to 150°W). Day 0 is the date of the peak melt area (day 138). Additionally, the mean maximum temperature over the day 0 melt onset area is plotted (orange dashed line; °C). Black markers on the orange dashed line denote days where above-freezing temperatures are present in the troposphere over at least part of the melt onset area.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
The area of melt at day 0 is associated with a broad area of meridional heat transport across the East Siberian and Chukchi Seas (150°E–150°W; Fig. 6a). The maximum meridional transport is found west of the melt onset area, suggesting that local zonal heat transport also contributed to the melt onset. In the sea level pressure fields (not shown), the increased meridional heat transport is associated within an area between a trough of low pressure centered around 150°E and an area of high pressure farther east. In between these two features, there is broad southeasterly flow that accounts for the poleward meridional heat transport. These features also appear in the 500-mb geopotential height field (not shown). Interestingly, when plotting the time–longitude evolution of total heat transport, the large positive value associated with the melt episode is at the eastern end of a diagonal feature on the plot. The diagonal feature begins around day 133 at 90°E, near the eastern part of the Kara Sea (Fig. 6b), and propagates eastward. These diagonal features suggest a persistent meridional heat transport as a result of one feature propagating eastward with time. In this case, the propagating feature is the southerly winds between the high and low couplet. Many of these diagonal features appear in Fig. 6b, suggesting that propagating large-scale features regularly affect the meridional heat transport across this latitude for this particular year. Interestingly, the magnitude of the meridional heat transport that is associated with the melt event on day 138 does not stand out relative to other positive meridional heat transport on other days or at other longitudes. This suggests that the magnitude of the heat transport may be less crucial to the triggering of the melt onset event compared to the temperature of the air transported. As seen in Fig. 4c, the trend in the date when the mean maximum temperature in the atmospheric column over the Siberian–Alaskan sector first reaches 0°C matched more closely the mean melt date trend than any trend in the meridional heat transport.
(a) Meridional heat transport (K m s−1) at day 0 of the melt event identified in Fig. 5 for the year 1990. Hatching denotes the area that exhibits the initial melt signature on this date. (b) Daily meridional heat transport (K m s−1) across 70°N at each longitude.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
Another significant melt event in 2011 on day 133 is highlighted in Fig. 7a. In this event, the total meridional heat transport in the sector of the melt is highly positive beginning 2 days before the melt event (Fig. 7b). Above-freezing temperatures in the atmospheric column begin to appear on day −1, and the mean maximum temperature in the atmospheric column above the melt area rises above 0°C on day +1. Similar to the 1990 case, the product of the high- and subseasonal-frequency wind with the mean temperature field are the largest components of the highly positive heat transport. Also like the previous case, the direct eddy heat transport does not significantly contribute to the peak heat transport (not shown), indicating that the increase of total heat transport is mainly driven by the cross terms representing the interaction of the HF and SF eddies with the mean temperature field. Further, the melt is collocated with an area of positive heat transport extending northward from just east of the date line (Fig. 8a). Unlike in the 1990 case, the increase in heat transport associated with the melt event is not associated with a long-lived heat transport anomaly (Fig. 8b). The diagonal banding of heat transport is still present in this year, but a band of increased heat transport is not directly associated with the melt event. However, similar to the 1990 event, the melt is found in a region of southerly flow between a high and low couplet, with the trough of low pressure in the SLP field centered around 180° longitude and the high located farther east (not shown).
(a) As in Fig. 5, but for 2011. (b) Centered on the event at day 133 and averaged from 180° to 150°W.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
As in Fig. 6, but for the melt event in 2011.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
In the analysis presented so far, the focus has been on the effects of eddy meridional heat transport, a dynamical component of atmospheric eddies directly driven by the interaction of the eddy-generated wind field and climatological temperature field. As hypothesized by Persson (2012), the next step in the chain of heat transfer processes that link the atmospheric temperature anomalies to the surface energy budget is the shortwave and especially longwave radiative flux anomalies generated by changes in temperature profile of the atmospheric column. Increased temperatures in the atmospheric column increase the amount of downwelling longwave radiation at the surface. Additionally, changes in the water vapor content in the atmospheric column as well as cloud cover and cloud properties contribute to anomalous downwelling longwave and shortwave radiation at the surface. The shortwave cloud forcing is a function of the solar zenith angle, the cloud transmittance, and the surface albedo, and the net shortwave cloud forcing generally cools the surface. The longwave cloud forcing is a function of cloud temperature, height, and emissivity, and the net longwave cloud forcing generally warms the surface (Shupe and Intrieri 2004). During the nonmelt season, the magnitude of the cooling shortwave cloud radiative effect is reduced due to the high zenith angle of incoming solar radiation and the high surface albedo; thus, the net effect of clouds is to warm the surface (e.g., Kay and L’Ecuyer 2013). In the early melt season over sea ice, the surface albedo is still high; thus, the net cloud radiative forcing is still positive despite a decreasing solar zenith angle. Later in the melt season, upon the onset of the melt, the shortwave cloud radiative forcing becomes more dominant as the surface albedo also decreases.
We explore the downwelling radiation anomalies associated with the melt in the May 2011 event by examining the anomalies in the downwelling shortwave (SW) and longwave (LW) radiation at the surface in the CERES data on day 0 and day −1 (Figs. 9 and 10, respectively). The anomalies are calculated relative to a 2000–12 climatology. The variability of downwelling SW and LW fluxes at the surface is primarily driven by atmospheric processes; therefore, we analyze the spatial distribution of the anomalies of the downwelling component. The SW (LW) downwelling flux bias of the CERES data is estimated to be −3.9 (7.0) W m−2, and the SW (LW) downwelling flux root-mean-square error (RMSE) for the CERES data is estimated to be 13.7 (12.3) W m−2. The estimate is based on a direct comparison of May 2011 CERES data and ground truth flux data at the U.S. Department of Energy Atmospheric Radiation Measurement (ARM) site at Barrow, Alaska (71.32°N, 156.61°W; Rutan et al. 2001). The site is representative, since it located close to the area of melt onset in the May 2011 event analyzed in Figs. 9 and 10. We exclude the event in 1990 from the analysis of the surface SW and LW anomalies, as the event occurs outside of the period covered by the CERES data.
Surface downwelling (a) shortwave, (b) longwave, and (c) longwave plus shortwave flux anomalies (W m−2) at day 0 for the 2011 melt event. Anomalies are calculated relative to a 2000–12 climatology. (d) Total cloud area, in terms of percent sky coverage, at day 0 for the 2011 melt event. Area that exhibits the initial melt signature on this date is outlined with the thick black contour. Area of interest for the daily flux calculations in Fig. 11 is outlined with the thin red line.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
In the region of initial ice melt, positive downwelling LW anomalies exist over the melt onset area on the day of the largest-area melt onset, associated with an area of increased cloud cover (Figs. 9b and 9d). Mixed slightly positive and negative SW flux anomalies exist over the melt onset area (Fig. 9a). The magnitude of the LW anomalies over the melt onset area is larger than the RMSE for representative LW fluxes in the CERES dataset; thus, confidence can be placed in the CERES estimates of the LW anomalies over the melt onset region. The sum of these two anomalies results in a total increase in downwelling SW and LW radiation of 10–30 W m−2 at the surface over the melt region (Fig. 9c). The coincidence of the melt onset with the surface downwelling LW anomalies and increased cloud cover is also present on the day before the melt (Fig. 10). Negative downwelling SW anomalies, which would not contribute to a surface warming, also appear over the melt onset area on that date. The fact that there is spatially a higher correspondence of positive downwelling LW anomalies and the area of melt onset supports the hypothesis that downwelling LW anomalies are crucial to melt onset. The increased downwelling LW radiation supports warming at the surface and is consistent with the existence of an area of initial melt.
To better understand the time evolution of the flux anomalies crucial to the melt onset, we analyze the time series of the mean net surface fluxes over the region of the melt onset event, along with key quantities that help explain the variability in the net surface fluxes. The daily evolution of the net SW and LW fluxes from day 75 (16 March) to day 150 (30 May), and the total net surface radiative flux Q are shown as the solid lines in Fig. 1. Over this period, the bias of the net SW (LW) fluxes is estimated to be 26.9 (−6.3) W m−2, and the RMSE is estimated as 35.4 (13.3) W m−2. The estimation method is identical as for the flux quantities shown in Figs. 9 and 10. Over the entire period shown, both before and immediately after the initial melt, the total surface radiative flux is increasingly positive (i.e., the flux into the surface is increasing). The overall increasing trend of the total flux is nearly matched by the trend in the net SW flux. The net LW flux is negative for the entire period, with a slight increase over the period shown. However, especially before the initial melt on day 133 (13 May), local maxima in the total net surface radiative flux are associated with increased net LW and decreased net SW (i.e., at days 108 and 115). As expected, given the effects of clouds and increased atmospheric water vapor on clear-sky downwelling LW radiation, the increased net LW flux is associated with higher precipitable water and cloud fraction. The net SW flux also decreases in response to increased cloud cover for this period, consistent with the increased reflection and scattering of incoming solar radiation by clouds. Qualitatively, the seasonal trend of net surface SW and LW radiative flux follows the increase in the mean maximum temperature in the atmospheric column in this period over this region. The slow increase of the mean maximum temperature in the atmospheric column toward the melting point follows the steady increase of net SW flux at the surface. Also, the high-frequency variability (on the time scale of days) of mean maximum temperature in the atmospheric column is associated with the high-frequency variability of LW net surface flux. Local maxima of the mean maximum temperature in the atmospheric column are associated with increased cloud cover and precipitable water. At the surface, local maxima of skin temperature similarly correspond to periods of increased cloud cover, increased net LW flux, decreased net SW flux, and increased precipitable water (not shown).
Immediately preceding the 2011 melt onset event (vertical line in Fig. 11a), the mean maximum temperature in the atmospheric column increased rapidly 2 days before the initial melt event (i.e., day −2 in Fig. 7). The increase in temperature is coincident with an increase in total heat transport (see Fig. 7b), which also commences on day −2, with the largest meridional heat transport on day −1 to day +1. As with other instances of rapidly increasing maximum column temperature before the melt onset date, the increase was associated with a rapid increase in precipitable water and an increase in net surface LW flux. As a response to the warming and moistening atmosphere, downwelling LW radiation on day 0 over the melt area was anomalously high (Fig. 10b). Above-freezing temperatures first appear over the melt onset area on day −1 (see Fig. 7b), and the mean maximum temperature in the atmospheric column first rises above 0°C on day +1. At the surface, the combination of these factors help sustain an increase in skin temperature on day −1 and day 0 that initialized on day −2 (not shown). On day 1, there is a sharp decrease in net LW and an increase in net SW at the surface, likely in response to a sharp decrease in cloud cover. The LW and SW flux changes cancel out; thus, the net effect of the total net surface radiative flux changes on day 1 is near zero. Another melt event in 2002, highlighted in Fig. 11b, shows similar changes in net LW, cloud cover, mean maximum temperature in the atmospheric column, and precipitable water associated with the melt onset event.
(a) Daily evolution of surface net shortwave (SWnet; W m−2) and longwave radiation fluxes (LWnet; W m−2), the net total surface radiative flux (Q; W m−2), mean maximum temperature in the atmospheric column over the melt area (mean TMax; °C), cloud fraction (CFrac; %), and precipitable water (PW; 10−1 mm). Values are averaged over the melt onset area in the region outlined in Figs. 9 and 10 (70°–79°N, 180°–150°W). Vertical line at day 133 denotes the date of the greatest area of initial melt over the area of interest. Radiative flux and cloud fraction are taken from the CERES SYN1deg dataset, and the PW and skin temperature are taken from the MERRA reanalysis dataset. (b) As in (a), but for the melt onset event in 2002 on day 140 over 70°–79°N, 150°E–150°W.
Citation: Journal of Climate 30, 3; 10.1175/JCLI-D-15-0366.1
The analysis results shown in Figs. 5–11 suggest two important processes that occur leading up to the melt onset events in 1990, 2002, and 2011. First, there is a sharp increase in precipitable water and the mean maximum temperature in the atmospheric column above the melt onset area a few days before the melt onset (Fig. 11). Positive total lower-tropospheric meridional heat (Figs. 5b and 7b) and moisture transport (not shown) accompanies these increases, with the high-frequency eddy cross term (i.e., VS×Tbar) being the largest-magnitude contributor to the total heat transport anomaly in 1990 and 2011. The increase of temperature and water vapor content of the lower troposphere is not unprecedented during the period leading up to the melt onset, but the melt onset occurs when the maximum temperature in the atmospheric column first reaches and exceeds 0°C, as noted by Persson (2012) and seen in Figs. 7 and 11. Second, associated with the increase in temperature and water vapor in the atmospheric column, there is an increase in cloud cover and surface downwelling and net longwave radiation, linking the increase in temperature and moisture in the atmosphere with an increase in temperature at the surface. The appearance of these processes coincident with the melt onset event and an increase in the surface skin temperature gives some confidence in that these related processes help support the melt onset event. However, we do not know whether these important processes are applicable or common to all large-area melt events across the Arctic based on the results presented. The analysis of more melt onset events across the Arctic is needed to determine the most important processes that drive large-scale melt onset across the Arctic sea ice.
4. Summary and conclusions
The purpose of this paper was to explain the decline in the mean initial melt date across the Arctic polar cap (i.e., all areas north of 70°N) and show the importance of atmospheric transient eddies in helping trigger the melt onset over Arctic sea ice. We calculated the lower-tropospheric meridional heat transport across the boundary of the polar cap using the NASA MERRA reanalysis dataset, bandpass filtering this quantity to isolate the high-frequency (HF), low-frequency (LF), and subseasonal-frequency (SF) eddy components. In addition, we briefly analyzed NASA CERES downwelling longwave and shortwave surface flux data to quantify the surface radiative flux anomalies associated with one case of initial melt over large areas of the Arctic.
In the first part of the results, we identified the regions in the Arctic in which trends in melt onset existed and showed the climatological characteristics of the total and eddy meridional heat transport in the sector where the trend was most pronounced. We also showed that the decline in the September minimum sea ice extent was coincident with an earlier occurrence of the initial melt of sea ice across the Arctic polar cap, especially later in the study period (2000–12; Fig. 1). In a plot of the mean melt date in each 5° longitude sector, the trend of earlier melt was primarily confined to the sector extending from 90°E to 130°W, which corresponds to the seas north of northeast Siberia and Alaska and is referred as the Siberian–Alaskan sector throughout the manuscript (Fig. 2). In this region, a decline in the mean melt onset date of 29 days from 1979 to 2012 existed. Climatologically, the direct meridional heat transport by eddies in all three frequency bands across 70°N was positive and greatest for SF eddies (Fig. 3). Cross terms that represent the interaction of the subseasonal- and high-frequency eddy winds with the climatological mean temperature field were the largest component of the total heat transport, a larger component than the direct product of the eddy wind and temperature fields.
In the second part of the results, we focused on the mechanisms that help explain the recent declining trend in the mean melt onset date across the Siberian–Alaskan sector. The steplike trend in earlier melt onset could not be explained by a trend in the total meridional heat transport into the entire Arctic polar cap or any individual longitude sector (Fig. 4b). However, there was a temperature increase at the surface and especially at the level of maximum temperature in the lower troposphere. The trend in the time series of the date of the first appearance of mean maximum temperatures above 0°C in the atmospheric column closely matched the trend in the mean melt onset date (Fig. 4c), supporting the hypothesis expressed by Persson (2012). Additionally, the melt events occur in a period of increasing cloud cover relative to earlier in the year (Fig. 4e), consistent with the climatology of Arctic cloud cover presented in Eastman and Warren (2010).
Individual large-area melt onset events common to the 1979–2012 period across the Siberian–Alaskan sector were similarly associated with the cross-term meridional heat and moisture transport anomalies. For example, when looking at two melt onset events in 1990 and 2011 that affected a large area across the Siberian–Alaskan sector, the melt onset events were immediately preceded by a large-magnitude positive total meridional heat transport across 70°N in the longitude sector where the melt onset event occurred. The cross terms VS×Tbar and VH×Tbar, which represent the interaction between the seasonal- and high-frequency eddies and the mean temperature field, respectively, were the largest magnitude heat transport terms at the time of the melt event in the decomposition of the total meridional heat transport (Figs. 5 and 7). In both melt events, the melt was located in a broad area of positive heat transport across the western Arctic Ocean (Figs. 6 and 8). In the 2011 event, the mean maximum temperature in the atmospheric column rapidly increased immediately prior to the melt onset event and rose above 0°C on the day of the melt onset. Anomalies in the downwelling shortwave and longwave radiation at the surface also accompanied the 2011 melt event, likely a consequence of cloud cover and atmospheric column temperature and water vapor content changes. A positive surface LW anomaly was collocated with the melt on the date of the maximum-area melt onset (Fig. 9). The crucial increase of the surface and atmospheric column maximum temperatures occurred on the previous day, with an increased net LW surface flux, precipitable water, and cloud fraction located over the area of melt onset (Fig. 11a). A similar evolution of the temperature and water vapor content of the atmospheric column, cloud cover, and the surface radiative fluxes was present before and during the 2002 melt onset event (Fig. 11b).
There are two important implications of these results. First, a primary finding was that the trend in earlier Arctic-mean sea ice melt onset date is not directly related to a trend in the magnitude of lower-tropospheric total meridional heat transport across the boundary of the Arctic polar cap. This suggests that the trend in the date of the Arctic-mean melt onset is a result of a mechanism that is not related to the magnitude of lower-tropospheric atmospheric heat transport but is instead related to other processes that contribute to polar warming amplification, such as surface albedo feedback and net cloud feedback (Taylor et al. 2013), affecting the base temperature over which the positive temperature anomaly forced by anomalous heat transport is superimposed. Second, these results show that in a given year, the melt onset of Arctic sea ice in the Siberian–Alaskan sector is a process that occurs in part due to the relatively slow increase in net SW flux at the surface and the increase in net LW surface fluxes, associated with episodes of increased cloud cover, atmospheric column water vapor, and temperature. Increased lower-tropospheric poleward eddy heat and moisture transport is associated with the increases in downwelling and net LW surface fluxes crucial to the melt onset, at least in the cases of large-area melt onset analyzed in detail in this study. The early melt initialized by the eddies can persist and affect the later melt season through the cumulative effects on downwelling SW radiation flux absorbed at the surface by the ice and sea surface (e.g., Persson 2012; Stroeve et al. 2014). If this mechanism is an important contributor to the variability of sea ice in a given melt season, then it is crucial to properly simulate the atmospheric eddy influence on temperature and moisture content in the lower troposphere, along with surface radiative fluxes, when modeling the melt of Arctic sea ice.
The results presented are an initial investigation into how the interaction between the dynamics (the atmospheric eddies) and thermodynamics (the LW and SW anomalies) play a role in initializing the melt of sea ice over a large area. However, a remaining question is whether this dynamic/thermodynamic interaction is important in melt onset events throughout the Arctic outside of the Siberian–Alaskan sector. The results of this study motivate a more complete quantitative surface energy budget calculation, including an analysis of turbulent sensible and latent heat fluxes, which have also been shown to contribute to the melt onset (Persson 2012). The results also motivate the collection of all large-area initial melt events across the Arctic, to better understand the complex interaction between atmospheric eddies and the surface processes involved with the melt onset of Arctic sea ice in all regions of the Arctic. It remains unclear the exact mechanism by which atmosphere variability is linked to the initial sea ice melt, and whether the presence of clouds, anomalous atmospheric water vapor, temperatures above 0°C in the atmospheric column above the melt area, or anomalous heat transport is most important, although this study suggests that all of these components might be important to the mechanism. Additionally, we explore only a few specific cases where there is a relatively large melt event. It is possible that the specific flux anomalies that are crucial to the melt onset may be different in different regions of the Arctic. We also do not consider in detail the duration of the melt after the melt is initialized by eddy influences. The crucial link between the initial melt and the September minimum extent at the end of the melt season is the cumulative effect of the melt after it is initialized, since the extra absorbed incoming shortwave radiation is the energy source for the extra melt of Arctic sea ice in the surface albedo feedback (e.g., Stroeve et al. 2014). Thus, the persistence of the melt and total absorption of excess net surface SW flux after it is initialized is likely the crucial process to linking early initial melt to a reduced September minimum sea ice extent.
Acknowledgments
This study is supported by the National Science Foundation under Grants AGS-1147601, AGS-1354402, and AGS-1445956. The CERES data were obtained from the NASA Langley Research Center Atmospheric Science Data Center, and the data describing the melt onset over Arctic sea ice were obtained from the National Snow and Ice Data Center. We thank three anonymous reviewers for their insight and suggestions, which helped improve the manuscript significantly.
REFERENCES
Belchansky, G. I., D. C. Douglas, and N. G. Platonov, 2004: Duration of the Arctic sea ice melt season: Regional and interannual variability, 1979–2001. J. Climate, 17, 67–80, doi:10.1175/1520-0442(2004)017<0067:DOTASI>2.0.CO;2.
Bliss, A. C., and M. R. Anderson, 2014: Snowmelt onset over Arctic sea ice from passive microwave data: 1979–2012. Cryosphere, 8, 2089–2100, doi:10.5194/tc-8-2089-2014.
Comiso, J. C., C. L. Parkinson, R. Gersten, and L. Stock, 2008: Accelerated decline in the Arctic Sea ice cover. Geophys. Res. Lett., 35, L01703, doi:10.1029/2007GL031972.
Dong, X. Q., B. J. Zib, B. Xi, R. Stanfield, Y. Deng, X. Zhang, B. Lin, and C. N. Long, 2014: Critical mechanisms for the formation of extreme Arctic sea-ice extent in the summers of 2007 and 1996. Climate Dyn., 43, 53–70, doi:10.1007/s00382-013-1920-8.
Drobot, S. D., and M. R. Anderson, 2001: An improved method for determining snowmelt onset dates over Arctic sea ice using scanning multichannel microwave radiometer and Special Sensor Microwave/Imager data. J. Geophys. Res., 106, 24 033–24 049, doi:10.1029/2000JD000171.
Eastman, R., and S. G. Warren, 2010: Interannual variations of Arctic cloud types in relation to sea ice. J. Climate, 23, 4216–4232, doi:10.1175/2010JCLI3492.1.
Else, B. G. T., T. N. Papakyriakou, R. Raddatz, R. J. Galley, C. J. Mundy, D. G. Barber, K. Swystun, and S. Rysgaard, 2014: Surface energy budget of landfast sea ice during the transitions from winter to snowmelt and melt pond onset: The importance of net longwave radiation and cyclone forcings. J. Geophys. Res. Oceans, 119, 3679–3693, doi:10.1002/2013JC009672.
Graversen, R. G., T. Mauritsen, S. Drijfhout, M. Tjernstrom, and S. Martensson, 2011: Warm winds from the Pacific caused extensive Arctic sea-ice melt in summer 2007. Climate Dyn., 36, 2103–2112, doi:10.1007/s00382-010-0809-z.
Herman, G., and R. Goody, 1976: Formation and persistence of summertime Arctic stratus clouds. J. Atmos. Sci., 33, 1537–1554, doi:10.1175/1520-0469(1976)033<1537:FAPOSA>2.0.CO;2.
Kay, J. E., and A. Gettelman, 2009: Cloud influence on and response to seasonal Arctic sea ice loss. J. Geophys. Res., 114, D18204, doi:10.1029/2009JD011773.
Kay, J. E., and T. L’Ecuyer, 2013: Observational constraints on Arctic Ocean clouds and radiative fluxes during the early 21st century. J. Geophys. Res. Atmos., 118, 7219–7236, doi:10.1002/jgrd.50489.
L’Heureux, M. L., A. Kumar, G. D. Bell, M. S. Halpert, and R. W. Higgins, 2008: Role of the Pacific-North American (PNA) pattern in the 2007 Arctic sea ice decline. Geophys. Res. Lett., 35, L20701, doi:10.1029/2008GL035205.
Lindsay, R., M. Wensnahan, A. Schweiger, and J. Zhang, 2014: Evaluation of seven different atmospheric reanalysis products in the Arctic. J. Climate, 27, 2588–2606, doi:10.1175/JCLI-D-13-00014.1.
Markus, T., J. C. Stroeve, and J. Miller, 2009: Recent changes in Arctic sea ice melt onset, freezeup, and melt season length. J. Geophys. Res., 114, C12024, doi:10.1029/2009JC005436.
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:10.1029/2007GL032043.
Nakamura, N., and A. H. Oort, 1988: Atmospheric heat budgets of the polar regions. J. Geophys. Res., 93, 9510–9524, doi:10.1029/JD093iD08p09510.
Ogi, M., and J. M. Wallace, 2012: The role of summer surface wind anomalies in the summer Arctic sea ice extent in 2010 and 2011. Geophys. Res. Lett., 39, L09704, doi:10.1029/2012GL051330.
Ogi, M., I. G. Rigor, M. G. McPhee, and J. M. Wallace, 2008: Summer retreat of Arctic sea ice: Role of summer winds. Geophys. Res. Lett., 35, L24701, doi:10.1029/2008GL035672.
Overland, J. E., and M. Wang, 2010: Large-scale atmospheric circulation changes are associated with the recent loss of Arctic sea ice. Tellus, 62A, 1–9, doi:10.1111/j.1600-0870.2009.00421.x.
Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate. American Institute of Physics, 520 pp.
Perovich, D. K., and C. Polashenski, 2012: Albedo evolution of seasonal Arctic sea ice. Geophys. Res. Lett., 39, L08501, doi:10.1029/2012GL051432.
Perovich, D. K., S. V. Nghiem, T. Markus, and A. Schweiger, 2007: Seasonal evolution and interannual variability of the local solar energy absorbed by the Arctic sea ice–ocean system. J. Geophys. Res., 112, C03005, doi:10.1029/2006JC003558.
Persson, P. O. G., 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:10.1007/s00382-011-1196-9.
Rienecker, M. M., and Coauthors, 2011: MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 3624–3648, doi:10.1175/JCLI-D-11-00015.1.
Rigor, I. G., J. M. Wallace, and R. L. Colony, 2002: Response of sea ice to the Arctic oscillation. J. Climate, 15, 2648–2663, doi:10.1175/1520-0442(2002)015<2648:ROSITT>2.0.CO;2.
Rutan, D. A., F. G. Rose, N. M. Smith, and T. P. Charlock, 2001: Validation data set for CERES Surface and Atmospheric Radiation Budget (SARB). GEWEX Newsletter, No. 1, International GEWEX Project Office, Silver Spring, MD, 11–12.
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:10.1029/2011JD015847.
Shimada, K., T. Kamoshida, M. Itoh, S. Nishino, E. Carmack, F. McLaughlin, S. Zimmermann, and A. Proshutinsky, 2006: Pacific Ocean inflow: Influence on catastrophic reduction of sea ice cover in the Arctic Ocean. Geophys. Res. Lett., 33, L08605, doi:10.1029/2005GL025624.
Shupe, M. D., and J. M. 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, doi:10.1175/1520-0442(2004)017<0616:CRFOTA>2.0.CO;2.
Stroeve, J. C., T. Markus, L. Boisvert, J. Miller, and A. Barrett, 2014: Changes in Arctic melt season and implications for sea ice loss. Geophys. Res. Lett., 41, 1216–1225, doi:10.1002/2013GL058951.
Taylor, P. C., M. Cai, A. X. Hu, J. Meehl, W. Washington, and G. J. Zhang, 2013: A decomposition of feedback contributions to polar warming amplification. J. Climate, 26, 7023–7043, doi:10.1175/JCLI-D-12-00696.1.
Tjernström, M., and Coauthors, 2012: Meteorological conditions in the central Arctic summer during the Arctic Summer Cloud Ocean Study (ASCOS). Atmos. Chem. Phys., 12, 6863–6889, doi:10.5194/acp-12-6863-2012.
Ukita, J., M. Honda, H. Nakamura, Y. Tachibana, D. J. Cavalieri, C. L. Parkinson, H. Koide, and K. Yamamoto, 2007: Northern Hemisphere sea ice variability: Lag structure and its implications. Tellus, 59A, 261–272, doi:10.1111/j.1600-0870.2006.00223.x.
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., 6, L05706, doi:10.1029/2008GL036706.
Wielicki, B. A., B. R. Barkstrom, E. F. Harrison, R. B. Lee, G. L. Smith, and J. E. Cooper, 1996: Clouds and the Earth’s Radiant Energy System (CERES): An Earth observing system experiment. Bull. Amer. Meteor. Soc., 77, 853–868, doi:10.1175/1520-0477(1996)077<0853:CATERE>2.0.CO;2.
Yamamoto, K., Y. Tachibana, M. Honda, and J. Ukita, 2006: Intra-seasonal relationship between the Northern Hemisphere sea ice variability and the North Atlantic Oscillation. Geophys. Res. Lett., 33, L22701, doi:10.1029/2006GL026286.
Zhang, J., R. Lindsay, M. Steele, and A. Schweiger, 2008: What drove the dramatic retreat of arctic sea ice during summer 2007? Geophys. Res. Lett., 35, L11505, doi:10.1029/2008GL034005.
Zhang, X., M. Ikeda, and J. E. Walsh, 2003: Arctic sea ice and freshwater changes driven by the atmospheric leading mode in a coupled sea ice–ocean model. J. Climate, 16, 2159–2177, doi:10.1175/2758.1.
Zhang, X., A. Sorteberg, J. Zhang, R. Gerdes, and J. C. Comiso, 2008: Recent radical shifts in atmospheric circulations and rapid changes in Arctic climate system. Geophys. Res. Lett., 35, L22701, doi:10.1029/2008GL035607.
