Abstract

The decreasing surface albedo caused by continuously retreating sea ice over Arctic plays a critical role in Arctic warming amplification. However, the quantification of the change in radiative forcing at top of atmosphere (TOA) introduced by the decreasing sea ice albedo and its feedback to the climate remain uncertain. In this study, based on the satellite-retrieved long-term surface albedo product CLARA-A1 (Cloud, Albedo, and Radiation dataset, AVHRR-based, version 1) and the radiative kernel method, an estimated 0.20 ± 0.05 W m−2 sea ice radiative forcing (SIRF) has decreased in the Northern Hemisphere (NH) owing to the loss of sea ice from 1982 to 2009, yielding a sea ice albedo feedback (SIAF) of 0.25 W m−2 K−1 for the NH and 0.19 W m−2 K−1 for the entire globe. These results are lower than the estimate from another method directly using the Clouds and the Earth’s Radiant Energy System (CERES) broadband planetary albedo. Further data analysis indicates that kernel method is likely to underestimate the change in all-sky SIRF because all-sky radiative kernels mask too much of the effect of sea ice albedo on the variation of cloudy albedo. By applying an adjustment with CERES-based estimate, the change in all-sky SIRF over the NH was corrected to 0.33 ± 0.09 W m−2, corresponding to a SIAF of 0.43 W m−2 K−1 for NH and 0.31 W m−2 K−1 for the entire globe. It is also determined that relative to satellite surface albedo product, two popular reanalysis products, ERA-Interim and MERRA, severely underestimate the changes in NH SIRF in melt season (May–August) from 1982 to 2009 and the sea ice albedo feedback to warming climate.

1. Introduction

Sea surface albedo in the Arctic Ocean has declined considerably over the past decades (Comiso and Hall 2014; Riihelä et al. 2013a) because of retreating sea ice coverage (Comiso et al. 2008; Kerr 2009; Parkinson and Cavalieri 2012), earlier melt onset (Markus et al. 2009; Stroeve et al. 2014), and decreasing ice thickness (Kwok and Rothrock 2009; Maslanik et al. 2007), which have forced the Arctic Ocean to absorb increasing amounts of solar radiation (Perovich et al. 2007a).

Sea ice albedo feedback (SIAF) has largely enhanced Arctic warming. Some studies suggest that SIAF played a central role in the recent Arctic warming amplification (Crook et al. 2011; Screen and Simmonds 2010; Serreze et al. 2009; Taylor et al. 2013), while other studies argue that compared to the temperature feedbacks (especially the lapse rate feedback), the contribution of SIAF to the Arctic warming amplification is not substantial (Pithan and Mauritsen 2014) or even negligible (Winton 2006). Therefore, a more accurate quantification of the Arctic SIAF is essential for understanding the physical mechanisms of accelerated sea ice loss and assessing the underlying evolution of Arctic warming amplification.

Most previous surface albedo feedback assessments have been based on model simulations (Colman 2003, 2013; Dessler 2013; Pithan and Mauritsen 2014; Taylor et al. 2013; Winton 2006), while estimates through satellite retrievals remain limited. Two recent typical studies based on satellite retrievals show large differences from one another. Flanner et al. (2011) used a synthesis of calculated sea ice albedo, with sea ice type derived from sea ice concentration and in situ measurements of sea ice albedo, and radiative kernels to estimate the sea ice radiative forcing (SIRF) in the Northern Hemisphere (NH). They found the change in SIRF from 1979 to 2008 was 0.22 (0.15–0.32) W m−2, and the corresponding SIAF was 0.28 (0.19–0.41) W m−2 K−1 based on surface temperature warming of 0.79 K reported by Goddard Institute of Space Studies (GISS) during that period. Pistone et al. (2014) estimated Arctic SIRF (although they did not use this concept directly) and SIAF from 1979 to 2011 with a combined time series of planetary albedo: For the period 2000–11, they used the observed planetary albedo from the Clouds and the Earth’s Radiant Energy System (CERES) product; for the period 1979–99, they used a derived planetary albedo from sea ice concentration. They found that the change in SIRF in the NH caused by sea ice loss north of 60°N was 0.43 ± 0.07 W m−2, nearly twice as large as that estimated by Flanner et al. (2011). The results of these two studies differ considerably, despite the fact that both studies used the same method to calculate the change in SIRF (expressed as linear trend multiplied by the time interval) over a similar time period.

Flanner et al. (2011) also compared their estimated SIAF to a Coupled Model Intercomparison Project phase 3 (CMIP3) multimodel-based estimate, and indicated that CMIP3 models substantially underestimated the change in NH SIRF because of a systematically slower decline in the simulated sea ice concentration compared to observed rates (Flanner et al. 2011; Stroeve et al. 2007). However, Dessler (2013) estimated the global surface albedo feedbacks using two reanalysis products—the Interim European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-Interim) and the Modern-Era Retrospective Analysis for Research and Applications (MERRA)—and compared them to the estimates from general circulation models (GCMs). He found that the global surface albedo feedbacks of 0.28 ± 0.15 W m−2 K−1 (ERA-Interim) and 0.24 ± 0.15 W m−2 K−1 (MERRA) are close to the values of 0.3 ± 0.12 W m−2 K−1 (control runs) and 0.28 ± 0.09 W m−2 K−1 (A1B) estimated by GCMs. Finally, Dessler (2013) drew a contradictory conclusion to Flanner et al. (2011), that there is no evidence GCMs underestimate surface albedo feedback.

In this study, CLARA-A1 (Cloud, Albedo, and Radiation dataset, AVHRR-based, version 1), a newly released satellite-retrieved long-term surface albedo product from the Advanced Very High Resolution Radiometer (AVHRR) (Riihelä et al. 2013b), which is different from the datasets used by Flanner et al. (2011) and Pistone et al. (2014), is used to estimate the change in NH SIRF and SIAF from 1982 to 2009. But, in order to make it comparable, the same GISS surface temperature record (Hansen et al. 2010) and the same linear change calculation method as the two previous studies (Flanner et al. 2011; Pistone et al. 2014) are applied. The disagreements between the previous studies are analyzed and reconciled by adjusting change in all-sky SIRF estimated from kernel method. In addition, we compare the satellite albedo-based estimates to those from ERA-Interim and MERRA reanalysis to evaluate the performance of reanalysis on the assessment of change in NH SIRF and SIAF.

2. Data and methods

a. Data

1) Surface albedo

Three monthly surface albedo products from 1982 to 2009 are used: the CLARA-A1 satellite retrieved surface albedo product and the ERA-Interim and MERRA reanalysis albedo products. The CLARA-A1 product was developed by the European Organisation for the Exploitation of Meteorological Satellites [EUMETSAT; Climate Monitoring Satellite Application Facility (CM SAF)] project from AVHRR data with a spatial resolution of 0.25° × 0.25°. A homogenization preprocessing was taken to remove intersatellite calibration differences in the imagery and make the retrievable albedo dataset internally consistent (Riihelä et al. 2013b). The retrieval accuracy for sea ice albedo validated with in situ measurements is approximately 10%–15% (Riihelä et al. 2013a,b). There are some gaps in the original CLARA surface albedo product around the North Pole in the large solar zenith angle months. Using the seasonal variation from previous studies (Flanner et al. 2011; Pistone et al. 2014) and the value of neighboring month, we filled in the missing pixels, although they did not significantly affect the estimation of SIRF and its changes because of little incoming solar radiation in these regions and months. The ERA-Interim product is the latest global atmospheric reanalysis dataset produced by the ECMWF based on an improvement over the ERA-40 dataset (Dee et al. 2011; Screen and Simmonds 2010). The monthly ERA-Interim sea ice albedo was calculated with the 0.25° × 0.25° clear-sky surface downward and upward shortwave fluxes. The monthly 0.67° × 0.50° MERRA sea ice albedo is produced by NASA’s Global Modeling and Assimilation Office (Rienecker et al. 2011).

2) Radiative flux

The CERES Single Satellite Footprint (SSF) 1.0° (SSF1deg) TOA observed broadband radiative flux, which is a recommended product for long-term climate trend evaluation by the CERES science team and was also successfully used by Pistone et al. (2014) to estimate the Arctic SIRF and SIAF, is introduced for comparison with the kernel method of synthesis of surface albedo and radiative kernels.

3) Cloud fraction

The CERES SSF 1.0° cloud fraction product, derived using CERES-MODIS cloud retrieval algorithm, from 2000 to 2009 and the CLARA-A1 0.25° cloud fraction, derived from the EUMETSAT Nowcasting Satellite Application Facility (NWC SAF) cloud-processing package (Karlsson et al. 2013), from 1982 to 2009 are used in adjusting the change in all-sky SIRF estimated by the kernel method.

4) Sea ice extent

The fourth version of the Northern Hemisphere Equal-Area Scalable Earth grid (NH EASE-Grid) 2.0 Weekly Snow Cover and Sea Ice Extent product, which was gridded to EASE-Grid from Sea Ice Concentrations and derived from the Nimbus-7 Scanning Multichannel Microwave Radiometer (SMMR) and Defense Meteorological Satellite Program (DMSP) SSM/I-SSMIS Passive Microwave Data, provided by the National Snow and Ice Data Center (NSIDC), is used to recognize the maximum sea ice coverage from 1982 to 2009.

5) Surface temperature

The GISS gridded monthly mean 2° × 2° combined land–surface air and sea–surface water temperature anomalies (Hansen et al. 2010) [the Land–Ocean Temperature Index (LOTI)] is used in the study to calculate the change in surface temperature in NH and the entire globe from 1982 to 2009.

b. Methods

Following the radiative kernel method, the time (t)-dependent SIRF within a region R (here, the NH) of area A and the SIAF can be estimated separately in two steps (Flanner et al. 2011; Qu and Hall 2006):

 
formula
 
formula

As shown in Eq. (1), the pixel-level SIRF can be calculated in three terms, representing individual contributions or processes: is the TOA incoming solar radiation, represents the change in planetary albedo resulting from a standard perturbation of surface albedo (usually specified as 1%), and is the sea ice surface albedo contrast, calculated as the sea ice surface albedo minus the open ocean water albedo (here, 0.0676). The term together with forms the radiative kernel, usually described as the TOA flux variation with surface albedo change (). There is a slight difference in Eq. (1) from that in the study by Flanner et al. (2011): the albedo products (CLARA, ERA-Interim, and MERRA) are the ocean surface albedo of both sea ice and open water, so another parameter to characterize the influence of sea ice cover fraction on SIRF is unnecessary.

There are currently three methods to generate the radiative kernels: 1) a physically based regression model that expresses planetary albedo as a function of different contributions, such as surface albedo, cloud cover, and cloud optical thickness (Qu and Hall 2006); 2) an analytical model that expresses planetary albedo as the sum of the contributions of surface and atmosphere, each with an analytical function (Donohoe and Battisti 2011); and 3) a simulated method using an offline radiative transfer code to calculate the change in planetary albedo associated with a unit (1%) perturbation of surface albedo based on climate models (Shell et al. 2008; Soden et al. 2008). It suggested that the simulated method is likely more accurate than the other two methods (Qu and Hall 2014). Two widely used monthly radiative kernels are applied, one generated with the Geophysical Fluid Dynamics Laboratory Atmosphere Model 2 (GFDL AM2), and the other with the National Center for Atmospheric Research Community Atmosphere Model version 3 (NCAR CAM3) (Shell et al. 2008; Soden et al. 2008), to estimate the SIRF separately, and are averaged to obtain mean SIRF.

The TOA shortwave radiative forcing and its changes from 2000 to 2009 are also calculated with CERES SSF planetary albedo and insolation, following the method applied by Pistone et al. (2014). The result is compared with those from the kernel method and used for adjusting the all-sky estimate.

Based on Eq. (2), SIAF can be calculated as the change in SIRF () divided by the change in surface temperature (). Both and are calculated as the linear trends multiplied by the time intervals; the uncertainty of the changes are given as the 95% confidence intervals of the fit multiplied by the time intervals. Owing to the onset of polar night, we consider only the months March through September for albedo, and calculate the annual mean (January–December) surface temperature over the globe and the NH.

All variables and the radiative kernels are regridded to a spatial resolution of 0.25° × 0.25° in the first step to estimate the SIRF on the pixel level. Then, the SIRF and cloud fraction maps are reprojected to equal area projection, and the maximum sea ice coverage area from 1982 to 2009 in the NH recognized by the EASE-Grid 2.0 Weekly Sea Ice Extent product are applied as a mask to statistically measure the regional average of all of the variables.

3. Results

a. Radiative forcing of sea ice albedo

Based on the surface albedo contrast, radiative kernels and sea ice extension mask, we calculate the multiyear averaged clear-sky and all-sky NH SIRFs with three albedo products (Table 1). All values in Table 1 are calculated as the average of SIRFs estimated using two radiative kernels. Uncertainties are the 95% confidence intervals of the multiyear averaged annual SIRFs; Min and Max are the minimum and maximum values of the SIRF time series.

Table 1.

NH sea ice radiative forcing (SIRF), in W m−2, averaged over 1982–2009 and two kernels’ estimates for all three albedo products (CLARA, ERA-Interim, and MERRA). Uncertainties are the 95% confidence intervals of the multiyear averaged SIRFs; Min and Max are the minimum and maximum values of the SIRF time series.

NH sea ice radiative forcing (SIRF), in W m−2, averaged over 1982–2009 and two kernels’ estimates for all three albedo products (CLARA, ERA-Interim, and MERRA). Uncertainties are the 95% confidence intervals of the multiyear averaged SIRFs; Min and Max are the minimum and maximum values of the SIRF time series.
NH sea ice radiative forcing (SIRF), in W m−2, averaged over 1982–2009 and two kernels’ estimates for all three albedo products (CLARA, ERA-Interim, and MERRA). Uncertainties are the 95% confidence intervals of the multiyear averaged SIRFs; Min and Max are the minimum and maximum values of the SIRF time series.

As seen in Table 1, the all-sky SIRF in NH averaged over 1982–2009 calculated with CLARA is −1.65 ± 0.14 W m−2, which is very similar to that of ERA-Interim, −1.71 ± 0.08 W m−2; both are larger than that of MERRA, −1.40 ± 0.07 W m−2. The satellite-observed SIRF here from CLARA is slightly larger than that given by Flanner et al. (2011) (similar to the upper bound), −1.34 W m−2 (−0.92 to −1.70 W m−2 in their Table 1). Even for estimates using the same radiative kernels, the all-sky SIRF given by Flanner et al. (2011) is still smaller, about −1.40 (−1.13 to 1.63) W m−2. The difference can be partly attributed to the parameterized sea ice albedo calculation method, which may underestimate the North Pole–centered sea ice albedo in melt season (Flanner et al. 2011; Perovich et al. 2007b). Additionally, a small portion of the difference could be due to the different time periods used for analysis in these studies. The clear-sky SIRFs are nearly twice that of the all-sky SIRFs for all three products, indicating that the cooling effect of sea ice albedo on the Arctic climate is weakened by cloud overspread.

The seasonal cycles of SIRF for all three products shown in Fig. 1 indicates that both all-sky and clear-sky SIRF occur mainly in spring and early summer, particularly March to June, and peak in May as a result of the large magnitude of premelt sea ice albedo and relatively higher insolation. Both satellite-retrieved and reanalysis data can capture the seasonal variation of SIRF, which is very similar to the estimate by Flanner et al. (2011). However, the SIRF estimated with MERRA is smaller than other two products from March to June because of its relatively lower sea ice albedo in this period, which results in the lower multiyear averaged SIRF. The seasonal variation of SIRF also indicates that it is reasonable to use the months from March to September to represent the entire year (Pistone et al. 2014). Based on data analysis, insolation from October through February contributes little (about 4% in total) to the annual averaged incoming radiative flux in the Arctic region.

Fig. 1.

Seasonal cycles (March–September) of the Northern Hemisphere (NH) sea ice radiative forcing (SIRF) averaged over 1982–2009 and two radiative kernels for all three products for (a) all sky and (b) clear sky. The whiskers depict the 95% confidence intervals of the multiyear averaged SIRFs.

Fig. 1.

Seasonal cycles (March–September) of the Northern Hemisphere (NH) sea ice radiative forcing (SIRF) averaged over 1982–2009 and two radiative kernels for all three products for (a) all sky and (b) clear sky. The whiskers depict the 95% confidence intervals of the multiyear averaged SIRFs.

The spatial distribution of NH SIRF averaged from 1982 to 2009 with CLARA and two other reanalysis datasets is shown in Fig. 2. It demonstrate the annual-mean NH SIRF occurs mainly in the region north of 70°N, which may be due to the longer annual coverage time of ice cover over this area. Therefore, studies focusing on the SIRF and its changes over the region of 60°–90°N can represent the entire NH to a great extent and are comparable to studies focusing only on 70°–90°N. As shown in Fig. 2, the annual-mean SIRF at TOA in the Arctic Ocean can be larger than 40 W m−2 for all-sky and 60 W m−2 for clear-sky. Overall, the magnitude of SIRF estimated with CLARA surface albedo is very similar to that of ERA-Interim, and both are larger than that of MERRA.

Fig. 2.

NH SIRF averaged over 1982–2009 for (a) CLARA all sky, (b) CLARA clear sky, (c) ERA-Interim all sky, (d) ERA-Interim clear sky, (e) MERRA all sky, and (f) MERRA clear sky. SIRFs here are the mean value of estimates with two radiative kernels.

Fig. 2.

NH SIRF averaged over 1982–2009 for (a) CLARA all sky, (b) CLARA clear sky, (c) ERA-Interim all sky, (d) ERA-Interim clear sky, (e) MERRA all sky, and (f) MERRA clear sky. SIRFs here are the mean value of estimates with two radiative kernels.

b. Changes in sea ice albedo radiative forcing (ΔSIRF)

The time series of annual-mean (March–September) spatial-mean SIRFs together with the statistical linear changes of SIRFs are shown in Fig. 3; all SIRFs have been averaged over the entire NH. The ΔSIRFs from 1982 to 2009 are calculated as the linear trend of the SIRF time series multiplied by the time intervals. Uncertainties of the changes are calculated as the 95% confidence intervals of the trend multiplied by the time intervals. The all-sky ΔSIRF over the NH with the CLARA albedo product is 0.20 ± 0.05 W m−2, which is equivalent to an energy increase of 0.10 ± 0.03 W m−2 over the entire globe. The clear-sky NH ΔSIRF with CLARA albedo is 0.46 ± 0.12 W m−2. Both all-sky and clear-sky NH ΔSIRFs estimated with satellite retrievals are larger (nearly 2 times) than those of the other two reanalysis products, implying that although satellite-retrieved product is not perfect, it is able to capture the change in surface albedo caused by sea ice loss. In light of these results and given the conclusions of the previous studies (Dessler 2013; Flanner et al. 2011), we can conclude that both reanalysis datasets and CMIP3 GCMs underestimate the NH ΔSIRF for the last three decades.

Fig. 3.

NH SIRF averaged over two radiative kernels and the estimated changes of SIRFs (ΔSIRF) from 1982 to 2009 (W m−2) for (a) all sky and (b) clear sky. The 95% confidence intervals of the ΔSIRF were also given. All linear changes have passed the 0.01 significance test.

Fig. 3.

NH SIRF averaged over two radiative kernels and the estimated changes of SIRFs (ΔSIRF) from 1982 to 2009 (W m−2) for (a) all sky and (b) clear sky. The 95% confidence intervals of the ΔSIRF were also given. All linear changes have passed the 0.01 significance test.

The estimated NH ΔSIRF of 0.20 ± 0.05 W m−2 with CLARA albedo is very close to the value of 0.22 (0.15–0.32) W m−2 averaged over 12 all-sky kernel estimates presented by Flanner et al. (2011). Considering the shorter time period (28 yr) in this study compared to the previous study (30 yr) and the method to calculate ΔSIRF (dependent on time intervals), nearly the same estimate is obtained here as Flanner et al. (2011). However, both estimates are lower than that by Pistone et al. (2014), who estimated the ΔSIRF from 1979 to 2011 to be 0.43 ± 0.07 W m−2 with TOA observationally based planetary albedo product.

Figure 4 shows the monthly NH ΔSIRF from 1982 to 2009, which indicates that the change in NH SIRF occurs primarily in the melt season (May–August), when both the TOA insolation and the change in surface albedo are relatively larger, and peaks in June for both clear-sky and all-sky conditions with the largest ΔSIRF values as 1.36 ± 0.50 W m−2 and 0.64 ± 0.25 W m−2 respectively. The seasonal variation of ΔSIRF appears similar to the results of Flanner et al. (2011). The large relative uncertainty of all-sky ΔSIRF in July is high, likely caused by the high cloud fraction during this period (Karlsson and Svensson 2013). It also shows both ERA-Interim and MERRA significantly underestimate the change in SIRF in the melt season, and consequently the annual averaged ΔSIRF.

Fig. 4.

Seasonal cycles of NH changes in sea ice radiative forcing (ΔSIRF) based on the average of two radiative kernels from 1982 to 2009 for (a) all sky and (b) clear sky. The whiskers depict the 95% confidence intervals of monthly changes of SIRFs. All of the changes, except value from ERA-Interim in July (0.1), have passed the 0.05 significance test.

Fig. 4.

Seasonal cycles of NH changes in sea ice radiative forcing (ΔSIRF) based on the average of two radiative kernels from 1982 to 2009 for (a) all sky and (b) clear sky. The whiskers depict the 95% confidence intervals of monthly changes of SIRFs. All of the changes, except value from ERA-Interim in July (0.1), have passed the 0.05 significance test.

The spatial distribution of NH ΔSIRF shown in Fig. 5 indicates that the change in sea ice surface albedo during the last three decades occurred mainly from 70° to 80°N in the Arctic Ocean. Three hot spots—Baffin Bay, North Barents–Kara Sea, and Chukchi Sea—play leading roles in the process of changing SIRF. Both ERA-Interim and MERRA can replicate the main spatial patterns of the ΔSIRF, but are numerically lower than the estimates with the CLARA surface albedo product. It should be noted that even though both reanalysis products underestimate the change in SIRF, the reasons are likely different. For ERA-Interim, the negative change in SIRF in the polar region (north of 83°N) and east of Greenland may bring down the total ΔSIRF, while for MERRA the nearly zero change in SIRF in the polar region (north of 80°N) combined with the lower magnitudes in the margin (70°–80°N) region may result in the smaller total ΔSIRF. More accurate data assimilation in the polar region (north of 80°N) would greatly improve the quality of reanalysis.

Fig. 5.

Changes in sea ice radiative forcing (ΔSIRF) from 1982 to 2009 based on two radiative kernels for (a) CLARA all sky, (b) CLARA clear sky, (c) ERA-Interim all sky, (d) ERA-Interim clear sky, (e) MERRA all sky, and (f) MERRA clear sky.

Fig. 5.

Changes in sea ice radiative forcing (ΔSIRF) from 1982 to 2009 based on two radiative kernels for (a) CLARA all sky, (b) CLARA clear sky, (c) ERA-Interim all sky, (d) ERA-Interim clear sky, (e) MERRA all sky, and (f) MERRA clear sky.

c. Sea ice–albedo feedback

The changes of annual-mean (January–December) surface air temperature (ΔTs) from 1982 to 2009 estimated with the GISS surface temperature product are 0.79 ± 0.16 K over the Northern Hemisphere and 0.54 ± 0.12 K over the entire globe (Fig. 6). By combining ΔSIRF and ΔTs, the feedback parameter associated with the change in NH sea ice albedo (NH SIAF) can be calculated with Eq. (2). Table 2 shows the SIAF and the calculated range (low and high bound) for each product (e.g., for the low bound, divide minimum ΔSIRF by maximum ΔTs; vice versa for the high bound).

Fig. 6.

Annual averaged (January–December) Goddard Institute of Space Studies (GISS) surface temperature anomaly over the NH (black) and the entire globe (GL; blue) from 1982 to 2009. GISS surface temperature anomaly is calculated based on 1951–80 climatology. The linear changes in surface temperature from 1982 to 2009 and the 95% confidence interval of the changes are given.

Fig. 6.

Annual averaged (January–December) Goddard Institute of Space Studies (GISS) surface temperature anomaly over the NH (black) and the entire globe (GL; blue) from 1982 to 2009. GISS surface temperature anomaly is calculated based on 1951–80 climatology. The linear changes in surface temperature from 1982 to 2009 and the 95% confidence interval of the changes are given.

Table 2.

NH and global sea ice albedo feedback (SIAF), in W m−2 K−1. The ranges (low and high) indicate the extreme minimum/maximum combinations of ΔSIRF and ΔTs considering the uncertainties of both ΔSIRF and ΔTs.

NH and global sea ice albedo feedback (SIAF), in W m−2 K−1. The ranges (low and high) indicate the extreme minimum/maximum combinations of ΔSIRF and ΔTs considering the uncertainties of both ΔSIRF and ΔTs.
NH and global sea ice albedo feedback (SIAF), in W m−2 K−1. The ranges (low and high) indicate the extreme minimum/maximum combinations of ΔSIRF and ΔTs considering the uncertainties of both ΔSIRF and ΔTs.

As shown in Table 2, the reanalysis-based estimates of global SIAF are 0.09 (0.05–0.17) W m−1 K−1 for ERA-Interim and 0.07 (0.04–0.13) W m−2 K−1 for MERRA, which are close to the model-based estimations of 0.10 ± 0.03 W m−2 K−1 from CMIP3 models and 0.11 ± 0.04 W m−2 K−1 from CMIP5 models reported in previous studies (Winton 2006; Pistone et al. 2014) within the range of uncertainties. This is consistent with Dessler’s (2013) findings. However, both ERA-Interim and MERRA underestimate nearly half of the NH SIAF to the global climate compared to the CLARA satellite-based estimate of 0.19 (0.11–0.30) W m−2 K−1. The SIAF for NH is 0.25 (0.16–0.40) W m−2 K−1, which is consistent with the Flanner et al. (2011) estimate of 0.28 (0.19–0.41) W m−2 K−1.

d. Adjustment of satellite based all-sky ΔSIRF and SIAF

As indicated above, the estimates of ΔSIRF with CLARA satellite-retrieved long-term time series surface albedo is nearly the same as that of Flanner et al. (2011). However, both all-sky estimates are only half of a recent estimate by Pistone et al. (2014), who used a derived planetary albedo directly. To determine the underlying causes of the difference between our estimate and that given by Pistone et al. (2014), the SIRF values estimated with kernel method and the TOA planetary albedo from CERES SSF product during the overlapping period from 2000 to 2009 were analyzed. The data processing for CERES SSF product is based on the same spatial domain (the maximum sea ice coverage area from 1982 to 2009 in the NH), and the same temporal domain (the months from March to September). The regional average statistics are also based on equal area map projection.

As shown in Fig. 7, the clear-sky ΔSIRF from 2000 to 2009 calculated using radiative kernels with CLARA surface albedo is 4.13 ± 1.84 W m−2, slightly lower than CERES SSF observed change in clear-sky TOA reflected shortwave flux of 4.56 ± 1.75 W m−2, which may be due to a small decrease in planetary albedo from increasing water vapor in the Arctic atmosphere (Dessler et al. 2013; Serreze et al. 2012). Pistone et al. (2014) also acknowledge that their estimate is an upper bound because of the decrease of the albedo caused by other factors. This implies that the clear-sky ΔSIRF estimated using the radiative kernels with the satellite albedo is more reliable for it has excluded the influences from other climate variables on the change in clear-sky planetary albedo. However, the all-sky ΔSIRF of 1.81 ± 0.92 W m−2 calculated using the radiative kernels is much lower than the change in all-sky TOA reflected shortwave flux (2.85 ± 0.99 W m−2) from CERES SSF products, indicating that the difference between the two methods may lie in the interaction of sea ice albedo and cloud radiation (the response of cloud albedo to the variation of surface sea ice albedo). It has been found that changes in surface sea ice albedo can result in a change in cloud albedo (and thus cloud radiative forcing), even if the cloud properties do not change (Shell et al. 2008). Inspired by this finding, we attempt to determine the causes to the variation of cloudy-sky albedo in Arctic Ocean.

Fig. 7.

Annual averaged (a) CLARA SIRF; (b) single satellite footprint (SSF) top of atmosphere (TOA) upward shortwave (SW) flux; and (c) calculated SSF cloudy-sky SW, adjusted all-sky SW flux, and cloud fraction. Figure 7d is the scatterplot of cloudy-sky upward SW flux and clear-sky upward flux. In subplots (a)–(c), the linear changes of each variable over 2000 to 2009 and the 95% confidence interval of the changes are given. For SSF cloud fraction, the multiyear average and the 95% confidence intervals of the mean value are also given.

Fig. 7.

Annual averaged (a) CLARA SIRF; (b) single satellite footprint (SSF) top of atmosphere (TOA) upward shortwave (SW) flux; and (c) calculated SSF cloudy-sky SW, adjusted all-sky SW flux, and cloud fraction. Figure 7d is the scatterplot of cloudy-sky upward SW flux and clear-sky upward flux. In subplots (a)–(c), the linear changes of each variable over 2000 to 2009 and the 95% confidence interval of the changes are given. For SSF cloud fraction, the multiyear average and the 95% confidence intervals of the mean value are also given.

Following the previous study (Pistone et al. 2014), all-sky planetary albedo is related to cloudy-sky planetary albedo, clear-sky planetary albedo, and cloud fraction (fc), as shown in Eq. (3). To estimate the changes of radiative flux, all albedo parameters have been converted to upward shortwave (SW) radiative flux [i.e., clear-sky upward SW (), cloudy-sky upward SW (), and all-sky upward SW flux ()] by multiplying the planetary albedo by a TOA climatology insolation:

 
formula

If all other terms are known, the TOA can be calculated with the above equation. Figure 7c shows the calculated based on the CERES SSF clear-sky, all-sky planetary albedo, TOA solar insolation flux, and the cloud fraction from the CERES–MODIS product. The figure indicates that the cloud reflected shortwave radiative flux decreases significantly. It is clear that most changes in clear-sky planetary albedo are due to the variation of sea ice (and thus sea ice albedo) (Koenigk et al. 2014; Pistone et al. 2014). Therefore, the high correlation between cloudy-sky and clear-sky upward shortwave radiative flux shown in Fig. 7d implies that the decrease of cloudy-sky planetary albedo over the Arctic Ocean is very likely to have arisen from the change in surface sea ice albedo as well, and is unfortunately underestimated by all-sky radiative kernels. The current CAM radiative kernel generated from the CAM5 model masks much less of the sea ice albedo effect on the TOA shortwave radiative flux which may offer a more realistic estimate of Arctic SIRF and its change (Perket et al. 2014). We also analyze the relationship between cloudy-sky planetary albedo and cloud optical thickness, and find no significant correlation between them. Based on the analysis above, we adjust the underestimated all-sky ΔSIRF by the kernels method.

Cloud coverage can significantly influence the variation of all-sky planetary albedo. To remove the effect of cloud fraction on the variation of , we fix the cloud fraction in the first year (2000) and then recalculate with Eq. (3) to get an adjusted all-sky upward shortwave radiative flux (, as shown in Fig. 7c). The year 2000 is chosen for fixing the cloud fraction because cloud fraction in this year is 72.2%, nearly the same as the multiyear average, 72.6%; a different cloud fraction would introduce a small uncertainty (about 1%) to the correction factor in Eq. (4). Afterward, the change in adjusted all-sky upward SW () is 3.30 ± 1.32 W m−2. Based on the radiative kernel method, a time-invariant cloud fraction results in a time-invariant ratio of all-sky ΔSIRF to clear-sky ΔSIRF, which means the ratio of to during 2000 to 2009 can be used to correct the underestimated all-sky ΔSIRF from kernel method with Eq. (4):

 
formula

Here, is the clear-sky ΔSIRF estimated with the radiative kernels, is the change in adjusted all-sky upward shortwave radiative flux, is the change in clear-sky upward shortwave radiative flux estimated with CERES SSF products, the ratio of to is the correction factor, and is the corrected all-sky ΔSIRF estimated with kernels. After correction, the all-sky ΔSIRF from 2000 to 2009 is 2.99 W m−2. The annual averaged (March–September) cloud fraction over Arctic Ocean from CLARA product shows no significant change over 1982 to 2009, and the multiyear mean value is about 72.5% (Fig. 8), which is the same as multiyear averaged cloud fraction of 72.6% estimated with the CERES SSF cloud fraction product (Fig. 7c) from 2000 to 2009. This means the all-sky ΔSIRF from 1982 to 2009 can also be corrected directly with Eq. (4) and the correction factor generated from CERES SSF TOA product during 2000 to 2009. The kernel-based all-sky NH ΔSIRF can be corrected from 0.20 ± 0.05 W m−2 to 0.33 ± 0.09 W m−2, which is larger than the lower bound estimate from Flanner et al. (2011) and smaller than the upper bound estimate from Pistone et al. (2014). Correspondingly, the all-sky SIAF can be adjusted to 0.42 W m−2 K−1 for the NH and 0.31 W m−2 K−1 for the entire globe.

Fig. 8.

Annual averaged (March–September) CLARA cloud fraction (CFC) over the NH sea ice–covered region from 1982 to 2009. The linear change in the cloud fraction time series (ΔCFC) from 1982 to 2009 and the 95% confidence intervals of the change are given. The multiyear average of CFC and the 95% confidence intervals of the mean value are also given.

Fig. 8.

Annual averaged (March–September) CLARA cloud fraction (CFC) over the NH sea ice–covered region from 1982 to 2009. The linear change in the cloud fraction time series (ΔCFC) from 1982 to 2009 and the 95% confidence intervals of the change are given. The multiyear average of CFC and the 95% confidence intervals of the mean value are also given.

4. Discussion

In this study, only the months March through September are considered for the estimation of annual averaged radiative forcing, because the insolation during October through February contributes very little (about 4%, which would lead to a small underestimation for the annual-mean SIRF) to the annual radiative flux in the Arctic region. For the calculated changes in SIRFs during last three decades, the influence from these months is even negligible, because the decrease of sea ice (and thus albedo) happens primarily in the melt season.

When we try to determine if the change in all-sky SIRF estimated with the kernel method is underestimated or not, the key issue is which climate variable caused the variation of cloudy-sky planetary albedo. Through data analysis with CERES SSF TOA products, we found that the variations (more than 90%) of cloudy-sky planetary albedo over Arctic Ocean are mainly caused by sea ice loss (and thus decreasing sea ice albedo), while cloud optical thickness has little effect on the change in cloudy-sky albedo. Based on this important finding, we adjusted the change in SIRF estimated using the kernel methods with surface albedo and obtain a more realistic estimate between the lower bound from Flanner et al. (2011) and the upper bound from Pistone et al. (2014). Considering the underestimation of change in all-sky SIRF by kernel method, a more accurate simulation on the response of cloudy-sky planetary albedo to the variation of surface albedo would greatly help improve the simulation of energy budget and the future sea ice loss in Arctic.

5. Conclusions

The role of SIAF in Arctic warming amplification continues to be a debated issue (Graversen et al. 2014; Kumar et al. 2010; Pithan and Mauritsen 2014; Screen and Simmonds 2010). An improved quantification of the SIAF is critical for better understanding of the physical mechanisms of the accelerated Arctic sea ice loss and assessing the underlying future evolution of the Arctic warming amplification. Most previous SIAF assessments were based on model simulations (Colman 2003, 2013; Dessler 2013; Pithan and Mauritsen 2014; Taylor et al. 2013; Winton 2006). Two recent studies using synthesized satellite-observed long-term products reached significantly different results (Flanner et al. 2011; Pistone et al. 2014). To take advantage of both satellite observation and model simulations, an approach was proposed to estimate the change in SIRF and SIAF to the climate over NH and the entire globe from 1982 to 2009.

Based on the assessment using the kernel method with the CLARA surface albedo product, the NH all-sky SIRF is estimated as −1.65 ± 0.14 W m−2, and the linear change of all-sky SIRF indicates that a 0.20 ± 0.05 W m−2 shortwave radiative flux was absorbed by the NH owing to the loss of sea ice within the 28-yr period, yielding an SIAF of 0.25 W m−2 K−1 for NH and 0.19 W m−2 K−1 for the entire globe. This result is the same as the estimate from Flanner et al. (2011), who also used the kernel method, but both are smaller than that reported by Pistone et al. (2014), who estimated the change in SIRF directly using planetary albedo. To reconcile the difference between two methods, further data analysis is performed, indicating that although the kernel method can separate the forcing of different climate variables, it is likely to underestimate the change in all-sky SIRF because of the poor representation of sea ice albedo and cloud–radiation interactions (too much of the surface albedo effect of sea ice is masked by all-sky radiative kernels). After correction, the change in all-sky SIRF can be adjusted to 0.33 ± 0.09 W m−2, yielding an adjusted SIAF of 0.43 W m−2 K−1 for the NH and 0.31 W m−2 K−1 for the entire globe. Three regions—Baffin Bay, the North Barents–Kara Sea, and the Chukchi Sea—played leading roles in decrease of NH SIRF during the period.

This study also determined both ERA-Interim and MERRA reanalysis products substantially underestimate the change in SIRF due to the poor replication of the change in sea ice albedo in melt season (May–August) and the polar region (north of 80°N). Considering the reports of previous studies (Dessler 2013; Flanner et al. 2011), we conclude that both reanalysis data and GCMs (CMIP3 and CMIP5) underestimate the change in SIRF over the last three decades. To achieve a realistic estimate of sea ice variation, more accurate data simulation and assimilation in melt season and the polar region are needed for reanalysis products.

Acknowledgments

This study is supported by the High-Tech Research and Development Program of China (2013AA121201) and the China Scholarship Council. The sea ice extent product was obtained from the National Snow and Ice Data Center (NSIDC), the CLARA-A1 data from the Climate Monitoring Satellite Application Facility (CM SAF) project, ERA-Interim data from the European Centre for Medium-Range Weather Forecasts (ECMWF), MERRA albedo product from NASA’s Global Modeling and Assimilation Office, the SSF1deg radiative and cloud products from the NASA Clouds and Earth’s Radiant Energy System, and the surface temperature from the NASA Goddard Institute for Space Studies. We thank Meredith G. L. Brown for polishing this manuscript. We also wish to thank two anonymous reviewers for their constructive comments, which have greatly improved the presentation of this paper.

REFERENCES

REFERENCES
Colman
,
R. A.
,
2003
:
A comparison of climate feedbacks in general circulation models
.
Climate Dyn.
,
20
,
865
873
, doi:.
Colman
,
R. A.
,
2013
:
Surface albedo feedbacks from climate variability and change
.
J. Geophys. Res. Atmos.
,
118
,
2827
2834
, doi:.
Comiso
,
J. C.
, and
D. K.
Hall
,
2014
:
Climate trends in the Arctic as observed from space
.
Wiley Interdiscip. Rev.: Climate Change
,
5
,
389
409
, doi:.
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:.
Crook
,
J. A.
,
P. M.
Forster
, and
N.
Stuber
,
2011
:
Spatial patterns of modeled climate feedback and contributions to temperature response and polar amplification
.
J. Climate
,
24
,
3575
3592
, doi:.
Dee
,
D. P.
, and Coauthors
,
2011
:
The ERA-Interim reanalysis: Configuration and performance of the data assimilation system
.
Quart. J. Roy. Meteor. Soc.
,
137
,
553
597
, doi:.
Dessler
,
A. E.
,
2013
:
Observations of climate feedbacks over 2000–10 and comparisons to climate models
.
J. Climate
,
26
,
333
342
, doi:.
Dessler
,
A. E.
,
M. R.
Schoeberl
,
T.
Wang
,
S. M.
Davis
, and
K. H.
Rosenlof
,
2013
:
Stratospheric water vapor feedback
.
Proc. Natl. Acad. Sci. USA
,
110
,
18 087
18 091
, doi:.
Donohoe
,
A.
, and
D. S.
Battisti
,
2011
:
Atmospheric and surface contributions to planetary albedo
.
J. Climate
,
24
,
4402
4418
, doi:.
Flanner
,
M. G.
,
K. M.
Shell
,
M.
Barlage
,
D. K.
Perovich
, and
M. A.
Tschudi
,
2011
:
Radiative forcing and albedo feedback from the Northern Hemisphere cryosphere between 1979 and 2008
.
Nat. Geosci.
,
4
,
151
155
, doi:.
Graversen
,
R. G.
,
P. L.
Langen
, and
T.
Mauritsen
,
2014
:
Polar amplification in CCSM4: Contributions from the lapse rate and surface albedo feedbacks
.
J. Climate
,
27
,
4433
4450
, doi:.
Hansen
,
J.
,
R.
Ruedy
,
M.
Sato
, and
K.
Lo
,
2010
:
Global surface temperature change
.
Rev. Geophys.
,
48
,
RG4004
, doi:.
Karlsson
,
J.
, and
G.
Svensson
,
2013
:
Consequences of poor representation of Arctic sea-ice albedo and cloud–radiation interactions in the CMIP5 model ensemble
.
Geophys. Res. Lett.
,
40
,
4374
4379
, doi:.
Karlsson
,
K.-G.
, and Coauthors
,
2013
:
CLARA-A1: A cloud, albedo, and radiation dataset from 28 yr of global AVHRR data
.
Atmos. Chem. Phys.
,
13
,
5351
5367
, doi:.
Kerr
,
R. A.
,
2009
:
Arctic summer sea ice could vanish soon but not suddenly
.
Science
,
323
,
1655
, doi:.
Koenigk
,
T.
,
A.
Devasthale
, and
K. G.
Karlsson
,
2014
:
Summer Arctic sea ice albedo in CMIP5 models
.
Atmos. Chem. Phys.
,
14
,
1987
1998
, doi:.
Kumar
,
A.
, and Coauthors
,
2010
:
Contribution of sea ice loss to Arctic amplification
.
Geophys. Res. Lett.
,
37
,
L21701
, doi:.
Kwok
,
R.
, and
D. A.
Rothrock
,
2009
:
Decline in Arctic sea ice thickness from submarine and ICESat records: 1958–2008
.
Geophys. Res. Lett.
,
36
,
L15501
, doi:.
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:.
Maslanik
,
J.
,
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:.
Parkinson
,
C. L.
, and
D. J.
Cavalieri
,
2012
:
Antarctic sea ice variability and trends, 1979–2010
.
Cryosphere
,
6
,
871
880
, doi:.
Perket
,
J.
,
M. G.
Flanner
, and
J. E.
Kay
,
2014
:
Diagnosing shortwave cryosphere radiative effect and its 21st century evolution in CESM
.
J. Geophys. Res. Atmos.
,
119
,
1356
1362
, doi:.
Perovich
,
D. K.
,
B.
Light
,
H.
Eicken
,
K. F.
Jones
,
K.
Runciman
, and
S. V.
Nghiem
,
2007a
:
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:.
Perovich
,
D. K.
,
S. V.
Nghiem
,
T.
Markus
, and
A.
Schweiger
,
2007b
:
Seasonal evolution and interannual variability of the local solar energy absorbed by the Arctic sea ice–ocean system
.
J. Geophys. Res.
,
112
,
C03005
, doi:.
Pistone
,
K.
,
I.
Eisenman
, and
V.
Ramanathan
,
2014
:
Observational determination of albedo decrease caused by vanishing Arctic sea ice
.
Proc. Natl. Acad. Sci. USA
,
111
,
3322
3326
, doi:.
Pithan
,
F.
, and
T.
Mauritsen
,
2014
:
Arctic amplification dominated by temperature feedbacks in contemporary climate models
.
Nat. Geosci.
,
7
,
181
184
, doi:.
Qu
,
X.
, and
A.
Hall
,
2006
:
Assessing snow albedo feedback in simulated climate change
.
J. Climate
,
19
,
2617
2630
, doi:.
Qu
,
X.
, and
A.
Hall
,
2014
:
On the persistent spread in snow–albedo feedback
.
Climate Dyn.
,
42
,
69
81
, doi:.
Rienecker
,
M. M.
, and Coauthors
,
2011
:
MERRA: NASA’s Modern-Era Retrospective Analysis for Research and Applications
.
J. Climate
,
24
,
3624
3648
, doi:.
Riihelä
,
A.
,
T.
Manninen
, and
V.
Laine
,
2013a
:
Observed changes in the albedo of the Arctic sea-ice zone for the period 1982–2009
.
Nat. Climate Change
,
3
,
895
898
, doi:.
Riihelä
,
A.
,
T.
Manninen
,
V.
Laine
,
K.
Andersson
, and
F.
Kaspar
,
2013b
:
CLARA-SAL: A global 28 yr timeseries of Earth’s black-sky surface albedo
.
Atmos. Chem. Phys.
,
13
,
3743
3762
, doi:.
Screen
,
J. A.
, and
I.
Simmonds
,
2010
:
The central role of diminishing sea ice in recent Arctic temperature amplification
.
Nature
,
464
,
1334
1337
, doi:.
Serreze
,
M. C.
,
A. P.
Barrett
,
J. C.
Stroeve
,
D. N.
Kindig
, and
M. M.
Holland
,
2009
:
The emergence of surface-based Arctic amplification
.
Cryosphere
,
3
,
11
19
, doi:.
Serreze
,
M. C.
,
A. P.
Barrett
, and
J. C.
Stroeve
,
2012
:
Recent changes in tropospheric water vapor over the Arctic as assessed from radiosondes and atmospheric reanalyses
.
J. Geophys. Res.
,
117
,
D10104
, doi:.
Shell
,
K. M.
,
J. T.
Kiehl
, and
C. A.
Shields
,
2008
:
Using the radiative kernel technique to calculate climate feedbacks in NCAR’s Community Atmospheric Model
.
J. Climate
,
21
,
2269
2282
, doi:.
Soden
,
B. J.
,
I. M.
Held
,
R.
Colman
,
K. M.
Shell
,
J. T.
Kiehl
, and
C. A.
Shields
,
2008
:
Quantifying climate feedbacks using radiative kernels
.
J. Climate
,
21
,
3504
3520
, doi:.
Stroeve
,
J. C.
,
M. M.
Holland
,
W.
Meier
,
T.
Scambos
, and
M.
Serreze
,
2007
:
Arctic sea ice decline: Faster than forecast
.
Geophys. Res. Lett.
,
34
,
L09501
, doi:.
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:.
Taylor
,
P. C.
,
M.
Cai
,
A.
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:.
Winton
,
M.
,
2006
:
Amplified Arctic climate change: What does surface albedo feedback have to do with it?
Geophys. Res. Lett.
,
33
,
L03701
, doi:.