Antarctic skin temperature warming related to enhanced downward longwave radiation associated with increased atmospheric advection of moisture and temperature

We used the ECMWF monthly mean ERA5 reanalysis data from 1950 to 2020 (from 1950/1951 to 2020/2021 for summer [December to February]) at a horizontal resolution of 0.5° (Hersbach et al 2020) (available from https://www.ecmwf.int). To investigate temperature variability and its causes, various atmospheric (e.g. sea level pressure (SLP), skin temperature, shortwave and longwave radiation, latent and sensible heat flux, amount of integrated water and ice cloud) and oceanic (e.g. sea ice concentration and SST) parameters were used. The ERA5 dataset comprises ten ensemble members. To evaluate the performance of air temperature in the ERA5, we used air temperature observation data at surface and pressure level (850, 700, 500, 300 hPa) in the Antarctic Climate Data, which are available from the SCAR READER project (Turner et al 2004) (available from https://legacy.bas.acuk/met/READER).

To investigate the causes in Antarctica skin temperature changes, we conducted surface budget analyses, which have been used previously for understanding Arctic surface climate change (Lesins et al 2012, Lee et al 2017). With this approach, we revealed how the various components of the energy budget contribute to variations in skin temperature.

The surface energy budget can be written as:

$$\begin{equation}G = {S_d} + {S_u} + {l_d} + {l_u} + {F_{{\text{sh}}}} + {F_{{\text{lh}}}} + R,\end{equation} \tag{ 1 }$$

where l and S are longwave and shortwave radiation, respectively, and u denotes upward and d downward, G is a storage term, F_lh is the surface latent heat flux and F_sh is the surface sensible heat flux. The residual term R includes heat conduction from below the surface (e.g. heat conduction through snow and/or heat loss through the melting of snow). When the fluxes are directed toward the surface (i.e. toward the interface), they are defined as positive. After taking the differential of equation (1), the trend in the surface energy budget is:

$$\begin{align}\Delta G = \Delta {S_d} \,{+}\, \Delta {S_u} \,{+}\, \Delta {l_d} \,{+}\, \Delta {l_u} \,{+}\, \Delta {F_{{\text{sh}}}} \,{+}\, \Delta {F_{{\text{lh}}}} \,{+}\, \Delta R,\end{align} \tag{ 2 }$$

where the differential operator Δ represents the trend. In this study, we assumed an infinitesimally thin air surface interface with a very small heat capacity (Lesins et al 2012, Lee et al 2017), and hence the storage term G can be dropped.

The upward longwave flux is expressed as −σT_s ⁴, where σ is the Stefan–Boltzmann constant, is the surface emissivity (herein set to 1) and T_s is the skin temperature. We may then write:

$$\begin{equation*}\Delta {l_u} = { } - \varepsilon \sigma { }\Delta {T_{\text{s}}}^4 = { } - \varepsilon \sigma { }4{T_{\text{s}}}^3\Delta {T_{\text{s}}}.\end{equation*}$$

Substituting this expression into equation (2) we derive:

$$\begin{align}\Delta {T_s}\, = {\text{ }}&(\Delta {S_d} + \Delta {S_u} + \Delta {l_d} + \Delta {F_{sh}} + \Delta {F_{lh}} \nonumber\\ & \quad + \Delta R){\text{ }}/{\text{ }}(4\varepsilon \sigma {T_s}^3).\end{align} \tag{ 3 }$$

Our equation (3) is identical to equation (3) in Lee et al (2017), except that we have included the trends of the upward and downward shortwave radiation because we examined all four seasons (they only studied the Arctic winter). The T_s in the denominator was the mean value of each season averaged over the study interval at each grid point. The parameter R is not provided in the ERA5 dataset. Therefore, the difference between ΔT_s and ((ΔS_d + ΔS_u + Δl_d + ΔF_sh + ΔF_lh)/4σT_s ³) was used as the ΔR/(4σT_s ³).

Figure 1(a) shows the time series of annual mean ERA5 skin temperature anomalies over Antarctica for each of the ten ensemble members and their mean. There was an overall warming trend over the entire continent from 1950 to 2020 in all members (0.30 ± 0.02 °C decade⁻¹). Although the skin temperature exhibited ensemble spread in all years; the spread became small from 1979 because of an increase in assimilated satellite data sets. Conversely, before 1978, there was relatively large ensemble spread, particularly in the early 1950s, compared with 1979. To reduce the parameter uncertainty in the ERA5, we used ten ensemble mean data. We calculated the September–November (spring), December–February (summer), March–May (autumn) and June–August (winter) linear trends of each parameter in the energy budget. In addition, skin temperature data in another four reanalysis data sets (ERA20C (available from https://www.ecmwf.int): Poli et al 2016; ERA-Interim (available from https://www.ecmwf.int): Dee et al 2011; CFSR (available from https://cfs.ncep.noaa.gov/cfsr): Saha et al 2010, 2014; JRA55 (available from https://jra.kishou.go.jp): Kobayashi et al 2015) were used to compare the amplitudes and trends of surface skin temperature change among these data (figure S1 (available online at stacks.iop.org/ERL/16/064059/mmedia)). There are robust long term warming trends among three reanalyses (ERA5, ERA20C and JRA55), and all reanalyses have Antarctica warming with statistical significance above the 95% level.

Zoom In Zoom Out Reset image size

Figure 1(b) shows the geographical distribution of the linear trend of annual mean skin temperatures over Antarctica (1950–2020). A sizeable portion of the continent exhibited significant positive trends including the Antarctic Peninsula, the eastern part of West Antarctica, the Ross Sea, the coast around 0°E and the interior of East Antarctica. Previous studies reported differences in the distribution and magnitude of surface air temperature trends for each season (Steig et al 2009, O'Donnell et al 2010, Nicolas and Bromwich 2014, Jones et al 2019). Our linear trends of skin temperature for each season are shown in figures 2(a)–(d). Although significant trends in summer were limited to the coastal area around 0°E and the southern part of the Ross Sea (figure 2(b)), widespread warming trends were observed in the other three seasons (figures 2(a), (c) and (d)). The magnitude, areal extent and significance are most marked over West Antarctica in spring, the Antarctic Peninsula in autumn and East Antarctica in winter.

Zoom In Zoom Out Reset image size

To explore the cause of Antarctica skin temperature change, we calculated linear trends of terms (i.e. l_d/4σT_s ³, (F_sh + F_lh)/4σT_s ³, (S_u + S_d)/4σT_s ³ and R/4σT_s ³) in equation (3) for the four seasons during 1950–2020 (figures 2(e)–(l) and S2(a)–(h)). The distribution and magnitude of linear trends of l_d/4σT_s ³ were very similar to skin temperature over most of Antarctica, except in summer and high elevations around 0°E (figures 2(a)–(h)). Although there was a widespread positive trend in (F_sh + F_lh)/4σT_s ³ over the interior of East Antarctica in summer, they were very weak in general (figure 2(j)). Very few trends in the other three seasons were significant (figures 2(i), (k) and (l)). The linear trends of (S_u + S_d)/4σT_s ³ were very small in all seasons because of the high albedo of the snow/ice cover over Antarctica (figures S2(a)–(d)). During spring and summer, the ozone hole reduces absorption of shortwave radiation in the ozone layer, influencing surface climate thorough changes in atmospheric circulation and increased downward shortwave radiation at the surface (Thompson and Solomon 2002, Thompson et al 2011). However, increases in surface downward shortwave radiation have a minor direct impact on surface warming (Chiodo et al 2017). The linear trends of R/4σT_s ³ were a little stronger but were still very modest (figures S2(e)–(h)). There were positive values in coastal areas around 0°E and the southern part of the Ross Sea; however, statistically significant change was limited to the southern Ross Sea in spring. Interpreting these results, we found that l_d/4σT_s ³ is the most important term in equation (3), which indicates that the increase in downward longwave radiation has a major impact on Antarctic skin temperature warming. Similar behavior of longwave radiation on surface budget is shown by Kumar et al (2021).

The downward longwave radiation is strongly influenced by change in atmospheric temperature and cloud properties. To understand the causes of changes in downward longwave radiation over our 71 year study period, we analyzed linear trends of air temperature at 2 m (T2m) and vertical integral of water vapor flux for each season from 1950 to 2020 (figures 3(a)–(d)). Over West Antarctica and the Antarctic coastal regions, the trends in T2m were very similar to skin temperature in all seasons, meaning that surface air temperature warming has key role in skin temperature warming. Increases in air temperature are induced by enhanced warm advection from the middle latitudes (Simmonds and Murray 1999, Simmonds and Keay 2000, Simmonds et al 2002, Simmonds 2003, 2015, Simmonds and King 2004, Hosking et al 2013, Clem et al 2017), in agreement with poleward advection increase from midlatitude (figures 3(a)–(d)).

Zoom In Zoom Out Reset image size

To illustrate the impact of local and large-scale changes on Antarctic climate, the linear trends of SLP and sea ice cover for each season are shown in figures 3(e)–(h). Although there were negative SLP trends around Antarctica in all seasons, the SLP pattern strongly resembled the positive SAM phase in summer, which is consistent with Jones et al (2019). Rudeva and Simmonds (2015) and Grieger et al (2018) also highlighted the tight connections between SAM, cyclones, and southward moisture transport. In autumn, negative SLP pattern is similar to negative phase of the Pacific–South American, which induces an east-west seesaw pattern of warm and cold advection over West Antarctica (Marshall and Thompson 2016). The circumpolar westerlies associated with these negative SLP patterns causes warm advection from the South Pacific to the high latitudes, which leads to air temperature warming in all seasons over the Antarctic Peninsula and West Antarctica (figure 3). In addition, sea ice concentrations decreased to the west of the Antarctic Peninsula in all seasons (figures 3(e)–(h)), which would lead to increasing evaporation and heat from the affected ocean (Turner et al 2013) and result in enhanced air temperature and water vapor. Over the southern Weddell Sea, the northerly warm and moist advection associated with negative SLP over the Atlantic sector high latitudes induced the increase in air temperature in spring and summer (figures 3(a), (b), (e) and (f)). Further to the west, deepening of the ASL drives northward cold and dry advection from Antarctica over the Ross Sea in summer and autumn (figures 3(b), (c), (f) and (g)). In addition, over the most of East Antarctica coastal region, the northerly warm and moist advection from Southern Ocean causes air temperature increases (figure 3). Therefore, increases in warm and moist advection from the middle latitudes have a direct effect on air temperature warming and an indirect effect on skin temperature increases in these regions (figures 2(a)–(d) and 3(a)–(d)).

Conversely, over the interior of East Antarctica, the magnitude of air temperature warming was smaller than skin temperature because of relatively weaker warm advection from lower latitudes (figures 2(a)–(d) and 3(a)–(d)). Over the high latitudes, long and shortwave radiation at the surface are strongly influenced by clouds. In fact, the magnitudes of linear trends of downward longwave radiation for clear sky (l_{d_sky})/4σT_s ³ without cloud forcing are smaller than those of l_d/4σT_s ³ for all seasons (figures 2(e)–(h) and S3(a)–(d)), meaning that change in cloud has also impact on downward longwave radiation change, in particular over the interior of East Antarctica. For that higher-elevation region, trends of middle level clouds between 800 and 450 hPa (MCC) are shown for each season from 1950 to 2020 (figures S3(e)–(h)). In all seasons, MCC showed statistically significant positive trends over the interior of East Antarctica, where there are relatively large differences in temperature between air and ground (figures 2(a)–(d), 3(a)–(d) and S3(e)–(h)). These results indicate that enhanced cloud cover increases downward longwave radiation and subsequent skin temperature increases in all seasons except summer over the interior of East Antarctica. Over this region, the weak wind and strong temperature inversion in winter reduces heat exchange between air and snow and influences the temperature difference between the air and surface (Heinemann et al 2019). Although warm and moist advection from lower latitudes is weak, there were significant increases in total column clouds water and moisture advection (figures 3(a)–(d) and S3(i)–(l)). However, cold conditions over interior East Antarctica is very conducive to the formation of clouds, in spite of relatively little moisture advection. Therefore, increase in moisture advection from lower latitude over East Antarctica interior would contribute to increasing clouds in low level.

Our analysis shows the dominant role played by downward longwave radiation in determining the overall trend in Antarctic skin temperature over our 71 year study interval. However, it is unclear whether this finding would hold true if only various intervals within the 71 year study period (hereafter segments) were considered, which would establish a physical connection between these two parameters. The structure of figure 1 suggests that any calculated trend is strongly dependent on the chosen start and end years, echoing similar comments by Nicolas and Bromwich (2014) and Gonzalez and Fortuny (2018). To investigate this, we used two-dimensional linear change diagrams to display temperature trends on a continuum (in steps of one year) of initial and end years. We constructed these diagrams for three key regions, West Antarctica, East Antarctica, and the Antarctic Peninsula (figure S4) as our analyses pointed to distinct behaviors of these subdomains.

Figure 4 shows the two-dimensional spatially averaged skin temperature trend diagram for these regions in all four seasons. In autumn and winter, the warming trends in the 1950s–1960s and early 1980s–2010s were most significant in all three regions with the greatest warming in East Antarctica (figures 4(c), (d), (g), (h), (k) and (l)). Over East Antarctica, significant warming also occurred in winter in the 1960s-early 1970s and early 1980s–2010s (figure 4(d)). East Antarctica exhibited the widest range of segments with significant temperature changes in winter. In spring, although significant warming was observed over West Antarctica in some segments starting before the late 1960s and ending after the early 2000s (figure 4(e)), there were no significant changes over East Antarctica or the Antarctic Peninsula even in longer segments (figures 4(a) and (i)). In summer, only the Antarctic Peninsula exhibited significant trends (figures 4(b), (f) and (j)). Significant warming was found for extended periods in the 1950s–early 1970s and early 1980s–2010s (figure 4(j)). We also found significant cooling of Antarctic Peninsula summer skin temperatures in both long and short segments starting after the 1980s and ending after the early 2010s, in agreement with an absence of warming since 1999 (Turner et al 2016). In addition, for short segments (<20 years) in the 21st century, skin temperature warming trends are not clearly seen over the all regions in all seasons except for in summer over East Antarctica (figure 4). Therefore, there is an absence of skin temperature warming over the entire Antarctic continent Antarctica since early 2000 (figure 1(a)). This corresponds to a period where, for the most part, there has been a progressive increase in Antarctic sea ice extent (Simmonds 2015). We do note, however, that 2020 established a new temperature record for Antarctica (figure 1(a)), a year in which the global average surface temperature tied for the hottest on record (Lenssen et al 2019). For more short segments (<10 years), skin temperature trend amplitudes were larger than long segments in all seasons in all regions, but these results were not significant (figure 4).

Zoom In Zoom Out Reset image size

We similarly investigated trends in l_d/4σT_s ³ (figure S5). It is striking that the magnitude and distribution of l_d/4σT_s ³ are very similar to skin temperature in all areas and seasons, except for spring over East Antarctica (figures 4 and S5). To explore the similarities in more detail, we calculated the ratios of l_d/4σT_s ³ and skin temperature changes (figure 5). When both skin temperature and l_d/4σT_s ³ exhibited significant changes, the ratio exceeded 0.7 in all regions in all seasons. In particular, the ratio exceeded 0.8 in autumn and winter over West Antarctica, with the highest values >1.0 in winter (figures 5(g) and (h)). In addition, high ratio values were found in short segments for the Antarctic Peninsula in summer when there was a negative trend in skin temperature changes (figure 5(j)). These results indicate that spatial and temporal Antarctic skin temperature changes are overwhelmingly explained by changes in longwave radiation.

Zoom In Zoom Out Reset image size