PUBLICATIONS. Journal of Geophysical Research: Atmospheres

Transcription

1 PUBLICATIONS Journal of Geophysical Research: Atmospheres RESEARCH ARTICLE This article is a companion to Bintanja et al. [2014] doi: /2013jd Key Points: Changes in Antarctica s surface winds are simulated with a climate model Simulated future katabatic winds over Antarctica will become weaker Changes in Antarctica s winds governed by the synoptic pressure gradients Correspondence to: R. Bintanja, bintanja@knmi.nl Citation: Bintanja, R., C. Severijns, R. Haarsma, and W. Hazeleger (2014), The future of Antarctica s surface winds simulated by a high-resolution global climate model. II. Drivers of 21st century changes, J. Geophys. Res. Atmos., 119, , doi:. Received 6 SEP 2013 Accepted 27 MAY 2014 Accepted article online 31 MAY 2014 Published online 19 JUN 2014 The future of Antarctica s surface winds simulated by a high-resolution global climate model: 2. Drivers of 21st century changes R. Bintanja 1, C. Severijns 1, R. Haarsma 1, and W. Hazeleger 1 1 Royal Netherlands Meteorological Institute (KNMI), De Bilt, Netherlands Abstract Antarctica s katabatic winds are among the strongest near-surface winds on Earth, and among the most consistent ones. As these winds are primarily due to the strong surface cooling, greenhouse warming of the surface may act to reduce the strength of these winds as well as their consistency. Here we use the atmospheric component of the global climate model EC-Earth in prescribed sea surface temperature (SST) simulations of the present day ( ) and future ( ) climates, using two model resolutions: (1) T159L62 (~100 km, 62 vertical levels), and (2) T799L91 (~20 km, 91 vertical levels) to investigate changes in Antarctica s surface winds and the reasons thereof. Circumpolar westerlies over the Southern Ocean strengthen and shift poleward because of the deepening of the circumpolar trough and the associated increase in Southern Annular Mode (SAM), especially in high resolution, causing weaker coastal easterlies. Generally, surface wind speeds over the Antarctica mainland exhibit a small decrease. According to the simulations, the temperature deficit (or inversion strength) and associated katabatic forcing exhibit only minor changes over the continent. Changes in the surface winds over Antarctica s slopes can thus be attributed mainly to changes in the synoptic forcing (large-scale pressure gradient). Hence, with modeled 21st century changes in the katabatic forcing being small, changes in zonal and meridional surface winds in and around Antarctica are largely decoupled from those over the Southern Ocean and are governed by changes in synoptic forcing and large-scale pressure gradients. As a result, these changes are largely independent on model resolution. 1. Introduction Antarctica s surface winds are among the strongest and most persistent winds on Earth. The main driver of these surface winds is the katabatic forcing, which is governed by cooling over a sloping surface. Another major contributor to Antarctica s surface winds is the synoptic pressure gradient, especially in the coastal regions. The downslope outflow by the katabatic winds is compensated for by an inflow higher up [e.g., Fortuin,1992;van Lipzig,1999;Parish and Bromwich, 2007], which closes a meridional overturning circulation that to a large extent determines the poleward transport of moisture toward the Antarctic continent and thereby modulates precipitation trends over Antarctica. It is therefore of great interest to assess possible future changes in this circulation. Another potentially important aspect governed by the surface wind strength is the surface mass balance. It is affected by drifting and blowing snow, or snowdrift, which can spatially redistribute the precipitated snow and induce sublimation of the airborne snow particles [Bintanja, 1998, 2001; Lenaerts et al., 2012] and by surface sublimation [Bintanja and van den Broeke, 1995; Bintanja et al., 1997]. To estimate future changes in surface mass balance of Antarctica, it is therefore vital to know the (reasons for) changes in Antarctica s surface wind field. However, while possible future changes in Antarctica s circumpolar winds have been studied extensively [e.g., Lynch et al., 2006; Yin, 2005; Fyfe and Saenko, 2006], future changes in the surface wind speed over the Antarctic continent and their possible dynamic connection with the projected increases in circumpolar winds have received relatively little attention. Analyzing a then high-resolution (low resolution in current standards) climate model, van den Broeke et al. [1997] found that future summertime surface winds over Antarctica will decrease in strength because surface warming will destroy the inversion and hence reduce the katabatic forcing. In winter, however, near-surface winds were found to increase as a result of stronger meridional pressure gradients associated with a deepening of the circumpolar trough. Annually averaged, the changes in surface wind speed were found to be small, with slight increases over the plateau regions and a slight decrease in the steeper escarpment regions. BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7160

2 Bracegirdle et al. [2008] used the Coupled Model Intercomparison Project phase 3 (CMIP3) multimodel ensemble to assess, among others, 21st century changes in the wind field over and around Antarctica. They found a strong and consistent (among various climate models) increase in circumpolar westerlies, but only minor changes in wind speed over the Antarctic continent. They explained the former in terms of the meridional temperature gradient and increases in the Southern Annular Mode (SAM) and their respective seasonal cycles, but left the latter largely unexplained. Hence, both studies addressing possible future wind changes over Antarctica do not present a clear answer as to why surface wind changes over the continent contrast so sharply with those over the adjacent circumpolar region. An important issue concerns the simulated changes (magnitude, position) in the circumpolar trough [e.g., Lynch et al., 2006; Yin, 2005; Fyfe and Saenko, 2006] and their potential impact on Antarctica s surface winds. Positive SAM trends are generally associated with cooler conditions over Antarctica, through a more efficient thermal isolation of the continent [Turner and Overland, 2009], so any increase in SAM will (partially) offset direct greenhouse warming of the surface. As such, a deepening of the circumpolar trough may potentially lead to enhanced katabatic forcing and fiercer surface winds (or, equivalently, to a smaller decrease). On the other hand, any increase in SAM will also affect horizontal large-scale (synoptic) pressure gradients over the continent, which will in turn also affect the surface winds. Hence, the interplay between dynamical changes in the circumpolar trough (which, in turn, are connected to stratospheric changes, see Bintanja et al. [2014]) and those over the Antarctic mainland is potentially complex and hard to predict a priori. A number of studies have found that high-resolution models simulate the current state of Antarctica s surface wind climate more accurately than reanalyses output based on relatively coarse resolution model data [e.g., Lenaerts et al., 2012]. Knowing this, why would one expect the response of the katabatic wind system over Antarctica to climatic changes to be resolution dependent? The main reason for this is that because katabatic forcing is proportional to the surface slope, and since the representation of the orography depends on horizontal resolution, the resulting katabatic winds might depend on the accuracy by which orography is implemented. We will here quantify the effect of resolution by applying two sets of global model simulations using a low (actually currently standard) resolution (T159L61) and a very high resolution (T799L91), which differ not only horizontally but also vertically (albeit only in the stratosphere), see Bintanja et al. [2014] for details. This may cause the Antarctic stratospheric vortex and stratosphere-troposphere interactions to differ between the two resolutions. Since resolution is the only difference between the two sets of simulations, this setup guarantees that any differences in simulated quantities and climate sensitivities will only be due to differences in resolution. We will address 21st century changes in Antarctica s katabatic winds in two related papers. In the accompanying paper [Bintanja et al., 2014], we study the ability of the model to simulate the present-day state by comparing in detail model output to direct and independent observations, both in standard and high resolution. In this paper, we will examine the 21st century climate response of Antarctica s katabatic winds using the climate model in standard and high-resolution mode and infer reasons for the respective changes in terms of the force budget. Geographical names over Antarctica and the location of the transect used later are depicted in Figure Methods 2.1. Model Specifics We use the global Earth system model EC-Earth, which is based on the European Centre for Medium range Weather Forecast (ECMWF) Integrated Forecast System (IFS), for the atmosphere and land part. EC-Earth is a fully coupled model, but here we use the atmosphere-only version in which we prescribe sea surface temperature (SST) and sea ice. We use EC-Earth version 2.2, which employs IFS cycle 31R1 [Hazeleger et al., 2010; Hazeleger and Bintanja, 2012]. EC-Earth was not specifically tuned with respect to its performance in the polar regions. The great advantage of a global model over a regional model is that it is not constrained by boundary conditions, which inevitably affect the mean state and the response of the model. To investigate the effect of model resolution on the ability of the model to accurately simulate katabatic wind over Antarctica in a global model, we employ two resolutions: T159L62 (low resolution (LR)) and T799L91 (high resolution (HR)), see Bintanja et al. [2014]. These resolutions are equivalent to about 125 and 25 km in the horizontal, respectively, and to 61 (top model layer 10 hpa) and 91 (top model layer 0.2 hpa) vertical layers (the additional layers in HR are in the stratosphere, extending the top layer of the model). These high-resolution simulations were carried out in the framework of the Future Weather project, the background BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7161

3 Figure 1. Map of the Antarctic continent with geographical names used in the text. The green line represents the transect along 135 E. of which is that in addition to (broad) climate changes, high-resolution weather changes (and extremes) arguably have a considerable impact on society. Localized weather phenomena with high-impact potential are better captured in high-resolution climate simulations, the obvious example being tropical hurricanes [Haarsma et al., 2013]. The same holds for orography-driven boundary layer katabatic winds, since the accuracy by which these are simulated may depend on the applied orography (which one would expect to be more accurate in high resolution). The only applied difference between the LR and HR simulations is the vertical and horizontal resolution (which will automatically introduce differences in certain resolutiondependent parameters in IFS that were used to optimize the model for each resolution). The model EC-Earth has taken part in the CMIP5 initiative [Taylor et al., 2012], which includes both coupled and uncoupled (i.e., prescribed SST) simulations. Both the uncoupled and coupled versions of EC-Earth have been subject to rigorous testing [Hazeleger et al., 2010], with global performance indices showing that the model generally simulates the current climate quite well. The main biases include a slightly too cold (upper) atmosphere, especially in the tropics, and too warm winters over the continental Northern hemispheric regions. The coupled model also exhibits a warm bias in the Southern Ocean, which is thought to result from to ocean physics [Sterl et al., 2012] and shortwave radiation inaccuracies, but this is not an issue in the uncoupled model used here Simulations and Forcings Using both the LR and HR resolutions in EC-Earth, we performed experiments for the present ( ) and end 21st century ( ) 5 year periods. For each of these two time frames, six members were simulated (each with the same 5 year SST forcing, but with slightly different initial atmospheric conditions, so that interannual variability is retained); hence, in total, we have 30 years of simulation for both the present and the future climates. Each simulation is preceded by a spin-up of a few months, the length of which was varied to obtain the various realizations by using a different start-date. Throughout the paper, we define the future-present differences as significant if they are larger than the sum of the standard deviations of the simulated interannual variability (based on the 30 years of simulation) of the present and the future climate. BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7162

4 Figure 2. Difference in (a) sea surface temperature (SST) (K) and (b) sea ice cover, between the periods and [Sterl et al., 2008]. The present-day SST and sea ice fields for were taken from NOAA [2012] (0.25 resolution) and interpolated to the required resolutions. SSTs for the future climate are not known, obviously. Differences between future and present-day SST fields were determined using the low-resolution (~1 ) ESSENCE ensemble [Sterl et al., 2008] 21st century simulations based on the SRES A1B scenario, as the average over 17 ensemble members. These (low-resolution) SST differences were subsequently added to the present-day SSTs to obtain the future SST fields (future sea ice cover was evaluated using an empirical relation between SST and sea ice cover from the present-day state; so we determined what SST values correspond to a certain sea ice cover in the present day and applied that relation to the future SST field to obtain the future sea ice cover). The HR future SSTs are thus a combination of the high-resolution present-day field and the lowresolution future-present difference based on the ESSENCE simulations. The forcing frequency of SST and sea ice cover is 6 h. Figure 2 shows the future-present difference in SSTs and sea ice cover. Evidently, the Southern Ocean exhibits only minor warming, except in areas where sea ice has retreated, where substantial surface BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7163

5 Figure 3. Twenty-first century annual mean changes in surface air temperature (2 m, in K) ( average minus average) for (a) low resolution (LR) and (b) high resolution (HR). Areas where the simulated changes are not significant with respect to the interannual variability (see text) are stippled. 3. Results warming is occurring. Greenhouse gases and other forcing agents (including stratospheric ozone), both for the present and future climates, were prescribed according to CMIP5 (the historical and RCP4.5 scenarios) [Taylor et al., 2012]. Hence, the future-present changes that will be reported on in this paper are due to (1) differences in SST and sea ice, and (2) changed greenhouse gases and other atmospheric forcing agents. We should note, however, that the SST fields based on the SRES A1B forcing are a somewhat larger (~700 ppm CO 2 in 2100) forcing than the RCP4.5 forcing used to prescribe changes in greenhouse gases (~550 ppm CO 2 in 2100). Preferably, we would have also used SST fields based on the RCP4.5 forcing, but these were not available at the time of doing the high-resolution simulations (which involved a project deadline). However, given the very large intermodel spread in the temperature responses to a particular forcing, especially in the Southern Ocean, our combination of SST and greenhouse gas forcingiswellwithintherealmof feasible forcings. Therefore, we believe that this combination of somewhat different forcings provides a realistic scenario of future changes. Nonetheless, in interpreting the results, one has to keep in mind that the applied SSTs are based on a somewhat larger climate forcing than the greenhouse gas forcing Geographical Changes in Surface Wind and Related Variables Here we focus on the 21st century changes in all relevant meteorological variables. Figure 3 shows the simulated distribution of annual mean change in surface air temperature, with annual mean changes indicating the mean over the 30 years of output available for each time slice. Maximum temperature changes are simulated in the belt immediately adjacent the Antarctic coastline and are related mainly to the (prescribed) changes in sea ice cover [Connolley and Bracegirdle, 2007]. Due to the relatively small changes in sea ice cover near the Adélie Land coastline, temperature changes are quite modest there, indicating that there is considerable zonal variation in the temperature response. Over Antarctica itself, temperatures show a moderate increase, which is significant almost everywhere except for some parts of the escarpment region in East Antarctica. Evidently, there is only a marginal effect of the strong temperature increases off-coast on continental temperatures, which may be attributed mainly to the katabatic winds blowing offshore, limiting the thermal anomaly to travel upstream. Most likely, the relatively modest increase in temperatures over the continent can be attributed to direct greenhouse forcing, possibly amplified slightly by a reduction of the surface inversion through which relatively warm air from the free atmosphere aloft can more easily mix down to BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7164

6 the surface [e.g., van den Broeke et al., 1997]. Also, the fact that the greenhouse forcing is based on a comparatively weak scenario (RCP4.5) as compared to the SST forcing (SRES A1B), as discussed above, might have exaggerated this difference to some extent. The interannual variability of the temperature changes increase going inland and peak over both large ice shelves, both in LR and HR. The interannual variability is relatively small over the regions with greatest warming. The relatively weak increase in surface temperatures over Antarctica (compared to those offshore) is in line with results from CMIP3 models [Bracegirdle et al., 2008], although a few models do show temperature amplification toward the pole. Note that our model as well as all current CMIP5 models does not take into account possible negative feedbacks related to melting ice shelves and sea ice increases [Bintanja et al., 2013], so the prescribed end of century sea ice change may well be an overestimation. Finally, despite the present-day climate being warmer in HR compared to LR [Bintanja et al., 2014], the differences in simulated 21st century surface warming between LR and HR are generally quite small. Hence, according to our climate model simulations, Figure 4. Same as Figure 3 but for sea level pressure (hpa). climate warming over Antarctica is hardly influenced by model resolution. Sea level pressure (SLP) drops over Antarctica, whereas it increases north of S (Figure 4), resulting in a stronger meridional pressure gradient. The strongest reduction in Antarctic SLP occurs in HR, and the peak drop seems to occur over the Ross and Amundsen Seas. Associated with this stronger large-scale meridional pressure gradient is a marked increase in the SAM. Annually averaged, the SAM (defined here as the zonal mean pressure difference between 40 S and 65 S) increases by 15% and 20% in LR and HR, respectively, with a tendency to peak in austral autumn and winter. Aside from the strength of the changes, the pattern of LR and HR is fairly similar. The modeled changes are significant in much of the escarpment region of East Antarctica, where the largest reductions in SLP occur. The increased meridional pressure gradient and the SAM induce stronger circumpolar westerlies, as shown in Figure 5. Increases of up to 3 m s 1 are simulated for both LR and HR, even though the increase is slightly stronger in HR, consistent with the stronger increase in synoptic pressure gradient [Lynch et al., 2006; Fyfe and Saenko, 2006]; the changes are significant in substantial parts of the zone of strong westerlies. Interestingly, there seems to be a strong zonal variation in wind speed increases, with the maximum increases occurring over the regions near smallest temperature increases (Figure 3). This may be associated with the fact that near the centers of greatest subpolar warming, the meridional temperature gradient north of that warming is actually reduced. This subdues the synoptic activity and baroclinicity, and hence the increase in surface winds diminishes. BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7165

7 Figure 5. Same as Figure 3 but for 10 m wind speed (m s 1 ). Just offshore, simulated wind speeds generally exhibit a sharp decline. Not only have the westerlies become stronger, the synoptic pressure gradient forcing these westerlies will shift poleward [e.g., Yin, 2005; Bracegirdle et al., 2008]. As a result of the reduced easterly synoptic forcing, the model results show that the predominant easterlies just offshore East Antarctica will strongly decrease. Over the Antarctic mainland, the surface wind speed exhibits a small decrease except over the high dome regions and some isolated coastal regions, which largely agrees with modeling results of van den Broeke et al. [1997]. Hence, zonal variations in offshore surface warming (Figure 3) appear to have little effect on the wind speed changes over the continent. The geographical pattern of the wind speed change is very similar for LR and HR, including the magnitude of the changes over the continent, suggesting that mechanisms that are potentially dependent on resolution (e.g., katabatic forcing and surface slope) play a minor role. Wind speed changes over Antarctica s mainland are largely insignificant compared to the interannual variability. Reasons for the simulated wind speed changes will be given below in terms of the (changes in) near-surface force budget. A possible indicator of the governing mechanisms behind the simulated 21st century wind speed changes is the wind directional constancy (DC), defined as the ratio of the mean to vector mean wind [Bintanja et al., 2014]. Figure 6 shows the change in annual mean DC for the LR and HR cases. Evidently, the geographical pattern largely mimics that of the change in surface wind speed, with a decrease in DC occurring virtually everywhere over the continent, meaning that Antarctica s surface winds generally become more variable and less consistent. Since the magnitude of the katabatic forcing (relative to other, more variable, forcings) is the main cause of the consistency and constancy of Antarctica s surface winds, this (slight) decrease in DC might be interpreted such that, generally, these winds will become somewhat less katabatic in nature, at least according to our simulations. Changes are small and not significant, thought, over the entire continent. Table 1 lists 21st century changes in the various variables over East Antarctica, confirming that modeled changes in surface wind and related quantities over the continent are minor but that, on average, the direction of the changes points to weaker winds that are less katabatic in nature Vertical Changes in Zonal Mean Temperature and Wind Speed Generally, surface wind changes are connected to climate changes higher up in the atmosphere. The change in zonal mean temperature is given in Figure 7. It shows reduced warming and even cooling in the polar upper troposphere and stratosphere, and warming in the upper tropospheric subtropics. Simulated changes are significant everywhere in HR, but in LR only in the area of greatest warming, that is, the upper BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7166

8 tropospheric subtropics. The increased meridional temperature gradient results in a strongly increasing upper-tropospheric jet (Figure 8) between 50 S and 60 S, which extends downward to the surface in conjunction with the increased meridional pressure gradient and the SAM [Turner and Overland, 2009], although the changes are largely insignificant (relative to the interannual variability) because of the strong interannual variability that is partly associated with the SAM. Also evident is the poleward shift of the jet, with increases in westerly winds peaking at the poleward side of the jet. This strong increase and poleward displacement of the southern hemispheric westerly jet are associated with continued greenhouse forcing, which is partially offset by the projected 21st century recovery of the Antarctic ozone hole [e.g., Shindell and Schmidt, 2004; Fyfe and Saenko, 2006]. Generally, the future strength and position of the jet depend on the rate of increase in greenhouse forcing as well as on the speed of stratospheric ozone recovery [Shindell and Schmidt, 2004], and are thus dependent on the precise scenario that is applied. Evidently, in the CMIP5 RCP4.5 forcing scenario used here, the diminishing effect of ozone recovery is insufficient Figure 6. Same as Figure 3 but for wind directional constancy (DC). to compensate for the increase due to greenhouse gas emissions, the net result being a strengthening westerly jet that is shifted poleward [Yin, 2005] as seen in most climate models. In some models, however, the greenhouse and ozone effects largely cancel each other [Shindell and Schmidt, 2004]. Note also that we use here a combination of scenarios, the SRES A1B for SSTs and RCP4.5 for greenhouse gases; this might affect the response such that changes in the jet might be slightly more influenced by changes in the meridional temperature gradient as compared to the effects of ozone changes. The temperature changes in the Antarctic lower to middle troposphere are generally mild and exhibit little difference between LR and HR. The same goes for the zonal winds, which over Antarctica show a small decrease throughout the troposphere, despite the differences in circumpolar jet changes between LR and HR. Hence, the dynamics associated with wind speed changes in the circumpolar jet region apparently have little influence on Table 1. Annual Averages Over East Antarctica (All Land Points Between 45 W and 150 E) of the Simulated Changes (Future Minus Present) in Various Quantities a Variable Unit Low Resolution (LR) High resolution (HR) Temperature at 2 m K Wind speed at 10 m m s Directional constancy Mean sea level pressure hpa Temperature deficit K Geostrophic wind speed m s Synoptic forcing 10 4 ms Katabatic forcing 10 4 ms a See section 3.3 for the definitions of temperature deficit, geostrophic wind speed, synoptic forcing, and katabatic forcing. BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7167

9 the wind climate over the Antarctic continent, which therefore is largely decoupled from the strong increases in wind occurring over the Southern Ocean that are governed by greenhouse and stratospheric ozone forcing. It can be concluded that future changes in surface winds over Antarctica are relatively insensitive to whatever greenhouse and ozone scenario is used as forcing, at least according to these model simulations. Figure 7. Twenty-first century annual and zonal mean change in air temperature (K) ( average minus average) for (a) LR and (b) HR. Areas where the simulated changes are not significant with respect to the interannual variability (see text) are stippled Changes in the Force Budget Along a Transect in East Antarctica (135 E) Since simulated changes in wind and associated variables over the Antarctic mainland are relatively small, we present here the changes in terms of the force budget along a transect in East Antarctica (denoted by the green line in Figure 1) because that will give a clearer (less noisy) insight into the processes behind the changes, while still being representative foratleasttherestofeastantarctica.we have chosen 135 E because of the relatively smooth elevation profile largely characteristic for the East Antarctic Ice Sheet, consisting of a relatively flat plateau in the interior and a steep escarpment region near the coast, and also because of the relatively undisturbed flow with a large catchment area in this region. Figure 9 shows the relevant meteorological surface variables along this transect at 135 E. Surface air temperature (Figure 9a) exhibits little change in the circumpolar westerly wind region. In the sea ice decline region, temperature change peaks, and after a sharp decline near the coast the change slowly increases toward the South Pole. The inland temperature differences between LR and HR, possibly affected by differential greenhouse forcing and thermal isolation because of increases in the SAM and more zonal flow [Turner and Overland, 2009], are small. The 10 m wind speed (Figure 9b) exhibits a notable increase in the westerly wind region as discussed before, and a decrease in the near-coastal easterlies. This is attributed to the poleward shift of the southern hemispheric storm tracks (see Figure 8) associated with the increase in the mid-tropospheric temperature gradient [Bracegirdle et al., 2008], which occurs in both LR and HR. Over the continent, wind speed changes are relatively small (roughly 5 10%) and tend to be negative. Except for the westerly wind region, where HR wind speed increase is larger than in LR because of the larger increase in SAM, wind speed changes along this transect exhibit little difference between LR and HR. The wind directional constancy (Figure 9c) largely follows the wind speed pattern with relatively high values over the circumpolar westerlies, with low values in the region with weak and variable winds just off-shore, and high values up to 0.9 and higher over the continent where steep slopes enable a high directional constancy by virtue of mainly the katabatic forcing. The increase in circumpolar westerlies is accompanied by a reduction in their variability, whereas the variability increases in the area of the diminishing coastal easterlies. Over the continent, changes in DC are slightly negative, suggesting a weak increase in variability. Figure 9d shows the associated profile of Weibull shape factor changes. The wind velocity distribution can generally be fitted to a BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7168

10 Journal of Geophysical Research: Atmospheres 1 Figure 8. Twenty-first century annual and zonal mean changes in zonal wind speed (colors, in m s, the light blue line denotes zero change) ( average minus average) for (a) LR and (b) HR. The black lines show the pre 1 sent-day zonal winds (m s ), with the black bold line denoting zero zonal wind. Areas where the simulated changes are not significant with respect to the interannual variability (see text) are stippled. Weibull probability density function, as explained by Bintanja et al. [2014], of which the Weibull shape factor (kw) characterizes the skewed wind speed distribution. Generally, our simulations indicate that kw decreases in magnitude, although the changes are not significant. Since a lower kw indicates more variable winds, suggesting a weakened katabatic regime [Sanz Rodrigo et al., 2012], the general conclusion may be that future surface winds over Antarctica will become less katabatic in nature. Changes between LR and HR are generally minor. To assess the reasons for these and other changes along this transect, we consider here the force budget for a flow in the zonal (x, west-east) and meridional (y, south-north) direction over a sloping surface [van den Broeke and van Lipzig, 2003]: ADVH ADVV THW COR SYN FRIC KAT g θ uw g ¼ þv þw þ þ f V f Vg þ Λθ sin αx t x y z θ0 x z θ0 V V V V g θ vw g ¼ þv þw þ þ f þ f g þ Λθ sin αy t x y z θ0 y z θ0 (1) (2) Here, U, V, and W are the zonal, meridional, and vertical wind components (positive eastward, northward, upward), θ the potential temperature, g the acceleration of gravity, Λθ the temperature deficit, and α the surface slope. We refer to van den Broeke and van Lipzig [2003] for a detailed explanation of all terms and variables. For the purpose of this study, we focus on the synoptic forcing (SYN) and the katabatic forcing (KAT), as these are the forcing agents most likely to change in a future climate. Obviously, the other terms will in principle also change as climate evolves, but mostly in response to the synoptic (large-scale pressure BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7169

11 Figure 9. Variation along the 135 E transect of present-day and future states (solid and stippled lines, respectively; upper panels) and the 21st century changes (bottom panels) in (a) surface air temperature, (b) 10 m wind speed, (c) wind directional constancy (DC), and (d) Weibull shape factor k w. In the bottom panels, the light blue (LR) and light red (HR) horizontal bars denote where along the transect the 21st century changes are significant with respect to the interannual variability (see text). gradient) and katabatic forcings. We will evaluate the SYN and KAT forcings at the surface (in the zonal and meridional directions, denoted by subscripts u and v, respectively) in order to assess how changes in surface forcing agents contribute to changes in the surface winds over Antarctica. The horizontal and vertical advection terms in (1) and (2) (ADVH and ADVV, respectively) are usually small over Antarctica [e.g., Parish and Waight, 1987]. COR denotes the Coriolis force acting on the flow, while FRIC is the vertical divergence of turbulent momentum flux, also commonly referred to as the friction term. THW is BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7170

12 Figure 10. Geographical distribution of the present-day mean ( average) ageostrophic meridional wind speed at the surface (actual meridional wind minus geostrophic meridional wind) in LR (m s 1 ). the thermal wind component (pressure gradient force due to horizontal gradients in temperature). van den Broeke and van Lipzig [2003] analyzed the present-day magnitude of the THW terms over Antarctica and found these to be small relative to the other terms. Based on this, we may expect climaterelated changes in THW to be even smaller, especially since the changes in the temperature gradients over the continent are minute (Figure 3). The zonal and meridional synoptic surface forcings fv g and fu g (equal to dz/dy and dz/dx, respectively, with Z being the geopotential height) depend on the Coriolis parameter f and the zonal and meridional geostrophic wind components (U g and V g, respectively) at the surface. However, the geostrophic wind at the surface is not known. We approximate the surface geostrophic wind components by linearly extrapolating the geostrophic wind profile in the free troposphere (sufficiently far away from the boundary layer) down to the surface. In sharply varying topography (hence especially in HR), this procedure will lead to strong spatial variations in surface geostrophic winds, which actually represent nothing else than vertical changes in geostrophic winds along steep topography. Having estimated the surface geostrophic winds over Antarctica, we can, as an example, evaluate the nongeostrophic wind component. Normally, surface winds are weaker than the geostrophic wind due to vertical momentum flux in the boundary layer associated with surface friction, but over the steep slopes in the escarpment region of Antarctica winds are actually stronger that the geostrophic winds (Figure 10) because of the katabatic forcing acting over sloping terrain accelerating the near-surface flow beyond their synoptic (geostrophic) magnitude. This confirms analyses from regional meteorological observations in Droning Maud Land [Bintanja, 2000a]. Figure 11. Schematic illustration of the vertical potential temperature profile and the value of the temperature deficit Δθ (see text). The dashed line represents the downward extrapolated potential temperature profile. The katabatic forcing (KAT) in (1) and (2) depends on the acceleration of gravity (g), the undisturbed background potential temperature (θ 0 ), the temperature deficit (Λ θ )defined as the difference between actual near-surface temperature and θ 0 (as schematically outlined in Figure 11), and the sine of the surface slope (sinα). θ 0 and Δθ are evaluated by linearly extrapolating the vertical potential temperature profile in the free troposphere (i.e., far away from the surface) down to the surface [Bintanja, 2000b]. In this paper, Δθ is defined positive when the actual surface air temperature is lower than θ 0. The most important aspect of KAT is its linear dependency on the temperature deficit and the surface slope; KAT peaks near the surface and decreases with height because the temperature deficit is largest close to the surface [Bintanja, 2000b]. Figure 12 shows the geographical distribution BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7171

13 Journal of Geophysical Research: Atmospheres of the simulated changes in temperature deficit (or inversion strength). Changes are generally slightly positive in the escarpment regions, and negative elsewhere, but not significant everywhere. This is a somewhat surprising result given that greenhouse forcing would in principle lead to warming that peaks at the surface, and thus to a decrease in inversion strength. Apparently, the reduced mixing because of the overall weaker surface winds opposes this direct forcing, leading to small changes in inversion strength. Figure 13a shows the temperature deficit and its simulated 21st century changes along the 135 E transect. Annually averaged, Δθ values of K are found over the Antarctic interior (both in LR and HR), with winter values peaking over 30 K (not shown), giving rise to strong katabatic winds over its sloping surfaces (Figure 13b shows the surface elevation profile along the transect). Future changes in the modeled temperature deficit, however, are small (within ± 0.5 K) and overall not significant over the Antarctic continent, both in LR and HR. As a next step, we will separately discuss the zonal and meridional force budget terms SYN and KAT and their impact on the respective surface winds. The zonal force budget terms and winds and their 21st century changes along the transect are depicted in Figure 14. Figure 12. Same as Figure 3 but for the temperature deficit or inversion SYNU forces westerly winds over the strength (K). circumpolar Southern Ocean, easterlies around the coast of Antarctica, westerlies further inland, and easterlies close to the South Pole. KATU is negative almost everywhere along the transect over the continent (and zero over the ocean owing to the absence of a sloping surface), indicating that the katabatic forcing drives easterly surface winds. Actual surface winds seem to be determined mainly by the synoptic forcing, which can be attributed to the fact that KATU is weak (note that KATU depends on the surface slope perpendicular to the transect, which is generally smaller than the meridional (pole to coast) slope depicted in Figure 13b). The differences between LR and HR are generally small. It is important to stress here that in this paper we focus on the mean (annual) forcings; Bintanja [2000a] analyzed detailed summertime meteorological observations in Dronning Maud Land and showed that the relative impacts of the synoptic and katabatic forcings have a pronounced diurnal cycle, with katabatic forcing dominating during the night and the synoptic forcing in daytime. Future changes in zonal winds over Antarctica along this transect are clearly governed by changes in SYNU. Zonal synoptic forcing increases over the ocean whereas it decreases over the continent, forcing stronger westerlies (or reduced easterlies) over the ocean, and strengthened easterlies (or reduced westerlies) over the continent. The changes over the ocean are mostly significant, but over the continent only in some locations in HR. The changes in zonal katabatic forcing are small and not significant, owing mainly to the small changes in temperature deficit. As a result, changes in zonal surface winds closely follow those in SYNU, indicating that changes in the large-scale synoptic pressure gradient dominate zonal surface wind changes over and around Antarctica (see also Figure 15); the two regimes indicate the increase in westerlies over the Southern Ocean as opposed to the stronger easterlies over the continent. Generally, the changes are very similar for LR and HR, with a somewhat stronger increase in circumpolar westerlies in HR as discussed before. BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7172

14 Figure 13. (a) Variation along the 135 E transect of present-day and future states (solid and stippled lines, respectively; upper panel) and the 21st century changes (bottom panel) in temperature deficit (Δθ), (b) surface elevation along the 135 E transect. In the bottom panel of a, the light blue (LR) and light red (HR) horizontal bars denote where along the transect the 21st century changes are significant with respect to the interannual variability (see text). BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7173

15 Figure 14. Variation along the 135 E transect of present-day and future states (solid and stippled lines, respectively; upper panels) and the 21st century changes (bottom panels) in (a) zonal surface synoptic forcing, (b) zonal surface katabatic forcing, and (c) zonal 10 m wind speed. In the bottom panels, the light blue (LR) and light red (HR) horizontal bars denote where along the transect the 21st century changes are significant with respect to the interannual variability (see text). The meridional force budget terms and surface wind along the transect are depicted in Figure 16. Over the continent, the large-scale pressure gradient forces downslope (northerly) winds associated with positive values of SYN V, peaking between 80 and 85 S, which is presumably associated with the persistent low pressure area over the Ross Sea region and possibly the proximity of the Transantarctic Mountains. The sign of KAT V closely follows the slope (Figure 13b) as per its definition, and peaks in the escarpment region between 65 S and 70 S where along the 135 E transect we find the steepest slope. As a result of this combined BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7174

16 meridional synoptic and katabatic forcing, the northerly meridional wind speed exhibits a double maximum, the inland one caused by the synoptic pressure gradient and the near-coastal one by the katabatic forcing. Figure 15. Zonal wind speed as a function of zonal synoptic forcing along the 135 E transect. The solid lines represent the best linear fits to the data, whose values are given in the lower right corner. There is a slight tendency for the meridional synoptic forcing to decrease over the continent, which would incur weaker northerly winds. The change in katabatic forcing is generally small over the continent (due to the slight changes in temperature deficit, Figure 13a), except over the steep escarpment region where it exhibits a moderate decrease. All taken together, simulated meridional winds over the continents tend to decrease, but the changes are small and not significant virtually everywhere, both in LR and HR. Average SYN and KAT values over East Antarctica are listed in Table 1 and confirm that continent-average changes in the force budget are small. Results indicate that changes in synoptic forcing over the Southern Ocean are largest in the non-summer part of the year (not shown), which is associated with higher SAM increases in those seasons [e.g., Bracegirdle et al., 2008]. However, this seasonal signature rapidly declines going south; over the Antarctic continent, there are no appreciable seasonally varying changes in simulated force budget terms and climate variables. 4. Conclusions and Discussion According to our climate model simulations, future warming causes a strengthening and southward migration of the southern hemispheric circumpolar westerlies associated with an increasing trend in the SAM, which is larger in HR than in LR. Roff et al. [2011] show that increased stratospheric resolution leads to much improved forecast skill in the high Southern latitude troposphere, with possible climatic implications for the SAM, which in turn impacts Antarctic temperatures [Thompson and Solomon, 2002]. Generally, the HR stratosphere may have a different impact on the troposphere compared to the LR stratosphere, presumably through a better representation of vertical stratospheric waves, the associated stratosphere-troposphere exchange, and sudden stratospheric warming events [Scaife and Knight, 2008]. The stronger increase in westerly circumpolar vortex strength in HR as compared to LR (Figure 8) implies a larger increase in SAM index in HR, and also, in theory, a slightly more moderate increase in Antarctic temperatures because the continent would then be more thermally isolated from lower latitudes [Turner and Overland, 2009]. In contrast to the LR-HR differences in circumpolar winds, simulated surface winds over the Antarctic continent exhibit only small and largely insignificant changes over the 21st century. The simulated small changes in Antarctica s surface wind speed are consistent with earlier modeling results [van den Broeke et al., 1997; Bracegirdle et al., 2008]. The slight reduction in surface winds over Antarctica can be attributed to a decrease in large-scale synoptic forcing, with katabatic forcing changes being small. Clearly, the significant 21st century changes in the tropospheric winds in the circumpolar region do not or hardly penetrate inland, so wind speed changes over Antarctica, if any, will mostly have a local cause. The only viable candidate for this is the katabatic forcing. The katabatic forcing depends primarily on the temperature deficit at the surface, which is found to change only slightly both in LR and HR. Why are the changes in simulated temperature deficit so small, given the BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7175

17 Figure 16. Variation along the 135 E transect of present-day and future states (solid and stippled lines, respectively; upper panels) and the 21st century changes (bottom panels) in (a) meridional surface synoptic forcing, (b) meridional surface katabatic forcing, and (c) meridional 10 m wind speed. In the bottom panels, the light blue (LR) and light red (HR) horizontal bars denote where along the transect the 21st century changes are significant with respect to the interannual variability (see text). warming of the surface? As explained by van den Broeke et al. [1997], intuitively one would expect the surface to warm under greenhouse warming because the surface would lose less heat through infrared radiation. The cold katabatic layer is influenced heavily by direct radiative forcing, with the strongly stable stratification (especially in winter) opposing vertical mixing. Hence, the direct greenhouse forcing and the reduced effectiveness of surface longwave radiation loss tend to warm the katabatic layer [van Lipzig, 1999]. BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7176

18 As a result of this surface warming, which would be greatest near the surface, the temperature deficit would decrease, and hence the katabatic forcing would be smaller and katabatically dominated surface winds would weaken, but our results show this to be a small effect. In principle, warming of the surface also leads to a reduction in static stability, and, in turn, to an increase in vertical momentum transfer. Changes in the geostrophic winds affect surface winds due to the reduced boundary layer stability and associated enhanced vertical mixing, but this effect depends on the magnitude and direction of the geostrophic winds (synoptic forcing) in relation to the katabatic forcing [van den Broeke and van Lipzig, 2003]. In other words, the surface flow would become less decoupled from that in the free troposphere aloft. In regions where the katabatic forcing is aligned with the (geostrophic) flow above the boundary layer, such as in most of the escarpment region of East Antarctica [Bintanja, 2000b], this increased coupling would strengthen the surface winds. At the same time, the increased vertical coupling would also mix down relatively warm air from the free atmosphere down to the surface, further reducing the temperature deficit and the katabatic forcing. Over the large part of the Antarctic continent these two effects apparently largely counteract each other, thus resulting in only minor changes in katabatically forced surface wind speeds. Overall, changes in inversion strength and vertical mixing are small, however, and changes in the katabatic forcing are thus minor. Indeed, the surface flow over the continent seems to weaken (albeit slightly, and not significantly almost everywhere) and exhibits a tendency to become less katabatic in nature as seen from the decrease in directional constancy and Weibull shape factor. Changes in the surface wind speed over Antarctica are governed by those in large-scale synoptic forcing, consistent with those in the circumpolar jet region. The van den Broeke et al. [1997] modeling results suggest that the small changes in surface wind speed over Antarctica are due to relatively small but opposing changes in summer and winter, while Bracegirdle et al. [2008] note that regions with increases largely cancel regions with decreases in wind speed. Here we have quantified the forcing terms of the force budget and can thereby attribute the (minor) changes in surface winds to mainly the synoptic pressure gradient, with the katabatic forcing changes being insignificant as discussed above. Even though the poleward shift in the westerly synoptic forcing [Yin, 2005] causes offshore and coastal easterlies to weaken [e.g., Bracegirdle et al., 2008], the wind speed changes over Antarctica s interior seem to be largely decoupled from those over the Southern Ocean. One should bear in mind that all results are based on model simulations, which can be model specific. However, in view of the earlier modeling studies cited above (confirming that future wind speed changes over the Antarctic continent are small), we believe that the results of this model are probably reasonable. Whereas the basic state and climate response of the circumpolar westerlies depend on model resolution (associated with (changes in) meridional tropospheric temperature gradient and the SAM), the changes over the Antarctic interior are largely independent of model resolution. The former can presumably be associated with the different vertical resolution in the stratosphere, as outlined by Bintanja et al. [2014], whereas the different horizontal resolution and the associated more accurate representation of the surface orography in high resolution generally have little effect on the climate response of Antarctica s surface winds. This would suggest that more accurate models in terms of model resolution will not necessarily lead to an improved or different climate response of Antarctica s surface winds. In general, modeled 21st century changes in wind speed over the Antarctic continents are small, with maximum changes of about 1 m s 1. These changes are generally smaller than the interannual variability, which implicitly includes variability due to the SAM. Van den Broeke and van Lipzig [2004] used climate model simulations to estimate the effect of the SAM on the surface climate over Antarctica. They found that while westerlies over the Southern Ocean increased from low to high SAM index (by about 1 m s 1 ), the SAM-related changes in surface wind speed over the continent are much smaller (i.e., weakening coastal easterlies by about 0.5 m s 1, and a very slight increase of up to about 0.2 m s 1 elsewhere). The direction of the change depends mainly on the alignment of the pressure gradient (or synoptic) forcing relative to the katabatic forcing [Van den Broeke and van Lipzig, 2003]. These SAM-related changes over the continent are somewhat smaller than the climate change related wind speed changes simulated here (Figure 5) but nevertheless confirm that the simulated climate signal is roughly of the same order of magnitude as the interannual variability in surface winds, meaning that climate-related wind speed changes over the continent are largely insignificant (see Figure 5) with respect to the interannual variability (of which the SAM is an important component). BINTANJA ET AL American Geophysical Union. All Rights Reserved. 7177