Abstract

Surface wind (U10) and significant wave height (Hs) response to global warming are investigated using a coupled atmosphere–wave model by perturbing the sea surface temperatures (SSTs) with anomalies generated by the Working Group on Coupled Modeling (WGCM) phase 3 of the Coupled Model Intercomparison Project (CMIP3) coupled models that use the Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4)/Special Report on Emissions Scenarios A1B (SRES A1B) scenario late in the twenty-first century.

Several consistent changes were observed across all four realizations for the seasonal means: robust increase of U10 and Hs in the Southern Ocean for both the austral summer and winter due to the poleward shift of the jet stream; a dipole pattern of the U10 and Hs with increases in the northeast sector and decreases at the midlatitude during boreal winter in the North Atlantic due to the more frequent occurrence of the positive phases of the North Atlantic Oscillation (NAO); and strong decrease of U10 and Hs in the tropical western Pacific Ocean during austral summer, which might be caused by the joint effect of the weakening of the Walker circulation and the large hurricane frequency decrease in the South Pacific.

Changes of the 99th percentile U10 and Hs are twice as strong as changes in the seasonal means, and the maximum changes are mainly dominated by the changes in hurricanes. Robust strong decreases of U10 and Hs in the South Pacific are obtained because of the large hurricane frequency decrease, while the results in the Northern Hemisphere basins differ among the models. An additional sensitivity experiment suggests that the qualitative response of U10 and Hs is not affected by using SST anomalies only and maintaining the radiative forcing unchanged (using 1980 values), as in this study.

1. Introduction

The Intergovernmental Panel on Climate Change Fourth Assessment Report (IPCC AR4) Working Group II has recognized that changes in the global wind wave climate is one of the main drivers for the assessment of the effects of climate change on coastal erosion and risks to coastal population and ecosystems (Nicholls et al. 2007). The interannual variability of waves and surges is, in some cases, exceeding the influence of projected sea level rise (Coelho et al. 2009).

Also, there are many other examples of the potential importance of changes in wind wave climate. As one example, since the ocean surface gravity wave field can penetrate far into an ice field, it can break up ice floes and accelerate ice melting during the summer, affecting estimates of the timing of the decay of the sea ice cover. One can also speculate about the importance of Langmuir circulations on ocean mixed-layer dynamics and heat and carbon uptake, especially in the Southern Ocean (Cavaleri et al. 2012).

Methods used to derive wave climate projections include statistical and dynamical approaches. A statistical approach uses current climate reanalyses (atmospheric and wave) to establish a statistical relationship between the predictor [e.g., mean sea level pressure (SLP) or surface wind] and the predictand [most commonly significant wave height (Hs)]. The statistical relationship established is then applied to projections of the predictor taken from the climate models to derive projections of the predictand (e.g., Wang et al. 2004; Wang and Swail 2006). Statistical projections of global seasonal mean and extreme Hs (20-yr return value) under chosen future climate scenarios have been carried out by Wang and Swail (2006) using the observed relationships between SLP and Hs from the 40-yr European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-40) dataset. These projections show Hs increases and decreases in many regions associated with the changes in midlatitude storms. However, the ERA-40 data tend to underestimate Hs at wind speeds above 14 m s−1 (Caires et al. 2004) due to its coarse resolution of the atmospheric model and its limited ability to resolve storm systems. The extreme wave heights using ERA-40 are severely underestimated (Hemer et al. 2010; Fan et al. 2012). Hence, statistical projections using the ERA-40 dataset may have similar biases as well.

In the dynamical approach, surface forcing is usually derived from climate models, typically based on the Coupled Model Intercomparison Project (CMIP) global climate models and used offline to force a spectral wave model. Because it is computationally expensive and highly dependent on the temporal and spatial resolution of the surface wind data, dynamical projections of the wave climate have generally been confined to regional domains, where selected downscaled CMIP emission scenarios are used to force regional wave models for time slice studies. Most of the works have been carried out for the northeast Atlantic (Kaas et al. 2001; Carretero et al.1998) and the North Sea (Debernard et al. 2002; Debernard and Røed 2008; Grabemann and Weisse 2008).

Fan et al. (2012) have developed a high-resolution global simulation system by coupling the operational wave model developed at the National Centers for Environmental Prediction (NCEP)/Environmental Modeling Center to the Geophysical Fluid Dynamics Laboratory (GFDL)'s High Resolution Atmospheric Model (HiRAM). They have evaluated a 29-yr wave climatology produced by this coupled system against National Data Buoy Center (NDBC) buoy measurements, satellite observation, and ERA-40 reanalysis and found the coupled atmosphere–wave model generated wave climate to be very reliable and able to resolve very realistic tropical storm climatology. In this study, we use this coupled atmosphere–wave model to investigate the wave climate change in response to the SST–sea ice anomalies in the late-twenty-first century simulated by global coupled models.

The outline of this paper is as follows. The coupled atmosphere–wave model is described in section 2. Section 3 discusses the wave climate change due to global warming. Summary and conclusions are given in section 4.

2. Methodology

The atmospheric model used in this coupled system is GFDL's HiRAM (e.g., Zhao et al. 2009; Fan et al. 2012), and the surface gravity wave model utilized is WAVEWATCH III developed and used operationally at the National Oceanic and Atmospheric Administration (NOAA)–NCEP (Tolman et al. 2002). This coupled model makes use of the Flexible Modeling System (FMS; http://www.gfdl.noaa.gov/fms/) coupler for calculating and passing fluxes between its atmosphere, land, sea ice, and wave components. The atmosphere model is built on a cubed sphere grid with 180 × 180 grid points on each face (corresponding to a horizontal resolution of ~50 km) and 32 vertical levels. The wave model is built on a latitude–longitude grid with a horizontal resolution of 0.5°. The wave spectrum of the model is discretized using 24 directions and 40 intrinsic (relative) frequencies extending from 0.0285 to 1.1726 Hz. Both the atmospheric model and wave model have a time step of 20 min. At every time step, the atmosphere model exchanges fluxes with the land, ice, and wave model, and the ice model passes (prescribed) ice coverage to the wave model. The coupler computes and passes fluxes between the component models and performs the necessary regridding so that each component receives inputs and supplies outputs on its own grid. All fluxes are conserved to within machine precision. More details on this coupled system can be found in Fan et al. (2012).

To generate more stable statistics and eliminate the possible complications in the response that might depend on the phase of ENSO, for example, we have chosen to follow the same strategy used in Zhao et al. (2009) in their study of tropical cyclone simulations by this atmospheric model, and use seasonally varying SSTs with no interannual variability as the lower boundary condition for the atmospheric model for all simulations.

For the control climatological SST simulation, time-averaged monthly varying SST over the time period of 1982–2000 was calculated using the NOAA optimum interpolation (OI) SST analysis dataset (Reynolds et al. 2002). We then perturb the control simulation with SST anomalies taken from various models of the climate near the end of the twenty-first century. We consider four anomaly patterns (Table 1), those obtained from the three models [GFDL's Climate Model version 2.1 (CM2.1), the Met Office (UKMO)'s third climate configuration of the Met Office Unified Model (HadCM3), and the Max Planck Institute (MPI)'s ECHAM5] that have been shown to be the most reliable models having the ability to simulate several different El Niño metrics in the current climate (van Oldenborgh et al. 2005) and that obtained by taking the ensemble mean for the simulations from 18 CMIP3 models (CMIP3 SST ensemble). All the results are taken from the Special Report on Emissions Scenarios A1B (SRES A1B) simulation in the CMIP3 archive (https://esg.llnl.gov:8443/index.jsp; Meehl et al. 2007) utilized extensively by the IPCC AR4 assessments. We compute the multimodel ensemble-mean SST anomaly by differencing the period 2081–2100 and the period 2001–20 from the historical simulations (labeled 20C3M) in the CMIP3 archive. For each of the three individual models, we first use one realization (run 1 in the CMIP3 archive) to compute the linear trend from 2000 to 2100 and then use this linear trend to calculate SST anomalies for the 80-yr period (for which we also have available the multimodel ensemble mean). The SST anomalies are computed separately for each month at each grid point. Because these trends are based on single realizations, a modest component of internal variability is likely mixed with the forced responses for the individual models, but this internal variability component should be much smaller in the multimodel mean.

Table 1.

Model experiments. All SST anomalies are taken from the A1B simulation in the CMIP3 archive (Meehl et al. 2007) utilized extensively by the IPCC AR4 assessments.

Model experiments. All SST anomalies are taken from the A1B simulation in the CMIP3 archive (Meehl et al. 2007) utilized extensively by the IPCC AR4 assessments.
Model experiments. All SST anomalies are taken from the A1B simulation in the CMIP3 archive (Meehl et al. 2007) utilized extensively by the IPCC AR4 assessments.

Figure 1 shows the seasonal-mean SST anomalies for the four patterns chosen to perturb the control run. As we can see that the warming patterns are similar between boreal winter [January–March (JFM)] and summer [July–September (JAS)] for each model with stronger warming during boreal summer. The HadCM3 anomaly is relatively large in the Pacific and relatively small in the Atlantic, while ECHAM5 has the largest average anomalies over the ocean domain. The CMIP3 SST ensemble anomaly is much smoother compared to individual realizations since many features are smoothed out through the ensemble averaging. In general, the warming is stronger in the Northern Hemisphere, and the SST anomaly in the Southern Ocean is much smaller compare to the rest of the global ocean. Notice that all four patterns show near 0- to −3-K (GFDL CM2.1) anomaly in the northernmost North Atlantic. This phenomenon is caused by the gradual weakening of the North Atlantic meridional overturning circulation during the twenty-first century due to increasing levels of greenhouse gas concentrations in the atmosphere (Dixon et al. 1999). We perturb sea ice distributions as well, as described in Zhao et al. (2009), but note that our wave model, which uses a latitude–longitude grid, sees a boundary at 72°N to avoid the numerical instability associated with the convergence of the meridians. In this initial study, we focus on the direct effects of SSTs on the wave climate, maintaining the same radiative forcing agents as in the control, which were set to be the 1980 values. This set up may give some overestimations on extreme wind and Hs (99th percentile) as Held and Zhao (2011) found that the global tropical cyclone frequency decreases about 10% as a result of the doubling of CO2. Polar ozone changes have an important impact on tropospheric circulation changes in the Southern Hemisphere. The stratospheric polar ozone depletion was shown to be the main contributor to the observed increase of summertime tropospheric westerlies in the Southern Hemisphere (Perlwitz et al. 2008) and shift of the Southern Hemisphere Annular Mode toward its positive phase (Arblaster and Meehl 2006). The model response to ozone recovery by 2100 shows that tropospheric circulation changes during austral summer caused by ozone depletion between 1970 and 2000 almost reverse, despite increasing greenhouse gas concentrations (Perlwitz et al. 2008). We do not expect significant differences in the wind and wave field responses at the end of the twenty-first century due to using the 1980 ozone concentrations in our projection experiments. A sensitivity experiment was conducted in the discussion section to explore more about the effect of ozone and doubling CO2 in our results.

Fig. 1.

JFM mean SST anomaly from (a) GFDL CM2.1, (b) Met Office HadCM3, (c) MPI ECHAM5, and (d) CMIP3 SST ensemble. (e)–(h) As in (a)–(d), but for JAS.

Fig. 1.

JFM mean SST anomaly from (a) GFDL CM2.1, (b) Met Office HadCM3, (c) MPI ECHAM5, and (d) CMIP3 SST ensemble. (e)–(h) As in (a)–(d), but for JAS.

To generate the initial condition for the wave-coupled system, a 1-yr integration of the coupled system is conducted for all experiments. For this 1-yr run, the wave model starts from calm sea; the atmospheric and land initial conditions at 1 January 1980 are taken from the end of a 10-yr run of the HiRAM model that uses climatological SSTs. Note that the spinup time needed for the wave model by itself is less than a month and is dominated by transit times of swell through the Pacific Ocean.

Both the control run and each of the perturbed runs is of 10-yr length, and we treat each year as statistically independent for the purpose of estimating the significance of the response. In other words, the 10-yr run can be treated as 10 ensemble members of a 1-yr run. Our comparisons (not shown) indicated that the control run produces slightly higher annual-mean 10-m wind speed (U10) and Hs in the Southern Ocean and lower U10 and Hs in the western North Pacific region than the annual mean of the simulations described in Fan et al. (2012). These differences are primarily caused by the difference between SSTs in the NOAA Reynolds OI and UKMO Hadley Centre Sea Ice and Sea Surface Temperature (HadISST) datasets (Zhao et al. 2009). Hemer et al. (2013) has also assessed the model skill of this control run using pattern correlation and root-mean-square differences (RMSD) between model wave fields and wave reanalysis data [from ECMWF Interim Re-Analysis (ERA-Interim) and corrected ERA-40 by Caires and Sterl 2005] and shown good agreements. Thus, we believe the qualitative response of the wave field to global warming SST anomalies is not significantly affected by the climatological SST starting point.

3. Global warming projection

a. Seasonal means

The responses of seasonal-mean U10 and Hs for each of the four perturbations are presented in Figs. 2 and 3 as the anomaly percentage (the difference of the projected ensemble mean and the control run ensemble mean expressed as the percentage of the control run ensemble mean). The black dots on the figures indicate the areas where the future ensemble means and present ensemble means are significantly different from each other at the 95% confidence level through the Student's t test (the  appendix). An example of the magnitude change of the two variables is given for the CM2.1 SST anomaly run in Fig. 4 for reference.

Fig. 2.

As in Fig. 1, but for mean U10 anomaly and CMIP3 SST ensemble expressed as a percentage of the control run JFM (JAS) mean. The black dots on the figures indicate the areas where the future ensemble mean and present ensemble mean are significantly different from each other at the 95% confidence level through the Student's t test as described in the  appendix.

Fig. 2.

As in Fig. 1, but for mean U10 anomaly and CMIP3 SST ensemble expressed as a percentage of the control run JFM (JAS) mean. The black dots on the figures indicate the areas where the future ensemble mean and present ensemble mean are significantly different from each other at the 95% confidence level through the Student's t test as described in the  appendix.

Fig. 3.

As in Fig. 2, but for seasonal-mean Hs.

Fig. 3.

As in Fig. 2, but for seasonal-mean Hs.

Fig. 4.

Absolute magnitude change of seasonal-mean (a),(b) U10 and (c),(d) Hs between the CM2.1 experiment and the control run during (left) JFM and (right) JAS.

Fig. 4.

Absolute magnitude change of seasonal-mean (a),(b) U10 and (c),(d) Hs between the CM2.1 experiment and the control run during (left) JFM and (right) JAS.

1) Boreal winter (JFM)

In the North Atlantic, dipole patterns are observed in both the responses of U10 (Figs. 2a–d) and Hs (Figs. 3a–d) for all four SST perturbations, with increases in the northeast Atlantic and decreases in the midlatitudes. Such changes are associated with more frequent occurrence of the positive phases of the North Atlantic Oscillation (NAO) under global warming (Monahan et al. 2000) and are consistent with the findings by Wang et al. (2004), who constructed Hs climate change scenarios for the North Atlantic using the observed SLP–Hs relationships through redundancy analysis. Three models show strong decrease of U10 (up to 10%) and Hs (up to 20%) at midlatitudes, while the responses to the HadCM3 SST are much weaker. The increases at the northeastern sector of the North Atlantic appear to be strongest in the CM2.1 model for both U10 (12%–13% or >1.2 m s−1 as shown by Fig. 4a) and Hs (up to 15% or >0.5 m as shown by Fig. 4c), while the responses from all the other three models are much weaker and within 7%–8%.

Because of the poleward shifting of the jet streams (Yin 2005; Lu et al. 2007), the SST anomalies from all four models resulted in U10 increase at high latitudes in both hemispheres. The U10 increase over the Southern Ocean (up to 10%–15%, ~1.1 m s−1 for the CM2.1 SST anomaly) is stronger than the Northern Hemisphere and is accompanied by weaker decreases around 40°S. As Kushner et al. (2001) has pointed out, the extratropical circulation response to global warming consists of a Southern Hemisphere summer half-year poleward shift of the westerly jet and a year-round positive wind anomaly in the stratosphere and the tropical upper troposphere. The tropospheric wind response projects strongly onto the model's southern annular mode (SAM), which is the leading pattern of variability of the extratropical zonal winds. The increases in greenhouse gas concentration will continue to shift the SAM toward its positive phase (Arblaster et al. 2011), which is marked by poleward displacements of the jet and thus the strengthening of the prevailing atmospheric eastward flow near ~60°S and weakening of the prevailing atmospheric eastward flow near ~40°S (Thompson et al. 2011).

Strong decreases (more than 20%, >1.5 m s−1 for the CM2.1 SST anomaly) of U10 in the tropical western Pacific Ocean are also observed consistently in all four models, which might be caused by the weakening of the Walker circulation (Vecchi et al. 2006; Vecchi and Soden 2007). The strong decrease in hurricane frequencies in the South Pacific basin for all four models (detailed discussions are given in section 3b) is another possible reason for such strong decreases in the mean U10 field.

The corresponding Hs responses show a very similar structure to the U10 responses in the Southern Ocean (10%–15% increase, ~0.35 m for the CM2.1 SST anomaly) and western Pacific (5%–10% decrease, <0.25 m for the CM2.1 SST anomaly), while the Hs anomaly in the eastern Pacific show much smoother structures with less variations than the U10 anomaly. This is because the eastern Pacific is dominated by swells from the Southern Ocean and the North Pacific midlatitude storm region. More than 75% of the total wave energy is from swells (Fan et al. 2013, manuscript submitted to J. Climate). Small variations in the wind sea due to local wind change are covered by the strong swell energy in the wave spectrum, especially in the equatorial region.

2) Boreal summer (JAS)

Compared to boreal winter, a larger magnitude of increases in U10 (10%–17%, ~1.2 m s−1 for the CM2.1 SST anomaly) and Hs (10%–20%, >0.5 m for the CM2.1 SST anomaly) are observed in the Southern Ocean for all four SST perturbations during boreal summer (austral winter). The increase areas have larger meridional extends but were accompanied by weak decreases (less than 7%) in the South Pacific sector between 160° and 75°W (the southern tip of South America). Similar changes are also found for the Hs responses in the Southern Ocean (Figs. 3e–h). Izaguirre et al. (2011) found that the positive phase of Niño-3 is correlated with wave height decrease in the South Pacific sector of the Southern Ocean, especially to the west of the South American coast and the Drake Passage region. Even though our experiments are designed to eliminate possible complications in the response that might depend on the phase of ENSO, the SST anomalies in the eastern and central Pacific exhibit a more pronounced warming with patterns that resemble El Niño SST anomalies. Thus, the projected changes in the wave field may be described as El Niño–like.

In the western North Pacific Ocean, the CMIP3 SST ensemble model SST anomaly show a strong decrease of U10 (>20%) in the tropics, as does the ECHAM5 SST, while the other two SSTs generate increases of the U10 (>20%, >1.5 m s−1 for the CM2.1 SST anomaly) in the western Pacific. This is mainly due to the different hurricane frequency responses among the models. The ECHAM5 and CMIP3 SST ensemble models project strong decrease in hurricane frequency in the western Pacific basin due to global warming, while the other two models project no change in frequencies (Fig. 6c), with southward shift in hurricane tracks. More details will be discussed in section 3b as we look at the extremes.

The Hs responses in the Southern Ocean look very similar to the U10 responses with larger magnitude. This is because the Southern Ocean is dominated by wind seas. The Hs for the wind sea part is roughly proportional to (U10)2, thus the responses in the U10 will be magnified in the wind sea part of the Hs responses. Large swells that emerge from these wind-sea generation areas will then cause a strong increase of Hs in the surrounding areas as well.

The Hs responses in the North Pacific look very different from the U10 responses at first, but if look more carefully, we find that the Hs responses are very similar to the U10 responses to the west of 150°E and very different to the east of 150°E. At 150°E is where the eastern boundary of the Australian continent is and where the land sheltering effect for the Southern Ocean waves ends. During boreal summer (austral winter), the large swells emerging from the Southern Ocean can propagate all the way to the very northern end of the Pacific Ocean but cannot enter any part of the Pacific Ocean west of 150°E because of the sheltering effect by the Australian continent. Thus, to the west of 150°E, the western Pacific wave responses are mainly controlled by local wind responses (hurricanes). While to the east of 150°E, the wave responses are controlled by the combined effect of both local wind responses (hurricane) and large swells from the Southern Ocean that reflect the large Hs increases in the Southern Ocean. The final outcome is that we observed a reduced decrease in Hs compared to U10 decrease for the ECHAM5 and CMIP3 SST ensemble model and enhanced increase in Hs for the other two models.

In general, the U10 and Hs responses are more robust across the four models during boreal winter than summer. This is because the U10 and Hs responses in the Northern–Southern Hemispheres are mainly controlled by the midlatitude storms during boreal winter. Robust poleward shifts of the midlatitude storms are observed across all four models. Especially in the North Atlantic, all four SST patterns show 0° to −3°C anomaly in the northeast with up to 3°C warming in the rest of the basin (Fig. 1). Such warming patterns are associated with the more frequent occurrence of the positive phases of NAO and the decrease of U10 and Hs in the midlatitudes. During boreal summer, the large swells emitting from the Southern Ocean propagate all the way across the entire globe, interact with waves on its path, and make the wave response pattern much more complicated than during boreal winter. Furthermore, the extreme wind and waves generated by hurricanes also alter the mean U10 and Hs fields in the tropics and subtropics, making their responses even more complicated.

b. The extremes (99th percentile)

The boreal winter (JFM) and summer (JAS) 99th percentile U10 and Hs were calculated from the 6-hourly model results on every grid point during the 10-yr calculation period for both the control run and the four perturbation runs. Their values for the control run are given in Fig. 5, and their comparison with the ERA-40 reanalysis can be found in Fan et al. (2012). We can see that the strongest winds (Figs. 5a,c) are mainly found at the Southern Ocean and the midlatitude storm-track region in the Northern Hemisphere. We can also clearly see the presence of strong tropical storms in all the ocean basins, especially the western North Pacific during boreal summer (Fig. 5c) because this region features the most numerous and intense tropical cyclones globally. The strong gap winds in the Gulf of Tehuantepec located in the east Pacific Ocean off the Mexican coast are also nicely resolved (Fig. 5a). The global structure of the 99th percentile Hs (Figs. 5b,d) follows the 99th percentile U10 in general pattern. The extreme wave heights generated by the tropical cyclones are more confined along the storm tracks due to their small size compared with midlatitude storms.

Fig. 5.

JFM 99th percentile (a) U10 and (b) Hs for the control experiment. (c),(d) As in (a),(b), but for JAS.

Fig. 5.

JFM 99th percentile (a) U10 and (b) Hs for the control experiment. (c),(d) As in (a),(b), but for JAS.

To better understand the behavior of the extreme wind and waves, we also analyzed the hurricane statistics in our control and projection runs. The tropical cyclone detection and tracking algorithm is adapted from Zhao et al. (2009). A tropical storm is categorized as a hurricane if the maximum surface wind speed at some point during its entire trajectory exceeds 33 m s−1. The total number of hurricanes summed over various ocean basins is shown in Fig. 6a for the control run. We can see that the number of storms in the North Atlantic and South Pacific is very close to that observed, while there is about a 20% underprediction of hurricanes in the east Pacific and a more significant (40%) overprediction of hurricanes in the west Pacific. The overprediction in the west Pacific is most pronounced in May–June, which fortunately is not during our analysis period for this study, while the underprediction in the east Pacific is the largest in August–September (Zhao et al. 2009). Also, although HiRAM produces a very realistic climatology of tropical cyclones, it generates few storms stronger than category 2 (Zhao et al. 2009). The response of storm frequency for each of the four perturbations is shown in Figs. 6b–e as the fractional changes in hurricane count per year for the four ocean basins.

Fig. 6.

(a) A comparison of observed (gray × symbols) and simulated annual-mean hurricane count in each ocean basin. The open circles stand for each member and the ensemble mean is represented by the black solid square. Fractional changes in annual hurricane count for the (b) North Atlantic, (c) west Pacific, (d) east Pacific, and (e) South Pacific basins from the four SST anomaly simulations. Error bars show the 90% confidence level through the Student's t test (see the  appendix), assuming the sampling distributions are normally distributed. The gray square and error bar show the results from the SST–ozone–CO2 experiment.

Fig. 6.

(a) A comparison of observed (gray × symbols) and simulated annual-mean hurricane count in each ocean basin. The open circles stand for each member and the ensemble mean is represented by the black solid square. Fractional changes in annual hurricane count for the (b) North Atlantic, (c) west Pacific, (d) east Pacific, and (e) South Pacific basins from the four SST anomaly simulations. Error bars show the 90% confidence level through the Student's t test (see the  appendix), assuming the sampling distributions are normally distributed. The gray square and error bar show the results from the SST–ozone–CO2 experiment.

The 99th percentile U10 and Hs responses to each of the four SST perturbations during JFM and JAS are presented in Figs. 7 and 8. Notice that the magnitude of these responses is twice as strong as the responses of the seasonal means in Figs. 2 and 3.

Fig. 7.

As in Fig. 1, but for 99th percentile U10 anomaly and CMIP3 SST ensemble expressed as percentage of the control run.

Fig. 7.

As in Fig. 1, but for 99th percentile U10 anomaly and CMIP3 SST ensemble expressed as percentage of the control run.

Fig. 8.

As in Fig. 7, but for the 99th percentile Hs anomaly.

Fig. 8.

As in Fig. 7, but for the 99th percentile Hs anomaly.

1) Boreal winter (JFM)

The major response for the 99th percentile U10 during boreal winter (austral summer) is that the SSTs from all four models show a robust reduction of up to 40% in the tropical South Pacific (Figs. 7a–d) where tropical cyclones are most active. All four models show a strong reduction of 40%–50% in hurricane frequency in this region (Fig. 6e). The relatively larger warming of the SSTs in the Northern as compared to the Southern Hemisphere tropics/subtropics, resulting in a more stable atmosphere in the south, is a plausible cause for this consistent reduction in activity (Zhao et al. 2009).

The strong reduction in U10 leads to a stronger reduction in the 99th percentile Hs (40%–60%) in the tropical South Pacific (Figs. 8a–d). Since the large swells predominately generated by the tropical cyclones in the tropical South Pacific can propagate throughout the entire South Pacific and tropical North Pacific (Alves 2005), robust significant Hs reductions are observed in the eastern South Pacific and tropical North Pacific for all four projections as well.

2) Boreal summer (JAS)

The 99th percentile U10 (Figs. 7 e–h) responses to the SST anomalies during boreal summer are mainly found in the Northern Hemisphere. First focusing on the North Atlantic, SSTs from two models (CM2.1 and ECHAM5) show patches of strong increases, while the HadCM3 model shows strong decrease. Such changes are consistent with the hurricane frequency response in Fig. 6b, where increases of storm frequencies are observed for the CM2.1 and ECHAM5 models and a strong decrease of storm frequency is shown for the HadCM3 model.

It is interesting to notice that although the CMIP3 SST ensemble model shows a ~30% decrease in hurricane frequency, we do not see significant changes in U10 compared to the other three models. Through analysis of the hurricane strength and duration (not shown), we found that the projected hurricane duration from the CMIP3 SST ensemble model is generally longer than the control run with relatively stronger intensity. Thus, the reduction in hurricane frequency did not result in a significant decrease in the 99th percentile U10 for this projection. We also found that the projected hurricanes barely enter the Gulf of Mexico and Caribbean region for all four models, which has resulted in a robust decrease of U10 in these regions.

The Hs responses in the North Atlantic are in general consistent with the U10 responses. The CM2.1 and ECHAM5 models show strong increases of 40%–60%, while the HadCM3 shows a strong decrease of more than 40%–50%.

In the western Pacific, both the CM2.1 and HadCM3 show roughly no change in hurricane frequency (Fig. 6c). But, interestingly, we see a decrease in U10 between 15° and 30°N and an increase in U10 between the equator and 15°N for both models. This is mainly because there are more (less) hurricanes south (north) of 15°N for both models, even the total number did not change much. The decrease in hurricane frequencies is observed for the ECHAM5 and CMIP3 SST ensemble model in the western Pacific, which is accompanied by decreases in U10 and Hs in the equatorial western Pacific.

All three models project an increase of hurricane frequencies in the eastern Pacific except the CM2.1 model, which indicates a small decrease (Fig. 6d). The HadCM3 model projected the largest hurricane frequency increase of 105% ± 20%, in contrast to the strong decrease it projected in the North Atlantic. The explanation very likely lies in the differential warming over the tropical Atlantic and tropical Pacific, with the ratio of Atlantic to Pacific warming in HadCM3 SST clearly smaller than in any of the other models considered (Fig. 1). Corresponding strong increases of U10 and Hs are observed in the eastern Pacific for all three models with the CM2.1 model showing mild decrease in these regions.

For all three basins discussed above, even though the Hs responses follow similar patterns as the U10 responses, the detailed spatial distribution and responding magnitude are noticeably different between the two variables. This is mainly due to the surface wave resonance effect on the right-hand side of the storm; that is, they were exposed to prolonged forcing from wind because the hurricane translation speed was comparable to the group speed of the dominant waves (Fan et al. 2009). In addition, the hurricane duration and curvature, which are not explicitly visible in the U10 responses, are also very important factors in determining the distribution and magnitude of the ocean surface gravity waves under the hurricanes (Moon et al. 2003).

c. The North American coasts

Figures 9 and 10 present the 99th percentile U10 and Hs changes due to global warming along the North American region in the boreal winter and summer season. During boreal winter (JFM), all four models show consistent projections of up to 15% decrease in the 99th percentile U10 across the North American region (Figs. 9 a–d). The decreases in the North Atlantic are associated with the weakening of westerlies in the midlatitudes due to the more frequent occurrence of the positive phases of NAO under global warming as we discussed in section 3a. Corresponding robust decreases of the 99th percentile Hs (10%–25%) are observed in the North Atlantic as well (Figs. 10a–d). Apparently the responses in the Hs field are much stronger than the U10 field. This is because the waves at the eastern coast of North America (NA) are mainly dominated by wind waves during boreal winter (Fan et al. 2013, manuscript submitted to J. Climate). As we discussed earlier, the wind wave part of the Hs is roughly proportional to (U10)2, thus the responses in the U10 are magnified in the Hs responses.

Fig. 9.

As in Fig. 7, but for the North American region 99th percentile U10 anomaly.

Fig. 9.

As in Fig. 7, but for the North American region 99th percentile U10 anomaly.

Fig. 10.

As in Fig. 7, but for the North American region 99th percentile Hs anomaly.

Fig. 10.

As in Fig. 7, but for the North American region 99th percentile Hs anomaly.

The decrease of U10 to the south of 30°N along the western coast of NA is accompanied by 5%–10% increases north of it. This phenomenon is caused by the intensification and poleward shift of midlatitude storm tracks associated with global warming (Yin 2005). The Hs responses in general agree with the U10 responses, but we also notice the different responses along the western coast of NA, especially for the ECHAM5 model and the CMIP3 SST ensemble model. This is because the waves here are dominated by swells from the North Pacific storm-track region. Since more than 75% of the wave energy is from swells in this region (Fan et al. 2013, manuscript submitted to J. Climate), the Hs response reflects a combination of local wind response and the response of remote midlatitude storms to the SST anomalies. Thus, we see a more robust response of both U10 and Hs decreases along the eastern coast of NA, while the Hs responses along the western coast of NA are quite complex because of the swell dominance.

The changes of the 99th percentile U10 and Hs are much stronger during boreal summer (JAS) compared to winter. The Hs responses are very similar to the U10 responses in both the Pacific and Atlantic part of the North American region with a bit stronger magnitude. Three projections show large U10 and Hs increases (>30%) dominate the western coast of NA with the HadCM3 model showing an up to 60% increase, while the CM2.1 model hardly shows any changes in U10 with mild decrease in Hs. These are apparently associated with the hurricane frequency increases for the three models in the eastern Pacific basin (Fig. 6d), where the CM2.1 projects no change in hurricane frequency. As for the North Atlantic, the CM2.1 and ECHAM5 SSTs show strong U10 and Hs increases of more than 40% due to the hurricane frequency increase in these two models, while strong decreases of similar magnitude dominate the HadCM3 projection due to the large hurricane frequency decrease. The CMIP3 SST ensemble model, on the other hand, shows patches of mild increases and decreases. Notice that all four models project robust decreases in U10 and Hs of more than 20% in the Gulf of Mexico and surrounding area, which is because the projected hurricanes rarely enter the Gulf of Mexico in all four models.

In general, the 99th percentile U10 and Hs responses are more robust across the four models during boreal winter than summer. This is because the extreme U10 and Hs responses are mainly controlled by the midlatitude storms in the Northern Hemisphere during boreal winter. Robust poleward shifts of the midlatitude storm are observed across all four models. Especially in the North Atlantic, all four SST warming patterns are associated with the more frequent occurrence of the positive phases of NAO and decrease of U10 and Hs in the midlatitudes. During boreal summer, the 99th percentile U10 and Hs are mainly dominated by extreme waves from hurricanes, which is more chaotic compared to midlatitude storms. Different hurricane frequency, track, and strength projection in different models have resulted in different increasing and decreasing patterns and strength in the North American region.

4. Summary and discussion

Fan et al. (2012) have produced a 29-yr (1981–2009) wind and wave climatology using a coupled atmosphere–wave global simulation system developed at NOAA–GFDL and evaluated it through extensive tests against NDBC buoy measurements, satellite observations, and ERA-40 reanalysis. In this study, we use the same coupled system to investigate the mean and extreme (99th percentile) surface wind and wave climate changes to the SST–sea ice anomalies in the late-twenty-first century by perturbing the SSTs with anomalies taken from various models of the climate near the end of the twenty-first century. SST anomalies from the ensemble of 18 CMIP3 models as well as from three individual models (GFDL's CM2.1, the UKMO's HadCM3, and MPI's ECHAM5) are chosen to perturb the control SST. All results are taken from the A1B simulation in the CMIP3 archive (Meehl et al. 2007) utilized extensively by the IPCC AR4 assessments. The radiative forcing agents are maintained as in the control, which were set to be the 1980 values.

For the seasonal-mean responses, the most robust response we observed is the increase of U10 and Hs in the Southern Ocean for both seasons due to the poleward shift of the jet streams. Similar changes were also observed by Wang and Swail (2006) through their statistical projections but with smaller magnitude. Mori et al. (2010) also show 6%–9% increases in the Southern Ocean in their annual-mean Hs projections. This indicates that the U10 and Hs increase in the Southern Ocean are robust across different studies (method, model, or scenario) regardless of the magnitude variations among them.

In the North Atlantic, we found dipole patterns with increases in the northeast sector and decreases at the midlatitudes during boreal winter (JFM) for the seasonal-mean U10 and Hs, which are associated with the more frequent occurrence of the positive phases of NAO under global warming. Wang et al. (2004) also reported the same response with lower magnitude in their study. The fact that our model resolves higher waves in the midlatitude storm region than ERA-40 (Fan et al. 2012) might be the reason that we see stronger responses in both the Southern Ocean and the North Atlantic.

Consistent decreases in U10 (>20%) and Hs (~10%) are observed at the tropical western Pacific Ocean for all four models during boreal winter. These decreases might be caused by a combined effect of the weakening of the Walker circulation (Vecchi et al. 2006; Vecchi and Soden 2007) and the strong decrease in hurricane frequencies in the South Pacific basin.

In general, the U10 and Hs responses are more robust across the four models during boreal winter. This is because the U10 and Hs responses in the Northern–Southern Hemispheres are mainly controlled by the midlatitude storms during boreal winter. Robust poleward shift of the midlatitude storms are observed across all four models. Especially in the North Atlantic, all four SST patterns show 0° to −3°C anomaly in the northeast with up to 3°C warming in the rest of the basin (Fig. 1). Such warming patterns are associated with the more frequent occurrence of the positive phases of NAO and decrease of U10 and Hs in the midlatitudes. During boreal summer, the large swells emitting from the Southern Ocean propagate all the way across the entire globe, interact with waves on its path, and make the wave response pattern much more complicated than boreal winter. Furthermore, the extreme wind and waves generated by hurricanes also alter the mean U10 and Hs fields in the tropics and subtropics, making their responses even more complicated. These results have emphasized the point that waves are nonlocal, and global dynamic projections are the most reasonable way of studying wave climate change. Regional dynamic projections of wave climate change are not able to represent the important effect of swells. The statistical approaches rely on the statistical relationship between the predictor [e.g., mean sea level pressure (SLP) or surface wind] and the predictand [most commonly significant wave height (Hs)] to generate wave climate projections. Since swells do not have a clear relationship with atmospheric variables, the swell effects may be underestimated or misrepresented in the statistical projections as well.

Changes of the 99th percentile U10 and Hs are twice as strong as the changes in the seasonal- mean fields, and the maximum changes are mainly dominated by the changes in tropical storms. More robust responses are observed again in boreal winter when all four models projected a decrease of hurricane frequencies in the South Pacific. The projected 99th percentile U10 and Hs changes along the North American coast are generally consistent across the four models during boreal winter with decreases of U10 (up to 10%) and Hs (up to 25%) dominating most the western and eastern coast. The projected 99th percentile U10 and Hs changes are much stronger and more complex during boreal summer and are mainly influenced by the hurricanes in the eastern Pacific and North Atlantic.

This study focuses on the influence of SST change on global wind and wave responses and maintains the radiative forcing unchanged. Since the southern annual mode (SAM) controls the wind field response in the Southern Ocean and the wave response globally, questions may arise on how different the projected shift in the SAM will be if we consider the effect of both ozone recovery and doubling CO2. To address this question, one more experiment, SST–ozone–CO2, was conducted (Table 1) by doubling CO2 and using ozone concentration values at the end of the twenty-first century in addition to the SST anomaly. For this experiment, we chose to use the SST anomaly from the ensemble of 18 CMIP3 models (as in the CMIP3 SST ensemble experiment). The seasonal mean and the 99th percentile U10 and Hs for boreal winter (JFM) and summer (JAS) for this experiment are given in Fig. 11.

Fig. 11.

JFM mean (a) U10 and (b) Hs; and 99th percentile (e) U10 and (g) Hs anomaly from the SST–ozone–CO2 experiment expressed as percentage of the control run. (c),(d) and (f),(h) As in (a),(b) and (e),(g), but for JAS. The black dots on the figures indicate the areas where the future ensemble mean and present ensemble mean are significantly different from each other at the 95% confidence level through the Student's t test as described in the  appendix.

Fig. 11.

JFM mean (a) U10 and (b) Hs; and 99th percentile (e) U10 and (g) Hs anomaly from the SST–ozone–CO2 experiment expressed as percentage of the control run. (c),(d) and (f),(h) As in (a),(b) and (e),(g), but for JAS. The black dots on the figures indicate the areas where the future ensemble mean and present ensemble mean are significantly different from each other at the 95% confidence level through the Student's t test as described in the  appendix.

For the SST–ozone–CO2 experiment, the global structures of seasonal-mean U10 responses (Figs. 11a,c) are very similar to the CMIP3 SST ensemble experiment (Figs. 2d,h) for both seasons with relatively small differences over the Southern Ocean. We observe weaker increase in the sector south of Australia during boreal winter (austral summer) and stronger increase in the Southern Ocean (mainly in the South Atlantic sector) during boreal summer (austral winter).

Since the 2100 ozone concentrations are very similar to the 1980 values, the differences we observe should be mainly caused by the effect of doubling CO2. The fact that we do not observe much difference between these two experiments suggests that the shift of the jet stream is determined not directly by the changes in the CO2 concentrations but mainly by the overall warming or cooling of the ocean surface caused by the CO2 changes (Lee 1999). Since the SST anomalies are the same for both experiments, we do not see much difference in the seasonal- mean U10 responses.

The corresponding seasonal-mean Hs responses for the SST–ozone–CO2 experiment (Figs. 11b,d) show similar patterns as the U10 responses with weaker increase in the Southern Ocean during boreal winter (austral summer) and stronger increase during boreal summer (austral winter) compared to the CMIP3 SST ensemble experiment (Figs. 3d,h).

Notice that the U10 responses during boreal summer (Fig. 11c) show stronger increase in the equatorial North Atlantic and weaker increase in the equatorial east Pacific compared to the CMIP3 SST ensemble experiment (Fig. 2h). This is caused by the hurricane frequency change in these two basins due to the doubling of CO2. We find that after we double the CO2 concentration, the decrease–increase in hurricane frequency was reduced in the North Atlantic–east Pacific basins (Figs. 6b,d), while the hurricane frequency decrease was enhanced for the west Pacific and South Pacific basins (Figs. 6c,e). As a result, we observe less increase of the 99th percentile U10 in the east Pacific during boreal summer for the SST–ozone–CO2 experiment (Fig. 11f), while the other basins do not show much changes compare to the CMIP3 SST ensemble experiment (Fig. 7h).

In summary, even though small differences are observed in the SST–ozone–CO2 experiment, for which we have specified the ozone concentrations for 2100 and doubled the CO2 concentrations, the results indicate that the qualitative response of the U10 and Hs fields to global warming still holds even though they are obtained from experiments using the SST anomalies only and that kept the radiative forcing agents unchanged as the 1980 values.

Even though this study is conducted under limited climate change scopes, it is a useful first step toward more comprehensive wave climate change studies using dynamical wave projections. This study has shown the importance of wave climate changes in the Southern Ocean on global wave climate change and demonstrated the close relationship between wave climate change and the climate indices. The extreme wave climate change corresponding to the tropical cyclone changes has raised concerns for the coastal community and emphasized the need to include wave climate projections in the IPCC CMIP simulations.

APPENDIX

Student's t Test

In this study, we use the Student's t test to assess whether the ensemble mean of the projected variables (U10 and Hs) are statistically different from the ensemble mean of control runs.

Let X and Y stand for the variables in the control and projections runs, then at every grid point we have two samples to investigate: (X1, X2, … XNx) and (Y1, Y2, … YNy). Here, the subscriptions denote ensemble members: Nx = Ny = 10 in this study. If the two samples are assumed to have the same variance , then the t-test statistic is given by

 
formula

which, under the null hypothesis (i.e., the two means are the same), follows a t distribution with (Nx + Ny − 2) number of degrees of freedom. In (A1), and are the mean of the two samples, and Sp is the pooled estimate of the common standard deviation

 
formula

Let t0.95(Nx + Ny − 2) denote the 95% critical value of the t distribution with (Nx + Ny − 2) number of degrees of freedom. If tstat > t0.95(Nx + Ny − 2), then the two means are significantly different from each other at the 95% confidence level.

REFERENCES

REFERENCES
Alves
,
J. G. M.
,
2005
:
Numerical modeling of ocean swell contributions to the global wind-wave climate
.
Ocean Modell.
,
11
,
98
122
.
Arblaster
,
J.
, and
G. A.
Meehl
,
2006
:
Contributions of external forcings to southern annular mode trends
.
J. Climate
,
19
,
2896
2905
.
Arblaster
,
J.
,
G. A.
Meehl
, and
D. J.
Karoly
,
2011
:
Future climate change in the Southern Hemisphere: Competing effects of ozone and greenhouse gases
.
Geophys. Res. Lett.
,
38
,
L02701
,
doi:10.1029/2010GL045384
.
Caires
,
S.
, and
A.
Sterl
,
2005
:
A new nonparametric method to correct model data: Application to significant wave height from the ERA-40 Re-Analysis
.
J. Atmos. Oceanic Technol.
,
22
,
443
459
.
Caires
,
S.
,
A.
Sterl
,
J.-R.
Bidlot
,
N.
Graham
, and
V.
Swail
,
2004
:
Intercomparison of different wind–wave reanalysis
.
J. Climate
,
17
,
1893
1913
.
Carretero
,
J. C.
, and
Coauthors
,
1998
:
Changing waves and storms in the northeast Atlantic?
Bull. Amer. Meteor. Soc.
,
79
,
741
760
.
Cavaleri
,
L.
,
B.
Fox-Kemper
, and
M.
Hemer
,
2012
:
Wind waves in the coupled climate system
.
Bull. Amer. Meteor. Soc.
,
93
,
1651
1661
.
Christensen
,
J. H.
, and
Coauthors
,
2007
: Regional climate projections. Climate Change 2007: The Physical Science Basis, S. Solomon et al., Eds., Cambridge University Press, 847–940.
Coelho
,
C.
,
R.
Silva
,
F.
Veloso-Gomes
, and
F.
Taveira-Pinto
,
2009
:
Potential effects of climate change on northwest Portuguese coastal zones
.
ICES J. Mar. Sci.
,
66
,
1497
1507
.
Debernard
,
J.
, and
L.
Røed
,
2008
:
Future wind, wave and storm surge climate in the Northern Seas: A revisit
.
Tellus
,
60
,
427
438
,
doi:10.1111/j.1600-0870.2008.00312.x
.
Debernard
,
J.
,
Ø.
Satra
, and
L. P.
Roed
,
2002
:
Future wind, wave and storm surge climate in the northern North Atlantic
.
Climate Res.
,
23
,
39
49
.
Dixon
,
K. W.
,
T. L.
Delworth
,
M. J.
Spelman
, and
R. J.
Stouffer
,
1999
:
The influence of transient surface fluxes on North Atlantic overturning in a coupled GCM Climate Change experiment
.
Geophys. Res. Lett.
,
26
(
17
),
2749
2752
.
Fan
,
Y.
,
I.
Ginis
,
T.
Hara
,
C. W.
Wright
, and
E. J.
Walsh
,
2009
:
Numerical simulations and observations of surface wave fields under an extreme tropical cyclone
.
J. Phys. Oceanogr.
,
39
,
2097
2116
.
Fan
,
Y.
,
S.-J.
Lin
,
I. M.
Held
,
Z.
Yu
, and
H. L.
Tolman
,
2012
:
Global ocean surface wave simulation using a coupled atmosphere–wave model
.
J. Climate
,
25
,
6233
6252
.
Grabemann
,
I.
, and
R.
Weisse
,
2008
:
Climate change impact on extreme wave conditions in the North Sea: An ensemble study
.
Ocean Dyn.
,
58
,
199
212
.
Held
,
I. M.
, and
M.
Zhao
,
2011
:
The response of tropical cyclone statistics to an increase in CO2 with fixed sea surface temperature
.
J. Climate
,
24
,
5353
5364
.
Hemer
,
M. A.
,
X. L.
Wang
,
J. A.
Church
, and
V. R.
Swail
,
2010
:
Modeling proposal: Coordinating global ocean wave climate projections
.
Bull. Amer. Meteor. Soc.
,
91
,
451
454
.
Hemer
,
M. A.
,
Y.
Fan
,
N.
Mori
,
A.
Semedo
, and
X. L.
Wang
,
2013
:
Projected future changes in wave climate from a multi-model ensemble
.
Nat. Climate Change
,
3
,
471
476
,
doi:10.1038/NCLIMATE1791
.
Izaguirre
,
C.
,
F. J.
Méndez
,
M.
Menéndez
, and
I. J.
Losada
,
2011
:
Global extreme wave height variability based on satellite data
.
Geophys. Res. Lett.
,
38
,
L10607
,
doi:10.1029/2011GL047302
.
Kaas
,
E.
, and
Coauthors
, cited
2001
: Regional storm, wave and surge scenarios for the 2100 century (STOWASUS-2100): Final report. [Available online at http://www.dmi.dk/pub/STOWASUS-2100/.]
Kushner
,
P. J.
,
I. M.
Held
, and
T. L.
Delworth
,
2001
:
Southern Hemisphere atmospheric circulation response to global warming
.
J. Climate
,
14
,
2238
2249
.
Lee
,
S.-Y.
,
1999
:
Why are the climatological zonal winds easterly in the equatorial upper troposphere?
J. Atmos. Sci.
,
56
,
1353
1363
.
Lu
,
J.
,
G. A.
Vecchi
, and
T.
Reichler
,
2007
:
Expansion of the Hadley cell under global warming
.
Geophys. Res. Lett.
,
34
,
L06805
,
doi:10.1029/2006GL028443
.
Meehl
,
G.
,
C.
Covey
,
T.
Delworth
,
M.
Latif
,
B.
McAvaney
,
J.
Mitchell
,
R.
Stouffer
, and
K.
Taylor
,
2007
:
The WCRP CMIP3 multimodel dataset: A new era in climate change research
.
Bull. Amer. Meteor. Soc.
,
88
,
1383
1394
.
Monahan
,
A. H.
,
J. C.
Fyfe
, and
G. M.
Flato
,
2000
:
A regime view of Northern Hemisphere atmospheric variability and change under global warming
.
Geophys. Res. Lett.
,
27
(8),
1139
1142
.
Moon
,
I.-J.
,
I.
Ginis
,
T.
Hara
,
H.
Tolman
,
C. W.
Wright
, and
E. J.
Walsh
,
2003
:
Numerical simulation of sea surface directional wave spectra under hurricane wind forcing
.
J. Phys. Oceanogr.
,
33
,
1680
1706
.
Mori
,
N.
,
T.
Yasuda
,
H.
Mase
,
T.
Tom
, and
Y.
Oku
,
2010
:
Projection of extreme wave climate change under global warming
.
Hydrol. Res. Lett.
,
4
,
15
19
.
Nicholls
,
R. J.
,
P. P.
Wong
,
V. R.
Burkett
,
J. O.
Codignotto
,
J. E.
Hay
,
R. F.
McLean
,
S.
Ragoonaden
, and
C. D.
Woodroffe
,
2007
: Coastal systems and low-lying areas. Climate Change 2007: Impacts, Adaptation and Vulnerability. M. L. Parry et al., Eds., Cambridge University Press, 315–356.
Perlwitz
,
J.
,
S.
Pawson
,
R. L.
Fogt
,
J. E.
Nielsen
, and
W. D.
Neff
,
2008
:
Impact of stratospheric ozone hole recovery on Antarctic climate
.
Geophys. Res. Lett.
,
35
,
L08714
,
doi:10.1029/2008GL033317
.
Reynolds
,
R.
,
N.
Rayner
,
T.
Smith
,
D.
Stokes
, and
W.
Wang
,
2002
:
An improved in situ and satellite SST analysis for climate
.
J. Climate
,
15
,
1609
1625
.
Thompson
,
D. W. J.
,
S.
Solomon
,
P. J.
Kushner
,
M. H.
England
,
K. M.
Grise
, and
D. J.
Karoly
,
2011
:
Signatures of the Antarctic ozone hole in Southern Hemisphere surface climate change
.
Nat. Geosci.
,
4
,
741
749
,
doi:10.1038/NGEO1296.
Tolman
,
H. L.
,
B.
Balasubramaniyan
,
L. D.
Burroughs
,
D.
Chalikov
,
Y. Y.
Chao
,
H. S.
Chen
, and
V. M.
Gerald
,
2002
:
Development and implementation of wind-generated ocean surface wave models at NCEP
.
Wea. Forecasting
,
17
,
311
333
.
van Oldenborgh
,
G.
,
S.
Phili
, and
M.
Collins
,
2005
:
El Niño in a changing climate: A multi-model study
.
Ocean Sci.
,
1
,
81
95
.
Vecchi
,
G. A.
, and
B. J.
Soden
,
2007
:
Global warming and the weakening of the tropical circulation
.
J. Climate
,
20
,
4316
4340
.
Vecchi
,
G. A.
,
B. J.
Soden
,
A. T.
Wittenberg
,
I. M.
Held
,
A.
Leetmaa
, and
M. J.
Harrison
,
2006
:
Weakening of tropical Pacific atmospheric circulation due to anthropogenic forcing
.
Nature
,
441
,
73
76
,
doi:10.1038/nature04744
.
Wang
,
X. L.
, and
V. R.
Swail
,
2006
:
Climate change signal and uncertainty in projections of ocean wave height
.
Climate Dyn.
,
26
,
109
126
.
Wang
,
X. L.
,
F. W.
Zwiers
, and
V. R.
Swail
,
2004
:
North Atlantic Ocean wave climate scenarios for the twenty-first century
.
J. Climate
,
17
,
2368
2383
.
Yin
,
J. H.
,
2005
:
A consistent poleward shift of the storm tracks in simulations of 21st century climate
.
Geophys. Res. Lett.
,
32
,
L18701
,
doi:10.1029/2005GL023684
.
Zhao
,
M.
,
I. M.
Held
,
S.-J.
Lin
, and
G. A.
Vecchi
,
2009
:
Simulations of global hurricane climatology, interannual variability, and response to global warming using a 50-km resolution GCM
.
J. Climate
,
22
,
6653
6678
.