Academia.eduAcademia.edu

Figure 2 – uploaded by ANNA ZACHARIOUDAKI

See full PDFdownloadDownload figure
independent event. This is interpreted, as in Debernard et al. (2008), as requiring a spacing of at least 48 h between events.

Table 2 independent event. This is interpreted, as in Debernard et al. (2008), as requiring a spacing of at least 48 h between events.

Related Figures (14)

The use of a large computational domain at a high spatial resolution (e.g. N1 at 0.2° resolution) is expected to provide the best possible results as it properly accounts for the swell component but also it well resolves the underlying bathymetry. However, this approach is very computationally demanding for long- term simulations. On the other hand, the alternative approach of using nested domains of progressively finer resolution can reach similar accuracy but at substantially reduced computing time. Nevertheless, achieving the required accuracy at the desired computational time is not straightforward and an optimal composition of domains and resolutions needs to be tested. Here, to design the nested grids, a benchmark simulation at 0.2° resolution was performed over the whole N1 domain, forced by spatially and temporally interpolated ECHAMS wind fields from January 1960. This was compared with the results obtained by combining nested domains with different resolutions and/or extents but the
The use of a large computational domain at a high spatial resolution (e.g. N1 at 0.2° resolution) is expected to provide the best possible results as it properly accounts for the swell component but also it well resolves the underlying bathymetry. However, this approach is very computationally demanding for long- term simulations. On the other hand, the alternative approach of using nested domains of progressively finer resolution can reach similar accuracy but at substantially reduced computing time. Nevertheless, achieving the required accuracy at the desired computational time is not straightforward and an optimal composition of domains and resolutions needs to be tested. Here, to design the nested grids, a benchmark simulation at 0.2° resolution was performed over the whole N1 domain, forced by spatially and temporally interpolated ECHAMS wind fields from January 1960. This was compared with the results obtained by combining nested domains with different resolutions and/or extents but the
Fig. 2. Maximum differences of hourly H, over January 1960 between: a Benchmark simulation (N1 at 0.20.2° resolution) and the set-up of Fig. 1, both forced with ECHAMS winds, (b/, 2) benchmark simulation and that when domain N4 is forced by CLM winds (b/, 2 show percentage and absolute H, differences, respectively) and ¢ the nested simulation over N4 and that when N4 is a stand-alone domain without forcing at its boundaries wave conditions are provided by the benchmark. The high sensitivity of the output to the spatial and temporal resolution of the forcing is evident in this figure. Specifically, it can be seen that the differences range from about 20-25% (0.2—0.8 m) off the Portuguese coast to 90% (0.8—2 m) or greater over the southwest of UK, west of Brittany and north of Spain. Greater difference, >90% and greater than 2 m, are also present in the Irish Sea, and those areas of the English Channel and of the Strait of Gibraltar included within the domain. These may be attributed to the more complex orography and land-sea distribution of these regions. The greatest benefit from RCM downscaling within the English Channel and the Strait of Gibraltar agrees well with the findings of Winterfeldt and Weisse (2009) for the Northeast Atlantic north of about 47° N, and Sotillo et al. (2005) for the Figure 2b shows percentage (bl) and absolute H, differences (b2) between the benchmark and a simulation when domain N4 is forced with CLM—RCM winds instead of the ECHAMS GCM data. In this case the boundary
Fig. 2. Maximum differences of hourly H, over January 1960 between: a Benchmark simulation (N1 at 0.20.2° resolution) and the set-up of Fig. 1, both forced with ECHAMS winds, (b/, 2) benchmark simulation and that when domain N4 is forced by CLM winds (b/, 2 show percentage and absolute H, differences, respectively) and ¢ the nested simulation over N4 and that when N4 is a stand-alone domain without forcing at its boundaries wave conditions are provided by the benchmark. The high sensitivity of the output to the spatial and temporal resolution of the forcing is evident in this figure. Specifically, it can be seen that the differences range from about 20-25% (0.2—0.8 m) off the Portuguese coast to 90% (0.8—2 m) or greater over the southwest of UK, west of Brittany and north of Spain. Greater difference, >90% and greater than 2 m, are also present in the Irish Sea, and those areas of the English Channel and of the Strait of Gibraltar included within the domain. These may be attributed to the more complex orography and land-sea distribution of these regions. The greatest benefit from RCM downscaling within the English Channel and the Strait of Gibraltar agrees well with the findings of Winterfeldt and Weisse (2009) for the Northeast Atlantic north of about 47° N, and Sotillo et al. (2005) for the Figure 2b shows percentage (bl) and absolute H, differences (b2) between the benchmark and a simulation when domain N4 is forced with CLM—RCM winds instead of the ECHAMS GCM data. In this case the boundary
Table 3 Locations of the analysis on extreme wave heights Figure 4 shows relative changes (Eq. 1) in annual mean (top row) and annual 99 percentile (bottom row) of wind speed, V;;, between the ECHAMS control and the respec- tive individual scenarios over the N3 domain. In this figure, as well as in the subsequent figures, the areas of statistically significant changes are shown in light grey shading and the relative percentage change in the samples’ mean is illustrated by contours. The relative changes in the annual mean wind speed are negative over almost the entire domain, i.e. on average, a less energetic future is projected. No change or very small positive changes (<2%) are mostly found for the higher emissions scenarios A1B and A2 and are confined to the Irish Sea, north and west of Ireland. The reduction in future mean Vz increases along a diagonal from the southwest of UK to the southwest of the N3 domain and becomes more pronounced when moving from
Table 3 Locations of the analysis on extreme wave heights Figure 4 shows relative changes (Eq. 1) in annual mean (top row) and annual 99 percentile (bottom row) of wind speed, V;;, between the ECHAMS control and the respec- tive individual scenarios over the N3 domain. In this figure, as well as in the subsequent figures, the areas of statistically significant changes are shown in light grey shading and the relative percentage change in the samples’ mean is illustrated by contours. The relative changes in the annual mean wind speed are negative over almost the entire domain, i.e. on average, a less energetic future is projected. No change or very small positive changes (<2%) are mostly found for the higher emissions scenarios A1B and A2 and are confined to the Irish Sea, north and west of Ireland. The reduction in future mean Vz increases along a diagonal from the southwest of UK to the southwest of the N3 domain and becomes more pronounced when moving from
extreme wave statistics (e.g. Van Heteren and Bruinsma 1981; Muraleedharan et al. 2007; Holthuijsen 2007; Li et al. 2008; Thompson et al. 2009), but here we extract annual maximum values at selected locations to fit a generalized extreme value distribution (GEV) and derive return level plots (RLPs) with confidence intervals for every wave climate. The stationarity of the annual maximum values of H,, an assumption underlying the GEV fitting, was confirmed at the 95% confidence interval. To get a better insight into how extremes might change in the future we complement the RLPs with quantile-quantile-plots (qq- plots) of the 100 most extreme values from the control against the 100 most extreme values from the scenario. Each extreme value is selected so that it represents an
extreme wave statistics (e.g. Van Heteren and Bruinsma 1981; Muraleedharan et al. 2007; Holthuijsen 2007; Li et al. 2008; Thompson et al. 2009), but here we extract annual maximum values at selected locations to fit a generalized extreme value distribution (GEV) and derive return level plots (RLPs) with confidence intervals for every wave climate. The stationarity of the annual maximum values of H,, an assumption underlying the GEV fitting, was confirmed at the 95% confidence interval. To get a better insight into how extremes might change in the future we complement the RLPs with quantile-quantile-plots (qq- plots) of the 100 most extreme values from the control against the 100 most extreme values from the scenario. Each extreme value is selected so that it represents an
Fig. 4 Relative changes (Eq. 1) in the annual mean (op) and annual 99 percentile (bottom) of wind speed, Vj. The changes are between the labelled scenarios and the control. Grey-shaded areas denote changes of statistical significance
Fig. 4 Relative changes (Eq. 1) in the annual mean (op) and annual 99 percentile (bottom) of wind speed, Vj. The changes are between the labelled scenarios and the control. Grey-shaded areas denote changes of statistical significance
changes of H, alone and the sensitivity of them on wind input resolution and model coverage. We also focus specifically on domain N4. observed for the A2 scenario (up to 10° shift clockwise towards the west and 15° anticlockwise towards the south- southeast). Results for scenario A1B are very similar to those of B1. Andrade et al. (2006) (cited in Andrade et al. (2007)), using control and A2 scenario wind fields from a coarse resolution GCM to force a third-generation wave model, have found a clockwise shift (up to about 10°) in future annual mean wave direction over almost the entire domain shown in Fig. 7.
changes of H, alone and the sensitivity of them on wind input resolution and model coverage. We also focus specifically on domain N4. observed for the A2 scenario (up to 10° shift clockwise towards the west and 15° anticlockwise towards the south- southeast). Results for scenario A1B are very similar to those of B1. Andrade et al. (2006) (cited in Andrade et al. (2007)), using control and A2 scenario wind fields from a coarse resolution GCM to force a third-generation wave model, have found a clockwise shift (up to about 10°) in future annual mean wave direction over almost the entire domain shown in Fig. 7.
Fig. 6 Control vs. A2 scenario wind directions of extreme winds at the BB location of Fig. 3
Fig. 6 Control vs. A2 scenario wind directions of extreme winds at the BB location of Fig. 3
Fig. 7 Control annual mean wind direction (/eft) and differences between control and scenarios AS a result, it 1s reasonable to compare the results obtained for the CLM B1 and A1B scenarios with those obtained for the ECHAMS5 A2 scenario (no CLM A2 scenario is available), keeping in mind the discrepancies that are likely to exist over certain regions and seasons, as shown in Fig. 9. In winter, the mean and 99 percentile of H, will increase offshore from France, UK and Ireland for all emissions scenarios. The relative increase is more pronounced for scenario CLM AI1B. It begins along the eastern boundary of the domain at about 45° N (offshore from Bordeaux, France) and is progressively larger towards the north. For mean H,, the increase is 3% west of Brittany, 4-8% south-southwest of UK, up to about 0% in the Irish Sea and statistically significant around UK. For extreme H,, the increase is greater. It ranges from 8% west of Brittany to 11% to the southwest of the UK. The area over which changes are statistically significant is also greater, extending further to the west and south, including now the waters around Brittany. The increase is smaller (up to 7%) for the ECHAMS A2 scenario. Also, in contrast to CLM AIB, the increase of the future 99 percentile of H, for this scenario (up to 5%) is less than that of the mean. Increases are now observed within the entire Bay of Biscay, being around 1—3% from the north of Spain to the west of France. These changes are significant only for the mean H, and over roughly the same area as for the CLM AIB scenario. For scenario CLM BI, the winter increase in H, is even smaller and insignificant. summer, a region of positive changes in mean and 99 percentile of H, found offshore from the north of Portugal for the CLM A1B scenario is displaced southwards for the ECHAMS AIB scenario. This, in turn, alters the area over which the changes are significant. In addition, in spring, positive changes in the Irish Sea found in the 99 percentile of H, for the ECHAMS AIB scenario coincide with negative changes in the CLM AJIB scenario. Not surpris- ingly, the differences between CLM and ECHAMS results are generally the greatest (25%) at the Irish Sea. This is because wind characteristics in this region are expected to vary within finer scales than those resolved by ECHAMS5 due to the proximity of the surrounding land. Furthermore,
Fig. 7 Control annual mean wind direction (/eft) and differences between control and scenarios AS a result, it 1s reasonable to compare the results obtained for the CLM B1 and A1B scenarios with those obtained for the ECHAMS5 A2 scenario (no CLM A2 scenario is available), keeping in mind the discrepancies that are likely to exist over certain regions and seasons, as shown in Fig. 9. In winter, the mean and 99 percentile of H, will increase offshore from France, UK and Ireland for all emissions scenarios. The relative increase is more pronounced for scenario CLM AI1B. It begins along the eastern boundary of the domain at about 45° N (offshore from Bordeaux, France) and is progressively larger towards the north. For mean H,, the increase is 3% west of Brittany, 4-8% south-southwest of UK, up to about 0% in the Irish Sea and statistically significant around UK. For extreme H,, the increase is greater. It ranges from 8% west of Brittany to 11% to the southwest of the UK. The area over which changes are statistically significant is also greater, extending further to the west and south, including now the waters around Brittany. The increase is smaller (up to 7%) for the ECHAMS A2 scenario. Also, in contrast to CLM AIB, the increase of the future 99 percentile of H, for this scenario (up to 5%) is less than that of the mean. Increases are now observed within the entire Bay of Biscay, being around 1—3% from the north of Spain to the west of France. These changes are significant only for the mean H, and over roughly the same area as for the CLM AIB scenario. For scenario CLM BI, the winter increase in H, is even smaller and insignificant. summer, a region of positive changes in mean and 99 percentile of H, found offshore from the north of Portugal for the CLM A1B scenario is displaced southwards for the ECHAMS AIB scenario. This, in turn, alters the area over which the changes are significant. In addition, in spring, positive changes in the Irish Sea found in the 99 percentile of H, for the ECHAMS AIB scenario coincide with negative changes in the CLM AJIB scenario. Not surpris- ingly, the differences between CLM and ECHAMS results are generally the greatest (25%) at the Irish Sea. This is because wind characteristics in this region are expected to vary within finer scales than those resolved by ECHAMS5 due to the proximity of the surrounding land. Furthermore,
Fig. 8 Relative changes (Eq. 1) in the seasonal mean (/eft four plots) and seasonal 99 percentile (right four plots) of H,. The changes are between the labelled scenario and the corresponding control. Grey-shaded areas denote changes of statistical significance
Fig. 8 Relative changes (Eq. 1) in the seasonal mean (/eft four plots) and seasonal 99 percentile (right four plots) of H,. The changes are between the labelled scenario and the corresponding control. Grey-shaded areas denote changes of statistical significance
Fig. 10 Same as Fig. 8
Fig. 10 Same as Fig. 8
At a broad level, this work complements previous work on the potential future changes in the wave climate of the Northeast Atlantic by introducing different climate change scenarios (i.e. different models and/or emissions scenarios) in the analysis. But most importantly, this work shifts the focus of the investigation from the northern seas to the continental shelf of the west Europe from the Irish Sea down to the Strait of Gibraltar. Unlike the North and Baltic Seas, swell is extremely important over this highly exposed domain. Therefore, for the sound investigation of the relative changes between control and scenario wave climates over this region, it is important to investigate changes in the swell compo- nent of the wave spectrum. To properly account for this component, wave generation and propagation over the greater part of the North Atlantic Ocean should be considered. Coarse resolution GCM wind fields may be used to force such a broad domain. However, these models are not ideal for assessing regions characterised by wind fields whose changes take place over finet scales than those resolved by the models. For example. they are expected to perform poorly over regions of complex coastal topography. To overcome this, RCMs of considerably finer resolution may be used _ instead. However, these models are very computationally de- manding, often prohibitively so for long-term predic- tions, and their application over large domains may be constrained. Moreover, in terms of output quality, they may provide no advantage compared with GCMs, ovet oceanic regions characterised predominantly by large- scale processes. Besides, publicly available RCM climate change experiments have boundaries that typically lie not very far away from the land. Consequently, they are more appropriate for investigations of potential future changes in locally generated waves rather than in the total spectrum of surface waves. Here, we have used publicly available combinations of GCM—RCM experi- ments, thereby allowing swell propagating into the west- European shelf seas to be included. At the same time, RCM forcing within the inner WW3 nested domain ensures that fine scale wind changes along the west- European coast are well represented.
At a broad level, this work complements previous work on the potential future changes in the wave climate of the Northeast Atlantic by introducing different climate change scenarios (i.e. different models and/or emissions scenarios) in the analysis. But most importantly, this work shifts the focus of the investigation from the northern seas to the continental shelf of the west Europe from the Irish Sea down to the Strait of Gibraltar. Unlike the North and Baltic Seas, swell is extremely important over this highly exposed domain. Therefore, for the sound investigation of the relative changes between control and scenario wave climates over this region, it is important to investigate changes in the swell compo- nent of the wave spectrum. To properly account for this component, wave generation and propagation over the greater part of the North Atlantic Ocean should be considered. Coarse resolution GCM wind fields may be used to force such a broad domain. However, these models are not ideal for assessing regions characterised by wind fields whose changes take place over finet scales than those resolved by the models. For example. they are expected to perform poorly over regions of complex coastal topography. To overcome this, RCMs of considerably finer resolution may be used _ instead. However, these models are very computationally de- manding, often prohibitively so for long-term predic- tions, and their application over large domains may be constrained. Moreover, in terms of output quality, they may provide no advantage compared with GCMs, ovet oceanic regions characterised predominantly by large- scale processes. Besides, publicly available RCM climate change experiments have boundaries that typically lie not very far away from the land. Consequently, they are more appropriate for investigations of potential future changes in locally generated waves rather than in the total spectrum of surface waves. Here, we have used publicly available combinations of GCM—RCM experi- ments, thereby allowing swell propagating into the west- European shelf seas to be included. At the same time, RCM forcing within the inner WW3 nested domain ensures that fine scale wind changes along the west- European coast are well represented.
Related topics:
GeologyOceanographyWestern EuropeClimate ChangeWave EnergyWave ClimateOcean DynamicsWind Wave
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved