Long-term hindcast for Wave Hub, Cornwall Validation and analysis of modelled results J. van Nieuwkoop-McCall, H. Smith, Prof. G. Smith, L. Johanning University of Exeter March 2012 Conducted for the TSB project: ‘Accelerated Development of the PB500 for Deployment at Wave Hub’ Hindcast Wave Hub Table of Contents 1 2 3 4 5 Introduction .................................................................................................................................... 3 1.1 Problem description................................................................................................................ 3 1.2 Project objectives.................................................................................................................... 3 Wave model .................................................................................................................................... 4 2.1 Introduction ............................................................................................................................ 4 2.2 Model description ................................................................................................................... 4 2.3 Model set-up ........................................................................................................................... 4 Validation with buoy data ............................................................................................................... 7 3.1 Introduction ............................................................................................................................ 7 3.2 Buoy data ................................................................................................................................ 7 3.3 Validation of hindcast data ..................................................................................................... 7 Wave resources............................................................................................................................. 17 4.1 Method ................................................................................................................................. 17 4.2 Data analysis location 1 ........................................................................................................ 19 4.3 Spatial variation in Wave Hub area....................................................................................... 30 Conclusions ................................................................................................................................... 34 Literature .............................................................................................................................................. 36 Appendix A Analysis corrected hindcast data ....................................................................................... 37 Appendix B Relation to previous studies .............................................................................................. 42 Appendix C Directional variation .......................................................................................................... 44 University of Exeter March 2012 2 Hindcast Wave Hub 1 Introduction 1.1 Problem description There are numerous measured and modelled datasets available within 20 km of the Wave Hub site covering different periods between 1989 and 2012. However, there is no single dataset covering the entire period and the measured datasets cover short periods of up to a year, consequently there is limited temporal overlap of the datasets. The different instruments and models used to produce each dataset, and the natural inter-annual variability of the site, make the assessment of wave resources and extreme wave heights more difficult and less reliable. Additionally, the datasets available cover a significant area, with differing bathymetry, so a variation in wave climate is to be expected. Due to change in the site location during development there is only one measured dataset at the final Wave Hub development area and no information about the variation inside this area. 1.2 Project objectives Reanalysis wave and wind information from large scale ocean models has been used as input for a high resolution regional SWAN model. In this way, a dataset for the Wave Hub development area with a 23-year period and more spatial information is created. The objective of this report is to present the analysis of this dataset. This entails: • To present the validation of this dataset against buoy data; • To present a wave resource assessment for the Wave Hub development area; University of Exeter March 2012 3 Hindcast Wave Hub 2 Wave model 2.1 Introduction This chapter describes the SWAN model that was used for the 23 year hindcast. First a description of the model physics is given in section 2.2. Subsequently, various model choices are discussed in section 2.3. Detailed information about the model set-up can be found in ‘A Wave model for the Cornwall coast’, August 2011. 2.2 Model description The spectral wave model SWAN (Booij et al. 1999) is used for the hindcast of the wave conditions over a 23 year period. The third generation mode for wind input, quadruplet interactions and whitecapping is used. SWAN computes the evolution of wave action density N using the action balance equation: With The terms on the left-hand side of equation 1 represent, respectively, the change in wave action over time, the propagation of wave action in geographical space, depth and current-induced refraction (with propagation velocity cθ in directional space θ) and the shifting of the relative radian frequency σ due to variations in mean current and depth (with the propagation velocity cσ). The right-hand side of equation 1 represents processes that generate, dissipate or redistribute wave energy, given by equation 2. These include the deep water processes of wind input (Sin), whitecapping dissipation (Swc), quadruplet nonlinear interaction (Snl4), and the shallow water processes of bottom friction dissipation (Sbot), depth-induced breaking (Sbr) and triad wave-wave interactions (Snl3). SWAN default parameterizations for wind, nonlinear quadruplet wave interactions, bottom friction dissipation, depth-induced breaking and triad wave-wave interactions are used. For whitecapping the formulation proposed by Rogers (Rogers 2003) is used. The weighting of the relative wavenumber term in the whitecapping formulation is altered (n = 2 instead of n = 1). By increasing n, the dissipation is reduced at lower frequencies and increased at higher frequencies compared to the default SWAN settings. 2.3 Model set-up The model covers the area of 4 to 7 degrees west and 49 to 51 degrees north. The model grid comprises the whole Cornwall coast and part of the Devon coast. Furthermore, the Isles of Scilly are included. A grid resolution of 1 km x 1 km is used for the model domain. Nests with smaller grid resolutions, down to 100 m x 100 m are used for nearshore areas of interest. The nest for the Wave Hub area is shown in figure 2.1. The nests for the area around Falmouth and for the Isles of Scilly are not shown in this figure. University of Exeter March 2012 4 Hindcast Wave Hub The bathymetry for the model is constructed from the bathymetry obtained from Marine DigiMap. Figure 2.1 shows the bathymetry in a selection of the model domain. Output is generated at various points along the model grid. The output points relevant to this study are shown in figure 2.1. The green squares indicate the validation points for validation against buoy data. Furthermore, various integral parameters are written to an output file for the entire Wave Hub area. For one point in this area spectral information is also given. PRIMaRE buoys Wave Hub Nest Wave Hub Perranporth Looe Bay Penzance Porthleven Figure 2.1: Bathymetry and output locations The global model hindcast of ECMWF (European Centre for Medium-Range Weather Forecasting), utilising the wave model WAM, provides the wave and wind input. The grid resolution of this model is 1.5 x 1.5 degrees and is therefore very coarse compared to the SWAN model resolution, see figure 2.2. The time resolution of the ECMWF global model is 6 hours. ECMWF provides the integral wave parameters Hm0, Tp, wave direction and wind velocity and wind direction. The wave parameters were interpolated to the SWAN model corner points using a matlab routine. The ECMWF wind data was interpolated to the SWAN model grid by SWAN itself. No water level variations and currents are taken into account. University of Exeter March 2012 5 Hindcast Wave Hub th Figure 2.2: ECMWF model grid and every 10 grid line of the SWAN grid (black dots), wave boundary input points (red circles) The period between 1 January 1989 to 1 November 2011 is hindcasted with a time step of 60 minutes. The spectral direction covers the full circle and a bin of 8° is chosen for the SWAN computations to provide a sufficiently fine resolution for swell waves. For the spectral resolution a range from 0.03 – 0.6 Hz is used, which results in 31 bins. SWAN version 40.81 has been used for the model computations. Use was made of a Beowulf cluster with 512 nodes. The hindcast has been split into periods of 2.5 years and was run in a period of 2 weeks. University of Exeter March 2012 6 Hindcast Wave Hub 3 Validation with buoy data 3.1 Introduction In this chapter the modelled wave data for specific sites are validated against buoy measurements. In section 3.2 an overview of the available measurements is given. In section 3.3 the validation against the buoy data is presented in various ways and error correlations with different integral parameters are analysed. 3.2 Buoy data The model set-up has been validated against buoy data from 6 different buoys over the time periods where data is available. The buoys include two of the PRIMaRE wave buoys situated near the Wave Hub and four Coastal Channel Observatory buoys: Perranporth, Penzance, Porthleven and Looe Bay, see figure 2.1. Prior to using any of the buoy measurements, each dataset was processed and subjected to a basic quality control test to ascertain the validity and reliability of the data. First outliers and data outside the physically realistic limits were removed. Subsequently, the wave parameters were averaged using a 3 hour moving average. The 3 hour time window was chosen to consistently compare data against the model outputs, corresponding to the temporal resolution of wind forcing and wave boundary input. Finally the data were interpolated on regularly spaced time vectors in order to be able to compare them with the modelled data. Table 3.1 gives details on location, depth and measurement periods for all buoy locations. Table 3.3.1: Wave buoy location details buoy name PRIMaRE A PRIMaRE D Perranporth Penzance Porthleven Looe Bay location 5.67°W;50.31°N 5.67°W;50.31°N 5.18°W;50.35°N 5.50°W;50.11°N 5.31°W;50.06°N 4.41°W;50.34°N depth 37m 37m 15m 10m 12m 12m period 02/2010-11/2010 10/2009-11/2010 11/2006-10/2011 03/2007-10/2011 10/2011-10/2011 05/2009-10/2011 08/2011-10/2011 08/2011-10/2011 3.3 Validation of hindcast data Statistical comparison The statistical results of the comparison between the measurement data and model results are shown in table 3.2. For each dataset the bias, the relative bias, the root mean square error (RMSE) and the scatter index (SI, which is the root mean square error normalized by the mean of observed values) have been calculated. It can be seen that the significant wave height is generally underestimated by a few centimetres by SWAN. Comparisons between model and measurement data reveal a root mean square error in the University of Exeter March 2012 7 Hindcast Wave Hub order of 0.3 m. This means that the scatter index is approximately 20% for the northerly locations. Penzance and Looe Bay have a relatively high scatter index due to the fact that average wave height is lower at these locations. These buoys are relatively close to the shore and in shallow water and therefore errors are more likely because of wind input errors on the land-sea boundary, shallow water processes and wave-current interactions. The wave period is also underestimated. In general, less underestimation of the wave period can be seen for south coast locations than for north coast locations. The scatter index varies between 20 and 30% depending on the location. Table 3.3.2: Wave buoy validation statistics buoy name N Bias PRIMaRE A PRIMaRE D Perranporth Penzance Porthleven Looe Bay 5006 7050 40003 38562 363 19159 -0.08m -0.05m -0.06m -0.09m -0.01m 0.05m Hm0 R. bias RMSE SI -4% -3% -4% -15% -1% 6% 0.29m 0.32m 0.28m 0.19m 0.27m 0.20m 17% 17% 19% 32% 19% 24% N 5006 7050 40003 38562 363 19159 bias -1.4s -1.2s -1.6s -1.1s -0.5s -0.5s Tm-1,0 R. bias RMSE -18% -15% -20% -17% -7% -9% 1.9s 1.6s 2.2s 2.2s 1.0s 1.7s SI 24% 20% 27% 35% 13% 28% Timeseries Figures 3.1 and 3.2 show the comparison between the measured and computed datasets for PRIMaRE wave buoy D and Looe Bay. The figures illustrate that the performance of the model compared to the measurements is best for medium range wave heights between 0.5 and 3 meters. Above and below these levels the wave height is often underestimated by the model. It can be seen that the wave period is almost consistently underestimated by the model at the PRIMaRE wave buoy D. The correspondence between measured and computed wave period is better for Looe Bay. The computed wave direction compares in general relatively well to the measurements for both locations. In addition, the lower two panels in figure 3.2 show the wind direction and velocity input compared to measured wind at the Looe Bay wave buoy. It can be seen that the ECMWF wind has a very smooth line, whereas the measured wind is very erratic. Moreover, the peaks in wind velocity are often underestimated by the ECMWF data whereas the lower wind velocities are often overestimated. An error in the wind field has large consequences for the computed significant wave height, as the wave height is strongly correlated to the wind speed. University of Exeter March 2012 8 Hindcast Wave Hub Figure 3.1: Timeseries comparison of computed and measured data PRIMaRE wave buoy D University of Exeter March 2012 9 Hindcast Wave Hub Figure 3.2: Timeseries comparison of computed and measured data Looe Bay University of Exeter March 2012 10 Hindcast Wave Hub Error dependence on wave height and period By looking at the error in more detail dependencies between the error and some parameters can be found. Figures 3.3 to 3.6 present the bias and root mean square error of Hm0 and Tm-1,0 binned by Hm0 and Tm-1,0. The number of observations in each bin is shown in figure 3.7. Figure 3.3 shows that the largest negative bias (larger than 0.5 meters) occurs either for very steep waves or long waves. The largest positive bias occurs for long waves with wave heights smaller than 1 meter. The smallest bias can be found for wave heights between 0 and 3 meters and wave periods between 4 and 10 seconds. The probability of occurrence in this range is also the highest, see figure 3.7. The root mean square error, shown in figure 3.4, demonstrates the same trend as for the bias. All bias on Tm-1,0 is negative, see figure 3.5. A clear correlation can be seen between the bias and the steepness of the waves. The smallest bias is found for the steeper waves, whereas the largest bias (circa 5 seconds) can be found for long small waves. The root mean square error, shown in figure 3.6, demonstrates a similar trend as the bias. Figure 3.3: Bias Hm0 [m] PRIMaRE wave buoy D Figure 3.4: Root mean square Hm0 [m] PRIMaRE wave buoy D University of Exeter March 2012 11 Hindcast Wave Hub Figure 3.5: Bias Tm-1,0 [s] PRIMaRE wave buoy D Figure 3.6: Root mean square Tm-1,0 [m] PRIMaRE wave buoy D Figure 3.7: Number of observations PRIMaRE wave buoy D Figures 3.8 and 3.9 give the same information as in the previous figure, but now presented in scatter plots. University of Exeter March 2012 12 Hindcast Wave Hub Figure 3.8: Scatter plot Hm0 PRIMaRE wave buoy D, for different values of Tm-1,0 [s], see colour bar Figure 3.9: Scatter plot Tm-1,0 PRIMaRE wave buoy D, for different values of Hm0 [m], see colour bar University of Exeter March 2012 13 Hindcast Wave Hub Error dependence on wave direction Figure 3.10 and 3.11 show respectively the bias and scatter of Hm0 and Tm-1,0 per wave direction. Most waves come from a westerly direction. It can be seen that most wave height errors are less than 1 meter. The positive and negative errors cancel each other out and therefore the bias is very small. As the wave heights are more often under- than overestimated, the bias is in most directions slightly negative. The largest errors are found for waves from south-westerly and north-easterly directions. Figure 3.10: Bias and scatter Hm0 versus wave direction PRIMaRE wave buoy D For Tm-1,0 the largest bias and root mean square errors come from waves with southerly directions. Errors up to 4 seconds are seen. However, most errors are in the order of 1 second. University of Exeter March 2012 14 Hindcast Wave Hub Figure 3.11: Bias and scatter Tm-1,0 versus wave direction PRIMaRE wave buoy D Error dependence on season Figures 3.12 and 3.13 show the variation in time of the bias in Hm0 and Tm-1,0. For this analysis the figures for Perranporth are shown, as Perranporth has the longest measured dataset and is also located on the north coast of Cornwall. It can be seen that there is some monthly variation in the errors. The larger errors are more likely to occur in the winter months than in the summer months. Figure 3.12 also shows the mean annual variation of the bias. Each year of the dataset was divided into spring (March, April, May), summer (June, July, August), autumn (September, October, November) and winter (December, January, February) datasets, and the mean power calculated. It can be seen that there is little annual variation of the bias. University of Exeter March 2012 15 Hindcast Wave Hub 2006: -0.05 m 2007: -0.07 m 2008: -0.07 m 2009: -0.07 m 2010: -0.03 m 2011: -0.07m 2006: -2.9 s 2009: -1.6 s 2007: -1.5 s 2008: -1.5 s 2010: -1.7 s 2011: -1.7 s Figure 3.12: Montly/annual variation of the errors in Hm0 and Tm-1,0 Perranporth Figure 3.13: Mean monthly error for Hm0 and Tm-1,0 Perranporth University of Exeter March 2012 16 Hindcast Wave Hub 4 Wave resources 4.1 Method In this chapter the wave resources in the Wave Hub area are assessed. Figure 4.1 shows the area and its bathymetry. The area is divided into four berth areas, as indicated in figure 4.1. 23-year time series of significant wave height, wave period and direction are available for all points in this area. Spectral output is available only at location 1, located in berth 3. centre Figure 4.1: Wave Hub area overview University of Exeter March 2012 17 Hindcast Wave Hub As spectral information is available at location 1, this point is used for a detailed assessment of the available resources. Section 4.2 presents the 23-year time series, the joint probability of wave height and wave period and the variability of the wave resource for time, direction and spectra. The power, P(f), is calculated using the spectral output, S(f), and the omni-directional power equation: Where the group velocity is And f is the frequency, ρ is the density of water (assumed to be 1025 kg/m3), g is gravitational constant, k is the wave number ( and d is the water depth. Subsequently, the spatial variation in the Wave Hub area is presented in section 4.3. To present the variation between the different berths and inside the berths, location A-D and M are used, see figure 4.1. The power for this area is calculated in a different way than has been done for location 1, as only the significant wave height and wave period are available. In this case an approximation is used: The spectral period Tm-1,0 is used to calculate the wave length that is needed to calculate cg. There are small differences in power between the two formulas. Figure 4.2 shows a scatterplot for location 1, for which the power is calculated with both equations. It can be seen that the power calculated with the integral parameters is approximately 2% smaller than the power calculated with the spectra. Therefore, the spatial power results, presented in section 4.3, are corrected by adding 2% power to the calculated wave power. 1 L is the wave length and is calculated iteratively with a matlab routine University of Exeter March 2012 18 Hindcast Wave Hub Figure 4.2: Comparison methods of power calculation It should be noted that the hindcast data used in this analysis hasn’t been corrected for any errors that were found in the validation shown in chapter 3. Appendix A illustrates the effect on the shown mean wave power, when data are corrected. The data are corrected using the regression formulas for PRIMaRE wave buoy D, see figure 1 and 2 in appendix A. As the spatial correlation between PRIMaRE wave buoy D and Wave Hub is high (shown in appendix A), it is assumed that the necessary error correction for Wave Hub is the same as for PRIMaRE wave buoy D. However, there could be some difference in the errors, as the waves are usually higher at PRIMaRE wave buoy D than at Wave Hub. The difference between the uncorrected and corrected wave power is large, approximately 7 kW/m for the mean wave power at location 1. This is due to the relatively large error in the wave period. 4.2 Data analysis location 1 In this section the wave resources are presented by means of: • time series of wave power, Hm0 and Tm-1,0 • Joint probability plot between Hm0 and Tm-1,0 • Power variation in time • Directional variation of wave power • Spectral variation Time series In figure 4.3 the 23-year time series for wave power, Hm0 and Tm-1,0 at location 1 are shown. The mean value for each parameter is given in the figure. At location 1 the mean power is 20 kW/m. This agrees with values found in the wave resource assessment that was conducted with the available measurement data, see ‘Assessment of the Wave and Current Resource at the Wave Hub Site’, January 2011. University of Exeter March 2012 19 Hindcast Wave Hub Figure 4.3: Timeseries power, Hm0 and Tm-1,0 at location 1, not corrected for hindcast error Joint probability Figure 4.4 shows the joint probability for Hm0 and Tm-1,0 at location 1. Every square shows the occurrence of the combination of Hm0 and Tm-1,0 in percent. It can be seen that the most occurring waves have a wave height of approximately 1-2 meters and a wave period of 4-7 seconds. University of Exeter March 2012 20 Hindcast Wave Hub Figure 4.4: Joint probability Hm0 and Tm-1,0 at location 1, not corrected for hindcast error Variation in time Figure 4.5, 4.6 and 4.7 show the annual and monthly variation in wave power, Hm0 and Tm-1,0 at location 1. In addition figure 4.8 presents the seasonal power variation, for which each year of the dataset was divided into spring (March, April, May), summer (June, July, August), autumn (September, October, November) and winter (December, January, February) datasets, and the mean power was calculated. The variation in power levels over each year is substantial, with average monthly power levels over 50 kW/m in some winter months, compared with frequent monthly averages below 10 kW/m in the summer months. The figure in the mid panel suggests an overall trend of decreasing power levels. The values agree with the values found for MetOffice location ‘U04’ in the wave resource assessment that was conducted with the available measurement data, see ‘Assessment of the Wave and Current Resource at the Wave Hub Site’, January 2011. This is shown in appendix B Averaged over the whole dataset, it can be seen that December, January and February are the most energetic months with mean power levels up to 40 kW/m. Wave heights vary from 1.3 meter in summer to 2.8 meter in winter. Wave periods vary from 6 seconds in summer to 9 seconds in winter. Finally, figure 4.9 shows the annual cumulative probability distribution of the wave power. The curve for the average distribution in the upper pannel of figure 4.9 shows that 95% of sea states have a power of 100kW/m or less. Furthermore it can be seen in the lower pannel of figure 4.9 that the University of Exeter March 2012 21 Hindcast Wave Hub proportion of sea states with a power of less than 15kW/m varies from approximately 56% for the most energetic year to 76% for the least. Figure 4.5: Montly/annual variation of power at location 1, not corrected for hindcast error University of Exeter March 2012 22 Hindcast Wave Hub Figure 4.6: Montly/annual variation of Hm0 at location 1, the bars are showing the standard deviation in the data, not corrected for hindcast error University of Exeter March 2012 23 Hindcast Wave Hub Figure 4.7: Montly/annual variation of Tm-1,0 at location 1, the bars are showing the standard deviation in the data University of Exeter March 2012 24 Hindcast Wave Hub Figure 4.8: Seasonal variation of power at location 1, not corrected for hindcast error Figure 4.9: Annual wave power cumulative probability distribution for location 1, upper panel: power up to 500 kW/m, lower panel: power up to 50kW/m, including the mean distribution (not corrected for hindcast error) University of Exeter March 2012 25 Hindcast Wave Hub Directional variation Figure 4.10 shows the directional wave power variation. The left-hand figure shows the occurrence of wave power smaller than 50 kW/m and the right-hand figure shows the occurrence of wave power larger than 50 kW/m. It can be seen that most wave energy comes from the west and in addition the most energetic waves come from the west. Figure 4.10: Wave rose for wave power [kW/m] at location 1, left-hand figure for power < 50 kW/m, right-hand figure for power > 50 kW/m (not corrected for hindcast error) Figure 4.11 shows the mean wave power binned by wave direction. The mean wave power is largest for wave directions between 240 and 300° North, with respectively 26 kW/m and 24 kW/m for the 240-270 and 270-300° bins. The joint probability between Hm0 and Tm-1,0 is shown for these bins in figure 4.12 and 4.13. Figures 4.12 and 4.13 show that the most occurring waves for the westerly wave directions are larger than when the joint probability for all directions is shown, as in 4.4. The joint probability diagrams for the remaining directions are shown in appendix C. University of Exeter March 2012 26 Hindcast Wave Hub Figure 4.11: Mean wave power [kW/m] binned by wave direction [°N] in 30 degree bins (not corrected for hindcast error) Figure 4.12: Joint probability Hm0 and Tm-1,0 for wave direction bin 240-270° (not corrected for hindcast error) University of Exeter March 2012 27 Hindcast Wave Hub Figure 4.13: Joint probability Hm0 and Tm-1,0 for wave direction bin 270-300° (not corrected for hindcast error) Spectral variation Figure 4.14 presents spectral data in a scatter diagram format. For each bin of the scatter diagram, all spectra with parameters that fall within the bin limits are plotted and the mean spectrum for the bin is calculated. The figure indicates that bi-model sea states occur. University of Exeter March 2012 28 Hindcast Wave Hub 2 Figure 4.14: spectral variation, with frequency [Hz] on the x-axis and energy density [m /Hz] on the y-axis (not corrected for hindcast error) University of Exeter March 2012 29 Hindcast Wave Hub 4.3 Spatial variation in Wave Hub area The spatial variation in mean wave power is shown in figure 4.15. Note that the wave power is calculated with integral parameters instead of spectra and subsequently corrected, see section 4.1. A difference of circa 3 kW/m is present in the area. Figure 3.16 shows the mean wave power relative to location 1. The results of section 4.2 can be translated to other locations in the Wave Hub area with the help of this figure. Finally, figure 3.17 gives a comparison of mean wave power between the berths and inside the berths. The mean power for each berth is shown in the figure. It can be seen that berth 4 is the most energetic, while berth 1 is the least energetic. However, mean wave power also varies inside the berths, with differences up to 0.5 kW/m. University of Exeter March 2012 30 Hindcast Wave Hub centre Figure 4.15: spatial variation mean wave power for Wave Hub area with bathymetry contours (not corrected for hindcast error) University of Exeter March 2012 31 Hindcast Wave Hub centre Figure 4.16: spatial variation mean annual wave power relative to power at location 1 for Wave Hub area with bathymetry contours (not corrected for hindcast error) University of Exeter March 2012 32 Hindcast Wave Hub Figure 4.17: Power variation between berths and inside berths and mean wave power per berth (not corrected for hindcast error) University of Exeter March 2012 33 Hindcast Wave Hub 5 Conclusions Hindcast validation • The agreement between computed and measured significant wave height is very good, in particular for operational wave conditions. • However, extreme wave heights are often underestimated by the wave model. This means that the analysis of extreme waves becomes more uncertain. By means of bias correction (see appendix A) and random error correction (not done for this study) the hindcast errors can be taken into account. • In addition, the wave period is often underestimated by the wave model. This has large consequences for the calculation of wave power. By means of bias correction the bias can be taken into account (see appendix A). Wave resources • The average wave power at Wave Hub is approximately 20 kW/m. When the bias on the hindcast results is corrected the average wave power is 7 kW/m higher. • The wave power varies with time, wave direction and space: o As can be expected, the wave power is highest in the winter months with mean yields of 40 kW/m and lowest in summer with mean yields of 10 kW/m. o Westerly waves have the largest wave power, with a mean of approximately 25 kW/m. o The mean wave power at Wave Hub varies with approximately 3 kW/m, depending on location. The variation is mainly due to bathymetry variation. Discussion The quality of the hindcast conducted for this study is largely depending on the resolution and quality of the wave boundary and wind input. In this study the freely available ECMWF2 global hindcast was used. As was shown in figure 2.2, the spatial and time resolutions of the ECMWF hindcast are very coarse. As a result extreme wind and wave occurrences are averaged out in time and space and this can explain why the SWAN hindcast conducted for this study tends to underestimate extreme waves. Ocean wave reanalysis data for the British Isles at a higher resolution (than the ECMWF hindcast) is available and can be purchased at various weather institutes, for example the Met Office. In case higher accuracy hindcast data is required, it is recommended to rerun the hindcast with the higher resolution input data. In addition, the wave boundary conditions for the SWAN hindcast were based on the wave parameters Hm0, Tm01 and wave direction for which a JONSWAP spectral shape was assumed. Spectral boundary input or partitioned wave boundary input can improve the quality of the hindcast as the occurrence of multi modal sea states on the wave boundary can be taken into account. 2 European Centre for Medium-Range Weather Forecasting University of Exeter March 2012 34 Hindcast Wave Hub Ocean wave reanalysis data for the British Isles with partitioned wave parameters is available and can be purchased at various weather institutes, for example the Met Office. In case higher accuracy hindcast data is required, it is recommended to rerun the hindcast with partitioned wave input. Finally, appendix A presents a correction on the hindcast data. The presentation of the corrected data is meant to show the effect of the hindcast errors on wave power. However, the correction was based on a simple regression technique. A more extensive study of the correction terms would be advisable, before the corrected hindcast values can be used. University of Exeter March 2012 35 Hindcast Wave Hub Literature Bathymetry, Marine Digimap Booij, R.C. Ris, N., Holthuijsen, L.H. (1999). A third-generation wave model for coastal regions, Part I, 7649-7666. Model description and validation. Brooker, D.C., Cole, G.K., McConochie J.D. (2004). The influence of hindcast modeling uncertainty on OMAE2004-51161 the prediction of high return period wave conditions. Nieuwkoop-McCall, J.C.C. van (2011). A wave model for the Cornwall coast: Phase 1: Wave model set-up and sensitivity study. University of Exeter. Rogers, W.E., Hwang, P. A., & Wang, D. W. (2003). Investigation of Wave Growth and Decay in the SWAN Model: Three Regional-Scale Applications. , !!(2), 366-389. Smith, H.M., Haverson, D., Smith, G.H., Cornish, C.S., Baldock, D. (2011). Assessment of the Wave and Current Resource at the Wave Hub Site. University of Exter. . WAFO group (2011). WAFO – a Matlab toolbox for analysis of random waves and loads. Lund $ % & ' University. " # University of Exeter March 2012 36 Hindcast Wave Hub Appendix A Analysis corrected hindcast data Regression of the hindcast error Separate corrections for Hm0 and Tm-1,0 are determined using regression techniques. Figures 1 and 2 show the error in Hm0 or Tm-1,0 relative to the computed value of Hm0 or Tm-1,0 for PRIMaRE buoy D. In addition the bin-averaged bias is shown and a regression line is drawn. The bias in Hm0 is modelled as quadratic, the bias in Tm-1,0 is modelled as constant. The trend lines of the errors in the computed values of Hm0 and Tm-1,0 can be used to correct the hindcast values at the PRIMaRE wave buoy location. Whether the correction is also valid at the Wave Hub location, depends on the spatial correlation between the locations. Figure 3 shows that the correlation between these locations is high. x2 x + Figure 1: Regression BIAS Hm0 PRIMaRE wave buoy D University of Exeter March 2012 37 Hindcast Wave Hub Figure 2: Regression BIAS Tm-1,0 PRIMaRE wave buoy D Figure 3: Spatial correlation between Wave Hub location 1 (ADCP) and PRIMaRE wave buoy D for Hm0 and Tm-1,0 Wave resources Figures 4 and 5 show the comparison between the corrected and uncorrected hindcast datasets. Figure 5 also shows the mean power, Hm0 and Tm-1,0 for the corrected dataset and the uncorrected dataset (in brackets). Furthermore, figure 6 shows the joint probability for Hm0 and Tm-1,0 at location 1. Every square shows the occurrence of the combination of Hm0 and Tm-1,0 in percent. University of Exeter March 2012 38 Hindcast Wave Hub Figure 4: Comparison Corrected (red) and uncorrected (blue) timeseries power, Hm0 and Tm-1,0 at location 1. University of Exeter March 2012 39 Hindcast Wave Hub Figure 5: Comparison monthly/annual variation of power at location 1 for corrected and uncorrected hindcast datasets University of Exeter March 2012 40 Hindcast Wave Hub Figure 6: Joint probability Hm0 and Tm-1,0 at location 1, corrected for hindcast error University of Exeter March 2012 41 Hindcast Wave Hub Appendix B Relation to previous studies Figure 2 compares the hindcast values with values calculated in the wave resource assessment that was conducted with the available measurement/model data, see ‘Assessment of the Wave and Current Resource at the Wave Hub Site’, January 2011. Both the corrected and the uncorrected hindcast results are shown in the figure. The locations of the aforementioned report are shown in figure 1. The ‘PWB’ are the PRIMaRE wave buoys, Met Office (indicated as MO_WW3 in figure 6) is the most recent Met Office wave model utilising WaveWatchIII, U04 is a grid point from the previous Met Office wave model and SWM is the Sea Watch Mini buoy. Figure 1: Measurement/ model locations ‘Assessment of the Wave and Current Resource at the Wave Hub Site’, January 2011 University of Exeter March 2012 42 Hindcast Wave Hub Figure 2: Comparison annual mean wave power with ‘Assessment of the Wave and Current Resource at the Wave Hub Site’, January 2011 University of Exeter March 2012 43 Hindcast Wave Hub Appendix C Directional variation Figures 1 to 12 show the joint probability diagrams for 30 degree directional bins. Figure 1: Joint probability Hm0 and Tm-1,0 at location 1 for 0-30°, for the uncorrected hindcast data Figure 2: Joint probability Hm0 and Tm-1,0 at location 1 for 30-60°, for the uncorrected hindcast data University of Exeter March 2012 44 Hindcast Wave Hub Figure 3: Joint probability Hm0 and Tm-1,0 at location 1 for 60-90°, for the uncorrected hindcast data Figure 4: Joint probability Hm0 and Tm-1,0 at location 1 for 90-120°, for the uncorrected hindcast data Figure 5: Joint probability Hm0 and Tm-1,0 at location 1 for 120-150°, for the uncorrected hindcast data University of Exeter March 2012 45 Hindcast Wave Hub Figure 6: Joint probability Hm0 and Tm-1,0 at location 1 for 150-180°, for the uncorrected hindcast data Figure 7: Joint probability Hm0 and Tm-1,0 at location 1 for 180-210°, for the uncorrected hindcast data University of Exeter March 2012 46 Hindcast Wave Hub Figure 8: Joint probability Hm0 and Tm-1,0 at location 1 for 210-240°, for the uncorrected hindcast data Figure 9: Joint probability Hm0 and Tm-1,0 at location 1 for 240-270°, for the uncorrected hindcast data University of Exeter March 2012 47 Hindcast Wave Hub Figure 10: Joint probability Hm0 and Tm-1,0 at location 1 for 270-300°, for the uncorrected hindcast data Figure 11: Joint probability Hm0 and Tm-1,0 at location 1 for 300-330°, for the uncorrected hindcast data University of Exeter March 2012 48 Hindcast Wave Hub Figure 12: Joint probability Hm0 and Tm-1,0 at location 1 for 330-360°, for the uncorrected hindcast data University of Exeter March 2012 49
© Copyright 2026 Paperzz