Partitionnement de spectres et statistiques sur l’acuité () des systèmes de vagues observés sur le site d’expérimentation EMR SEM-REV J-Baptiste SAULNIER Ecole Centrale de Nantes, LHEEA (France) Ile de Berder – 05/07/2013 (Comm. OMAE2013-11470) Introduction • Marine Renewable Energy needs fine characterisation of environmental parameters, and sea state ones in particular (design, survivability, commissioning/decommissioning…) Wave spectra from in situ measurements (wave buoys, ADCPs…) • Statistics of Hs, Tp… and spectral peakedness (bandwidth/narrowness) required in particular for simulating extreme sea states (fatigue and survivability) using e.g. JONSWAP spectra effect of wave groups • A sea state is the combination of several independent wave systems (swell(s) and wind-sea) Sea state partitioning for considering wave systems individually and the peakedness characterising each system -I- WAVE SYSTEM IDENTIFICATION AND MODELLING Goal Complex sea state Individual components ‘i’ Simple (swells, wind-sea) methodology Hm0 i, Hm0,i Tp, T02… Tp,i, T02,i… θp, θm… θp,i, θm,i… γ (shape)...??? γi … … More relevant physically Not relevant if more than 1 peak in the spectrum (simulations, design…) STEP 1 Partitioning of the discrete spectral matrix Ŝ(fi,θj) (source: dir. wave buoy, ADCP, array of sensors… or numerical models) Simplified watershed technique [Hanson et Phillips, 2001] SWELL Ŝ(f,θ) WIND-SEA Bimodal directional spectrum estimated from buoy measurements (with smoothing) Watershed partitioning algorithm = path of steepest ascent technique (e.g. Hanson & Phillips, 2001) STEP 2 fs = separation frequency Partitions grouping: Partitions with fp > fs Partitions with fp <= fs & Hm0 >= Hmin Partitions with fp <= fs & Hm0 < Hmin PARTITION 1 = WIND-SEA PARTITION = SWELL j GROUPED WITH SWELL WITH CLOSEST fp STEP 3 P partitions identified (1 wind-sea + (P-1) swells) Fitting of analytical shapes (least-squares minimisation) JONSWAP for Sj (f) (∫partition_j(f,θ) dθ) [Hasselmann et al., 1973] Cos^2s for Dj (θ) (∫partition_j(f,θ) df) [e.g. Mitsuyasu et al., 1975] JONSWAP Set of parameters for each identified wave system Cos^2s =1 = 3.3 =7 10 Frequency fitting shapes… 8 JONSWAP spectra (gamma = 1, 3.3, 7) 6 4 2 0 0.05 0.1 0.15 0.2 0.25 Frequency[Hz] Cos^2s function (s = 2, 10, 50) … Directional fitting shapes (not crucial here) Dir. spreading function[1/rad] Wave spectral density[m2/Hz] 12 0.3 2 s=2 s = 10 s = 50 1.5 1 0.5 0 0 45 90 135 180 225 270 315 360 Direction[°] STEP 4 Correction of mutual influences [Kerbiriou et al., 2007] Correction of Hm0,j so as to minimise the area difference of the total reconstructed density S(f) with target Ŝ(f) SEM-REV - 05-Jan-2011 07:00:00 1.6 Original Not corrected fit 1.4 Corrected fit S1 f1 S2 f1 Sp f1 S f S f S f 1 2 2 2 p 2 Σ S f S f S f 2 n p n 1 n 2 /Hz] 1.2 1 Spectral density[m Σ C Sˆ 0.8 0.6 e ~ 25% 0.4 0.2 0 0 0.1 0.2 0.3 0.4 0.5 Frequency[Hz] n Goodness-of-fit estimator: e Sˆ f S f f i i 1 i n Sˆ f f i 1 i i i 100% - II - SEM-REV WAVE DATA SEM-REV location Nantes (50km) Loire estuary SEM-REV location BMTO2 E WAVE BUOY W WAVE BUOY ADCP Datawell directional buoy and spectral processing: • • • • • • Measurements of {x,y,z} motions (continuous) 1.28Hz sampling rate HF radio transmission + onboard storage 1h-based signals for cross-spectral analysis 36 non-overlapping 100s periodograms (72 dof) Cos^2s directional reconstruction (based on 1stand 2nd-order dir. Fourier coefficients) • Δf = 0.01Hz, Δθ = 10° • Spectral smoothing (3x3 cell moving average) 8748 hourly directional spectra in 2011 (easternmost buoy) over 8760 expected (99.9% success rate) - III - RESULTS AND DISCUSSION Processing of SEM-REV 2011 hourly dir. spectra Separation frequency swell/wind-sea: Interpolated (1h) ECMWF ERA-Interim 10m-height wind speed for location (4.75°N, 3.0°W) close to SEM-REV In practice here: fs = min(g/2πβU10 , 0.20Hz) Min. threshold for swell partition grouping : Hmin = 0.20m Algorithm performance e mean = 17,7% (95% | e ≤ 30%) f , f [Hz] s 0.3 p 0.2 0.1 H m 0 [m], U /2[m/s] ECMWF wind data p , w,10 [º] w,10 4 3 2 1 0 360 270 180 [-] Correlation to ECMWF wind data (ERA-Interim) [º] No time tracking 0.4 0 90 0 5 4 3 2 1 0 60 45 30 15 0 60 e[%] Time evolution of wave system parameters (~18600 systems extracted, ~2.1 syst./s.s.) 0.5 45 30 15 0 07/02 08/02 09/02 10/02 11/02 12/02 13/02 14/02 15/02 Sea states type in SEM-REV (2011) for different Hm0 thresholds (i.e., wave systems with Hm0 lower than this value are disregarded in the counting) /8748 Sea states may be considered as unimodal only 25% to 64% of time! (according to threshold) Peakedness statistics (γ < 10, -3%) [0.08;0.12Hz[ Hm0 > 0.5m, 0.04Hz<f<=0.08Hz Hm0 > 0.5m, 0.08Hz<f<=0.12Hz Hm0 > 0.5m, 0.12Hz<f<=0.15Hz 40 40 5 [-] 0 0 10 Hm0 > 1m, 0.04Hz<f<=0.08Hz 5 [-] Hm0 > 1m, 0.08Hz<f<=0.12Hz 40 0 0 5 [-] 20 0 0 10 Hm0 > 3m, 0.04Hz<f<=0.08Hz 5 [-] = 1.36 0 0 5 [-] 10 40 20 0 0 SWELLS Occurrence rate[%] 0 0 10 5 [-] Hm0 > 1m, 0.2Hz<f<=0.5Hz Hm0 > 1m, 0.15Hz<f<=0.2Hz 30 20 10 0 0 10 10 = 2.28 5 [-] 10 20 10 0 0 5 [-] 10 Hm0 > 3m, 0.12Hz<f<=0.15Hz = 1.90 Occurrence rate[%] 10 10 5 [-] 10 = 1.65 20 = 1.21 Occurrence rate[%] 20 Hm0,i > 3m 5 [-] 20 60 60 30 10 0 0 10 30 0 0 10 Hm0 > 3m, 0.08Hz<f<=0.12Hz 40 20 = 1.38 Occurrence rate[%] 10 40 = 2.14 30 = 1.27 Occurrence rate[%] 20 5 [-] 40 = 1.38 Hm0,i > 1m 10 Hm0 > 1m, 0.12Hz<f<=0.15Hz 60 30 20 0 0 10 Occurrence rate[%] 10 30 = 1.81 30 Occurrence rate[%] 10 20 Hm0 > 0.5m, 0.2Hz<f<=0.5Hz Hm0 > 0.5m, 0.15Hz<f<=0.2Hz = 1.46 Occurrence rate[%] Hm0,i > 0,5m 30 [0.20;0.50Hz[ 30 = 1.47 Occurrence rate[%] Occurrence rate[%] 20 0 0 Occurrence rate[%] [0.15;0.20Hz[ 30 = 1.60 Occurrence rate[%] [0.12;0.15Hz[ Occurrence rate[%] [0.04;0.08Hz[ f 5 [-] 10 40 no data 20 0 0 5 [-] no data 10 ? WIND-SEAS • Again, statistics vary according to Hm0 threshold • Mean peakedness values found within [1;2] (except HF) Values range from 1 to 5 mostly, even for swells • In ]0.04; 0.12Hz] (swells) γ decreases with fp on average consistent with theory of swell evolution [e.g. Gjevik et al., 1988] • Above 0.15Hz γ (wind-seas) increases with fp on average consistent with JONSWAP observations as peakedness decreases during sea growth [Hasselmann et al., 1973] • [5% bias to be deducted from γ here approx. due to sampling variability in the spectral estimation with 72 dof (see paper OMAE2013-10004, same author)] Peakedness in severe sea states fatigue, survivability, certifications… Severe sea state: Hm0 > 3m (> 8m Joachim storm in December 2011) 6 Regression line: γ (biased) against Hm0 for Hm0 > 3m 5 [-] (100% sea states are unimodal) More data required ]0.04;0.08Hz] ]0.08;0.12Hz] ]0.12;0.15Hz] ]0.15;0.20Hz] ]0.20;0.50Hz] 4 3 2 Storms with low fp within ]0.04Hz;0.12Hz] 1 0 = 0.219*H 2 4 6 Hm0[m] +0.43 m0 8 10 - IV - CONCLUSIONS & FURTHER WORKS • On average, JONSWAP peakedness γ decreases and increases with peak frequency within [1;2] – from swell to wind-sea frequency range (most values within [~1;5] for both) • Partitioning algorithm successful: In SEM-REV in 2011, sea states could be considered as unimodal 64% of time at best partitioning required for metocean and engineering studies • Further work 1: JONSWAPs adapted to the spectral modelling of swells?... (preliminary results available now) • Further work 2: dynamic tracking of wave systems for better system type identification Merci de votre attention Contact: jbsaulni@ec-nantes.fr (< août 2013) toupaixil@yahoo.fr (ensuite) Interval 0.04-0.08Hz 0.08-0.12Hz 0.12-0.15Hz 0.15-0.20Hz 0.20-0.30Hz 0.30-0.50Hz 0.20-0.50Hz Hm0 min 0.2m 9.2% 29.6% 10.5% 8.5% 16.5% 15.4% 31.9% 0.5m 7.0% 21.6% 7.1% 6.0% 10.1% 2.3% 12.4% 1m 3.9% 10.0% 4.8% 9.9% 1.1% 0 1.1% 3m 0.2% 0.6% <0.1% 0 0 0 0 SEM-REV location Cable route Le Croisic town SEM-REV Salt evaporation ponds of Guérande