Strong winds in a coupled wave-atmosphere model during a North - PowerPoint PPT Presentation

Strong winds in a coupled wave-atmosphere model during a North Atlantic storm event: evaluation against observations Lucia Pineau-Guillou (Ifremer/LOPS) , Fabrice Ardhuin, Marie-Noëlle Bouin, Jean-Luc Redelsperger, Bertrand Chapron, Jean Bidlot ( ECMWF ), Yves Quilfen AMS 21 st Conference on Air-Sea Interaction, 11-15 June 2018, Oklahoma City 1

1. Introduction 2. Case study: storm description 3. Model & observations 4. Comparison between simulated winds & observations 5. Sensitivity to wind stress parameterization 6. Conclusions 2

1. Introduction ● Motivations ➔ Underestimation of large wave heights in wave models (up to 15% Rascle & Ardhuin, 2013), as well as storm surges in ocean models (Muller et al., 2014) ➔ Could be partly due to: ➔ underestimation of strong winds in atmospheric models ➔ inappropriate representation of wind stress in numerical models • Objectives 1) Evaluate strong winds against observations (buoys, platforms & satellites) 2) Test how alternative wind stress parameterizations could reduce the bias and lead to a more accurate model ● Method ➔ Use of ECMWF Integrated Forecasting System (IFS CY41R1), which provides a drag as a function of the sea state ➔ Numerical simulation of two extratropical storms 3

2. Case study: storm description ● Event selection : analysis of ERA-Interim database over the last 10 years (2005-2015) of North East Atlantic (30°W 10°E 30°N 65°N) ● 10 more energetic events classified ● Kaat and Lilli storms, which crossed the North Atlantic in January 2014 Kaat and Lilli storm tracks 4 In black dotted line, principal tracks from Hoskins and Hodges (2002)

3. Model & observations ● Model: ECMWF atmosphere model IFS coupled with ECWAM (cycle CY41R1), simulations without data assimilation ● Wind measurements Buoys and platforms come from GTS. Correction to 10m based on a logarithmic law. scatterometers radiometers altimeter 20 buoys 59 platforms 5

3. Model & observations Wind measurements ASCAT winds from RSS (right) stronger than those from KNMI (left) Wind field on 26 January 2014 estimated from ASCAT-KNMI Metop-A (a), ASCAT-RSS Metop-A (b), 6 AMSR2 (c), WindSat (d) for descending passes, and for SMOS (e) and JASON-2 (f)

4. Comparison between simulated winds and observations Wind biases Negative bias, except for ASCAT-KNMI and buoys Wind correlations from 23 to 27 of January 2014 between default ECMWF parameterization (CY41R1) and 7 ASCAT-KNMI (a), ASCAT-RSS (b), AMSR2 (c), WindSat (d), SMOS (e), JASON-2 (f), buoys (g) and platforms (h)

4. Comparison between simulated winds and observations Wind biases Group 1 : ASCAT-KNMI/buoys 7 m/s Group 2 : satellites (except ASCAT-KNMI)/platforms 5-20 m/s 20-40 m/s good agreement underestimation ● Strong winds are generally underestimated compared with observations ● Biases exist between observations: group 1 gives lower strong winds than group 2 8 ● Difficult to conclude which dataset should be used as a reference

5. Sensitivity to wind stress parameterization [1] Default ECMWF parameterization (CY41R1) Janssen (1991) z 1 Effect of wave through roughness length 2 π z 0 wave = ∞ τ w =g ∫ k ∫ ⃗ √ 1 − τ w β ( f,θ ) E ( f,θ ) / ( 2 πf ) dfdθ ⃗ τ 0 0 [2] WW3 physics β given by Ardhuin et al. (2010) and E(f,θ) influenced by a different dissipation term [3] Wave age dependent parameterization given by Oost et al. (2002) − 2.5 α= 50 ξ with ξ= C p / u * wave age [4] Empirically-adjusted Charnock parameterization T o obtain a maximum C d of 2.5 10 -3 for wind speed around 30 m/s 9 [5] Constant Charnock 0.018

5. Sensitivity to wind stress parameterization [1] Default ECMWF parameterization [1] Default ECMWF parameterization (CY41R1) [5] Empirically-adjusted Charnock parameterization 10 Drag values more consistent with observations

5. Sensitivity to wind stress parameterization The wave age formulation gives too high drag values (due to unrealistic extrapolation to high winds of a relation based on winds 6-18 m/s) 11

5. Sensitivity to wind stress parameterization Results Larger Charnock → larger z 0 , higher Cd and stress and lower wind Wave age parameterization New parameterization reduces the wind bias ~2 m/s at 30 m/s 12

6. Conclusions ● Moderate simulated winds (5-20 m/s) match well with observations ● For strong winds, mean differences appear as much as -7 m/s at 30 m/s ● Significant differences also exist between observations ➔ Buoys and ASCAT winds generally lower than platforms and other remote sensing data (AMSR2, ASCAT-RSS, WindSat, SMOS and Jason-2) ➔ Difficult to conclude which dataset should be used as a reference ➔ However, buoy and ASCAT-KNMI are likely to underestimate the real wind ● Concerning parameterizations ➔ Wave age not appropriate for coupling ➔ New parameterization leads to higher winds ● To go further: looking at the ocean response through storm surges in the North Sea 13