Dynamic validation of Globwave SAR wave spectra data using an observation-based swell model. R. Husson and F. Collard
Context 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 Why? Ocean sampling from orbiting station is intermittent - Difficulty for cal/val & users Partial information: wind sea is not seen by SAR - Difficulty to compare with other full-range data Data quality is heterogeneous, depends on observing scene and metocean conditions - Variation of swell measurement error estimation - Difficult to use (assimilation, e.g. ECMWF, 2006) Latest developments SEASAT ERS-1 ENVISAT Sentinel-1 A ERS-2 Globwave WHAT IS IT? CFOSAT Sentinel-1 B SAR wave mode observations Swell field reconstruction (Holt et al., 1998; Heimbach et al., 2000; Collard et al., 2009) Swell energy decay (Ardhuin et al., 2009) Swell field energy distribution (Delpey et al., 2010) Latest scientific results and European projects propose favorable canvas for further developments
Contents Input Data SAR wave mode measurements of swell spectra - error estimation Swell reconstruction Gathering SAR observations into coherent swell fields Synthetic swell field model Quality-controlled swell model with regular space-time grid Validation results
Contents Input Data SAR wave mode measurements of swell spectra - error estimation Swell reconstruction Gathering SAR observations into coherent swell fields Synthetic swell field model Quality-controlled swell model with regular space-time grid Validation Results
1/5. Input Data : Globwave SAR L2P Very few co-locations with in situ measurements despite dynamical co-location method Description Partitioned swell spectra with three integral parameters (for 2 most energetic partitions): - Significant swell height - Peak period - Peak propagation direction Data are corrected for systematic errors on T p and H ss Modulation of sea surface roughness on a SAR imagette caused by a long swell (left) and associatted swell spectrum (right) Error estimation (Monaldo, 1988; Johnsen et al. 1999; Johnsen et al., 2006; Collard, 2009) Thorough estimation of parameters impacting measurement errors Co-location against Globwave in-situ database Best quality flag data: - Normalized variance - Azimuth cut-off wrt. swell azimuth wavenumber - Confidence in propagation direction Dynamical co-location: - Using linear propagation theory, positions of swell observations are estimated up to 1000 km away (assuming deep ocean and absence of current)
1/5. Input Data Performances Remaining error estimation Large values of T p and H ss - Cases with fewest co-locations (gen. occurrence + buoy locations) H ss above 2.5m under-estimated by 0.25m T p above 15.5s under-estimated by 1s Quality flags Effect of azimuth cut-off still very present (0.20m bias) Additional impacting parameters Sea surface wind speed Time of propagation Swell partition contrast (RMSE doubles from small to large contrasts) Partitioned wave spectrum with four wave systems Error correction is limited and additional parameters should be used for more quality flag levels
Contents Input Data SAR wave mode measurements of swell spectra - error estimation Swell reconstruction Gathering SAR observations into coherent swell fields Synthetic swell field model Quality-controlled swell model with regular space-time grid
2/5. Swell reconstruction - Principle Problematic How can observations belonging to the same swell event be gathered if isolated in space and time? Exploit the coherence of the swell measurements belonging to the same swell field (Holt et al., 1998; Heimbach et al., 2000; Collard et al., 2009) Hypothesis 1. Swell observations of the same swell field all originate from the same region? Swell observations whose estimated past-positions converge to the same region belong the same swell field 2. Propagation in open ocean, absence of currents Principle All swell observations are retropropagated using linear theory travel at group speed along great circle: - swell period - swell direction - Position at t 0 lon/lat, D p, T p before & after t 0 Analysis of density maps of retropropagated swell positions (2x2 grid). 9 avril 2008 11 avril 2008 10 avril 2008 12 avril 2008 Density maps at successive time steps showing the displacement of a refocusing region related to a storm event Very strong hypothesis - Additional modifications are needed to insure validity
2/5. Swell reconstruction - Example All swell observations on acquisition time Swell observations of same swell field on acquisition time Propagated swell observations of the same swell field Fireworks Ocean sampling is not intermittent anymore but sparse (time/space/quality)
2/5. Swell reconstruction - Results Fireworks Using Wave mode data only as input Using Wave mode and Image mode data as input
Contents Input Data SAR wave mode measurements of swell spectra - error estimation Swell reconstruction Gathering SAR observations into coherent swell fields Synthetic swell field model Quality-controlled swell model with regular space-time grid
3/5. Synthetic swell field - Principle Problematic The spatio-temporal distribution of each integral parameter is described for each swell field. How to combine these heterogeneous measurements (space/time/quality) of the same coherent phenomenon? Main challenges Correct for the irregular swell sampling? Filter data outliers? Quantify the accuracy of our estimations? Previous research (Delpey et al., 2010) Assessment of the swell field energy distribution in longitudinal and transverse directions (r, ) Assumption of point source storm for T p and D p From previous results, not possible anymore! Need a reference source of information (buoy or model) We want to be totally independent No filter on input propagated swell observations Estimate the spatial distribution of each integral parameter for each time step using an iterative surface fits + outlier rejections
3/5. Synthetic swell field - Methodology - Region of valid SF Propagated observations without land-blocking Valid regions from a simulated isotropic storm with land-blocking Valid regions from minimum density of propagated observations Valid regions over which the synthetic swell field is estimated
3/5. Synthetic swell field - Results Wavelength Input Propagated swell measurements Output Synthetic swell field
3/5. Synthetic swell field - Results Significant swell height Input Propagated swell measurements Output Synthetic swell field
Contents Input Data SAR wave mode measurements of swell spectra - error estimation Swell reconstruction Gathering SAR observations into coherent swell fields Synthetic swell field model Quality-controlled swell model with regular space-time grid Validation Results
Synthetic swell field error trend analysis SAR L2P and L3 synthetic fields vs Globwave in-situ database Observations - L2P Synthetic swell field - L 3
Synthetic swell field RMS error analysis Inter-comparaison données SAR/in-situ/modèle SAR L2P SAR L3 Hauteur significative [cm] Période [s] Direction [deg] 30 0.7 16 < 30* 0.4 11 Wavewatch III Model < 30** 0.8 12 *Performance dependent on SAR sampling ** Performance dependent on specific criteria (swell or wind sea)
Conclusion & perspectives
dynamic validation of SAR Globwave products Why? Ocean sampling from orbiting station is intermittent Difficulty for data use & cal/val Goal Exploit the SAR data to its full potential by developing and validating an independent observation-based quality-controlled swell model.
2/5. Swell reconstruction outcomes Yearly analysis (2008) Over 700 storms, more than 60% indicate different refocusing times and places dpg. on the swell wavelength - In certain cases, difference reaches 48h -In average: - longest swell refocuses soon after the storm reaches its maximum strength - shortest swell during the long storm decay - In such cases: Δt = 14h, σ = 12h Explanations 1. Wave vs. storm speed: Longer waves outrun the storm and leave their generation area (at refocusing time, mean wave age = 1.4, σ = 0.3) 2. Free-propagation: To start to propagate freely, wave-wave interactions, driven by active wind forcing, have to diminish. For shorter swell, active wind forcing requires less stronger winds This condition is fullfiled later for shorter than for longer swell Consequences Mis-positioning of storm source can modify the swell energy decay by approx. 10%
2/5. Swell reconstruction Outcomes Input data SAR L2P swell spectrum observations Achievements Solve issue of swell field intermittent observation by SAR Propose a robust & automated methodology for swell reconstruction Point source storm hypothesis is an over-simplified hypothesis A more realistic data-driven approach is proposed Provide a rush of swell field observation for further studies (~700/year) Perspectives Use the long SAR archive (15 years) for a climatology study of storms distribution & energy
Global analysis over year 2008 23 700 storms detected, mainly extra-tropical storms
Global analysis over year 2008 24 Specific focus on storms events that have generated the swell field with the longuest wavelengths
Global analysis over year 2008 25