A US NOPP project to stimulate wave research Focus on operations and deep water Hendrik L. Tolman Chief, Marine Modeling and Analysis Branch NOAA / NWS / NCEP / EMC Hendrik.Tolman@NOAA.gov Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 1/26
NOPP The National Oceanographic Partnership Program (NOPP) is a collaboration of federal agencies to provide leadership and coordination of national oceanographic research and education initiatives. http://www.nopp.org/about-nopp/ Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 2/26
Motivation NOPP project: Improving Wind Wave Predictions; Global and Regional Scales Considering progress in understanding of wave model physics, particularly dissipation and economical interaction approximations, Linwood, Don and Hendrik started pushing for this project after the 2008 Ocean Sciences meeting. Buy-in from: ONR, BOEM (was MMS) with funding. NOAA, USACE, NRL with in-kind contributions. Focus on operational modeling, basin and shelf scale. Several surf-zone proposals also funded. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 3/26
Outline Outline of paper: NOPP teams Validation data. Validation techniques. 30 year hindcast. Code management. Outlook. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 4/26
NOPP teams PI-s Topics Focus areas Ardhuin in+ds Dissipation (breaking, swell, bottom) Unstructured grids in WW III. Babanin in+ds Observations based + swell diss. Banner in+ds Extreme conditions, explicit breaking prediction, fluxes including sea spray Perrie nl Two-Scale Approximation. Tim Janssen nl (shal) Combine quads & triads, field data sets. Zakharov / Pushkarev Kaihatu / Sheremet nl+in+ds Advanced statistical and dynamical nonlinear models + input and dissipation shal Traditional mud and vegitation models Two-layer Boussinesq mod. Field data. Van Vledder shal Shallow water models and obs., including surf beats. Hanson shal Duck data sets, spatial partitioning. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 5/26
NOPP teams Organization PIs In-kind contributions USACE NOAA/NCEP NRL/Stennis Resio Smith Tolman Alves Chawla V/d W. Rogers Campbell FRF + Currituck Sound data. New source terms / studies (in+nl+ds) IMEDS + Additional model metrics WAM/STWAVE + ADCIRC coupling Partitioning + tracking (with NCEP) Source terms + QS model for WW III. Code management for WW III. NOPP data server. Pre-operational testing of new ST. 30 year wave hindcasts. Curvilinear grids in mosaic (WW III). ESMF wrapper (SWAN, WW III). Automated regression testing (WW III). Code management best practices. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 6/26
Validation data Two type of validation / data important for operational wave models. Model needs to work all the time: Bulk long term validation / development against routine observations. In-situ, altimeter, SAR (?). Model needs to make physical sense. Directed measurement campaigns focused on specific physics of waves. Individual campaigns. Data mining of routine observations. Select data set types and conditions, rather than campaigns. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 7/26
Validation data Conditions Long term validation Wind Sea and Swell Non-Aligned winds Extreme conditions Diminishing winds Shallow water Data sources In-situ, altimeters, SAR. JONSWAP, Great Lakes, Lake George, SAR, Tehuantepec, Duck. Spectral partitioning of buoy data. Slanting fetch, Tropical cyclones (Duck, SRA, WSRA). TCs, data mining. FAIRs, data mining, tradewind and monsoon data (INCOIS, ) Data sets provided by teams, older bottom friction data sets. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 8/26
Validation techniques Going beyond the traditional validation techniques used for wave models. Traditional: bulk error measures: Mean, std, SI, r 2, scatter/pdf, sometimes qq Event-based statistics: Peak values, timing, shape of signal. Spectral wave model validation: Ridge plots identifying individual swell events. IMEDS analysis using spectral partitioning. For forecasting, hit-miss statistics are very important, but rarely used in scientific papers. Use additional relevant physical parameters, mss, peakedness, etc. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 9/26
Validation techniques Using more concise presentations. Taylor diagram. Variance Correlation Error 0 1 Wind speed during January 2010 based on 75 buoys 0.1 0.2 0.3 0.4 0.5 Co r r e l a t 0.6 Target plot Bias Error Standard deviation 0.5 S D 1 0.5 i on 0.7 Coe f f 0.8 BC D E i c i en t 0.9 0.95 M 0.99 R 0 A 1 Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 10/26
30 year hindcast The Climate Forecast System (CFS) reanalysis and re-forecasting project (CFSRR) provides a 30+ year high resolution wind field 0.5 hourly wind and temperature fields. Associated 0.5 daily ice analyses. This data set appears ideal to be the basis of a wave reanalysis over the same 30+ year period. There is insufficient data in any period to obtain a datadominated analysis, therefore It makes more sense to do a hindcast without assimilation, and use data possibly later for bias corrections of the hindcast only. Ideal as basis for long-term validation in NOPP project. http://cfs.ncep.noaa.gov/cfsr Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 11/26
30 year hindcast NH biases Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 12/26
High wind speed issues with CFSR in SH 30 year hindcast NH SH Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 13/26
Annual mean biases against altimeter data 2008, Jason 1 2002, Envisat 30 year hindcast Annual signal, After 2005 Persistent signal Seasonal (djf) Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 14/26
Code management WAVEWATCH III community modeling code management environment Traditionally code distributed as tarball, code delivered back to NCEP the same way. Does not work with many teams working on same code Subversion (svn) server for version control. Each team has code manager with access to server and latest developmental model versions thereon. Code managers at NCEP merge individual contribution into trunk version of code. Henrique, Arun, (Andre, Hendrik). Best practices guide for community model development of WAVEWATCH III as deliverable for NOPP project. NCEP intends to maintain this environment after project is finished. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 15/26
Outlook WAVEWATCH III added capabilities: Curvilinear and unstructured grids. Quasi-stationary model version. New source terms. GMD and nonlinear filter. Two moveable bed bottom friction terms. Ifremer physics packages. Massively expanded output options (coupling). Wave system tracking. Post-processing tools: Re-gridding. NetCDF output. Coupling interfaces: ESMF. PALM. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 16/26
Snl in hurricanes Wave height and relative error(%) -20% WRT x 20% > 12m < 15% G13d - WRT 3 x DIA Tolman, 6/25/2012 G11d - WRT optimum DIA < 10% G35d - WRT 5 x new quad ECMWF Workshop on Ocean Waves, 17/26
Spectral parameters 100km behind eye (minimal wave height errors) spectrum (2D) spectrum (1D) steepness S nl in hurricanes direction 60 (f ) 45 dir. spread (e) Snl ( o ) 30 15 0 0.05 0.1 0.15 0.2 0.25 f (Hz) green: WRT dashed green: G35d blue: G13d red G11d Tolman (2011, 2013) Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 18/26
Wave system tracking raw partitioning Raw partitioned data wave system tracking Spatially tracked systems Hs Tp Hs Tp Van der Westhuysen, Hanson and Devaliere (2013) Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 19/26
Outlook In the pipeline: Spatial and temporal tracking of wave fields: Porting internal partitioning to SWAN. Space-time tracking (external / internal). NCEP planning first physics upgrade in operational wave models based on NOPP project in 2012 (following slides). NCEP planning NOPP consensus upgrade in 2014/5. Replacing DIA and other deep source terms. New products for SOLAS, specifically wave breaking. Full polar coverage (tri-polar / curvilinear Artic grid). Unstructured coastal grids (2-3 km resolution) Upgrades shelf physics. Multiple well performing physics packages Multi-physics ensembles. Estimate of uncertainty in wave model physics. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 20/26
Outlook NCEP physics upgrade based on Ifremer results, tested with NCEP global and Great Lakes winds. old biases Dec. 2009 Feb 2010 new Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 21/26
Outlook Monthly global errors 2009 Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 22/26
Outlook Great Lakes buoy 45007, 2009 old physics H s Total (m), Measurements vs. st2default 4 3 H s (m) 2 1 0 01/05/09 01/06/09 01/07/09 01/08/09 01/09/09 01/10/09 01/11/09 4 4 3 3 WW3 H s (m) 2 WW3 H s (m) 2 1 1 0 0 1 2 3 4 Measured H (m) s 0 0 1 2 3 4 Measured H (m) s Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 23/26
Outlook Great Lakes buoy 45007, 2009 new physics H s Total (m), Measurements vs. st4t500 4 3 H s (m) 2 1 0 01/05/09 01/06/09 01/07/09 01/08/09 01/09/09 01/10/09 01/11/09 4 4 3 3 WW3 H s (m) 2 WW3 H s (m) 2 1 1 0 0 1 2 3 4 Measured H (m) s 0 0 1 2 3 4 Measured H (m) s Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 24/26
Outlook Taylor diagram and alternative version for several GL buoys for 2009 0 1.25 0.1 0.2 0.3 0.4 0.5 Co r r e l a Standard deviation (normalized) 1 0.75 0.5 1 0.8 M 0.6 D S 0.4 0.6 t i o n 0.7 C o e f 0.8 f i c i e n t 0.9 0.95 0.6 R MS D 0.4 0.2 95 90 50 60 70 80 Co a t r r e l i on 30 0.25 R 0.2 0.99 99 0 0 1 1.25 1 0.75 0.5 0.25 0 Normalized STD / m o GLERL-Donelan (1-G) WW III Tolman and Chalikov (1996) WW III Ardhuin et al. (2010) Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 25/26
Conclusion Conclusion: Making great progress toward improved operational models. Laying ground work for community model development and modeling environment. Tolman, 6/25/2012 ECMWF Workshop on Ocean Waves, 26/26