Validation of a thirty year wave hindcast using the Climate Forecast System Reanalysis windsby Arun Chawla, Deanna M. Spindler, Hendrik L. Tolman

Ocean Modelling




Wind wave models

Validation dcas ice gene n of abov ys. graphic/bathymetric) as well as issues related to interpolating wind fields at the land-sea margins. There bases s deve r engin ics. Th onmen

There are not enough available data to develop a traditional reanalysis for wind waves. Furthermore, wave dynamics are different from atmospheric dynamics in the sense that they represent a forced and damped problem rather than a (chaotic) initial value problem, with the wind forcing being the dominant process driving wave dynamics. Due to the forced and damped nature of wind reanalysis of the 2000), as well as sing traini on of early ysis data sets is provided in Caires et al. (2004).

A new NCEP Climate Forecast System Reanalysis (CFS recently been developed and entails a coupled reanalysis atmospheric, oceanic (only circulation), sea-ice and land data from 1979 to 2010 (Saha et al., 2010). This reanalysis has much higher spatial and temporal resolutions than previous reanalyses, and thus provides a valuable resource to develop a long-term hindcast database for wind waves. NCEP has chosen to perform wave hindcasts without data assimilation to avoid inhomogeneities in quality of the product associated with severe sparsity of observation data.

We believe that a more homogeneous ‘assimilative’ product can be q MMAB Contribution No. 299. ⇑ Corresponding author.

Ocean Modelling 70 (2013) 189–206

Contents lists available at

Ocean Mo elsE-mail address: (A. Chawla).the database should not be used for climate studies. In atmospheric modeling, a statistically more homogeneous dataset can be generated by performing a reanalysis with a consistent model setup for the entire period covered (e.g. Saha et al., 2010). been made to correct for this using a kinematic winds and subsequent hindcast (Swail and Cox, non-parametric corrections to the wave field u (Caires and Sterl, 2005). A detailed inter comparis1463-5003/$ - see front matter Published by Elsevier Ltd. sets reanalR) has of themaintains a wave hindcast database that extends from 1999 to the present. This database uses the archived analysis winds from the GFS atmospheric model (Moorthi et al., 2001) to drive the waves. However, this database is statistically inhomogeneous because numerical and physical upgrades to the models (both wave and atmosphere) are responsible for trends, and therefore atmospheric and wave data (Uppala et al., 2005). Using a separate reanalysis wind field fromNCEP/NCAR, Cox and Swail (2001) developed a global 40 year wave hindcast. However, due to resource limitations, historically reanalysis winds have been developed on temporal and spatial grids that are too coarse to resolve some of themajor events that drive the strongerwaves. Some attempts have1. Introduction

Long-term global wind wave data and engineering applications, such a gies, long-term statistical analysis fo studies, and validation of model phys at the National Center for Envirare some concerns about the wave climate in the Southern Hemisphere due to the over prediction of winds (early part of the database) as well as the lack of wave blocking due to icebergs (in the model).

Published by Elsevier Ltd. have multiple scientific loping wave climatoloeering design, scenario e wave modeling group tal Prediction (NCEP) waves, it is possible to produce accurate hindcastwithout assimilating any wave data, for instance using a wind field from a long-term reanalysis project. There are several such examples in literature.

Sterl et al. (1998) used a 15 year reanalysis wind field from the

European Center for Medium-RangeWeather Forecasting (ECMWF) to build a hindcast wave database. This was further expanded into a coupled 40 year reanalysis that included assimilation of bothReanalysis


At most buoys there is excellent agreement between model and data out to the 99.9th percentile. The agreement at coastal buoys is not as good as the offshore buoys due to unresolved coastal features (topo-Validation of a thirty year wave hindcast

System Reanalysis windsq

Arun Chawla a,⇑, Deanna M. Spindler b, Hendrik L. To aNOAA/NCEP, Environmental Modeling Center, College Park, MD 20740, United States b IMSG at NOAA/NCEP, Environmental Modeling Center, College Park, MD 20740, United a r t i c l e i n f o

Article history:

Available online 17 August 2012



Hindcasts a b s t r a c t

A thirty one year wave hin

Reanalysis (CFSR) wind and generated using the third nested grids. The resolutio wave height Hs and 10 m ( eter records and NDBC buo journal homepage: www.sing the Climate Forecast an a tes t (1979–2009) using NCEP’s latest high resolution Climate Forecast System database has been developed and is presented here. The hindcast has been ration wind wave model WAVEWATCH III with a mosaic of 16 two-way the grids ranged from 1/2 to 1/15. Validation results for bulk significant e Mean Sea Level) wind speeds U10 have been presented using both altimIn general the database does a good job of representing the wave climate.

SciVerse ScienceDirect delling evier .com/locate /ocemod generated by using observations to produce bias corrections for a hindcast, but the latter is considered outside the scope of this study.

The wave model used at NCEP is the third generation wind wave model WAVEWATCH III (Tolman, 2009). In 2007, the model was expanded to run as a mosaic of two-way nested grids (Tolman, 2008). The nested grid driver is described in Tolman (2007a,b). To drive the waves the wave model requires two input fields: ice and winds (including the air–sea temperature difference). The high resolution global winds at 10 m height used here have an hourly temporal and 1/2 spatial resolution. The reanalysis daily ice concentration fields have a 1/2 spatial resolution, and are derived from passive microwave from the SMMR and SSM/I using the NASA

Team algorithm.

The hindcast database has been developed taking advantage of the multi-grid features in WAVEWATCH III with finer resolution grids in coastal waters and semi-enclosed basins like the Mediterranean Sea, to provide adequate resolution for wave evolution, or to spatially resolve areas with in situ observations. This paper deregion (for regional error metrics). because they are representative of the region and for the length of their records. Validation statistics are primarily limited to U10 and