Turbulent fluxes

To examine differences among datasets for each turbulent flux component over the Atlantic Ocean (70°W-30°E, 45°S-45°N), data for 1996-2005 at daily, weekly, monthly and seasonal timescales are used. During this period observing systems, both satellite and in situ, were optimal. Differences between IFREMER and WHOI as well as each of them against independent in situ data (PIRATA/FETCH/Romeo) are examined statistically. To examine the deviation between each of the two flux estimates and independent ground truth, the approach of Bourras (2006) is applied to both latent and sensible heat fluxes (Santorelli et al., 2011).

There is evidence that wind stress and turbulent heat fluxes can be improved directly by the use of scatterometer and radiometer high-resolution wind data. Although scatterometer wind data are already assimilated in some NWP systems, the meteorological assimilation generally truncates most of the small spatial and temporal scales which are critical for ocean dynamics. This priority on scatterometer and radiometer data is also justified by the availability of global high-resolution (25/12.5 km) wind data from SeaWinds (on board QuikScat and for a short period from ADEOS II) and WindSat (launched in January 2003). The latter exploits a new concept for remotely sensed wind vectors and microwave polarimetry.  Furthermore, it provides estimates of sea surface temperature, rain rate, and of water vapor necessary for flux calculations.

We use buoy measurements from the Prediction and Research Moored Array in the Atlantic (PIRATA) (Servain et al., 1998) as primary source of ground truth. The placement of the buoys was chosen to provide coverage along the equator for regions of strong wind forcing in the western part of the basin and significant seasonal-inter-annual variability in SST in the central and eastern parts of the basin. PIRATA buoys provide wind speed at 4 m and air temperature and specific air humidity at 3 m.

Since the PIRATA data are assimilated   in the WHOI product, the evaluation of this product against PIRATA data is not independent.    Independent buoy observations that have not been assimilated by WHOI are also used in the evaluation. Data from an ASIS buoy in the “flux, etat de la mer, et teledetection en conditions de fetch variable” (FETCH, Hauser et al., 2003) experiment was used. During FETCH it was moored at 42° 58' 56'' N, 4° 15' 11'' E by the University of Miami. Observations were collected between March 18 and April 10, 1998 at 28.5 minute intervals, but averaged daily for comparison with IFREMER and WHOI data. The buoy provides surface wind speed at 7 m, sea surface temperature at 2m below the surface, as well as air temperature and specific air humidity (calculated from relative humidity) at 5 m.

Shortwave Radiative fluxes

The following was done:

1.   Produce longer term estimates at best available resolution with historical data and evaluate them. For SW radiative fluxes used are the various ISCCP data such as ISCCP D1 at 2.5⁰ and ISCCP DX gridded to 0.5⁰ resolution using inference schemes with improved treatment of aerosols.

2.  Produce short term estimates with newer satellite systems such as MODIS on TERRA and AQUA. The year 2004 was selected since during that year both METEOSAT-7 (3 channel instrument) and METEOSAT-8 (12 channel instrument) were simultaneously operational. METEOSAT-7 provides the link to the long term data sets (ISCCP D1 and DX) which are based on METEOSAT-7 observations at reduced temporal and spatial resolution.

Longwave Radiative fluxes

A new approach for calculating downwelling surface longwave (DSLW) radiation under all sky conditions is used (Nussbaumer and Pinker, 2011). The   DSLW/UMD) model is driven with: 1) a synthesis of the latest 1⁰ resolution Moderate Resolution Imaging Spectroradiometer (MODIS) level-3 cloud parameters and information from the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-Interim model; 2) same as in 1) using ISCCP DX instead of MODIS observations. The DSLW/UMD’s clear sky contribution is based on the Rapid Radiative Transfer Model (RRTM), while a statistical cloud structure model and parameterization determine the cloud contribution to DSLW.