April 16, 2013

At the 13th GHRSST Science Team Meeting in Tokyo, Japan (June 4-8, 2012), Dudley Chelton and co-authors presented preliminary results of a proposed test for evaluating the analysis procedures used by various GHRSST L4 datasets. In their presentation (that can be found at https://www.ghrsst.org/files/download.php?m=documents&f=120611160329-04CheltonGHRSSTTokyo.pdf) and recent Journal of Climate article that is now in press ("Objective Determination of Feature Resolution in Two Sea Surface Temperature Analyses" by Reynolds, Chelton, Roberts-Jones, Martin, Menemenlis and Merchant; http://dx.doi.org/10.1175/JCLI-D-12-00787.1), the test was applied to the analysis procedures of the NCDC Two-Stage OI and OSTIA. There was an expressed interest among GHRSST L4 data producers to perform this test on their analysis procedures. To that extent, Dudley Chelton and Dick Reynolds have been working closely with the GDAC/PO.DAAC to provide this capability.  

Their procedure uses simulated datasets created from daily complete SST fields from a high-resolution global model, ECCO2. The gridded fields of model-simulated SST were subsampled based on approximate times and locations of the actual satellite SST data from one MW instrument and one IR instrument to represent microwave and infrared satellite data, respectively. To study the impact of noise in satellite measurements of SST, realistic noise was added to the full and subsampled model data. These simulated datasets were then used in the NCDC Two-Stage OI and OSTIA GHRSST L4 analysis procedures.  Power spectral densities (both auto-spectra and cross-coherence spectra) were computed over six regions in the global ocean to evaluate the analysis procedures (e.g., in terms of smoothness imposed onto the data).

Such simulated datasets, for input in GHRSST L4 analysis procedures, are now provided on the PO.DAAC ftp site (ftp://ghrsst@podaac.jpl.nasa.gov; contact Ed Armstrong at Edward.M.Armstrong@jpl.nasa.gov for the login/password). A high-resolution (i.e., “Simulated_Datasets/hires”) directory exists, and within each high-resolution file there are four variables including, high-resolution full (hifull), high-resolution reduced (hired), high-resolution full noise (hifulnos), and high-resolution reduced noise (hirednos).  The spectral estimation procedure used in the paper by Reynolds et al. (2013) is provided (in Fortran) along with the numerical results from the NCDC and OSTIA analysis methods on the ftp site. (i.e., “spectest” and “Results”). A “HowTo” text file is provided on the ftp site that details the procedure, consisting of:
 

  1. Downloading the simulated data,
  2. Downloading the analysis software package,
  3. Running the spectral analysis codes, and
  4. Sending the results (i.e., data files, spectra data and plots) to PO.DAAC.

 
The GDAC/PO.DAAC requests that the L4 datasets produced by each interested GHRSST L4 analysis producer, as well as the spectral data and plots generated, are then sent to the GDAC/PO.DAAC to make them available for all interested participants.