MSFPP-WP3000 : Data products


SPACE OCEANOGRAPHY MFSPP WP 3000 Near Real Time Remote Sensing Data Collection and Analysis
Objectives Participants Planning Task structure Data processing Data products Publications Links

Altimetry

SST

Ocean color

Altimetry

The different modules of the MFSPP NRT data processing system are described below (see figure). The MFSPP processing system uses the CLS global altimeter data acquisition and processing module developed as part of the DUACS project. The processing steps specific to MFSPP are the extraction of along-track Sea Level Anomaly relative to a new mean sea surface .(see subtask 3120 results), the correction of orbit error using a local inverse method (see subtask 3110 results) and the mapping of Sea Level Anomaly using an improved optimal interpolation method (see subtask 3110 results).

Validation
Input files for the MFSPP NRT processing software are level-2 altimeter data organized in chronological order and by half-orbits. At T0, every 7 days on Wednesday, the validation module of the NRT processing software processes 7 days of those data. Nominally, the processed data cover the period T0 - 9 days to T0 - 2 days (2 days is the mean acquisition delay). The processing is performed to remove bad altimeter data and to check to data quality. Only valid data are kept. The validation first uses existing relevant flags in the altimeter products. It then uses data thresholds on altimeter range, altimeter corrections and on differences relative to a mean sea surface. An outlier detection algorithm using spline functions is also used. The next modules use validated data from the current week and from the two previous weeks. Nominally, the processed data thus cover the period T0 - 22 days to T0 - 2 days.

Crossover analysis and orbit error reduction
Global crossover minimizations are performed to reduce orbit error. The first step is to calculate the single T/P and ERS crossover differences and the dual T/P and ERS crossover differences. These calculations are performed on a time window of 21 days using validated data from the current week and from the two previous weeks. T/P orbit error is then calculated through a global minimization of T/P - T/P crossover differences. The ERS orbit error is estimated through a global minimization of T/P-ERS and ERS-ERS crossover differences. As T/P orbit error is much smaller than ERS orbit error and has already been partially corrected for in the previous step, T/P data are used as a reference for the ERS-2 orbit error calculation. The method is described by Le Traon et al. (1995) and Le Traon and Ogor (1998).

Calculation of corrected sea surface heights
The corrected sea surface height files are created by applying altimetric corrections and orbit error correction to the sea surface height. The mean sea surface MSS_MED is also removed to correct for cross-track geoid gradient errors.

Repeat-track analysis
From several corrected sea surface height files, a conventional repeat-track analysis is performed to extract the Sea Level Anomaly (SLA) relative to a mean profile : data are resampled along the mean profile using cubic splines and differences relative to the mean profile are calculated. T/P and ERS mean profiles used for MFSPP correspond to a 5-year mean (1993 to 1997). If differences relative to the mean sea surface are needed, data are only resampled along the mean profile and no mean is removed. Sea Level Anomaly files are organized by satellite only (multi-cycles).

Filtering and sub-sampling of SLA files
SLA along-track data are then filtered and subsampled. The filter is a non linear median over 3 points (roughly 21 km) followed by a low pass along track linear Lanczos filter. The Lanczos filter cut-off wavelength is 42 km. SLA data are then subsampled by keeping 1 point over 2.

Local inverse method for long wavelength error correction
The global crossover minimization cannot remove all long wavelength errors and a local adjustment needs to be performed. The local adjustment method is based on the optimal interpolation method discussed in Le Traon et al. (1998) (see section 2.1.1). The main difference is that, instead of mapping the oceanic signal, the along-track bias is directly calculated for each track given an a priori knowledge of the covariance functions of the oceanic signal and of the variance of long wavelength errors. The local adjustment method uses then simultaneously T/P and ERS data over (nominally) the period (T0- 22 days - T0-2 days) to reduce along-track correlated errors. It allows a correction of residual errors due to orbit errors and mainly inverse barometer effects. The covariance functions for the oceanic signal and long wavelength errors are the same as for the mapping method (see below). In practice, the bias is calculated every 200 km. Smoothing cubic spline are then used to estimate a bias for each point along the track and to produce corrected along-track SLA files.

Mapping of Sea Level Anomaly
Finally, mapping of Sea Level Anomaly is performed at the date (T0-7 days) (using T/P and ERS data over (nominally) the period (T0- 22 days - T0- 2 days). It consists of a sub-optimal space/time interpolation method which is detailed in Le Traon et al. (1998) (see section 2.1.1). Long-wavelength errors are directly taken into account in the mapping procedure; the mapping thus uses as input SLA files uncorrected for along-track long wavelength errors. Mapping is done on a regular grid (model grid) with a resolution of 1/8 of a degree. The space correlation scales of the ocean signal (zero crossing of correlation function) are set as 150 km. The temporal correlation scale (e-folding scale) is set at 15 days. The long wavelength error is assumed to be a bias uncorrelated for different tracks and cycles. The a priori long wavelength error is 30% and 50% of the ocean signal variance for T/P and ERS respectively. Note that ERS is adjusted onto T/P in the crossover minimization, thus only a slightly larger long wavelength error is assumed for ERS. The instrumental noise variance is set to 10% and 15% of the signal variance for T/P and ERS respectively.

SST

SST are operationally produced and validated.

The Western Mediterranean and Eastern Mediterranean
SST data are respectively processed by CMS and CNR centers. At CMS, only AVHRR NOAA-14 nigth data are acquired and weekly computed (median value), every day with a resolution of about 2 km, and then averaged with a resolution of 1/8 of a degree. At CNR, AVHRR NOAA-15 nigth and morning data are acquired, computed and daily averaged with a resolution of 1/8 of a degree.

The merging of Western and Eastern
data is done at CMS center. Firstly, the bias between CNR and CMS data is removed in order to have the NOAA-14 night data as a common reference. Finaly, SST data are interpolated over the whole Mediterranean Sea on the model grid (1/8°x1/8°) using an objective analysis method.

Ocean color

Ocean color data processing consist in