Observed data are subjected to a battery of checks to identify and remove unrealistic points― which are revealed by viewing property profiles as a function of depth and theta. A combination of statistical and visual methods identifies both systemic problems (malfunctioning equipment) and random errors. "Bad" points are removed and stored in a separate file. Detailed Description

Property Profile Plots

The profiles are plotted in both pressure and theta space for an initial assessment of the data. If necessary, they are divided into subgroups based on geography or characteristic T-S profiles.

Systemic Problems are Identified

Profiles with common deviation (salt shift, for example) are removed. These will often hail from a single cruise. If necesarry, the entire cruise is extracted (with hb_extract) to a separate file and plotted in order to determine whether the entire cruise should be eliminated.

Range Checking

A preliminary sieve of data eliminates values that are obviously extreme. Acceptable ranges are defined for temperature and salinity as a function of depth in a user-supplied file.

Statistical Fitting and Checking

Here we compute a characteristic T-S curve and standard deviation envelope as a function of density for a group of data. This is then used to identify and remove outliers from the observed data in the next step.

A set of potential density bins appropriate for the area is specified. These sigma bins divide up the T-S curve into segments that together characterize the total T-S relation. For depths 0-1000, sigma-0 densities are used. Sigma-2000 bins are used for depths 1000-3000 m, and sigma-4000 bins are used for depths > 3000m. The statistical checking is done with hb_statchk_ts which identifies points that lie outside of a standard deviation envelope. The" bad" points and "good" points are output to separate files and plotted to get a visual picture of what is being identified. The bottle, ctd, and float data are stored in separate files.