Temperature and Salinity Profiles

Profile observations are first checked for duplicate depths and strictly increasing depths. Reported levels that fail these tests are flagged and not used in the following profile quality control procedures.

4.4.6.1 Instrumentation Error Checks

Special instrument specific error tests are applied to profile observations to identify errors that have unique profile signatures. These errors include temperature inversions at the bottom of the profile, spikes in the temperature profile, and positive temperature gradients (warm bulge) in the mixed layer. The instrumentation error checks are applied iteratively until all errors are found, since a profile may have one or more of these types of errors. Reported temperature-depth levels that contain instrumentation errors are flagged and not used in the next iteration of the instrumentation error checks. One difficulty with the current suite of profile instrumentation error checks is that the tests are designed to detect errors specific to expendable bathythermographs (XBT) (Bailey et al. 1994). Other profile data types, such as Argo floats, gliders and CTD probes, are likely to have failure modes that are different from a XBT. Automated quality control tests need to be developed to detect instrumentation errors in these data types as more experience is gained with their assimilation.

4.4.6.2 Static Stability

A static stability test is performed to detect density inversions in profile observations. The reported in situ temperature and depth data pairs are first converted to potential temperature and pressure, and then potential density is computed at each pressure level using observed or derived salinity values. Salinity observations are generated for profiles that report only temperature. Salinity is computed from observed temperature values using bi-monthly climatological temperature-to-salinity regression models that have been computed on a global 0.25° resolution grid. The potential density profile is examined for inversions (higher density shallower than lower density), and observed temperature and salinity profile levels with inversions that exceed a minimum specified inversion threshold of 0.025 kg m-3 are flagged. For profiles with derived salinities, static instabilities are corrected by iteratively adjusting the derived salinity until the resulting profile is neutrally buoyant. Salinity is removed from the top of the permanent thermocline upward and added from that depth downward in the adjustment. The salinity correction algorithm is not applied to density inversions for profiles that observe both temperature and salinity levels, since it is difficult to determine a priori if the cause of the density inversion is due to the reported temperature or the reported salinity value. In this case profile levels with density inversions are simply flagged.

4.4.6.3 Vertical Gradient Checks

A global climatology of vertical mean temperature differences and standard deviations about these means has been computed from the historical profile archive. The climatology is used to test observed vertical temperature gradients for outliers. First, the climate temperature differences and variability are interpolated to the observation location and sampling time. Second, the vertical temperature differences are converted to vertical temperature gradients and interpolated to the observed profile levels. Observed vertical temperature gradients are computed, and the difference between the observed and the expected mean vertical gradient from the climatology is standardized by the expected gradient variability, z = (ATo • m-1 - ATc • mTl)lo (4.5)

where ATo • m-1 is the observed vertical gradient, ATc • m-1 is the climate mean vertical gradient, a is the variability about that mean, and z is the standardized vertical gradient variate. If the observed profile gradient exceeds 0.2°C-m-1 and |z| > 4, then the profile level is flagged. Experience has shown that the vertical gradient test based on climate statistics tends to spuriously flag as erroneous profile levels associated with a strong thermocline. This problem is particularly acute in the tropics. Truly erroneous profile vertical gradients are often associated with bad temperature or salinity observations, which are detected in the spike test and background field checks described previously. Hence, at the present time, quality control flags set by the vertical gradient check are flagged and only used for informational purposes (Sect. 5).

4.4.6.4 Profile Shape Comparisons

Observed profiles are compared to profiles extracted from the various background fields using a profile shape quality control procedure. This procedure has the advantage of taking an overview of the entire profile. Profile levels that have previously been determined to be unreliable based on other profile quality control data checks are excluded in the profile shape quality control procedure. The shape quality control procedure computes an integrated observed-minus-predicted statistic that takes into account level thicknesses. The test statistic is calculated as, n = ^((°k - Pk) • (Zk+1 - Zk-1 )/£ (Zk+1 - Zk-l) (4.6) k k where Ok is the observed value at level k, Pk is the prediction (background) value at level k, ak is the prediction error standard deviation at level k, and zk is the depth of level k. The probability of n being greater than zero is computed assuming a normal probability distribution function. The shape comparison statistic is analogous to a goodness-of-fit test of two cumulative distribution functions. It identifies observed profiles with large errors relative to the background profiles. Profiles that have large temperature or salinity differences over narrow depth ranges, such as dissimilar mixed layer depths, will be considered similar. Observed profile shape must be consistent with forecast and climate background profiles in order for the profile to be accepted into the analysis.

4.4.6.5 Gliders

Ocean gliders are autonomous platforms which fly in a saw-tooth-sampling pattern in the upper ocean by changing their buoyancy. Depending upon configuration, gliders sample profiles of pressure, temperature, and conductivity. The gliders surface at regular intervals to transmit their observations to shore or satellite based receivers. Gliders provide both downward and upward profiles of temperature and salinity, with glider position and time varying with depth during the dive. Quality control of glider data is similar to that of single profile data, other than relaxation of the strictly increasing depths check. However, several glider specific tests are performed that are, in most cases, functions of the vertical velocity of the glider. These tests are applied to gliders with a non-pumped CTD, where flow through the conductivity cell depends upon the speed of the glider, making the thermal inertial correction speed-dependent.

0 0

Post a comment