What is the data quality?

Historically, the various WOCE Data Assembly Centers (DACs) had specific criteria for evaluating and flagging data. These procedures are carried over into the CLIVAR program. The Hydrographic DAC, for example, has well-defined WOCE standards for data reduction and flagging of CTD data. For the shipboard ADCP, the processing and quality assessment have been left to the data originator. For the many cruises, the data have been processed and documented by the E.Firing ADCP Laboratory at the University of Hawaii, or using the Common Oceanographic Data Access System (CODAS) designed by this group. Profiles are screened objectively and subjectively, but personnel and methods change over time. Data originators are instructed to contribute only calibrated, processed, and documented data sets. Once these data arrive at the ADCP DAC, a general review of the data is undertaken, but no specific flagging system is added.

The heterogeneity of operational conditions--ship speed, sea state, and distribution of acoustic scatterers--makes it difficult to quantify the data quality for a given cruise. Conditions and data quality can vary greatly during a single cruise. Data quality also varies with depth in each profile, with problems being most common in the top bin and towards the end of the profiling range. The data user is therefore referred to the documentation file accompanying each cruise for notes about adverse conditions and corresponding data quality problems at specific times during a cruise. Additional quality information is available in the high-resolution data set stored in CODAS blockfiles. For each depth of each profile, a flag byte shows the results of editing, partly by hand and partly mechanized. A "percent good" byte records the percentage of single ping samples accepted by the ADCP for averaging at each depth bin in each ensemble-average profile (typically 5 minutes, the shortest interval for which data are saved). Both the flag byte and "percent good" (> 30%) are used when extracting data for the ASCII standard subset. For more insight and control, users may wish to extract data from the CODAS directly, using their own criteria.