International Comprehensive Ocean-Atmosphere Data Set (ICOADS)
Moisture Variable Corrections for 1988 and 1992-1993
The new 1992-93 data represent an extension of Release 1a through 1993, and a reprocessing for 1992 (referred to as the "1992-93 extension"). Data for 1992 were reprocessed to take advantage of additional ship logbooks (foreign- and US-keyed) that became available since completion of Release 1a. During reprocessing, however, the COADS project discovered an error in dew point temperatures (DPT) from US-recruited ship logbook data keyed at NCDC (deck 927), such that erroneous DPT values averaged too high by approximately one degree Celsius. On a global basis, this deck contributed about 20% of all DPT estimates, and the error was estimated to bias global mean relative humidity (R), for example, by about +1%.
This DPT error impacted Release 1a (MSTG) statistics of specific humidity (Q) as well as R (and variables derived using the humidities), predominantly during May through November 1992 in the North Pacific and to a lesser extent North Atlantic. The error was corrected by recomputing DPT in deck 927 for all of 1992-93 by assuming the keyed observations of wet bulb temperature (WBT) were correct. Post-correction analyses have confirmed significant improvement in the observational records and in the monthly statistics involving humidities.
A distinct deck 927 problem yielding DPT errors of similar magnitude and spatial coverage was discovered in Release 1a data for approximately January-May 1988. In this case the keyed WBT value appears to have been inadvertently substituted into the DPT field position, from which a spurious value for WBT was calculated. From this spurious WBT a "DPT" was recalculated according to a biased formula, yielding results biased too high by up to several tenths of a degree Celsius.
As an approximate correction within deck 927, the contents of the DPT field were transferred into the WBT field, and DPT was recomputed from the new contents of WBT. The erroneous handling of the moisture variables appears to have also resulted in an unrealistic number of reports with saturation conditions (air temperature, WBT, and DPT all equal). WBT and DPT were thus deleted in reports containing saturation conditions. All of these 1988 corrective measures were further constrained to apply only to reports with a quality control date of December 1988 (as assigned by NCDC).
Here again, post-correction analyses of the 1988 period have shown significant improvement in the monthly statistics of Q and R, but also evidence for residual positive biases (e.g., order 0.5% in global mean R), apparently due to the aforementioned biased recalculation of "DPT." We are recommending that NCDC consider re-digitizing the data for this period because a more accurate correction based on the available digital records appears to be extremely difficult. Since it seems impossible to reliably distinguish reports affected by the problem, for example, in some cases the correction probably corrupted good data, including legitimate records of saturation conditions.
During analyses of these problems other less significant but still suspicious- looking transitions in the global humidity records were discovered, possibly related to WMO code changes and data mixture issues. For example, a drop in 1982 (order -0.5% in global mean R) may be partially explained by an uncorrected truncation error that exists in dew point depressions in the basic archive of data from the National Meteorological Center (NMC). This error, which was corrected by NMC sometime before 9 May 1984, appears to impact only mixed precision temperatures, i.e., air temperature reported to tenths Celsius and DPT to whole degrees (in accordance with the WMO Ship Code before 1982). The 1982 change in the Ship Code to report DPT to tenths of degrees Celsius, reduced but did not eliminate mixed precision temperatures, because some ships (e.g., US-recruited) continued to report DPT in whole degrees. These additional problems are being looked into as time permits. Documentation will be updated as more becomes known.
U.S. National Oceanic and Atmospheric Administration hosts the icoads website privacy disclaimer
Document maintained by firstname.lastname@example.org
Updated: Nov 8, 2005 23:50:33 UTC