This presentation focuses on methods to infer complete metadata and determine instrument adjustments at over 1200 stations (including ships) from 1973 to 2002. The inferred metadata is to be provided to the Comprehensive Aerological Reference Data Set project at NCDC. A companion presentation in the Symposium on Observing and Understanding Water Vapor Variations discusses global precipitable water trends incorporating these preliminary instrument adjustments.
To infer complete metadata, common characteristics of each instrument type are sought using monthly time series of more than 200 variables at each station. Moisture variables (primarily the average dew point depression at 500 hPa) are the most useful for distinguishing instrument types, especially combined with measures of processing and reporting practices (such as the lowest relative humidity reported, lowest pressure or temperature with a reported dew point, and number of reported levels per sounding). Signatures of instrument types at well-documented stations are also seen at stations with little or no documentation, allowing reliable inference of instruments and transitions. Initial examination of 1973 to 1996 data found 40 distinguishable instrument types, but few instrumental tropospheric or stratospheric temperature discontinuities.
After identifying instrument types, biases can be removed (but lost information cannot be restored) by adjusting each instrument to a "reference" instrument. Here, the "reference" is the average of Vaisala RS80 and recent VIZ (Sippican) instruments. This research focuses on tropospheric moisture. It develops dew point adjustments for each instrument type by comparing cumulative distributions of dew point depressions in different pressure and temperature ranges at a large number of stations (before and after an instrument change, or where adjacent stations simultaneously use different instruments). A short "chain" of adjustments may be needed to adjust to the reference, such as instrument "Type A" to "Type B" to the reference. Globally, the average correction reduced precipitable water by about 4 percent in the 1970s, but by less than 1 percent in the 1990s.