Upper air temperature and moisture trends derived from historical radiosonde data are questioned because of frequent instrument changes which cause false trends, mainly cooling and drying. This presentation summarizes work in progress to produce complete global historical radiosonde metadata and develop adjustments for temperature and dew point from each distinct instrument type to a "reference." The goal of this project is to determine unbiased global and regional trends of total precipitable water, eventually back to the late 1950s.

This project examines archived soundings (currently 1973 to May 2004) from over 2500 stations, including 200 ships, to identify signatures of instrument types at well-documented stations. Similar signatures can identify instruments and changes at stations with incomplete or inaccurate documentation. Inferring instruments is not random because there are only a few candidate instruments in each case. Elevations are also checked hydrostatically because almost all changes are reported at an incorrect time.

Bias adjustments from each distinct instrument to a "reference" instrument will be developed from appropriate comparisons of pairs of instruments, including the same station before and after a change, a station with alternating instrument types, and adjacent stations simultaneously using different instruments. Temperature and dew point depression adjustments at each level of each sounding will be stratified by factors such as pressure, temperature, and sun angle.

A preliminary project covering 1973-1996 shows distinct global-scale regimes of total precipitable water: Dry in 1974-1976, moist in 1979-1988, and almost as dry since 1992 as 1974-1976. Transitions coincide with documented climate shifts. The dew point adjustments reduce precipitable water by about 4 percent in the middle 1970s (an average station was 4 percent wetter than the "reference") but by less than 1 percent since 1990. Most countries switched to drier instruments between about 1977 and 1986. China, Russia, and India are finally starting transitions to drier instruments.

The outstanding feature of recent (unadjusted) global averages is extreme ENSO-related moistening in 1998 and drying through 2000, with 2000 averaging 7 percent drier than 1998. Instrument adjustments will have little effect on the size of this event or on the overall lack of a moistening or drying trend since the early 1990s.

Most researchers consider the Russian radiosonde history to be very problematic. Steps and challenges of developing complete Russian radiosonde metadata are illustrated by examining 239 stations starting 1973, including 6 in Antarctica, 3 Arctic ice islands, and 27 ships. Preliminary findings are as follows:

1. Of the 6 radiosonde series used in this period, the 4 older types using goldbeaters skin hygrometers (A22, RKZ, MARS, and MRZ) are not distinguishable by differences in temperatures or dew points, but are usually distinguishable by data reporting practices such as the coldest temperature with dew point reported. MARS soundings can usually be exactly identified.

2. The 2 newer types use capacitive hygristors (MRZ-3AM domestically-produced and RF95 from Vaisala). MRZ-3AM is drier than the A22 through MRZ series and RF95 is drier than MRZ-3AM, with differences most noticeable at upper levels. MRZ-3AM and RF95 are used at only a few stations by 2004.

3. Published metadata for Russian land stations appears about 80 percent complete (some transitions omitted) and about 70 percent correct (some transitions are to a different instrument type, or more than a few months from the stated date).

4. Most station histories are quite complex, with frequent brief or long-term transitions back to the preceding instrument type. Recently, some stations have used up to 4 instrument types in the same period.