|Clockwise from bottom left: June Wang, Hal Cole, Katy Beierle and ATD director Dave Carlson. Cole heads the Surface and Sounding Systems Facility within ATD. Beierle oversaw corrections for more than 8,000 Vaisala soundings. (Photo by Carlye Calvin.)|
As far back as 1997, ATD researchers began probing into a mystery caused by unexpected humidity readings taken from Vaisala radiosondes. The readings, collected in TOGA COARE (the Tropical Ocean and Global Atmospheres Coupled Ocean-Atmosphere Response Experiment), showed moisture levels in the lower atmosphere a few percentage points lower than expected from independent measurements at the surface. The resulting effect on instability indices was even more pronounced, hinting that large stretches of the western Pacific shouldnt have been experiencing clouds and storms when they actually were.
Now, after almost five years of meticulous work, a team from ATD and Helsinki, Finland-based Vaisala has solved the mystery of the too-dry radiosonde data. ATDs June Wang is lead author of an upcoming paper in the Journal of Atmospheric and Oceanic Technology that details the interwoven factors that led to the bias and provides algorithms that researchers can use to fix archived data. The corrections were produced through teamwork with engineers Ari Paukkunen and Tapani Laine at Vaisala, which makes just over half of the radiosondes used for daily global measurements.
Tests at a Vaisala plant found one source of error: nonwater molecules from the radiosonde packaging gradually seeped onto the polymer-based humidity sensor, reducing its ability to absorb moisture. Even when a radiosonde is stored for only three or four months, as is typical in operational settings, this factor can induce a dry bias. In a field project, researchers may use sondes that have been sitting on the shelf for years, creating even larger errors than those found in TOGA COARE.
Another problem was the equation used to derive humidity at the ambient temperature from the humidity at the calibration temperature. This formula assumed a linear relationship, while Vaisala found that the actual response was nonlinear. A third source of error involved the basic calibration model used to derive humidity from the sensor capacitance, and yet another bias arose from the humidity sensor arm, which can heat quickly before launch on hot days and produce a misleadingly high saturation temperature.
Two generations of Vaisala sensors were analyzed (the older one has been in use since 1980), and each had a different bias profile. Based on the tests conducted at Vaisala, ATD came up with the algorithms to address these sources of bias. The Vaisala soundings collected in TOGA COAREmore than 8,000 in allhave been scrutinized and fixed one by one, says June, making it one of the most-examined and highest-quality radiosonde data sets ever collected. Meanwhile, Vaisala added a new type of protective shield around the temperature humidity sensors, and it changed the dessicant in the sonde packaging from clay to an impurity-absorbing mix of charcoal and silica gel. June stresses the importance of careful storage and launch techniques, including a ground-truth observation taken before launch.
For their work, the ATD team of Hal Cole, Erik Miller, Katy Beierle, and Junealong with Vaisalas Paukkunenwere nominated for a 2001 Outstanding Accomplishment Award for Technical and Scientific Achievement.
At this summers field campaign of the International H2O Project (IHOP 2002), to be based in Oklahoma, June and her colleagues plan to examine data from high-end reference radiosondes. Each will carry three humidity sensors, including an extra-precise Swiss device that uses a chilled mirror on which moisture condenses.
Deluxe sondes like these may be invaluable in detecting climate change, says June, since the standard global network now includes a mix of radiosondes from roughly a dozen manufacturers, each with their own biases. June also plans to compare sondes with other humidity sensors at IHOP. We want to validate some of the new instruments and get their accuracy for data assimilation and modeling. Eventually we want to integrate all the data.
Edited by David Hosansky,
Prepared for the Web by Carlye Calvin
Last revised: Tues Mar 14 17:08:40 MST 2001