An Evaluation of Problems Affecting the Measurement of Low Relative Humidity on the United States Radiosonde

Abstract

This paper explores the low-humidity problem that has plagued the radiosonde hygristor for nearly 30 years and that makes the hygristor appear to become insensitive at relative humidities (RH) below about 20% RH. The problem led the National Weather Service (NWS) in 1973 to begin a practice of terminating the radiosonde humidity measurement at 20% RH and to begin reporting any humidities evaluated at less than 20% using a 30°C dewpoint depression in the coded radiosonde message. This practice remains in effect today and has resulted in a permanent, 20-year gap in the radiosonde humidity archive for the United States. This study examines a number of factors that can potentially affect the accuracy of NWS radiosonde low-humidity data, including 1) characteristics of the hygristor response curve, 2) characteristics of the NWS analog radiosonde and methods used to transmit and record the radiosonde signal, and 3) reduction techniques used to convert the radiosonde signal into RH. It is shown that the primary factor that biases the NWS low-humidity data, and that makes the hygristor appear to lose its sensitivity below 20% RH, is an error in the placement of the 10% and 15% RH lines on the humidity evaluator used to convert the radiosonde signal into RH. This error has propagated forward to the present because of an NWS requirement that current humidity reduction algorithms match the evaluator within 1% RH. The paper reviews the past work of Brousaides, which initially described the error in the evaluator, and presents results of recent tests conducted by the manufacturer of the hygristor that corroborate Brousaides's earlier work. The humidity reduction algorithm currently used by the NWS is described, and it is shown that by changing the coefficients used to derive the low-humidity data, the sensitivity of the hygristor can be restored. An example of data correction using soundings collected in a dry-microburst environment is presented. The paper discusses the decreased resolution inherent in the radiosonde recorder trace in the low-humidity region, but shows that current automated sounding systems have eliminated this device as a recording medium. Limitations in the accuracy of the low-humidity measurement stemming from uncertainties in the sensor's lock-in resistance are also discussed. The paper recommends a change in the low-humidity reduction algorithm used by the NWS, and a reversal of the 20-year-old practice of truncating the humidity report at 20% RH.