Paper presented at the 1999 American
Meteorological Society Conference
15th International Conference on Interactive Information and Processing Systems (IIPS)
-------------------------
ENVIRONMENTAL CHAMBER TESTS OF NWS RADIOSONDE
RELATIVE HUMIDITY SENSORS
W. Blackmore and B. Taubvurtzel
National Weather Service, Office of Systems Operations
Note: Section 5.1 of this paper was updated to include the correct chamber RH setting (98% RH) for the H-humicap test and text was added to discuss the dry bias with the sensor. This version of the paper does not appear in the AMS conference preprints.
1. INTRODUCTION
Since October 1, 1993, the National Weather Service (NWS) has reported radiosonde relative humidity (RH) below 20% and at temperatures below -40O C. The reporting of RH below 20% ended a 20 year practice of reporting such values using a 30OC dewpoint depression in the coded message and 19% in the data archive. Before 1993, NWS reported all RH measurements below -40OC as missing.
The history of NWS radiosonde RH sensors, reporting practices, and the reasons for the cutoff in RH measurements in dry and cold conditions was discussed by Elliot and Gaffen (1991) and Wade (1994). Their studies focused on the carbon "hygristor" RH sensor, because at the time it was in use at all NWS upper-air stations. They noted that the RH in dry or cold environments was terminated because the hygristor could not provide accurate data in such conditions.
Wade (1994) showed that the inaccuracy of hygristor RH measurements below 20% was largely because of faulty coefficients used in the algorithm for converting radiosonde signals to RH for values below 20%. These incorrect coefficients (known as "1B" act to make the hygristor lose its sensitivity in dry conditions. Based on Wade's studies and requests from the data user community, NWS reinstated the reporting of RH below 20% and at temperatures below -40OC. However, it would be up to data users to determine the accuracy of the RH data and apply corrections to the data.
Since the change in RH reporting practices, NWS has implemented improvements in the measurement of hygristor RH below 20% (Blackmore and Lukes, 1998) and is using radiosondes that measure RH with different technology. Lastly, NWS has been assessing the accuracy of radiosonde RH sensor measurements in dry and cold environments. To aid in this effort, NWS has been testing RH sensors in environmental test chambers.
The intent of this paper is to provide some results of chamber tests conducted to date.
2. RADIOSONDES USED BY NWS
NWS currently uses three models of radiosondes in its upper-air network: the Vaisala RS-80-57H (60 stations), Sippican VIZ-B2 (29 stations), and Sippican Microsonde (3 stations). For about 40 years, the NWS has used Sippican Incorporated (formerly VIZ Manufacturing Company) radiosondes; the latest model was implemented in 1997 (Blackmore and Lukes, 1998). Since 1995, NWS has used radiosondes built by Vaisala Incorporated at a portion of its upper-air network.
2.1 Radiosonde RH Sensors
Sippican radiosondes use a carbon "hygristor" to measure RH. The sensors are similar in design and performance as those discussed by Stine (1965). It consists of a small strip of plastic dipped in a mixture of carbon particles, celluloid resin, and other chemicals, and then dried. As electric current passes through the strip, the carbon particles allow it to act as a resistor. The celluloid absorbs (diffuses) water vapor and expands (contracts) with changing RH. This varies the spacing between carbon particles, changing the resistance. High (low) RH results in high (low) resistance.
The "humicap" sensor (Salasmaa and Kostamo, 1975) is used to obtain RH measurements with Vaisala RS-80 radiosondes. The sensor is comprised of a glass substrate that supports a porous, dielectric material placed between two electrodes. When AC voltage is applied to the sensor, it acts as a capacitor. The dielectric material absorbs and diffuses water vapor as RH is varied, causing the capacitance of the dielectric material to change. Low (high) capacitance corresponds to low (high) RH.
3. CHAMBER TESTS OF RH SENSORS
One way NWS determines if radiosondes meet their specifications is by environmental chamber testing of the instruments. The radiosonde, or just a sensor, is placed in the chamber and data are collected while the chamber simulates the atmospheric conditions that radiosondes are exposed to during a typical sounding. When NWS started reporting radiosonde RH below 20% and -40OC, it wanted to determine the accuracy of these measurements. NWS obtained additional chamber equipment and developed techniques to test RH sensors in as dry and cold conditions as possible.
3.1 Thunder Scientific Chamber model 2500 (TS2500)
The TS2500 has a working volume of 31 by 25 by 31cm with a maximum air flow rate of 20 L/min. Its temperature range is +70O to -10OC and RH can be adjusted from 10% to 95%, at any temperature. The TS2500, which utilizes the proven "two-pressure" principle for generating RH, was used for tests of hygristors, especially their lock-in values. It was also used to test humicap sensor packs.
3.2 Thunder Scientific Humidity Generator model 3900 (TS3900)
Based on the combined Atwo-pressure@ and Atwo-temperature@ principle, the TS3900 supplies a continuous humidified gas stream within the dew/frost point range of !95O to +10OC. It has a flow rate up to 5 L/min, which is sufficient to change the atmosphere of the mini-vessel in about 2 minutes and the micro-vessel in a matter of seconds.
3.3 Mini and Micro Vessels
The mini-vessel is a cylinder 13 cm in diameter by 20 cm deep made of electro-polished stainless steel. A hermetically sealed lid with a pipe for cables prevents leaks. When submerged in the liquid of a temperature bath (range: +60O to !80OC), the air from a humidity generator passes through a 6mm stainless steel tube coiled around the outside of the vessel to equalize the temperature inside it with that of the bath. Two Platinum Resistance Thermometers (PRT) measure this temperature and two reference standard hygrometers sample the dew/frost point temperature. All RH calculations are made with respect to water. A data acquisition system graphically displays these parameters to determine stability of the readings, which are recorded at least 50 samples every six seconds. One humicap sensor pack or up to six hygristors can be tested at a time. The TS3900 is used mostly with this vessel. However, the TS2500 is used in some tests as a dew point generator and its output was instantly switched with that of the TS3900 to rapidly change RH from one test point to another at the same fixed temperature.
The micro-vessel is similar in design, but is only 2.5 cm in diameter by 5 cm deep and holds one hygristor and one PRT. It is used to evaluate response time of hygristors to very rapid changes in RH. For such tests, the TS3900 humidity generator pumped air with known RH into the vessel. After a few minutes of stabilization, the input is instantly switched to dry nitrogen, which purges the vessel in a few seconds. This procedure can be reversed, if desired. The resistance of the hygristor during this transition is recorded every second.
4. HYGRISTOR TEST RESULTS
Most of the tests conducted to date have focused on Sippican hygristors, because until recently, Sippican radiosondes were used at most NWS stations. Hygristor RH values were determined by using the measured hygristor resistances and applying them to the operational data conversion algorithm used with VIZ-B2 radiosondes (Potts 1996; Blackmore and Lukes 1998).
4.1 Sensor Calibration
The algorithm for converting hygristor resistance to RH includes calculating the ratio of the observed resistance to the sensor Alock-in@ resistance at 33% RH and 25OC. The lock-in value is measured at the factory and typical values are around 11,000 Ohms. The algorithm assumes that if the RH is held constant at 33% and the temperature varied, the lock-in resistance remains fixed. However, Stine (1965) reported that the lock-in resistance from sensors tested in that study increased significantly with decreasing temperature. Wade (1994) showed that errors in the lock-in value will cause increasingly larger RH errors in the sensor measurements as the ambient RH nears 0 percent.
To determine if the lock-in value from Sippican sensors increases with deceasing temperature, NWS conducted chamber tests measuring lock-in values ranging from +35O to -60OC. In each test the chamber RH was kept at 33% (+ 1% to 2%) as the temperature was varied. The mini- and micro-vessels and TS 2500 chambers were used to conduct these tests.
Figure 1 shows results from one test where six hygristors were simultaneously tested in the mini-vessel chamber from +25O to -25OC. Note that as the temperature decreased to -25OC, all six sensors showed an increase in their lock-in resistance by more than 1,500 Ohms. Note that for two sensors, the lock-in value increased by more than 2,000 Ohms. Other lock-in tests conducted in this temperatures range showed increases in resistance with decreasing temperature, but not always to the extent as shown in Figure 1. This increase in resistance is contrary to the what is found in the RH data conversion algorithm, where it is assumed that the lock-in value remains constant with temperature. Such increases result in the hygristor reporting RH values higher than the actual RH. As an example, an increase in lock-in value of 1,500 Ohms at -25OC introduces a positive bias near 10% as the ambient RH approaches 0%.
Figure 1. Graph showing electrical resistance (y-axis, kilo-Ohms) versus temperature (x-axis, degrees C) for six hygristors. Chamber RH was kept constant at 33%. Values seen at + 25OC are the sensor lock-in values.
From Figure 1 one might conclude that at -25OC or colder the lock-in resistance would increase further. Likewise, at warmer temperatures, it appears the lock-in value should remain constant or decrease. Lock-in resistance tests conducted at -40O, -50O, and -60OC with individual hygristors did not show increases in resistances exceeding 1,500 Ohms. In fact, tests at -60OC showed a slight decrease in the lock-in resistance. Lock-in tests conducted at temperatures warmer than +25OC were inconclusive. It should be noted, however, that only a few sensors were tested at these temperatures.
4.2 Time Response in Cold Temperatures
Studies by the World Meteorological Organization (WMO, 1996) and Schmidlin (1998) indicate that the hygristor responds quickly (less than 1 second) to RH changes between 70% and 30% at +10OC and warmer. As the temperature becomes colder, the WMO reported that the time constant of response increases significantly and is near 10 minutes at -50OC when the RH is decreasing.
NWS investigated the time response of hygristors from -20O to -60OC. However, NWS did not have the capability to accurately measure RH changes (e.g., 70% to 30%) at these temperatures. To solve this problem, hygristors were tested in the micro-vessel where the RH was rapidly switched from 0% to 50% and vice versa. By flushing the vessel with dry nitrogen gas, 0% RH was achieved.
Time constants could not be accurately determined, because in all the tests, the hygristor RH did not decrease to 0% RH when the chamber was rapidly purged with dry nitrogen. At -20OC, after being exposed to dry nitrogen for more than 10 minutes, the lowest hygristor RH reported was near 12%. At -40OC, the lowest RH reported was more than 20%, even after the sensor was exposed to dry nitrogen for more than 30 minutes. Although the time constants could not be determined, the change in sensor resistance with time was still measured. It slowed considerably with decreasing temperature, in agreement with the WMO findings. At -60OC , the sensor showed no indications of functioning as the RH was rapidly changed between 0% and 40%.
4.3 Hysteresis
Blackmore and Lukes (1998) showed that the current algorithm for computing hygristor RH provides improved sensitivity for RH measurements below 20%. However, the algorithm can provide negative RH values, which are set to 1%. Blackmore and Lukes showed an example of a VIZ-B2 RH profile with 1% layers in the lower troposphere. The dry layers were likely real, but the 1% values were probably a result of negative RH occurring in the layers rather than a measure of the true RH.
Further investigation of VIZ-B2 soundings indicated that negative RH values appear to result if the hygristor is exposed to high RH followed by rapid drying, which usually occurs when exiting a cloud. Such changes in RH may show hysteresis of the sensor, which results in erroneous RH values as reported by Brousaides (1975) and the WMO (1996).
NWS conducted chamber tests to verify that sensor hysteresis is the cause for the observed 1% (i.e., negative) RH values. A chamber test with the mini-vessel was configured to duplicate as close as possible a portion of a VIZ-B2 radiosonde sounding where 1% RH values were observed after the radiosonde exited a cloud. The RH data conversion algorithm was modified so that any negative RH values would be recorded. Six hygristors were exposed to a chamber RH of 95% and temperature of +15OC for 8 minutes, followed by drying to 10% RH. Figure 2 shows a plot of the results
Note from Figure 2 that all 6 sensors reported negative RH when the chamber RH was 10%. The magnitude of the errors would be less if the negative values were set to 1%, which is the current operational practice. When the chamber RH and temperature were set to 20% and - 5OC, respectively, the sensors continued to show a dry bias that ranged from about 10% to 20%. The test began and ended by measuring the lock-in resistance values for the sensors. Hysteresis caused the lock-in values to decrease by as much as 3,000 Ohms, resulting in the dry bias seen (last chamber test point) in the figure.
Other hysteresis tests, where the chamber RH was decreased more slowly from near saturation, showed the opposite effect; the hygristor RH was higher than the chamber RH by as much as 7%.
Figure 2. Graph showing RH (y-axis, %) from six hygristors (thin, dashed lines) versus chamber RH (Thick line) at +25O, +15O and -5OC chamber settings (x-axis).
5. HUMICAP TEST RESULTS
Unlike hygristors, humicaps can not be easily tested if they are removed from the radiosonde. This is due to the electrical properties of the sensor and radiosonde. To solve this problem, NWS tested the radiosonde with the transmitter electronics removed. With this configuration, humicap RH could be recorded directly. NWS uses Vaisala radiosondes equipped with the "H" model humicap and testing was done with this sensor type. Chamber tests of the H-humicap are still ongoing and only limited test results are currently available.
5.1 Hysteresis
To investigate H-humicap hysteresis, NWS exposed a sensor to similar chamber RH and temperature values used in the hygristor hysteresis tests. However, the sensor was exposed to 98% RH for only 3 minutes. The mini-vessel was used during the test, but only one sensor could be tested because of limited space in the vessel. Results of the test are shown in Figure 3. The results indicate that the H-humicap exhibited minimal hysteresis when the chamber RH was changed from near saturation to 10%. However, the H-humicap RH reported a dry bias of 6% at a chamber RH of 98% and 3% or less at the other test points.
Figure 3. Graph showing RH (y-axis, %) from one H-humicap (thin, dashed line) versus chamber RH (thick line) at +25O, +15O and -5OC chamber settings (x-axis).
6.0 CONCLUSIONS
NWS chamber tests of the hygristor showed that the lock-in resistance value does not remain constant. Cold temperatures and sensor hysteresis change the value significantly, causing RH errors that may exceed 10%. The time response of the sensor slows considerably at very cold temperature, in agreement with previous studies. Test results suggest that the sensor no longer functions at temperatures near -60OC or colder. Lastly, when it is exposed to high RH, followed by drying to low RH , sensor hysteresis tends to result, causing RH errors that may exceed 10%. Again, these findings are in agreement with earlier studies. It should be noted that Sippican has developed algorithms to help correct RH errors caused by hysteresis and lock-in value variations. However, NWS has found these corrections to be unsatisfactory at the present time.
When the H-humicap was exposed to high chamber RH, followed by drying to low RH, minimal hysteresis resulted. However a dry bias of 6% was observed with the sensor. NWS will continue testing of the H-humicap to help identify any deficiencies or biases in the sensor.
5. ACKNOWLEDGMENTS
We wish to thank Charles Wade, NCAR, for his assistance in developing the chamber tests, and Carollyne Hutter for her editorial assistance.
6. REFERENCES
Blackmore, W.H., and M.M. Lukes, 1998: Implementation of the VIZ-B2 radiosonde at NWS upper-air stations. Preprints, Tenth Symp. on Meteorological Observations and Instrumentation. Phoenix, Amer. Meteor. Soc., 22-27.
Brousaides, F.J., 1975: The radiosonde hygristor and low humidity measurements. Bull. Amer. Meteor. Soc., 56, 229-234.
Elliot, W.P., and D.J. Gaffen, 1991: On the utility radiosonde humidity archives for climate studies. Bull. Amer. Meteor. Soc., 72, 1507-1520.
Potts, L.W., 1996: Humidity sensor equations. VIZ Tech. Publication No 80415 (E), Rev. 960904, 6 pp.
Salasmaa,E., and P. Kostamo, 1975: New thin film humidity sensor. Preprints, Third Symp. on Meteorological Observations. and Instrumentation. Phoenix, Amer. Meteor. Soc., 33-38.
Schmidlin, F.J, 1998: Radiosonde relative humidity sensor performance: The WMO intercomparison-Sept 1995. Preprints, Tenth Symp. on Meteorological Observations and Instrumentation. Phoenix, Amer. Meteor. Soc., 68-71.
Stine,S.L.,1965: Carbon humidity elements: manufacture, performance and theory. Vol. 1, Humidity and Moisture, Robert E. Ruskin, Ed., Reinhold Pub Corp., 316-330.
Wade, C.G., 1994: An evaluation of the problems affecting the measurement of low relative humidity on the United States radiosonde, Journal of Atmospheric and Oceanic Technology. 11, 687-700.
World Meteorological Organization, 1996: Guide to meteorological instruments and methods of observation Sixth Edition, WMO No. 8.
----------------------------------------------------------------------------------------------------------------------------------------
Back to the Upper-air Observations Main Page