Temperature and humidity measurements during icing conditions
Michel Leroy*, Bengt Tammelin**, Reijo Hyvönen**, Jacques Rast***, Marc Musa***
*Météo-France, Direction des systèmes d’observation, BP 202, 78195 Trappes, France
**Finnish Meteorological Institute
***MeteoSwiss
Introduction
Experimental data on the effect of icing on meteorological sensors have been published in the literature. The EUMETNET project ‘Specifications on severe weather sensors’ (Tammelin and al., 1999) summarized most of the knowledge in this field. Following this report, SWS II, a new EUMETNET project on the effect of icing on sensors was launched in 2000. Field tests have been conducted at three test sites during the winter 2001-2002: Luosto in northern Finland, Mont Säntis in the Swiss Alps, Mont Aigoual in south of France (Tammelin and al, 2001, 2002). The effect of icing is generally studied for wind measurements. Temperature and humidity measurements are also affected by icing. The Eumetnet SWS II project included the test of different sensors and screens used for temperature and relative humidity (RH) measurements.
Instruments tested
The following instruments have been installed and tested.
Three types of naturally ventilated multi-plate screens :
· Vaisala (Luosto, Säntis), including a Pt100 probe and a Vaisala HMP45 humidity sensor.
· Socrima (Aigoual), a screen widely used in the Meteo-France network, including a Pt100 probe and a Vaisala HMP35DE humidity sensor.
· IRDAM (Aigoual, Luosto) is a hot film anemometer. It includes at the bottom of the sensor body a small multi-plate screen with a Pt100 probe and a RH sensor.
Four types of artificially ventilated screens :
· Rotronic (Aigoual, Luosto, Säntis) : The air is aspirated at the bottom centre and leaks along the protection cap. The screen includes a Pt100 and a Rotronic capacitive RH sensor.
· Eigenbrödt (Luosto) : A ventilator is installed at the top of a multiplate screen, including a Pt100 probe and a RH sensor (a Vaisala HMP45 has been used).
· THYGAN (Aigoual, Luosto, Säntis) is a chilled mirror dew point sensor. The air is aspirated at the bottom and rejected at the top. There is a measurement cycle of 10 minutes, including some internal and external heating process. The sensor outputs dew point and the air temperature.
· Vaisala HMP243 (Luosto) is a mixed system. Dew point is calculated using a capacitive RH sensor in an artificially heated air flow, where the RH and the temperature are measured. The air temperature is also measured under a naturally ventilated multi-plate screen.
Three wind sensors, delivering also the air temperature :
· Metek USA-1 (Aigoual, Lusoto, Säntis) is a 3D sonic, calculating the sonic temperature.
· Thies (Luosto) is a 2D sonic, , calculating the sonic temperature.
· NOWA (Säntis) is a hot wire anemometer. Thermocouples in the sensor body measure the temperature.
Choice of a reference
Considering past tests and experiences, no instrument was considered as an absolute reference known to be ice free under all circumstances. Nevertheless, the THYGAN was selected and used as a relative reference for the following reasons :
· A long and successful experience from MeteoSwiss.
· The physical principle used for the measurement of dew point (chilled mirror), widely used in laboratories. The Thygan has been designed for field measurements.
· The artificially ventilation minimizing radiation errors.
· A heating cycle, including the sensor body, which minimizes the risk of ventilation blocking by ice or snow.
· The installation of this sensor on each test site.
Measurement errors occur with this sensor when the air flow is totally blocked by ice or snow. Such extreme circumstances have been identified, using visual observation of the sensor (when available), an internal status of the Thygan and comparison with temperature measured by a sonic anemometer. When the Thygan airflow is blocked, the indicated temperature is likely to be warmer than the true air temperature. The RH is also lowered, both RH and temperature exhibiting unusual rapid variations. The sonic temperature has been seen to be biased with a near constant value. When the Thygan air flow is blocked, the temperature differences between Thygan and sonic increase and this may be a good indicator of a problem. Performance of sonic temperatures during the SWS II tests have been studied (Musa, 2002).
Data analysis
Data were acquired every 10 minutes from 1-Sept-2001 to 30-April-2002. Numerous icing events were experienced at all three sites. Visual observations of the sensors were made at Mont Säntis and Mont Aigoual, while a video camera was permanently observing the sensors at Luosto. From these observations, a classification of the icing state of each sensor has been manually done : 0 for no ice to 3 for sensitive elements completely covered with ice.
Scatter plots of a given sensor compared to the Thygan have been drawn, taking in account the different ice classifications (one different colour for each icing class). Such plots have been drawn for each parameter (T, RH), each sensor and each test site. The final SWS II report will include detailed and exhaustive results. An example is shown.
On the whole test period, the majority of the temperature measurements taken under a screen are within ±2°C of the Thygan values. Larger differences, up to 5°C, may occur in specific conditions, which will be detailed below.
A 10 days period with a long icing period at Mont Aigoual is shown below and is a good illustration of the major effect of icing on the different types of measurement. For RH, the measurement of Thygan (reference), Socrima (multi-plate screen) and Rotronic (artificially ventilated screen) are shown. For temperature, the sonic temperature from Metek is added. Measurements from IRDAM (small multi-plate screen) are not shown to avoid confusion on the graphs.
· During period 1, no icing occurs. The temperature measurements are close to the Thygan reference. There is a constant bias of about –2.4 °C for the Metek sonic temperature. The Rotronic screen exhibits warmer temperature (~1°C) around noon. This is an effect of solar radiation on this type of screen. The RH measurements are so close together that they cannot be separated on the graph.
· 2 indicates the beginning of icing. Photos show that the screens are rapidly covered with ice.
· Quite quickly, during period 3 and 12 hours after 2, the Rotronic begins to indicate higher temperature values (~3°C). The IRDAM temperature also indicates higher values. Photos show that both screens are blocked with ice, including the aspiration of the Rotronic. An explanation of higher Rotronic temperature values, both during day and night (thus excluding solar radiation errors), is that the internal air, not replaced by outside air, is heated by the ventilator itself. Being heated, the RH of the internal air is lowered as shown on the RH graph. For the IRDAM, the small multi-plate screen being blocked by ice, the temperature inside is increased by conduction of the heat in the body, which is large enough (300 W) to keep the body free of ice. The IRDAM RH stays close to 100%. The Socrima screen is also blocked with ice. But there is no internal heat source and the measured temperature stays very close to the Thygan. The ice on (and inside) the screen is at the same temperature than the air. The Socrima RH stayed close to 100%, as the reference Thygan RH, fog occurring during this whole period.
· Period 4. Photos show that the Thygan is blocked with ice. Though icing conditions continue, the Thygan RH decreases and exhibits rapid fluctuations. The Thygan temperature also increases, compared to the Metek sonic temperature (and Socrima). The physical explanation is the same that for the Rotronic during period 3 : the aspiration inlet is blocked with ice and the internal air is warmed, increasing the measured temperature and decreasing the RH. The differences between Socrima and Metek temperatures remain constant, indicating that these two measurements remain unaffected. The sunny morning of the 19 February did not generate radiation errors with the Socrima.
· Period 5. the Thygan ventilation is no longer blocked with ice. Thygan, Socrima and Metek (with bias) temperatures are close together. Rotronic progressively recovers as the aspiration inlet becomes ice free, Rotronic RH increases. At the end of this period 5, the Thygan and Rotronic RH decrease down to 45%. The Socrima RH remains very high (>90%), due to remaining ice on the screen, which is still disturbing the air close to the RH sensor.
Some conclusions
A deeper analysis and detailed results for each type of screen tested will be part of the final SWS II project report.
With the exception of the Thygan, the screens are not artificially heated, so ice may remain on the screen during long periods, the ice disappearance depending on solar radiation and/or positive air temperature.
A first conclusion is that the effect of icing on the temperature measurements is much less spectacular and more limited than the effect of icing on wind measurements. The majority of measurements are within ± 2 °C of the Thygan. Larger errors, up to 5°C, may occur when the aspiration intake of artificially ventilated screens is blocked with ice, due to internal heat sources or solar radiation. Naturally ventilated multi-plate screens may also exhibit large temperature errors (5°C), when completely covered with ice, in presence of solar radiation. But surprisingly, some multi-plate screens have less frequent large errors than some artificially ventilated screens. This is not the case of RH measurements. Very large errors occur when a screen is covered with ice, except during icing events, when both outside RH and inside (the screen) RH are close to 100%. RH measurements inside artificially ventilated screens are correct, as soon as the ventilation inlet is not blocked.
References
· Musa M., 2002. Measurements of temperature with wind sensors during severe winter conditions. WMO/CIMO TECO2002.
· Tammelin B., Joss J. And Haapalainen, J., 1999. Specification on severe weather sensors. Second Edition, Finnish Meteorological Institute, Helsinki. CD-ROM.
· Tammelin B. and al., 2002. Ice-free sensors. The Eumetnet SWS II project. WMO/CIMO TECO2002.
· Tammelin B., Heimo A., Leroy M., rast J., Säntti K., 2001. Meteorological measurements under icing conditions. Eumetnet SWS II project. Finnish Meteorological Institute, Helsinki.
Example of scatter plot from Luosto.
Variations with time of several temperature and RH measurements, discussed in text.