The successful use of sensors for objective weather measurement

This article will give an overview of the most widely used meteorological sensors and their strengths and weaknesses. The word ‘sensors’ is here being used a little loosely and will sometimes include transducers, describing the conversion of the measured variable to another. There are innumerable sensor types in use and it is impossible to cover more than a fraction of them.

This article does not go into detail concerning more elaborate measurements using optical sensors of parameters such as cloud or visibility nor does it deal with data logging or display, instead the focus is on the sensors commonly used in monitoring weather at or near the surface.

Temperature

There are many techniques for measuring temperature based on physical properties that change with temperature. The most familiar temperature sensor to most people is the liquid in glass thermometer which exploits the expansion of a material with temperature. Fig 1 shows liquid in glass thermometers in a radiation screen. Meteorologists often use mercury as the measuring liquid although alcohol in glass thermometers are needed for temperatures below the freezing point of mercury, -39°C.

The expansion of material with temperature is also exploited for temperature measurement in the form of bi-metallic strips. These are formed of two strips of different metals fixed together. The differing expansion with temperature causes the strip to bend. This is typically used to move a pen across paper to leave a trace to form a device known as a thermograph.

Materials that change their electrical properties in response to temperature changes can be used to make temperature sensors. The most simple is the change in resistance of metals with temperature. Possibly the most widely used temperature sensor for weather monitoring applications is the ‘Pt100’ sensor. This is a platinum resistor which has a resistance of 100ohms at 0°C. Platinum is widely used in this application because its resistance has a large variation with temperature, a Pt100 will have a resistance of about 103.9 ohms at 10°C. These changes are still rather small so some care is needed in use to eliminate or compensate for the effects of lead resistance. The Pt100 is something of an industry standard but Pt1000 sensors, with a resistance of 1000ohms at 0°C are also available.

Semiconductor materials can be manufactured with useful resistance variation with temperature. A thermistor is a semiconductor device with a resistance which varies with temperature. The form of the variation is:

R=a x exp (b/T) Where R is resistance, T is absolute temperature and a and b coefficients for the sensor.

Thermistors are available in a huge variety of sizes, shapes, resistances and coefficients. Their main disadvantage is the non linearity of the resistance variation with temperature. This can be allowed for in subsequent data processing.

Other electronic devices can be used for temperature measurement. Capacitive temperature sensors are used in specialist fields such as radiosondes where a low mass sensor to measure a wide range of temperatures is required. Temperature sensors based on the current in a back-biased silicon diode are widely used for applications that do not require high levels of accuracy. Thermocouples, based on the small voltages generated by dissimilar metals in contact are used in various industries. However this technique can only measure temperature differences not absolute temperatures so that a real or synthetic ‘reference junction’ has to be created and unless the temperature differences are huge only very small voltages are created, typically less than a millivolt for a 20°C difference.

Whatever technique is used the installation and location of the sensor is likely to be more important to the quality of the measurement than the sensor itself. Any temperature sensor only measures its own temperature. The key to good quality measurements of atmospheric temperature is therefore to make sure this is the same as the sensor temperature. The major problem is the effect of radiation on the sensor. In daytime sensors have to be protected from sunlight and at night they have to be prevented from cooling by their own thermal radiation.

The usual way to achieve this is to put the sensor inside a radiation shield which is white or polished metal to scatter or reflect solar radiation. The traditional ‘Stevenson screen’ with its white louvres is familiar to most people. They are made in various sizes according to the size of the sensors to be housed. Most Stevenson screens in use are wood and are difficult to maintain and clean with their multiple wooden slats. Now similar designs are available in durable and easy to maintain plastic. This is not just a matter of maintenance costs, a dirty screen will lose much of its effectiveness.

An interesting development in recent years concerns the internal colour of radiation screens. Traditional Stevenson screens were white inside and out but many screens are now black inside. This is because of evidence that the benefits of eliminating the scattering of light inside the screen outweighs any additional heating by radiation inside. Fig 1 shows a Stevenson screen in modern materials and the sensor in fig 2 incorporates a modern compact screen based on an array of plates. Both are black inside.

Another way to reduce the effects of radiation on temperature sensors is aspiration, which is drawing air over the sensor. Aspirated sensors do have a lot of advantages especially in light wind and also have faster response times, albeit at the cost of requiring power. It’s also worth noting that very small or thin sensors suffer from radiation errors less than larger ones.

Relative humidity (RH)

The measurement of humidity in general was covered comprehensively in the December/January issue of AWE by J.C.Jones so I will here only deal with sensors applied to weather monitoring.

Most modern weather monitoring systems use capacitive sensors. They work well but their accuracy usually falls at RH values above about 90%. An interesting application of these sensors involves using a warmed sensing element. The temperature of this element is measured and used with the measured RH to calculate a dew point. Since dew point is invariant with temperature change this dew point is a good measure of the dew point of the ambient air. Additionally if the whole probe head is warmed the risk of adjacent condensation is reduced. This not only keeps the measurement range of the sensor away from 100% RH but also allows a faster response and recovery in saturating conditions such as in fog.

An additional measurement of the actual air temperature allows other variables such as RH and vapour pressure to be calculated. The traditional method for meteorologists to measure relative humidity is the wet and dry bulb thermometer or ‘psychrometer’. This is based on comparing the temperature from a normal thermometer, the ‘dry bulb’ with that from one covered with a damp wick, the ‘wet bulb’. Note the thermometers can be electronic sensors as well as liquid in glass.

The psychrometric formula below allows vapour pressure to be calculated from the wet and dry bulb temperatures: e = e s – Ap(T-T w )

Where e is vapour pressure, es is saturated vapour pressure at the wet bulb temperature, T w . The atmospheric pressure is p, the dry bulb temperature is T and A is the psychrometric constant. Unfortunately A is not very constant in practice and varies with the airflow over the sensors and sensor geometry. Values for A in widespread use range from about 0.7 to 0.8 x 10 -3 K -1 .

Wet and dry bulb thermometers can still be found in use today despite serious limitations. The wick needs careful maintenance and should be kept clean and a source of distilled water for the wick presents problems for automated measurement. Some applications may require a ‘wet-bulb temperature’ but even in these cases it is probably best to use a more modern technique and calculate it from standard formulae or tables. It is doubtful if wet and dry bulb thermometry has any place in modern serious professional applications except in unusual circumstances.

Other sensors sometimes used are chilled mirror devices and devices using absorption of light by water vapour. The most popular of the latter use either the ultra-violet Lyman-Alpha band or absorption bands in the infra red. All these are mostly used for research applications.

Rainfall and other precipitation

Rainfall information falls into two distinct types. Firstly there is information on the amount or ‘accumulation’ of precipitation that has fallen over a particular period of time. Accumulation is measured in terms of the depth of water, or water equivalent for frozen precipitation, that has fallen. Secondly there is information on whether precipitation is occurring at a particular time and its type and intensity.

Accumulation is measured with some sort of collecting bucket or rain gauge. The simplest form is just a cylinder with a funnel leading to a collecting vessel. The amount of water in the vessel gives a measure of the accumulation of precipitation. Alternatively rain gauges can be fitted with a mechanism to automatically record when particular amounts of precipitation have fallen. The most widely used example of this is the tipping bucket rain gauge. These have a funnel feeding collected water into a bucket which ‘tips’ over when a particular amount of water has collected.

This tipping point is calibrated to represent a particular accumulation usually around 0.1 or 0.2 mm. The tipping of the bucket operates a reed relay allowing an electrical signal to record and/or transmit information. Alternatively the water in the collecting vessel can be weighed continuously. These techniques allow for some information on the rate of accumulation to be obtained.

Rain gauges are not as simple to use as might be imagined. The aerodynamic properties of gauges are very important as turbulence around the rim can lead to precipitation missing the inlet. To avoid this some rain gauges are made in an hourglass or similar shape which has been shown to reduce these errors. These may be made in the form of two cones point to point but the ultimate example of this concept is the rather elegant gauge in fig 3.

An alternative approach is a wind shield such as that in fig 4. There are various designs for these and their use is especially important if the precipitation is in the form of snow. The location of rain gauges is critical to their accuracy and various schemes have been tried involving gravel surrounds or grids to prevent water splashing into the gauge.

Capacitive or resistance based sensors are available that rely on liquid on the sensor surface altering the electrical properties to an extent dependent on the rate of precipitation. These sensors have to be heated to prevent the build up of liquid but are generally straightforward and widely used.

Another interesting technique is the so-called ‘acoustic disdrometer’. This is basically just a matter of listening to the sound of raindrops or other precipitation particles and using a relationship between the particle size and type and the sound made to calculate precipitation intensity. One practical form of these sensors is a plate resting on piezo electric material. The device in fig 2 has such a sensor on top. As might be imagined they have limitations for snow or drizzle but are very effective for applications requiring quick warning and assessment of heavy rain showers.

Solar radiation and sunshine

Solar radiation and sunshine are distinct and different measurements. Solar radiation refers to the amount of energy received from the sun. It may be measured in various ways and different components are used in different applications, for example total solar radiation is the energy over all wavelengths received at a horizontal plane usually measured in watts per m2. This is probably the simplest and most widely used measurement.

Other variations include ‘diffuse solar radiation’, which is limited to that outside the direction of the sun, ‘net radiation’ which is a measure of radiation balance and includes long wave radiation from the underlying surface and photosynthetically active radiation or PAR which is limited to wavelengths used by plants in photosynthesis. All these are usually measured using one of two basic techniques. The most accurate involves exposing an array of thermocouple junctions to solar radiation, which may be filtered or shaded according to the particular measurement.

The output voltage for a given level of radiation is increased by using a number of junctions in series but even so outputs are still low, typically 10 μV/Wm -2 giving voltages of only a few mV in bright sunshine.

The measurement of sunshine refers to recording whether ‘the sun is shining’ and recording the duration of sunshine. This is not a straightforward parameter to define. Older measurement techniques using lenses to focus sunlight onto cards which would burn at some threshold were widely used for many years.

However they are not easily incorporated into automatic measurement and are being progressively replaced by devices whereby sunlight operates bimetallic strips or semiconductor sensors to indicate if the sun is shining. The World Meteorological Organisation recommends that the sunshine threshold should be 120 W/m 2 in the direct beam. There are numerous ingenious sensor designs available by which diffuse radiation is excluded or compensated so that a simple ‘yes/no’ output is available.

Wind

The most widely used sensors for wind speed and direction measurement are so called ‘mechanical sensors’. These include the familiar cup anemometer and wind vane wind direction sensor and also propeller anemometers. However all mechanical sensors have serious drawbacks. They are poor at low wind speeds. They will stall or stop moving below a threshold speed.

This is quite serious for air quality applications since it is at low wind speeds that pollutant concentrations are highest. At the other end of the measurement range the gust response of a cup anemometer has serious limitations. The cups will speed up in response to a sudden increase in speed faster than they slow down in a sudden reduction. This results in an over reading of wind speed in gusty conditions. Much research on the ideal characteristics of cup anemometers has shown that conical cups are superior to hemispherical, three cups are best for a steady rotation and there is an optimum ratio of cup diameter to the sensor radius.

Cup anemometers suffer from poor ‘cosine response’. This means that as the wind direction changes to hit the sensor from below or above the horizontal the response of the sensor deviates from the ideal response, which is spinning at a rate proportional to the cosine of the angle of the wind above or below horizontal. For most designs the cups will actually spin faster for a wind a few degrees above or below horizontal. Again this results in an over reading in turbulent conditions.

Propeller anemometers have different drawbacks. The design is usually a perfect helical shape which has a good cosine response close to its axis overcoming the cosine response problems of cup and vane designs. Unfortunately propeller sensors suffer very poor cosine response when the wind is nearly at right angles to the sensor axis.

Mechanical wind sensors respond faster to change in stronger winds. For this reason their performance in response to change is usually given as a length or distance constant, typically a metre, rather than a time constant. This is usually defined as the length of the run of wind that will pass for a 63%, 1/e, response to a step change. The response of mechanical sensors is largely independent of the air density, at least for winds well above the threshold velocity.

Mechanical sensors are now being replaced by ‘sonic anemometers’, sometimes called ‘ultrasonic anemometers’. These operate by sending pulses of ultrasonic sound between transducers and measuring the time of flight. This allows a calculation of the component of wind velocity along the line between the transducers.

They all have the advantage of working down to very low wind speeds, in principle down to zero wind speed. This is very important in air quality applications since it is in low wind speeds that pollutant concentrations are highest. Issues of response time and averaging are much more manageable. Sonic anemometers can provide measurements with short averaging periods, typically measured in tenths of a second or less, over the whole wind speed range.

Barometric pressure

Barometric pressure is an important parameter for meteorologists but is less significant for other applications. However it is often used in estimating the density of the air which is obviously directly related to pressure. There are any number of pressure sensors available, the majority of which are still based on the principle of an aneroid capsule, an evacuated container which expands and contracts as the pressure falls and rises.

These days the capsule may be in the form of a machined cavity bounded by silicon membranes on a chip and measurement of the expansion and contraction will be based on the capacitance of the device rather than the mechanical linkages of traditional barometers and barographs.

One neglected issue for accurate barometric pressure measurement is the effect of wind. The pressure required is the so called static pressure of the air. However unless installed carefully the measured pressure will be over-read by a factor given by Bernoulli’s equation: –

P = Ps + 1/2ρv 2 Where P is total pressure, P s is static pressure, ρ is density and v velocity.

Taking ρ as 1.25 kg/m 3 and v as an unexceptional 12ms -1 the potential error is 90Pa or 0.9hPa as meteorologists would prefer. This is a sizeable error, larger than the accuracy of good quality sensors. To get round this problem a static pressure head such as that shown in fig 5 should be used. These are designed to allow measurement of static pressure without contamination by the dynamic pressure.

Location, location, location

For good quality data the quality of the installation and their location is even more important than the sensors themselves. A mediocre instrument intelligently used will give better results than a state of the art sensor in a poor location. This is an important point as most users of weather monitoring equipment are not meteorologists themselves and may not appreciate the scale of local variation or possible errors.

A wind sensor incorrectly mounted on a building could indicate a wind direction literally 180° in error and a temperature sensor not correctly protected from radiation errors could be 25°C or more in error. The starting point for any installation must be to consider exactly what the weather data is being collected for. If the information will be compared with existing reference data, for example maps of expected frequency of gales or average rainfall or is to be used with climate records as part of an environmental impact assessment then the installation should match as closely as possible those used to collect the original statistics.

Where there is no need to compare local data with other data then sensor location should be based on the application. For example if there is a need for wind speed at the top of a chimney or crane or other structure then this is the height that the wind sensor should ideally be installed and if there is a need to measure the temperature of air entering an intake at some level then that is the place to mount the sensor.

If in doubt users should consult an accredited meteorologist through their suppliers or an organisation such as the Royal Meteorological Society, www.rmets.org. The RMetS web page also has links to other organisations and sources of information, including the WMO.

References

WMO Guide to Meteorological Instruments and Methods of Observation, seventh edition, WMO-no.8; Secretariat of the World Meteorological Organisation -Geneva-Switzerland 2006.

Published: 01st Mar 2008 in AWE International