Chapter 8:

Meteorological Networks

Fred V. Brock June 1993

Networks of meteorological surface stations in the U.S. are exemplified, in the synoptic scale by the NWS-FAA-DoD Automated Surface Observing System (ASOS) (NOAA, 1992), and in the mesoscale by the NCAR Portable Automated Mesonet (PAM) (Brock and Govind, 1977 and Brock et al., 1986) and the Oklahoma Mesonetwork (Mesonet). These networks are characterized by slow response instrumentation (no flux measurements) and short towers (10 m or less). These three networks will be described in the following text. ASOS, PAM and the Mesonet were designed to meet quite different objectives. This is reflected in the design philosophy, site selection criteria, communications systems, power sources, measurement site configuration, instrumentation and local data processing employed. This section ends with a brief overview of data quality assurance procedures.

8.1. Introduction

ASOS is a synoptic scale network that will eventually provide 900 to 1700 reporting stations at airports across the United States. It will be the primary surface weather observing system of the U.S. It is designed to provide both synoptic data and surface aviation observations. In contrast to prior measurement practice on the synoptic scale, ASOS automates many observations from wind speed and gustiness to present weather and cloud height. To an outside observer, it seems that the ASOS design philosophy was dominated by the need to make aviation observations and to preserve continuity with the preceding manual observations. Apparently all ASOS sites are on airports as close to the touchdown zone as possible. This sometimes requires more than one observing site per airport. Continuity with manual observations requires specialized data processing procedures not found in other surface observing systems.

PAM is a research-oriented mesoscale system. It was designed to support short-term, intensive research projects. A typical term was one to three months. The scale ranged from locating 30 stations in 1 km²to 60 stations spread out over two states. There were two PAM systems and soon there may be a third. PAM I was used from 1976 to 1982 and PAM II was used from 1985 to 1993. As the name suggests, PAM was designed to be portable, readily moved from one area of the county to another. This meant that set-up time had to be short; crews could install 3-5 stations per day. It also meant that neither commercial power not telephone communications could be used.

The Mesonet is the largest and most comprehensive of the state networks (Meyer and Hubbard, 1992). It provides 108 stations distributed across the state of Oklahoma. It is designed to support agricultural, hydrological and meteorological goals as well as emergency management and energy conservation needs. Design of the Mesonet was constrained by cost, available data communications systems and above all, by the need to use rural sites. The cost and site constraints meant that, in contrast to the ASOS, the Mesonet could not use commercial power or telephone communications. They were generally not available at the preferred sites and it would have been prohibitively expensive to bring them in.

8.2. Site selection criteria

Since the ASOS primary site criterion was to be near the touchdown zone on an airport, other site criteria were not given much weight. The advantage of this scheme is that commercial power and commercial telephone communications are readily available. Airport sites generally have good fetch, level terrain and excellent security. Vandalism should not be a problem. Part of the ASOS data processing system and one sensor, the barometer, is located in a building where exposure to the elements and to a wide temperature range is not a consideration. As noted below, the availability of plentiful power eases some sensor exposure problems. The primary disadvantage is that aircraft flight safety is a predominant consideration. This can limit the height of towers to less than the nominal 10 m and compromise the site in other ways.

Research investigations employing PAM had various objectives, time durations, and terrain types. Criteria for site selection changed accordingly. PAM stations have been located from sea level to nearly 4000 m, in deserts, grasslands and forests. Sometimes they were located in valleys precisely to monitor the valley effect.

Mesonet site standards were established considering the diversity of potential applications. Rural sites were required to avoid anthropogenic factors present in urban and suburban sites and to make the sites representative of agricultural applications.

Physical characteristics of a site, including soil properties, must be representative of a large area. Since the USDA Soil Conservation Service has conducted extensive surveys of the soil types in the state, soil types characteristic of each region were known. The Soil Conservation Service will determine soil types at each Mesonet site to a depth of 1 m. Sites were required to be as far away as possible from irrigated areas, lakes and forests to minimize those influences.

Site slope was specified to be less than 5^o. There should be a minimum of obstructions that impede ventilation. The rule of thumb is that the distance between a wind obstruction and the site should be at least 20 times the height of the obstruction (Fujita and Wakimoto, 1982, and Wolfson and Fujita, 1989). Precipitation measurements are best made in clearings surrounded by brush and trees to reduce the wind effect. This conflicts with the requirement for good ventilation. To minimize land usethe rain gauge was located near the tower. The good ventilation requirement prevailed, i.e., priority was given to wind exposure at the expense of an ideal precipitation measurement site. To partially compensate for this, each rain gauge is equipped with a wind screen, as described below.

Each site must be accessible by vehicle in all weather. Vegetation at the site should be uniform and low. Short grasses were preferred. Bare soil should not be visible except where that is a characteristic of the surrounding region.

Other essential requirements included a reasonably uniform distribution of sites over the state, with at least one site in each county. There was a stringent communications requirement that will be discussed below. Land owners were asked to provide the site free of charge. Site security was an additional constraint; the sites were selected to minimize the probability of vandalism, if possible. However, there was no requirement to locate sites near telephone or power lines.

In the real world, some of these requirements are mutually exclusive and so many sites failed to meet all of the above specifications. Finally, all sites are documented with photographs to show the terrain, especially the wind fetch. Fig. 8.1 shows the locations of the 108 Mesonet sites:

Fig. 8.1. Map of Oklahoma showing county lines and Mesonet sites.

8.3. Communications Systems

All three of these systems require real-time data reporting at intervals of 1 minute, 5 minutes, or 15 minutes. Common desirable features of a communications system are reliability, relatively low-cost, wide-area coverage and modest bandwidth of 1200 to 3000 bits per sec. Different solutions to the communications problem were found for each of these systems. Each was powerfully molded by the chosen communications system.

8.3.1 Telephone

Commercial telephone systems provide a bandwidth of at least 2700 Hz, are reliable except where lines are strung on poles in rural area (far from repair facilities), and cover most land areas with reasonable population density. The cost is prohibitive if one must pay for running lines to each station, especially for a short term project. Telephone communications is ideal for the ASOS since lines would already be run to airports. Even if an additional line were needed, there may be spare lines in place and it would not be necessary to pay for installing poles, etc. Having designed the system to utilize the infrastructure of an airport, it would be difficult to put an ASOS station in an isolated area or even at a very small airport.

8.3.2 Direct radio.

Direct radio links from the remote stations to a central base station are desirable and flexible. But earth curvature limits line-of-sight links, as shown in Fig. 8.2 where the maximum line-of-sight link between two stations is shown if the remote station antenna is at a height of 10 m. For a base station or repeater antenna height of 200 m,the line-of-sight link is only a little more than 60 km. PAM I used direct radio with a single repeater. This severely limited the size of the network and caused immense difficulties in complex terrain. If the path of the signal from a remote station to the repeater or to the base station is too close to the ground, the signal could be trapped in an inversion layer and ducted away from the intended destination. To avoid these problems, PAM II used a different communications system. The Mesonet could not rely on direct radio alone as coverage of the entire state would require many repeaters that would be difficult to manage.

Fig. 8.2: Maximum line-of-sight distance from a 10-m tower to a base or repeater tower.

8.3.3 Satellite.

The only communications satellite that permits an inexpensive uplink (low power transmitter and simple antenna) is the Geostationary Operational Environmental Satellite (GOES). An inexpensive uplink is essential due to the number of remote stations involved. Since stations may be powered by batteries and a solar panel, the radio transmitter output power must be low. Severe limitations exist with the GOES technology, such as a limited data rate of 100 bits per second and one-way communications. Transmissions are permitted to the remote stations but they are fairly clumsy and completely inappropriate for a polling protocol. Some restrictions will be lifted in the next series of GOES. PAM II has used GOES despite the bandwidth restriction by designing the system with very tight timing tolerances and very short message lengths. The one-way limitation has meant that a technician had to go to a remote station to reset the clock, reload the program or restart the station. With other communications systems, these things could be done from the central site. The advantage is that PAM stations can be deployed almost anywhere in the Western hemisphere except in the polar regions where the satellite would appear very low on the horizon.

Two-way communications and much longer message lengths were deemed essential for the Mesonet so it does not use GOES. And the Mesonet will be fully operational by the time the next GOES has been put into orbit.

8.3.4 Oklahoma Law Enforcement Telecommunications System.

The Oklahoma Law Enforcement Telecommunications System (OLETS) is a statewide communications network composed of city, county, state, federal and military law enforcement and criminal justice agencies. Its purpose is to provide for the interstate, intrastate, and interagency exchange of criminal justice related information. The OLETS mission also involves access to real-time weather information for emergency management of community response to floods, fires and storms. OLETS is a very reliable₁high-speed, digital communications system that operates continuously with redundant main computer systems. It has the capacity to switch 2.5 million messages per day. There are over 200 terminals statewide that are potential sites for Mesonet traffic.

OLETS is an ideal vehicle for the Mesonet communications traffic as it provides the reliability, bandwidth and full two-way services needed (Crawford and Long, 1993). The Mesonet message load represents only a small fraction of the OLETS system capability. To get data to OLETS, the Mesonet had to provide a direct radio link from each remote site to the nearest OLETS terminal, typically in a sheriff's office or police station. The resulting communications system is shown in Fig. 8.3. An RF base station, located at the chosen OLETS terminal, is capable of servicing several remote sites directly, or through repeaters.

Fig. 8.3: Mesonet communications system utilizing the Oklahoma Law Enforcement Telecommunications System.

By using a combination of direct radio and OLETS, the Mesonet gained considerable flexibility. It was possible to cover the whole state but sites had to be in line-of-sight range of an OLETS terminal or a repeater. It is possible that communications will be disturbed occasionally by inversion ducting or interference from other transmitters.

8.4. Power systems

Commercial power is both necessary and readily available for the ASOS. The visibility and cloud height sensors would be seriously compromised by stringent power requirements. And power is always available at airports. This is a luxury for the system designer. Sensors can be selected without concern for power consumption or supply voltage requirements. Heaters can be used as necessary.

Systems, such as PAM and the Mesonet, which must use alternate power sources are constrained to select sensors with low power consumption and then to switch the sensors on only as needed to further conserve power. Usually, sensors must operate from an unregulated supply (typically 10 to 30 V). Heaters are not available and local computational capability is severely limited. Therefore, all components must be rated for operation over the temperature range of -30 to 50^oC.

PAM and the Mesonet use a solar power system comprising a solar panel, a regulator, a battery and an enclosure. In the Mesonet, the data logger, the modem, transceiver and the sensors consume less than 1 W. The battery and solar panel were sized to provide this level of power for a minimum of 10 days under the worst expected conditions (continuously cloudy and cold) in Oklahoma.

8.5. Measurement site configuration

A typical ASOS basic configuration at an airport is shown in Fig. 8.4:

Fig. 8.4: Layout of the ASOS sensors near the runway and the ACU in an airport building.

Note that the Acquisition Control Unit (ACU) is located in an airport building. All of the local processing, including formatting, quality control, storage and retrieval of data, is done here. The barometers are located in the ACU. Data are telemetered the short distance from the runway site to the ACU by radio. Wind speed and direction are measured on top of a nominal 10 m tower at the runway site. A Data Collection Platform (DCP) in the sensor array collects sensor data and transmits them to the ACU.

Fig. 8.5 shows a PAM II station. The 10 m mast is self-supporting; no guy wires are required. The tripod was simply staked down. This meant that no site preparation was required (no concrete pad needed), set-up and tear-down could be accomplished quickly. The wind sensors were at the top of the mast, at 10 m.

Fig. 8.5: PAM II station showing the self-supporting 10 m tower and the instrumentation.

The layout of a Mesonet remote station is shown in Fig. 8.6 and in Fig. 8.7. The lightning rod, shown in Fig. 8.6, is used at only 27 of the 108 sites. This reflects some uncertainty as to the need for lightning rods in Oklahoma. Perhaps after a decade of operation we will know whether they were needed.

Fig. 8.6: Mesonet remote station site showing the tower and instrumentation.

Fig. 8.7: Plan view of a remote station showing rain gauge and soil sites relative to the tower.

The data logger enclosure is not shown. It contains the data logger, the RF modem, the radio transceiver, and the barometer. The radio transmits in the VHF range at 169.425 MHz with an RF power output of 5 watts.

The pyranometer is mounted on a free standing tripod south of the tower so that there would be good access to the pyranometer for cleaning and to minimize reflections from the tower. There are two soil temperature plots, one for sensors installed under bare soil and one for sensors installed under native sod cover.

The towers are guyed so a concrete base was not needed. A galvanized iron tower was selected for low cost. Aluminum towers are much lighter and easier to transport but, because of their higher cost, they are justified only if a tower is to be moved frequently. The wind sensor is mounted on the top of the tower. Temperature and relative humidity sensors and the 2 m wind sensor are mounted on booms projecting about 1 m from the tower.

Due to the frequent presence of livestock, the stations are enclosed by a cattle panel fence approximately 10 m by 10 m. This protects the tower, the guy wires, the rain gauge and wind screen, the soil plots where soil temperature measurements are made and the pyranometer stand.

8.6. Instrumentation

The ASOS sensor specifications are listed in Table 8.1. Extra sensors are sometimes installed at large airports; the number of sensors is shown in the "number" column. ASOS reports cloud height in feet, visibility in statute miles, temperature and dew point in degrees Fahrenheit, pressure in hectopascals, wind speed in knots, and precipitation in inches but, for internal consistency, SI units are used in the following discussion.

Table 8.1 ASOS Basic Sensor Specifications.

Variable	Number	Sensor	Range	Inaccuracy¹
Cloud height	1 or 2	Laser ceilometer	30...3700 m	30 m or 5%
Visibility	1 or 2	Forward scatter meter	< O.4...16+ km	0.4...2 km; up to 3 km at 16 km
Present weather precipitation identification	1	LEDWI²	Detect light, moderate, and heavy rain or snow; or mixed precipitation	0.25 mm or 10%
Freezing rain occurrence	1	Vibrating element	Accumulation over 0.25 mm/hr	1% in precipitation rates as low as 1.3 mm/hr
Ambient temperature	1	Resistive temperature device (RTD)	-60⁰C...54⁰C	RMSE 0.5^oC...1.0^oC
Dew point	1	Chilled mirror	-34⁰C...30⁰C	RMSE O.6⁰C...4.4⁰C
Pressure	2 or 3	Redundant aneroid cells	570...1060 mb	0.7 mb
wind direction	1	Vane	0...359^o	5^o (above 2.6 m/s)
wind speed	1	Cup anemometer	0.. .60 m/s	1 m/s or 5%
Liquid precipitation	1	Heated tipping bucket	0...250 mm/hr	5 mm or 4% of hourly total

¹when inaccuracy is stated as a certain value or a certain percent, use whichever is greater. Thus the error in cloud height is (+/-)30 m or 5% of the reading, whichever is greater.
²LEDWI = Light Emitting Diode Weather Indicator.

Each PAM station measures wind speed and direction, temperature and relative humidity, pressure and rainfall. The stations can be readily expanded to measure many other parameters. The PAM data are reported at 5-minute intervals.

A Mesonet remote station includes a 10 m tower, a suite of sensors, a solar power system, a data logger, an RF modem, a transceiver and an antenna are enclosed in a fence for protection from livestock and other animals. Every station is equipped with a standard set of core sensors. In addition there are a number of supplemental sensors that are installed on about half the stations. The sensors are listed in Table 8.2³.

Table 8.2: Mesonet Sensor Characteristics³

Sensor	Model³	Height	Characteristics
[Core Sensors: (All Stations):]
Propeller-vane Wind speed and direction	R. M. Young 5103	10 m	Speed: 1 to 60 m/s Direction: 0 to 360^o
Air temp. Thermistor & Sorption humidity sensor Radiation shield	Vaisala HMP35C Coastal Climate	1.5 m	Air temp: -30 to 50^oC Humidity: 0 to 100% Unaspirated 12-plate type
Barometer	Vaisala PTB202	0.75 m	700 to 1100 mb
Rain gauge Wind screen	MetOne 099M Internal constr.	0.6 m	Tipping bucket 30 cm gauge. 0.25 mm bucket Alter style
Pyranometer,insolation	LiCor 200	1.8 m	Silicon cell
Thermistors (2) soil temp.⁴	Fenwal	-10 cm	-30 to 50^oC Stainless steel jacket
[Supplemental Sensors (About 50 stations):]
Thermistors (3) Soil Temperature	Fenwal	-5 cm -30 cm	-30 to 50^oC Stainless steel jacket
Thermistor, Air Temp. & Radiation shield	Thermometrics Coastal Climate	9 m	-30 to 50^oC, fast response Unaspirated 12-plate type
Cup anemometer	R.M. Young 3101	2 m	Speed 0.5 to 50 m/s
Leaf wetness sensor	Internal constr.	0.5 m	Electrical impedance grid

³Mention of specific brand names does not constitute endorsement by NCAR.
⁴Soil temperatures at the 5 cm depth (supplemental parameter) and at the 10 cm depth are measured under both bare soil and under natural sod.

Inaccuracy and resolution are listed for each sensor in Table 8.3. The inaccuracy specification was provided by the instrument vendor. It does not include errors due to insufficient coupling with the atmosphere. As noted below, air temperature sensors can have large radiation induced error due to the radiation shield. Resolution is that of the data in the Mesonet archive. It is determined, in most cases, by the resolution of the analog-to-digital converter in the data logger. Resolution is listed here because there are applications which can reasonably use some of this resolution, even though it is far smaller than the accuracy specification. For example, the barometer can detect gravity waves. A gravity wave of 0.1 mb amplitude could be detected and the amplitude measured even though the inaccuracy of the sensor is 0.4 mb.

Table 8.3. Mesonet sensor inaccuracy and resolution.

Sensor	Inaccuracy	Resolution
[Core Sensors: (All Stations):]
Wind speed and Direction	2% of Reading 3^o	0.03 m/s 0.05^o
Thermistor, air	0.35^oC	0.01^oC
Humidity	3%	0.03%
Barometer	0.4 mb	0.01 mb
Rain gauge	1% of reading	0.25 mm
Pyranometer,insolation	5% of reading	0.23 W m^-2
Thermistors, soil	0.5^oC	0.03^oC
[Supplemental Sensors (About 50 stations):]
Thermistors, soil	0.5^oC	0.03^oC
Thermistor, Air	0.4^oC	0.03^oC
Cup anemometer	2% of reading	0.25 m/s
Leaf wetness sensor	(not established)

PAM and the Mesonet do not measure cloud height, visibility, present weather or freezing rain. The ASOS does not measure solar radiation, soil temperatures or leaf wetness.

8.6.1 Cloud height

Cloud height is measured by the ASOS using a vertically pointing laser ceilometer. It operates in the infrared at 9 mm. The pulse repetition frequency varies from 620 to 1120 pulses per second. The nominal prf is 770 Hz at room temperature. A laser beam is very nearly parallel. The beam width is 5 mrad, about 0.29^o, which produces a beam width of just 18 m at a height of 3600 m. Since the light pulse travels at the speed of light in the atmosphere, the cloud height can be determined from the time required for the pulse to travel to the cloud target and back. Thus h = ct/2 where C is the speed of light and t is the travel time. The vertical resolution is 15 m and the maximum height is 3600 m. The cloud base may be obscured by rain, snow and fog. Then the cloud height indicator measures vertical visibility. This instrument determines cloud height by averaging measurements, made at a single point, over time. This is in contrast to the human observer who averages over space and makes an estimate of cloud height at one instant in time.

8.6.2 Visibility

Visibility is the one measurement where a human is the standard. In all other measurements, the standards are other instruments, not the human observer. So the visibility sensor measures Sensor Equivalent Visibility (SEV) which is defined with respect to human visibility. The human observer determines visibility by observing objects at various known ranges so the determination that the visibility is, say 3 km, is made by integrating over a 3 km range. The ASOS instrument uses a short path, on the order of 1 m, to measure visibility. (Previous instruments used longer paths of up to 450 m that caused problems with alignment of the source and detector.)

The ASOS visibility sensor is a forward scatter meter. It uses visible (mostly blue) light from a pulsed Xenon flash lamp. An optical detector observes a portion of the beam from an angle of about 400 from the center of the beam. No light from the beam will reach the detector unless there are particles in the beam which scatter the light out of the beam and into the detector. The advantage of the short path is that the source and detector can be mounted to a common, rigid frame and thus preserve their respective alignment. Further, it is easier to monitor the source strength. The primary disadvantage is that the instrument can be seriously affected by local variations in visibility such as a patch of fog. This is countered by averaging over time. The instrument is sampled once every 30 s and the measurement is converted to an extinction coefficient which, in turn, can be converted to visibility. A one-minute mean is calculated from the 30 s data. Then a 10-minute harmonic mean is calculated from the 1-minute data and updated every minute. The harmonic mean given by

(8.1)

is used because it responds more rapidly to deteriorating conditions.

8.6.3 Present weather

Present weather or precipitation identification is measured by a Light Emitting Diode Weather Identifier (LEDWI), also known as an optical rain gauge. The LEDWI transmits a beam from an infrared light emitting diode over a 1 m path to an optical detector. The beam is shaped, with two lenses and a mask, to be rectangular with much greater width than height. Because of this shape, it is much more sensitive to the vertical velocity of particles in the beam than the horizontal velocity. Rapidly falling particles (rain) produce high frequency scintillations in the detector output while slowly falling particles (snow) produce lower frequency scintillations.

The detector output is filtered into two band-pass regions. The low band is 75 to 250 Hz while the high band is 1 to 4 kHz. Signal power is measured in each band. Rain/snow discrimination is determined from the ratio of power in the high band to that in the low band. Rain falls faster than snow and produces more power in the high band. For a given precipitation type, such as rain, the scintillation frequency is proportional to the vertical fall speed which, in turn, is a monotonic function of the particle diameter. Thus the power in the band is related to the precipitation rate. While the tipping-bucket rain gauge rain rate is the one reported by the ASOS, the LEDWI rate can be used to indicate precipitation intensity on a coarse scale: light, moderate, heavy.

Fog, dust and smoke move with very small vertical speeds so don't affect either band. When blowing snow is carried to a height greater than the LEDWI, about 3 m, the instrument can mistakenly report falling snow. Ice pellets may be reported as rain.

8.6.4 Freezing rain

Freezing rain is detected by a small cylindrical probe, stimulated to vibrate at its resonant frequency of 40 kHz. The vibration frequency is proportional to the mass of the probe. As ice accumulates on the probe, its mass increases and so the resonant vibration frequency decreases. There is a well defined relationship between ice accretion and the probe frequency. The instrument can measure accumulation rates as low as 0.25 mm/hour. When freezing rain accumulates to a depth of 6 mm, a heating cycle is initiated to melt the ice. This takes 5 to 10 minutes.

The cylinder is oriented vertically to avoid false bird signals. Snow can accumulate on the probe when the temperature is near freezing and generate a false signal. Snow can be differentiated from freezing rain by checking the output of the LEDWI.

8.6.5 Wind vector

A cup anemometer and wind vane are used by the ASOS to measure wind speed and direction. The starting threshold of each is 1 m/s and the cup distance constant is 9 m. The mounting height is a nominal 10 m but may be less in some cases due to aircraft safety restrictions.

The primary wind sensor of the Mesonet is the R.M. Young Model 5103 Wind Monitor which is a propeller, wind vane combination. It is mounted at the World Meteorological Organization (WMO, 1983) standard height of 10 m above ground level.

Sensor specifications were selected to balance cost, performance and durability. The wind speed sensor, a propeller, has a starting threshold of 1 m/s, a range of 0 to 60 m/s, with gust survival to 100 m/s, and a distance constant of 2.7 m. The signal output is a magnetically induced a.c. voltage which produces 3 cycles per propeller rotation or 10.2 Hz per m/s. The vane threshold is 1.0 m/s, the damping ratio is 0.25 and it has an undamped natural wavelength of 7.2 m. See ASTM, 1987 for a definition of these terms and for methods of testing wind sensors. The associated transducer is a pot which has a range of 0 to 355^o (with 5^o open). The propeller vanes are oriented to true North.

Local processing, in the data logger, generates six products from these two measurements: average wind run, average vector wind speed and direction, standard deviation of both wind speed and direction, and the maximum wind speed within the averaging period. These measurements are expected to be in compliance with the emerging standard for wind measurement to be issued by the Federal Coordinator for Meteorology (private communication with Thomas Lockhart).

Some stations have a cup anemometer at the 2 m level. This anemometer has a starting threshold of 0.5 m /s a range of 0 to 50 m/s with gust survival to 60 m/s, and a distance constant of 2.3 m. The signal output is a magnetically induced AC voltage which produces 1 cycle per cup wheel rotation or 1.33 Hz per m s^-1.

8.6.6 Temperature and Relative Humidity

Air temperature and dew point are measured by the ASOS using an aspirated chilled-mirror dew point instrument and a platinum Resistance Temperature Detector (RTD) mounted in the incoming air stream. Dew point is determined by cooling a small, front-surface mirror until dew or frost condenses on the surface. By definition, this is the dew point. The mirror is cooled (heated) by a thermoelectric or Peltier cooler (heater). Dew is detected optically with a Light Emitting Diode (LED) and two photodetectors. Light from the LED is directed at a 45^o angle to the mirror surface. One photodetector is positioned to receive the direct beam of reflected light when the mirror is clear of dew. When dew accumulates on the mirror, light is scattered away from this "direct" photodetector and into the "indirect" photodetector. In operation, a feedback controller uses the "direct" and "indirect" photodetector signals to control the electronic heat pump.

A RTD is mounted in the mirror, as close to the front surface as possible, to measure the mirror temperature. The assumptions are that the RTD is at the mirror front surface temperature, that the RTD is properly calibrated, and that the mirror is being held at the dew point by the controller. The latter assumption fails when there is dirt on the mirror, when there are salts on the mirror, and when the ambient dew point changes faster than the controller can change the mirror temperature. Dirt on the mirror can cause light scattering which may confuse the photodiode signals. This is controlled by activating an automatic mirror cleaning procedure once a day. In this procedure, the mirror is heated to drive off all of the moisture. As the condensate drops evaporate, their radius shrinks and dirt particles held within the drops are pulled together. When the mirror is dry, the dirt particles will be in small, isolated clumps and will have little effect on the reflected light. Salts on the mirror can act as condensation nuclei, reduce the vapor pressure around the drops and cause dew to form at a higher temperature than the true dew point. Conversely, if the mirror is too clean, dew may form at a temperature lower than the true dew point. Therefore, while the chilled-mirror dew point instrument seems to be a direct implementation of the dew point definition, there are a number of ways that it can be in error.

PAM used a specially designed psychrometer for humidity measurements for years and than gradually moved to sorption sensors. In the early years of PAM, sorption sensors were not nearly as good as they are today. Most PAM projects that required accurate humidity measurements were summer projects. The psychrometer, shown in Fig. 8.8, was designed for continuous operation with a double water bottle that could keep the wick moistened for a week even in very dry weather.

Fig. 8.8. PAM psychrometer showing the two water reservoirs that keep the wick wet for at least one week of continuous operation.

The advantage of a psychrometer is that the sources of error have been well documented and are readily controlled, as noted below:

Sensitivity, accuracy and matching of the temperature sensors. A psychrometer is less sensitive to the absolute error in the temperature sensors than to the relative error, or matching error, between wet- and dry-bulb sensors.

Ventilation rate. Typically, the ventilation rate should be at least 3 m/s to maximize the heat transfer by convection and evaporation and to minimize heat transfer by conduction and radiation. The minimum ventilation rate needed is a function of the sensor thermal mass. Sensors made from small diameter thermocouple wire with a fine cloth wick have been successfully used without forced ventilation (Stigter and Welgraven, 1976.)

Radiation incident on the temperature sensors. The sensors must be shielded from direct and reflected solar radiation and from long-wave or earth radiation. This is a major source of error in the field that is not usually a factor in the laboratory.

Size, shape, material and wetting of the wick. Specially prepared psychrometer wick, available from instrument vendors, should always be used and not ordinary cotton cloth. Most commercial cotton cloth contains hydrophobic, or anti-wetting chemicals, that will eventually impede wetting of the wick. When used in continuously operating psychrometers, even standard psychrometer wicking should be boiled in a solution of lye and detergent, then boiled in distilled water and flushed with distilled water before use.

The relative positions of the wet- and dry-bulb sensors. The air must not flow from the cooled wet-bulb to the dry-bulb.

Purity of the water used to moisten the wick. Only distilled or deionized water should be used.

The disadvantage is that the wick must be kept clean which limits the performance in areas where salts are present, most especially the seashore. Even in optimum conditions, the wicks must be changed weekly. Another problem is that the sensitivity to small errors in the wet-bulb depression increases with decreasing temperature. And, of course, a psychrometer designed for continuous operation fails in freezing temperatures.

Mesonet air temperature and relative humidity are measured at 1.5 m with a combination thermistor-sorption probe mounted in an unaspirated multiplate radiation shield. The sensor is a HMP35C probe which is a Vaisala HMP35 sorption probe for humidity modified by Campbell Scientific to include a thermistor for air temperature and a power switching transistor to reduce power consumption. Relative humidity is sensed by a capacitive sorption sensor with a range of 0 to 100% RH, and errors of 2% for 0 to 90% and 3% in the range 90 to 100%. The air temperature and the humidity sensor are enclosed in a filter to keep dust off of the humidity sensor. This has the effect of slightly reducing the air flow around the sensors and therefore, increasing their time constant. This can be significant in light winds, perhaps increasing the time constant to as much as 10 minutes.

Since the temperature sensor is in close proximity to the humidity sensor, the relative humidity output can be readily converted to dew point, which is more useful for many applications. Conversion of a constant 2% relative humidity error to dew point error is nonlinear and is a function of the ambient air temperature, as shown in Fig. 8.9. The Mesonet maximum tolerable humidity error goal was specified in terms of dew-point temperature. The specified error was <0.1^oC for dew points above 0^oC; the tolerance is somewhat larger for lower temperatures as shown by the dashed line in Fig. 8.9. A 2% error in relative humidity, when converted to a dew point error, meets this goal except when the dew point depression is large. For example, when the air temperature is 30^oC, the sensor dew point error is less than 1.1^oC when the dew point is greater than 10^oC, or the dew point depression is less than 20^oC.

Fig. 8.9. Conversion of a 2% error in relative humidity to an error in dew point over a range of ambient temperatures from -10 to 50⁰C. The dashed line is the specified error tolerance.

A sorption-type humidity sensor was chosen based on cost and power considerations but also because this type of sensor is typically well-behaved in the field. A well-maintained chilled-mirror dew point instrument is capable of superior performance in the laboratory but can show large, transient errors due to a failure to maintain control after a sudden change of dew point.

In general, an aspirated air temperature sensor is superior to one mounted in a passive radiation shield; errors in a passive shield can be as large as several degrees Celsius in calm wind (< 1 m/s) and strong radiation (> 800 W m^-2) conditions. Extreme conditions of insolation coupled with a highly reflective ground surface were used by Gill (personal communication with R. M. Young) to obtain the data shown in Fig. 8.10. The upper curve in Fig. 8.10 shows the effect of a sun angle approximately 70^o above the horizon with an irradiance of 1080 W m^-2. The lower curve is for the same intensity but with a sun angle of 90^o above the horizon. Evidently the shield is more sensitive to radiation coming from an angle of 70^o, probably due to internal reflections. This figure cannot be used to correct Mesonet air temperature data because the plotted data were obtained in a wind tunnel under conditions not necessarily representative of actual sites and because the Mesonet radiation shields are similar to the Gill shield but not exactly the same. However, the figure can be used to indicate conditions of light winds and strong radiation when the Mesonet air temperatures should be used with caution.

Fig. 8.10. Maximum temperature error induced in the temperature sensors by the incident radiation.

A passive shield was specified, with some reluctance, because of power and cost constraints. Calm winds and a strong radiation combination is somewhat less frequent in Oklahoma than in other parts of the country. The ideal solution would be an aspirated shield designed to be an efficient radiation shield even when the fan is turned off. Then the fan could be turned on only when needed. Until such a shield is available, the Mesonet may flag the air temperature data as questionable when the ambient wind and radiation conditions are not favorable. It should be noted that an aspirated shield is not necessarily free from radiation induced error.

Some of the Mesonet stations have another air temperature sensor at 9 m. It is a thermistor identical to the one described above and has the same type of radiation shield.

8.6.7 Barometric Pressure

Barometric pressure is measured by PAM, the ASOS and the Mesonet using similar aneroid sensors with comparable performance. All three use capacitive sensors where the space between the plates is permanently evacuated to a vacuum so the capacitance is proportional to the absolute, or barometric, pressure. Each uses a dedicated microprocessor to control the sensor, apply the calibration equation, and generate a digital output. The Mesonet uses one instrument while the ASOS uses two or three instruments for additional reliability.

The barometer selected for the Mesonet was the Vaisala PTB 202 which utilizes a silicon capacitive integrated circuit sensor. To minimize power consumption, the barometer is switched on only when needed to make a measurement. The sensor range is from 700 to 1100 mb with an uncertainty of 0.4 mb. This is an acceptable error level for pressure measurement given the range of pressures and temperatures (-30 to 50^oC) to which the barometer is exposed. This does pose a problem for those wishing to compute pressure gradients over small horizontal distances, as illustrated in Fig. 8.11, which shows the geostrophic wind error due to using barometers with a given level of uncertainty. For example, if the pressure gradient between two stations, separated by 30 km, was in error by 0.4 mb, the error induced into the geostrophic wind would be 13.3 m/s. Another source of error, which can be even larger, is due to correction of the station pressure to a reference height. There is some uncertainty in the actual station height (where an 8 m height error is approximately equivalent to a 1 mb pressure error) and the correction algorithm may introduce another uncertainty. Any computation of pressure gradients must be made with great care keeping these error sources in mind.

Fig. 8.11. Geostrophic wind error as a function of barometer error between stations, and station spacing, from 30 km to 500 km.

Ideally, the station pressure is the total static pressure uncontaminated by the dynamic pressure due to wind flow. The classical way to minimize dynamic effects is to use a static pressure port which is usually some sort of flat plate with a hole, or static port, in the middle connected to the barometer with a tube. The ASOS barometer is located inside a building and is simply vented into the building unless the building is pressurized. In that event, the barometer is vented by a tube to a static pressure port outside of the building.

The PAM static port is mounted on a boom projecting out about a meter from the tower. One of the problems with a flat-plate port is that if it is not horizontal, perhaps due to accidental misalignment, large errors can be induced. An alternative, used in the Mesonet, is to run the pressure tube from the barometer down the tower leg to a point about 15 cm above the ground and to terminate it there with a filter to keep insects out. This is a compromise since the average wind speed is only truly zero at ground level, but there the tube would be subject to flooding and possible mechanical damage.

8.6.8 Rainfall

PAM, the ASOS and the Mesonet use tipping-bucket rain gauges to measure rainfall. The ASOS gauge is not equipped with a wind screen but it is heated to melt frozen precipitation.

The objective in the Mesonet is to measure rainfall rather than precipitation. This distinction is made in recognition that the sensor chosen, an unheated tipping bucket gauge, is not suitable for snow or freezing precipitation measurement. Cost and power considerations dictate the use of an unheated gauge. Wet snow may clog the gauge and cause underreporting while dry snow frequently will be blown out of the gauge, also causing underestimation of precipitation. The gauge selected has a 30 cm diameter orifice and generates a bucket tip for every 0.25 mm of rain. The counts are accumulated over a 5-minute interval. It was necessary to locate the rain gauge close to the 10 meter tower and the tower location was selected to optimize wind fetch. The top of the gauge is 0.6 m above ground level. Since the largest source of rain gauge error, after sampling error, is due to wind induced precipitation loss, an Alter-style wind screen is used to reduce the wind effect.

8.6.9 Solar radiation

Solar insolation data are used in agricultural models, hydrology, renewable energy and in meteorology. The Mesonet provides spatial resolution of measurements not previously available. A silicon photodiode type pyranometer, the Licor Model 200, was chosen to measure solar radiation. It has the advantage of ruggedness and relatively low cost. However, as shown in Fig. 8.12, the spectral response does not extend uniformly over the full solar radiation range, usually taken to be 0.3

m to 4.0

m. The calibration factor was obtained by comparison with an Eppley Precision Spectral Pyranometer and the uncertainty of the calibration, relative to the Eppley, is claimed to be 5%. The Licor pyranometer is useful to the extent that the spectral distribution of solar radiation in the 0.3 to 4

m band does not change from clear to cloudy days. This would invalidate the use of the Licor sensor under plant canopies, in a greenhouse or in artificial light. The pyranometer is mounted on a separate tripod south of the tower at 1.75 m. The purpose is to provide nearly complete freedom from obstructions above the horizontal plane of the sensor, easy access for cleaning and to minimize reflections from the 10 m tower.

Fig. 8.12. Spectral sensitivity of a LICOR pyranometer.

8.6.10 Soil Temperature

Soil temperatures are measured by the Mesonet with thermistor probes at a depth of 10 cm under both bare soil and under natural sod. There are supplemental soil temperature sensors at 5 cm (bare soil and sod) and at 30 cm (sod). The thermistors are enclosed in stainless steel jackets and inserted horizontally into undisturbed soil. Cables are routed down into the soil and then back up to the sensors. This was done to minimize heat conduction and water flow along the cable to the sensor. At the 5 cm depth, it is sometimes possible to detect the shadow of the solar panel as it passes over the sensor.

8.6.11 Leaf Wetness

A leaf wetness sensor was developed by Fisher, Stone and Elliott (1992) and fabricated at Oklahoma State University. It is a printed circuit board that forms a voltage divider. One arm is a 100 K-Ohm resistor and the other is the grid printed onto the board. The board is covered with paint to mimic the wetting and drying characteristics of a leaf. The paint allows condensate on the board to penetrate to the printed grid underneath and change the electrical impedance of the circuit. The board is tilted above the horizontal to allow runoff. The tilt angle is yet to be determined. The recorded signal output ranges from 0 (totally dry) to 1 (totally saturated).

8.7 Local data processing

Data processing at the local level is much more complex and sophisticated in the ASOS than in PAM or the Mesonet. Therefore, this discussion of ASOS processing will be somewhat simplified. For more details, consult the ASOS User's Guide (NOAA, 1992). Local data output from the ASOS includes airport displays and computer-generated voice radio broadcasts to pilots. The principle off-site messages are the One-Minute Observation (OMO) and the hourly Surface Aviation Observation (SAO). The OMO and SAO messages contain sky condition (cloud height and amount), visibility, present weather (rain, snow, freezing rain), obstructions to visibility (fog and haze), pressure (sea level and altimeter setting), air temperature and dew point, winds (wind direction and speed, gusts and squalls), remarks (pressure change, etc.), and precipitation accumulation.

Data sampling and averaging are performed at two levels in the ASOS, at the sensor level and at the ACU level as shown in Table 8.4. For example, the temperature and dew point are sampled at 30 s intervals and 1 minute averages are generated at the sensor level. Then the ACU algorithms take the 1-minute averages, perform some quality assurance checks (see below) and generate 5-minute averages which are updated every minute. The 5-minute average is reported every minute in the OMO and the most recent 5-minute average is reported in the SAO.

Table 8.4. Data sampling and averaging in the ASOS.

Variable	Sensor Level		ACU Algorithms
Variable	Sampling	Averaging	Sampling	Averaging	Reporting
Cloud height	1.3 ms	12 s	30 s	30 min	1 min
Visibility	30 s	1 min	1 min	10 min	1 min
Present weather	1 s	1 min	1 min	10 min	1 min
Freezing rain	1 s	1 min	1 min	15 min	1 min
Temp. & dew point	30 s	1 min	1 min	5 min	1 min
Wind	1 s	5 s	5 s	2 min	1 min
Pressure	10 s	1 min	1 min	1 min	1 min
Precipitation	1 s	1 min	1 min	15 min, 1 hr	15 min, 1 hr

Wind speed and direction are sampled at 1 s intervals at the sensor level in the ASOS where 5 s averages are formed. At the ACU level, the 5 s averages are further averaged over 2 minutes which are updated every 5 s. The ASOS reports wind character as gusts, squalls, variable wind direction, wind shift, peak wind, daily peak wind, and fastest 2-minute wind. A simplified version of the wind gust definition is that a gust is the maximum 5-s average observed over the last 10 minutes if that gust is greater than 7 m/s and the 2-minute average is greater than 1 m/s . ASOS does not perform vector averages nor does it compute standard deviations of wind variables.

Local data processing capability in the PAM is intermediate between the ASOS and the Mesonet. PAM has a more powerful processor and is easier to program.

The data logger, located in each remote station, is the heart of the measurement system since all data flow through it. The Mesonet uses Campbell Scientific CR1 OT data loggers. They perform analog to digital conversion, on-site data processing and play an essential role in the digital data communications link. The logger also limits the number of sensors of each type (up to 12 analog, two event or counter, and eight digital input-output lines) that can be installed at a given site. (It is possible to install more analog sensors by adding a multiplexer expansion module.)

The analog subsystem accommodates the sensors whose output is a voltage signal and those whose output can be readily converted to a voltage. These include the wind vane, the air and soil temperature sensors, the relative humidity sensor and the pyranometer.

A digital input-output port (or line) can be used with a tipping-bucket rain gauge by programming the data logger to enter a subroutine each time the port senses a low-to-high transition. The subroutine counts the number of times it has been entered and, therefore, the number of tips of the bucket. These ports control the barometer, which has its own microprocessor, and to read the serial digital signal.

The data logger provides sufficient programming capability to perform the local processing: setting the sampling and averaging periods, determining which input channels are to be read and how they are to be processed, and what data are to be output and in what order. Local memory is used to store the operating program, and to provide temporary storage of intermediate operands and of data waiting for transmission. In the Mesonet configuration, the data logger has sufficient memory to store data for a week without loss in the event of a communications failure.

Each Mesonet remote station generates a data message every 15 minutes. Each message is a composite of three 5-minute averages of most variables and a 15-minute average of the remaining variables as indicated in Table 8.5. In addition, some housekeeping variables are reported such as the battery voltage.

Table 8.5. Mesonet sampling and averaging times. Supplemental variables are indicated with an (S).

Variable	Sampling	Averaging	Quantity Reported
Wind speed 10 m	Count 3s	5 min	Wind run (scalar average) Wind speed (vector average) Wind direction (vector average) Wind direction standard deviation Wind speed standard deviation Wind speed maximum
Wind direction 10 m	3 s	5 min
Air temperature 1.5 m and 9 s (S)	3 s	5 min	Average
Relative Humidity 1.5 m	3 s	5 min	Average
Barometric Pressure	12 s	5 min	Average
Rainfall	--	5 min	Accumulative rainfall
Solar Radiation	3 s	5 min	Average
Wind speed 2 m (S)	3 s	5 min	Average
Soil temperatures	30 s	15 min	Average
Leaf wetness (S)	30 s	15 min	Average

Wind direction is sampled every 3 s while wind speed is counted or integrated over a 3-second period. Thus there are 100 samples of wind speed and direction in each 5-minute averaging period. The maximum wind speed is defined to be the maximum 3-second speed within a 5-minute period. Wind speed max is not comparable to the ASOS wind gust. Wind run is the scalar average of the wind speed. Wind speed and direction are the vector average of the speed and direction. Standard deviations of the speed and direction are taken over the scalar quantities.

Power to the barometer is turned on only during a measurement. It takes the barometer 1 s to stabilize and generate a reading after the power is turned on. In the Mesonet, the barometers are sampled at 12 s intervals. They are turned on and allowed to stabilize for 3 s, then the pressure is read and the barometer is turned off. The power duty cycle is approximately 0.25.

Rainfall is the count of bucket tips over the averaging period. Pulses from the 2 m wind speed anemometer, like the 10 m wind speed propeller, are counted over the averaging period. Sample averages for the other variables in Table 8.3 are scalar averages of the instantaneous readings taken at the sampling period (3 s or 30 s) during the averaging period.

8.8 Data Quality Assurance

Data quality assurance in the ASOS is a three level plan, conducted at individual sites, at the regional level and at the national level. Each site has self-diagnostics built into the system, as part of the local processing. For example, the 1 minute average temperature, T, and dew point, T_d are checked each minute. A simple range test is applied. If -60 < T < 55^oC then the temperature is accepted, otherwise it is missing. If -30 < T_d < 30^oC then the dew point is accepted else it is missing. For both T and T_d, the rate of change must be less than 3.3^oC/minute. Finally, if T_d > T + 1^oC, the dew point is missing.

At the regional level, the ASOS messages are checked for error indicators and technicians are dispatched as needed to the individual sites. There are no technicians permanently resident at the local level so all maintenance comes from the regional level.

Messages are checked at the national level to determine whether they arrive within an allotted time window. This central facility synchronizes clocks at all the ASOS stations and maintains data such as station elevation and magnetic declination. At the National Meteorological Center (NMC) the data are checked for consistency with the NMC optimal interpolation analysis. The National Climatic Data Center (NCDC) may perform additional quality control of the ASOS data before they are archived. This last check would not be done in real time so could not be used for the usual maintenance but would serve as a long term check on performance.

The objective of the data quality assurance (QA) program is to maintain the highest possible data quality in a network (Brock and Fredrickson, 1993). To achieve this goal, data faults must be detected rapidly, corrective action must be initiated in a timely manner, and questionable data must be flagged. The data archive must include provision for status bits associated with each datum. The QA system is designed to never alter the data but only to set status bits that indicate the probable data quality. It is possible the QA system will flag data that actually is valid but represents unusual or unexpected meteorological conditions; with this QA system, the actual reported data always will be available. In addition, flagged data will be available to some users but will not be available for routine operational use.

The major components required for this program are laboratory calibrations, field intercomparisons, real-time monitoring of the data, documentation, independent reviews, and publication of data quality assessment.

8.8.1 Laboratory Calibrations

Laboratory calibration facilities are required to verify the calibration of new sensors, to check the calibration of suspect instruments and to obtain new calibrations for instruments that have drifted out of calibration or have been repaired. Most sensors require annual recalibration to correct for drift, if nothing else. Screening of new sensors is essential since even the best vendors will occasionally ship a defective or out-of-calibration sensor.

8.8.2 Field intercomparisons

Laboratory calibrations are necessary but not sufficient for sensor calibrations because some field conditions are very difficult to simulate in the laboratory. For example it is difficult to replicate the effects of solar radiation and wind speed on temperature sensors.

There are two types of field intercomparisons that must be done. The Mesonet has a field intercomparison station located near Norman. And when technicians visit a Mesonet station, they carry portable transfer standards and make a routine comparison check.

1) Field Intercomparison Station

The field intercomparison station is a special remote station, referred to as a reference station and equipped with better sensors, that is located adjacent to a standard station. Two Mesonet stations are collocated with ASOS stations. These stations are used for intercomparison to provide an on-going check of sensor performance in the field. In addition, the reference station can be used to test proposed new sensors.

2) Routine Intercomparisons

Several sets of portable transfer sensors are available; each time a technician visits a station, reference measurements are made. This can be used to detect drift or other failures of sensors that might otherwise go undetected. These sensors can be standard sensors that have been carefully calibrated and checked frequently to insure against drift. Technicians carry a lap-top computer to read current data and to read the portable reference instruments. Some problems can be detected in the field. Regardless, the usual fix is to replace sensors or modules and not attempt repairs in the field. Routine visits for preventative maintenance will be scheduled at intervals of three months. Real-time, automatic monitoring of the data, described below, will detect many problems and other visits to stations will be scheduled as needed.

8.8.3 Data Monitoring

Neither laboratory calibration nor routine field intercomparisons can provide a completely positive indication of current problems in the network. Since the data are collected and reported in real time, bad data will be transmitted to users unless detected and flagged. The volume of data flow is far too great to allow human observers to effectively monitor the data quality. Therefore, a real-time, automatic monitoring system (QA) has been implemented.

Some elements of the QA system are fairly simple and can be used to test each datum as it arrives in the central computer. The range test, as used by the ASOS, is an example of a very simple but quite effective algorithm. Other tests are run periodically as sufficient data accumulates - these use time series or time series in conjunction with spatial analysis.

In general, the QA programs utilize knowledge of the instrumentation system, knowledge of the atmosphere and statistical techniques.

8.8.4 Documentation

Several kinds of documentation (metadata) are needed: a file of remote station characteristics, a station descriptor file, and a sensor data base. Documentation of remote station characteristics includes the latitude and longitude and elevation of each station, panoramic photographs showing the fetch in all directions and the nature of the land, etc.

An instrumentation data base should contain sensor serial number, current location of the sensors, and the operational status. Some sensors have individual calibration coefficients so there must be a method of accounting for sensors to insure that the correct calibration coefficients have been entered.

8.8.5 Independent Review

For much the same reason that proposals and papers are reviewed, we should invite periodic, independent reviews of the network performance. It is always possible for people in constant close proximity to a project to become blind to problems. Inviting outside reviews of the network and its operation will help to remove the blinders.

8.8.6 Publication of Data Quality Assessment

There will be frequent data faults in a network of this size, even with the data quality assurance program outlined above. Toassist critics in making a realistic assessment, it would be desirable to periodically publish an honest appraisal of the network performance including all data faults, causes when known, and action taken. This network newsletter also should contain the results of the various intercomparisons undertaken which will show that we are seriously evaluating the network relative to other stations.

8.9 REFERENCES

American Society for Testing and Materials, 1987:1987 Annual Book of ASTM Standards, Atmospheric Analysis: Occupational Health and Safety. Section 11, Vol 11.03, Philadelphia, Amer. Soc. for Test. and Mater.

Brock, F.V.₁and S. Fredrickson, 1993: Oklahoma mesonet data quality assurance. Eighth Symp. Meteor. Obser. lnstr., Anaheim, Amer. Meteor. Soc., 311-316.

Brock, F.V., and P.K. Govind, 1977: Portable Automated Mesonet in operation. Jour. AppI. Meteor., 16, 299-310.

Brock, F.V., G.H. Saum, and S.R. Semmer, 1986: Portable Automated Mesonet II. Jour. Atmos. Ocean. Tech., 3, 4, 573-582.

Crawford, K.C., and B. Long, 1993: The design of a digital data communications system for the Oklahoma mesonet. Ninth Inter. Conf. on Inter. Infor. and Proc. Syst., Anaheim, Amer. Meteor. Soc.

Fisher, D.K., M.L. Stone, and R.L. Elliott, 1992: Design and testing of a leaf-wetness sensor. Nashville, Amer. Soc. Agric. Engr., Paper Nr. 923545.

Fujita, T.T., and R.M. Wakimoto, 1982: Effects of miso- and meso-scale obstructions on PAM winds obtained during Project NIMROD., J. AppI. Meteor., 21, 840-858.

Meyer, S.J., and K.G. Hubbard, 1992: Nonfederal automated weather stations and networks in the United States and Canada: A preliminary survey. Bull.. Amer. Meteor. Soc., 73, 4, 449-457.

National Oceanic and Atmospheric Administration, 1992: ASOS User's Guide.

Stigter, C.J. and A.D. Welgraven, 1976: An improved radiation protected differential thermocouple psychrometer for crop environment. Arch. Met. Geoph. Biokl., Ser. B., 24, 177-187.

Wolfson, M.M., and T. T. Fujita, 1989: Correcting wind speed measurements for site obstructions., Jour. Atmos. and Ocean. Tech., 6,2, 343-352.

World Meteorological Organization. 1983. Guide to meteorological instruments and methods of observation. WMO - NO.8, 5th edition, Geneva, Switzerland.