Potter T.D., Colman B.R. (co-chief editors). The handbook of weather, climate, and water: dynamics, climate physical meteorology, weather systems, and measurements

Подождите немного. Документ загружается.

Voltage Measurements

While environmental dataloggers can have analog inputs capable of measuring either

voltage or current signals directly, Campbell Scientiﬁc dataloggers have always

measured voltages and use an external shunt resistor to measure current signals.

Voltage measurements are made by switching the signal from one of the analog

channels into the analog-to-digital (A=D) circuitry. After waiting for the signal to

stabilize at its correct value, the signal voltage is allowed to charge up a capacitor for

a ﬁxed amount of ‘‘integration’’ time. The voltage on the capacitor is then held for

the A=D conversion. The A=D conversion is made using a successive ap proximation

technique. For example, to measure a signal of þ1000 mV on a 2730-mV range

the microprocessor ﬁrst compares the signal to 0.0 mV to determine the sign (þ or )

of the signal. Since þ1000 mV is greater than 0.0 mV, a value would be stored

indicating a positive signal and the 12-bit digital-to-analog converter (DAC)

would then output þ1365 mV, half of the remaining range. The signal is now less

than the DAC output, so a 0 would be stored for the ﬁrst bit and the DAC would next

output 682.5 mV. The process repeats until all the bits have been set, each time

closing in on the sensor’s voltage by half of the remaining range.

A datalogger typically has one DAC that outputs voltages across the largest full-

scale range (FSR). For a datalogger with a 12-bit DAC and a 2500-mV FSR to

measure a thermocouple with a 2.5-mV full-scale output, it must measure the small

signal by switching in a gain circuit that electronically multiplies the voltage by a

factor of 1000. Thus a 2.5-mV thermocouple signal is really measu red on the same

2500-mV range with the same 12-bit resolution DAC as the largest full-scale

voltage. Having multiple FSRs with voltage multipliers preserves the resolution

and accuracy of the measurements. Good system design requires that the datalogger

have multiple FSRs that have been carefully chosen to cover the voltages produced

by the sensors it is to measure.

The resolution of a voltage measurement is the smallest incremental change

detectable in the signal and therefore represents the ﬁneness of the measurement.

The greater the number of bits for a given full-scale voltage range, the ﬁner the

resolution of the measurement. An uncertainty of 1 bit must be assumed in the

measurement. Continuing the previous voltage measurement example out to

12 bits would yield a signal of þ1000.089 with a resol ution of 0.666 mV.

A 13-bit and a 16-bit measurement would yield þ999.925 0.333 mV and

þ999.985 0.042 mV, respectively. The number of bits a measurement is divided

into is determined by the number of bits in the DAC plus a couple of tricks. A 12-bit

DAC (2

7 1 ¼4095) is one that can change its output voltage in increments of

1=4095 of its full-scale output. An extra bit of resolution is gained by determining if

the signal is positive or negative before the successive ap proximation starts. A

second extra bit is gained when a measurement is the average of two integrations

(e.g., a differential measurement includes two integrations). To convert the resolution

of a measurement range in millivolts into engineering units, multiply the resolution

of the appropriate FSR by the ratio of the engineering unit range to the signal range.

For example, given a resolution of 0.333 mV on the 2500 mV range, and a 40

786 COMMERCIAL RESPONSE TO MEASUREMENT SYSTEM DESIGN

to þ60



C temperature signal represented by a 0- to 1000-mV signal, the resolution

would be

0:333 mV ½60 ð40Þ



C

ð1000  0mVÞ

¼0:0333



When the resolution approaches the magnitude of the input noise, the latter

determines the measurement uncertainty. Input noise is statistical and is speciﬁed

in terms of the root-mean-square (rms) value. Numerical averaging of N measure-

ments reduces the input noise by a factor of N

1=2

. One method of reducing the

effect of noise on the measurement of a signal is to integrate the signal over a longer

time period. Integration time is the time over which the signal is being electronically

averaged. The integration time can be long for greater averaging, which smoothes

the noise-induced peaks and valleys, short if more frequent measurements are

required, or it can be broken into two integrations, one half of an ac cycle apart

to remove either 60- or 50-Hz noise. The use of shielded sensor leads with the shield

connected to system g round helps reduce high-frequency external noise.

Accuracy is often expressed in terms of the FSR, e.g., 0.1% FSR, and should be

stated for a speciﬁc temperature range because of the temperature dependency of

most errors. The accuracy of voltage measurements depends upon the accuracy of an

internal voltage reference diode and the self-calibration ability of the datalogger over

time and temperature. Input offset voltages also affect accuracy, but a good micro-

processor-based system corrects for this error by shorting the input to ground and

measuring the input offset voltage as part of the signal measurement sequence. The

offset error is then removed from the result numerically.

The accuracy of each datalogger is carefully measured over several hot=cold

temperature cycles in automated precision environmental test chambers using regu-

larly calibrated National Institute of Standards and Technology (NIST) traceable

voltage sources. The dataloggers are tested at the environmental extremes quoted

in the speciﬁcations and then tested at points 10



C beyond those extremes to ensure

their reliability and accuracy. Even in the ‘‘startup’’ days of the company when

money for test equipment was limited, environmental testing was so important

that dataloggers were manually calibrated while still cold from ‘‘soaking’’ in a

chest freezer with dry ice and then again after soaking in a ‘‘hot box’’ made from

an old refrigerator and a heat lamp.

In the ﬁeld, the accuracy of the internal voltage reference is maintained over time

and temperature by an automated self-calibration procedure that utilizes a precision

reference diode to calibrate the rest of the system. To avoid interfering with the

routine measurements, the calibration is done in the background in small segments

throughout the 3-min self-calibration cycle. The calibration results are then averaged

across a 15-min period, which keeps the logger calibrated across most naturally

occurring environmental temperature changes .

‘‘Single-ended’’ and ‘‘differential’’ ampliﬁers are commonly used for voltage

measurements. With single-ended measurements the signal is measured with respect

to instrumentation ground. Input connections therefore have one active side and one

ground connection. With differential measurements, one side of the signal is

3 DATALOGGER 787

measured with respect to the other side, requiring that both inputs be active. Often

these connections are labeled a s ‘‘high’’ and ‘‘low’’ or þ and . In some designs

either the high or low side of the differential channel can be used to make two single-

ended measurements. Most sensors require that the voltage channels have high

impedance (use very little current).

Some signals can be measured either single-ended or differentially. Others, such

as resistive full bridges, must be made differentially because the signal is referenced

to one side of the bridge and not to ground. Single-ended measurements are

adequate for high-level voltages, thermistors, potentiometers and similar signals.

Differential measurements provide better rejection of noise common to both

sensor leads, and for this reason they should be used for sensors having low-level

signals (e.g., thermocouples). In addition, the accuracy of a differential measurement

is improved by internally reversing the inputs and making a second measurement.

The second measurement with the inputs reversed cancels out thermal or noise-

induced offset errors in the following manner: Suppose a small change in the

datalogger’s temperature since the last self-calibration is maki ng the high input

read 1 mV high so with a 10-mV signal, the high input reads 11 mV, the low

reads 0 mV, and the ﬁrst differential voltage measured is 11 mV. When the inputs

are reversed, the high input reads 1 mV, the low input reads 10 mV, so the second

differential voltage measured is 9 mV. When the sign of the second is reversed and

the two measurements are averaged, the 1-mV offset cancels yielding a differential

voltage of 10 mV. Good system design requires careful attentio n to even the smallest

measurement details.

Thermocouples

One of the common sensors used to measure temperature is the thermocouple. In

addition to the need for an accurate measurement of a voltage signal usually less than

5 mV, thermocouple measurements require a reference junction temperature and

polynomial equations to convert the voltage to temperature.

Precision Excitation

Resistive sensors, those whose resistance change in response to the parameter they

are sensing, require an excitation voltage to output a voltage that corresponds to the

changing parameter. The excitation voltage needs to be selectable so the output

voltage ﬁlls one of the full-scale ranges. It also needs to have accuracy and resolu-

tion characteristics that match those of the input voltage range. Power consumption

is reduced by shutting off the excitation after the sensors have been measured.

Most resistive sensors can be measured ratiometrically, which means that the

measured voltage is divided by the excitation voltage. Because both the measure-

ment of the signal voltage and the generation of the excitation voltage utilize the

same voltage reference, dividing one by the other cancels out errors in the voltage

reference. As a result, the accuracy of the measurement increases from 0.1% FSR

across the 25 to þ50



C range to 0.02% across the same temperature range.

788 COMMERCIAL RESPONSE TO MEASUREMENT SYSTEM DESIGN

Some of the sensors included in this group are thermistors, RTDs, potentiometers

as used in wind vanes, weighing rain=snow gages, pressure transducers, and load

cells.

Pulse Measurements

Several important sensors output low-level ac signals, square-wave pulse s, or even

just a switch closure. The datalogger needs a voltage supply and a bounce elimina-

tion circuit to measure switch closures. Ampliﬁers are required for the low-level ac

signals. Dedicated counters that accumulate while the logger is sleeping in between

measurement intervals keep the current drain low. The processor must be able to

detect occasions when it is too busy to measure and reset the pulse accumulators at

the designated time. The pulses from the excessively long interval must be discarded

if the signal is a rate or accumulated if it is not. Vibrating wire transducers,

commonly used for long-term monitoring of water pressure or deformation, require

special circuitry and software to determine the frequency of their limited-duration

signal.

Serial Inputs and Digital I= O Ports

In an effort to increase their accuracy or to even be able to make some of the more

difﬁcult measurements, several sensors are now microprocessor based. These

sensors make the requi red measurement, digitize the value, and then transmit

the value in digital form to a datalogger via a serial protocol such as SDI-12,

RS232, RS485, etc. In the case of the SDI-12 protocol, the datalogger tells

SDI-12-based sensors when to make a measurement and then requests the data

after the wait time speciﬁed by the sensor. Some smart sensors simply transmit

their data after the measurements are made on their programmed time interval.

Most serial signals are transmitted and received using the digital input=output

(I=O) port.

Digital ports are needed to input status signals or to output control signals. This

capability allows the datalogger to measure, control, or alarm based on time or

measured values.

Programming Versatility

A well-built datalogger is not much good if it cannot be easily told what needs to be

done. A less versatile datalogger programming language works ﬁne in markets

where there is a lot of repetition. A more versatile programming language is

needed in diverse markets, especi ally where research is the primary objective.

A powerful datalogger language consisting of canned measurement instructions

and canned mathematical processing instructions take the work out of doing good

complicated measurements. If the versatile programming language is backed up with

good PC support software to facilitate the program development, routine programs

become easy and complex programs become possible.

3 DATALOGGER 789

On-Site Data Processing

The advantage of processing measured values to obtain more efﬁcient data storage

has been mentioned. Data compression is particularly important in remote applica-

tions where site visitations are costly. Processed results such as averages, stand ard

deviations, extremes, and values recorded conditionally all reduce data storage and

handling logistics.

Linear calibration constants are entered into the datalogger to convert measure-

ments into engineering units immediately. The datalogger displays the sensor signal

in engineering units that enable the user to verify the correctness of the signal and its

conversion. Field calibration of sensors is possible. User-entered polynomial coefﬁ-

cients are used to linearize nonlinear measurements. Linearization and the scaling of

sensor outputs into engineering units make it possible to correct sensor readin gs

on-site (e.g., correcting a piezometer reading for barometric pressure yields pore

pressure). Nonlinear signals converted to engineering units can then be averaged, but

in their nonlinear form they cannot. The ability to convert sensor signa ls into correct

engineering units is extremely useful when verifying the performance of the sensors

and the datalogger in the ﬁeld.

The ability to compare values or time with programmable limits and make deci-

sions provides useful control functions such as sampling at a faster rate, measuring a

selected sensor, or initiating a radio or phone co mmunication for an alarm message.

Internal Data Storage

The last 25 years of computer hardware innovation has greatly increased the amount

of memory stored per chip. On-board data storage that used to be in increments of 2

kilobytes now comes in increments of 1 and 4 megabytes. To keep up with the rapid

advances in memory storage technology, dataloggers are designed and built with the

ability to add the next generation of memory chips as soon as they became available

or cheap enough to justify their use.

The limited on-board memory of the past required one or more of the following:

compression of the data into hourly or daily summaries, immediate data transfer to

an external data storage device and even those had limited capability, or immediate

data transfer to a compu ter via one of the telecommunication methods. Today’s

extensive memory often eliminates the need for on-site external data storage devices,

allows for more frequent recording of the data, and permits data retrieval on less

frequent intervals.

Today’s on-board data storage is not only more extensive but it is also more

reliable. Part of the data, the system program, and the clock are maintained by an

internal lithium battery that takes over the instant the system voltage drops below the

operational level of 9.6 V. The extended memory is stored in FLASH, a memory chip

that only erases or records data when power is present.

790 COMMERCIAL RESPONSE TO MEASUREMENT SYSTEM DESIGN

Reliability

At ﬁrst glance, reliability and low cost work against each other. For example,

military-grade (mil-spec) parts have higher reliability than industrial-g rade compo-

nents but they cost much more. Only when the full costs of a datalogger, including

warranty repai r, support, and marketing (ever try to sell a lemon?) are understood,

can one properly balance cost and reliability. For example, there are ways to signiﬁ-

cantly reduce the cost of a datalogger without sacriﬁcing reliability. Industrial-grade

parts can be tested upon arrival. Where possible, speciﬁc parts that have proven

reliable in previous dataloggers can be designed into new dataloggers. Rigorous

environmental testing of the completed units result in mil-spec or better reliability

at a fraction of the cost.

Production and repair personnel keep careful records of the parts that fail, hoping

to detect a bad batch of parts before many of the dataloggers ship. This method

catches most of the bad par ts but, unfortunately, there are those rare occasions when

a part only fails after time or after exposure to humidity or temperature variations in

the ﬁeld. This happened to the very popular CR10 datalogger. In the spring of 1998,

a sharp increase in capacitor failures in CR10s returned for warranty repair was

noted. Exa mination of the carefully kept inter nal records ﬁnally linked the cap and

datalogger failures to a change in the cap manufacturing process early in 1997. The

cap manufacturing change caused an internal crack in the capacitor that eventually

led to its failure. Although nearly 2500 dataloggers were shipped with the faulty

parts, the carefully kep t records helped resolve the problem. New capacitors from a

new vendor were installed in units that began to ship in the fall of 1998. A recall was

issued in the fall of 1999 when the rising failure rate indicated the seriousness of the

problem.

Good system design requires good personnel in the test and repair groups. Feed-

back from these groups on the problems and parts that fail must make its way back to

the system design group so that problems can be solved and better parts used in

revisions of current dataloggers or in the design of new dataloggers. A monthly

meeting of a quality control group made up of production, repair, testing, and

engineering personnel helps maintain high quality. Reliability is a direct resul t of

quality. High reliability requires that quality be an issue of continuous concern.

In the ﬁeld, dataloggers are frequently exposed to voltage surges and RF noise.

Dataloggers are best protected from lightning by providing it a low-resistance path to

a single good earth ground. Datalogger lightning protection design has improved

with each new datalogger. The most recent dataloggers are built with separate

grounds for power while the rest of the circuit board is covered with a ground

plane, which serves as the ground for analog signals and sensor shields. Up to

2 A of current can be run through the ground plane without inducing voltage gradi-

ents, which affect analog measurements. A good beefy connection from the ground

plane to an earth ground gives lightning a good low-resistance path to travel. Spark

gaps protect the analog input channels.

3 DATALOGGER 791

The datalogger both creates and receives RF noise. The amount of RF noise is

greatly reduced by the metal package surrounding the electronics. Tighter RF stan-

dards and better test equipment have lead to much better RF shielding.

Inexpensive rugged packaging that offers portability, protection from humidity,

dust, and temperature extremes helps increase system reliability. The ability to power

the datalogger from a variety of power supplies facilitates its use in many environ-

ments.

In the competitive datalogger market, where a company’s livelihood is earned by

small sales to private and publicly funded researchers, most on limited budgets,

success requires good technical support, word-of-mouth advertising, a reasonable

price, and great reliability.

4 DATA RETRIEVAL

The reliability and convenience with which measurements are transferred from the

ﬁeld to the computer are important design goals in any measurement system. Most

of the datalogger design requirements, especially low power and environmental

ruggedness, apply to the data retrieval components. Whether the data are stored

on-site and retrieved during site visits or retrieved remotely using various telecom-

munication options, the manufacturer should provide the necessary hardware and

software tools to accomplish this task. The challenge has been to keep hardware and

software current given the rapid changes in the ﬁelds of electronics, telecommunica-

tions, and personal computers. The following sections brieﬂy describe most of the

data retrieval options currently available. Examples of some speciﬁc beneﬁts result-

ing from new or enhanced data retrieval methods are included.

On-Site Data Storage

An important part of almost all data retrieval systems is the storage of data on-site.

On-site storage holds data between site visits or data transmissions. Should the

telecommunication system fail, data may be retransmitted once the system is

repaired. Printer and magnetic tape methods of on-site data storage have been

replaced by data storage in memory either inside the datalogger or in an external

memory module.

Our ﬁrst fully digital datalogger, the CR21, stored data internally in random-

access memory (RAM) for telecommunication and externally on either thermal

printer paper or a magnetic cassette tape for data retrieval during a sit e visit. The

cost of memory limited internal data storage to only 600 processed data values. The

printer paper held 35,000 data values, but it had to be manually transcribed into

the computer. Cassette tapes held 180,000 values on one side of a 60-min tape.

A tape reader transferred the data to a computer. While the printer stuck in high

humidity and the tape recorder failed below 5



C, they were the only cost-effective

on-site data storage options available at the time.

792 COMMERCIAL RESPONSE TO MEASUREMENT SYSTEM DESIGN

In 1983 Sequoia National Park requested a solid-state data storage module so it

could gather winter precipitation data at high elevations in the Sierra Nevada. The

desire to meet this customer’s need combined with the falling cost of RAM led to the

development of a RAM module powered by a small lithium battery. Due to its

portability and reliability in cold and hot conditions, the memory module and its

successors have become the preferred method of external on-site data storage and

site visit data retrieval. Memory module development led to convenient data storage

under cold winter conditions.

The increased portability and ruggednes s of laptop computers and personal digi-

tal assistants (PDAs) allows them to be used to retrieve internally or externally stored

data. Expansion of the dataloggers internal memory has contributed to this trend by

reducing the frequency of site visits.

Data Retrieval via Telecommunication

Data can be transferred from remote ﬁeld stations to a computer through wire, radio

waves, or a combination of both. Variations on these options have greatly increased

in number and quality in recent years. The new methods have increased data transfer

speed and reliability. Cost, reliability, operating distances, and data rates are impor-

tant in determining the usefulness of a telecommunication system for a particular

application. Remote data retrieval provides timely reporting, early detection of

equipment malfunctions, and, for more isolated sites, may be the only practical

means of data recovery. The expense of manual data collection, particularly in

larger networks, often justiﬁes automated techniques even when near-real-time

reporting is not required. Telecommunication improvements have changed data

transmission methods from teletype or voice into a fully automatic digital transfer

from the datalogger in the ﬁeld to the computer running the synoptic forecast model

or generating the reports.

Where only one-way communication is supported, such as with satellites or

certain RF-based designs, redundant transmissions are often used to obtain high

data capture rates. Interrogated networks based on two-way communication allow

error checking and retransmission of imp roperly received data. Two-way commu-

nication also allows the user to remotely change the datalogger’s program, clock, and

control ports. The following information details advantages and disadvantages of

most currently available telecommunication options.

Dedicated Cable

Short-haul or multidrop modems operate up to distances of several kilometers, the

expense of the cable and its installation being the more practical distance limitation.

Short-haul modems require a PC COM port and an independent cable (two twisted

pairs) for each remo te site accessed by the base station. A local area network (LAN)

links multiple remote sites to a single PC COM port by a single two-conductor cable

(typically COAX). Individual stations within the network are accessed through

addresses set in the LAN modems. Direct cable links are useful in more permanent

4 DATA RETRIEVAL 793

operations and for combining several local sites into a network (e.g., on a

watershed), which can be accessed remotely by a single telephone or RF link.

The communication rate is dependent upon both the cable characteristics and its

length. Model SRM-5A short-haul modem manufactured by RAD Data Commu-

nications, Inc. speciﬁes operation at 1200 and 9600 bit per second up to distances of

10.5 and 8 km, respectively, for 19 AWG wire. A network of 10 of Campbell

Scientiﬁc’s LAN modems (Model MD9) operates at 9600-bit per second up to a

distance of 2 km assuming the loss characteristic of the coaxial cable is 0.02 dB=m.

Switched Telephone

Where available, standard voice-g rade telephone lines provide a simple, reliable

method for retrieving data. A typi cal PC and modem calls and retrieves data from

each remote site. System support software facilitates the scheduling and setup of the

call parameters. The modem at the measurement site needs to run on minimal dc

power and function under hot and cold ﬁeld conditions. Communication rates of

300, 1200, 4800, and 9600 bits per second are standard. Standard phone company

procedures for transient suppression should be followed at the remote site. Suppres-

sion devices included in the modem may be augmented by external suppression

devices.

Campbell Scientiﬁc’s (CSI’s) ﬁrst rugged, low-power 300-bit per second phone

modem (DC103A) combined nicely with the CR21-based automatic weather station

(AWS). These telephone-linked weather stations were quickly utilized to form AWS

networks in New Mexico, Nebraska, and California. In the CIMIS project in the

Sacramento Valley of California, a computer polled each of 40 sites daily, providing

data for evapo-transpiration (ET) models. The introduction of telephone telemetry to

the AWS beneﬁted farmers with daily crop water usage information.

Cellular Telephone

The cellular phone systems established in many countries provide vast networks

of cellular repeaters in most urban areas and along major highways in rural areas.

A cellular phone provides the radio link from a remote site to the nearest cellular

repeater where the message is converted to telephone signals and passed down a

standard phone line to the PC. Stationary sites able to utilize a cellular yagi antenna

can communicate across distances of 20 km if there is line-of-sight to the cellular

repeater site. The cellular link eliminates the installation costs of a standard phone

line. In addition, the mont hly ‘‘air time’’ fee for a cellular site accessed once a day is

typically less than the monthly service fee for a standard phone line. The current

used when transmitting is typically 1.8 A. The stand-by current drain of 170 mA

requires either a larger power supply or that the cellular phone be turned on only a

few hours during the day. Data throughput is limited by the bandwidth of the

frequency to 4800-bit per second. Though the ﬁrst cell phone systems were

analog, overlapping digital networks are rapidly being installed.

794 COMMERCIAL RESPONSE TO MEASUREMENT SYSTEM DESIGN

Internet

Where Internet-g rade T1 phone lines a re available, an Internet modem is used to

access data from one datalogger or a radio-linked network of dataloggers.

Single-Frequency RF

Single-frequency ultrahigh frequency (UHF) or very high frequency (VHF)

networks with individual stations accessed by unique addresses are in general less

expensive and logistically simpler than older style, multifrequency networks.

Designs based on low-power transceivers (2 to 5 W) are available. Operating

distances are limited to line-of-sight, typically 25 to 35 km, but this range is

extended where individual stations may be used as repeaters to access more distant

stations. Access of an RF network through telephone lines fur ther extends the data

collection range. Average power consumption of approximately 80 mA dc (1 A when

transmitting) is typical for many low-powered radios. In standard UHF and VHF

applications, a Federal Communications Commission (FCC) license is required. The

bandwidth is regulated, limiting the RF bit per second rate. Data ‘‘through-put’’ rates

depend upon the communication protocol, the overhead times required to establish

the link, and the length of the transmitted data block. For example, data through-put

rates of around 30 data values per second are reasonable.

At many sites, the cost of installing and maintaining a phone line are prohibitive.

To obtain data from remote mountain tops economically, a radio modem and

VHF=UHF radios were added to meet the needs of customers such as the military

who were doing tests across the 8300 km

White Sands Missile Range. Ski resorts

doing avalanche control with remote, mount ain-top data also beneﬁted from the

radio links.

In 1992 a faster polling scheme was developed for radios. The increased speed of

the new method allows data to be retrieved every 5 min from a 35-station network

monitoring meteorological conditions across the 1400-km

research complex at

Idaho National Engineering and Environmental Laboratory (INEEL). Kennecott

Utah Copper smelter used the same setup to obtain air quality data from pollution

monitors at 20 ambient sites and 3 stack sites every 2 min.

The same technology was taken one step further in the Oklahoma Mesonet

Project (B rock et al., 1995). In that project, 108 weather stations uniformly distrib -

uted across the 180,000 km

of the state measure air temperature, humidity, baro-

metric pressure, wind speed and direction, rainfall, solar radiation, and soil

temperatures. Each station is polled every 15 min for its 5-min data. From a typical

site, the data is transmitted several kilometers by VHF radio to the nearest police

station or sheriff’s ofﬁce. From there it is relayed by the Oklahoma Law Enforcement

Telecommunications System (OLETS) on a 9600-bit per second synchronous dedi-

cated line to the central computer at the Oklahoma Climatological Survey ofﬁce in

Norman, Oklahoma. The central site ingests the data, runs quality assurance tests,

archives the data, and disseminates weather information and warnings in real time to

4 DATA RETRIEVAL 795