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.

home or the thermistor outside wi ll be burned out by a nearby lightning strike and

result in data loss until the ﬁeld manager either makes a site visit to replace the unit

or ships the affected part(s) to the observer for their own replacement. Other causes

of a data loss from the MMTS resulted from the connecting cable being cut by

nearby utility construction or in the western states by various energetic burrowing

animals. Electrical power interruptions to the units either by thunderstorm or winter

snow and ice have also caused data loss over the years. These data losses can be

signiﬁcant when the power outages are widespread and extended, such as was the

case in the New England ice storm of Januar y 1998.

As undesirable as the above-mentioned data losses are, they are at least apparent

and recognizable to the user. This, however, is not true when RF, lightning, or the salt

air of the marine environment cause small changes in the resistance in the electrical

loop between the thermistor and the readout in the observer’s home or business.

Such a resistance change caused abnormally cool temperatures, some of which were

below freezing, to be recorded at a low elevation site in Hawaii. In another case, a

lightning strike caused a resistance change equal to about 4.5



F that went undetected

by the observer, NWS, and NCDC for 6 months. In these cases the errors were not

detected by NCDC’s automated QA program but rathe r by quality assurance

specialists.

3 PRECIPITATION

In terms of identifying global climate changes, anthropogenic or natural, quality

precipitation data is equally as important as temperature; however, because of the

noncontinuous temporal nature and spatial variability of precipitation, its quality

assurance is more difﬁcult. As was the case with temperature data, errors in preci-

pitation data can be introduced at any stage from the original observation to the ﬁnal

quality assurance and archive.

Before continuing with the discussion of precipitation data, it should be noted

that there are many more uses of meteorological data than just long-term climate

analysis. In fact there are thousands of users that depend on high-quality data, and

since many of those users are interested in the condition at a single point and time,

they cannot afford the perceived luxury expressed by some climatologists that errors

in the data sets will likely be random and thus average out over the long run. These

‘‘single point and time’’ uses of the data range from the meteorological conditions at

the site of an auto accident, aircraft mishap, or the structural failure of a bridge or

building. Even though the original reasons for the collection of meteorological data

did not envision such applications, they are none the less vital to our understanding

of the effects of extreme meteorological conditions on human activities and their

supporting infrastructure.

As in the case of temperature measurements, one of the best assurances of quality

precipitation data at the observation site is to have some element of redundancy or

reality check. The high costs of even the basic manual precipitation gages employed

816 DATA VALIDITY IN THE NATIONAL ARCHIVE

by the government-sponsored networks proved to be too expensive to install more

than one identical gage at most cooperative stations. There are, however, about 850

of those stations that are equipped with two gages, o ne manual and the other auto-

matic. We will discuss this subject later in the chapter. There also was more than one

precipitation gage located at most primary NWS sites that were staffed by trained

government or contract weather observers.

Observational procedures at these sites, however, did not require a comparison of

the measured amounts of each gage. The high quality of precipitation data from

these sites came primarily through the aspect of their reporting of hourly, 6 hourly,

and 24 hourly amounts, which provided for cross reference at the station as well as at

NCDC when processing the data for archive and publication.

Unfortunately this cross-referencing process was signiﬁcantly compromised with

the advent of the Automated Surface Observing System (ASOS) in the mid 1990s.

ASOS is equipped with a heated tipping bucket precipitation sensor that does not

always dependably and accurately measure precipitation. Its greatest difﬁculty

occurs during periods of frozen or freezing precipitation. In the years prior to

ASOS, human observers monitored conditions so they could resolve such problems

through the use of a backup gage or make an estimation of precipitation amounts;

however, now with no observers at most ASOS sites to augment the automated

system, the weather indicator Light Emitting Diode Weather Identiﬁer (LEDWI)

may report snow while the precipitation gage does not report any corresponding

meltwater equivalent and, if temperat ures warm to near or above freezing, the snow

will melt and produce a false time and intensity for the precipitation event.

Even in those cases where observers are in place to augment the ASOS observa-

tions, they do not have the ability to alter the one-minute data ﬁle from which the

hourly precipitation ﬁles and publications are produced, the net effect being that no

longer does the hourly precipitation data from the ASOS sites always agree with the

6-hourly, daily, or monthly totals. Needless to say such discrepancies seriously

undermine the integrity and use of the data, particularly in litigation or insurance

adjustment cases.

As was earlier mentioned about 850 of the cooperative weather stations are

equipped with both manua l and automatic precipitation gages. One of the gages is

the ‘‘standard 8,’’ which has an 8-in. diameter receiving funnel that directs the

precipitation into a measuring tube that is one-tenth the area of the receiving

funnel and thus expands the true depth of the rainfall by a factor of 10 and provides

for an accurate measurement of the precipitation to one hundredth (0.01) of an inch.

Today the vast majority of the 2250 automatic gages are Fischer–Porter type of

gages, which also have an 8-in. diameter receiving chute that directs the precipitation

into a large basin where the precipitation is weighed. That weighed amount is

recorded onto a binary punched paper tape every 15 min to a maximum precipitation

resolution of one tenth (0.10) of an inch.

The use of a weighing-type recording precipitation gage in the cooperative

network began in the mid-1930s and conti nued until the mid-1960s when the

Fischer–Porter type began to replace it. This earlier weighing gage was known as

the Universal and was not only used at about 3000 cooperative stations but also

3 PRECIPITATION 817

became the primary gage for the primary NWS stations from the early 1960s until

the mid-1990s when they were phased out with the commissioning of ASOS.

There were two major advantages of the Universal-type gage over the Fischer–

Porter. First the Universal recorded data to the hundredth of an inch (0.01), and

secondly it made a const ant trace of the precipitation event on a chart that allowed a

time resolution of as little as one minute. The accepted world record 1 min rainfall of

1.23 in. (3.12 cm) was extracted from one of the Universal autographic charts.

For quality assurance purposes at NCDC it has been unfortunate that for many

years the data from the co-located standard and automated gages at the cooperative

stations could not be compared during quality assurance and publication because the

receipt and reduction of the punched paper tapes and or autographic charts lagged

about a month or more behind that of the manually reported daily data forms.

The resulting inability to compare the data from these co-located gages prevented

NCDC from effecting a higher level of QA for those sites and in some cases allowed

signiﬁcantly different precipitation amounts to be published for the same location.

The worst known difference was published for a site in Puerto Rico. A heavy tropical

rain event moved into the south central part of the island in October, 1985. The

heavy rains washed out a major highway bridge during the night hours and allowed

29 motorists to drive to their death, one vehicle at a time.

A $65 million lawsuit was later initiated. Obviously, the true amount of rainfall

was critical to the case as well as the knowledge of the true return period rainfall

potential for the area. Unfortunately, the ofﬁcial published daily total from the

manual 8-in. gage was 8.80 in. greater than the sum of the hourly values later

published for the co-located automated Fischer–Porter gage.

In this case the automated gage was equipped with a self-siphoning overﬂow

feature that not only drained the collecting bucket when it reached its capacity, but

because of the extremely heavy rain, continued to siphon until the heavy rain ended,

and thus the gage never accumulat ed or recorded the balance of the precipitation

during the storm.

Fortunately, a more timely receipt and processing of the Fischer–Porter punched

paper tapes now allows NCDC to do a comparison of the co-located station’s

precipitation records. Results of this QA process have also aided NWS cooperative

network ﬁeld managers to identify and correct gage or observer problems in a more

timely fashion.

During the twentieth century various automated precipitation gages have been

employed not only to measure and report precipitation in real time but also in a

retrospective mode to help identify the frequency and nature of extreme precipitation

events throughout the United States and its territories. Any user of these data ﬁles

should pay particular attention to the signiﬁcant changes that have occurred in the

measurement and data extraction over the years.

The most signiﬁcant of these changes occurred in 1973 when the derived extreme

values changed from that of ‘‘excessive precipitation’’ to ‘‘maximum short duration

precipitation.’’ Excessive precipitation data began to be summarized in 1896 and

continued (with several procedural changes) until 1973 when it was terminated in

favor of the simpler maximum short duration precipitation (MSDP), which continues

818 DATA VALIDITY IN THE NATIONAL ARCHIVE

to this day. Both the excessive and MSDP systems used the same time periods (5, 10,

15, 20, 30, 45, 60, 80, 100, 120, 150, and 180 min) for which amounts were

determined; however, there were two fundamenta l differences in their determination.

The ﬁrst was that in determining values for excessive events there were certain

precipitation intensity criteria that had to be met before any values were computed.

There were some variations in the criteria over the early years, but most centered

around that of .01 t þ 0.20 in. (where t ¼time in minutes) except in the south-

eastern third of the United States where .02 t þ 0.20 in. was used until 1949 when

these southeastern states reverted back to .01 t þ 0.20, which placed the whole

nation on the same criteria. Anytime a precipitation event met the above criteria at a

given station, excessive precipitation values were determined for that event, and, if

there were 10 excessive events during a station-month, then there would be 10 events

documented.

This brings us to the second primary difference between the excessive and MSDP

systems. Unde r MSDP there is no intensity criteria that must be met before compu-

tation. Therefore, if the maximum intensity for an event were only 0.01 in. in a

station-month, it would be summarized and become the extreme for the month.

That is because under MSDP procedures there is one extreme summarized, recorded,

and published for each station for each month. As was earlier indicated, the exces-

sive method might have any number (10, 12, 15, or none) excessive events in a

station-month, while under MSDP if there were multiple excessive events only the

most extreme precipitation of each time period (i.e., 5, 10, 15, ..., 180 min) would

be summarized, recorded, and published for that station. Anyone who might

combine the data from these two ﬁles would truly be mixing meteorological

apples and oranges. NCDC has a digital ﬁle (TD 9656) of excessive precipitation

data for the period of 1962–1972, and maximum short duration precipitation (TD

9649) for the period 1973 to the present. Prior to 1962 all excessive precipitation is

available only in manuscript for m. The MSDP data is also available in manuscript

form in the monthly Local Climatological Data (LCD) publications from January

1982 through present.

There are also some similar extreme precipitation values that are published in the

back pages of the NCDC monthly publication Hourly Precipitation Data (HPD).

These data are primarily derived from the automated Fischer–Porter gages in the

cooperative observing network but also include data from the primary NWS stations

as well. Since the maximum time resolution for the automated gages in the coop-

erative network is 15 min, the ‘‘precipitatio n maxima’’ summary pages only report

precipitation values for 15, 30, and 45 min as well as 1, 2, 3, 6, 12, and 24 h for each

station-month.

Once again caution should be exercised when comparing these values with those

recorded=published in either the excessive or MSDP systems. The reason being that

all values presented in the precipitation maxima tables of the HPD publication are

based on data derived from automated gages that are conﬁned to the ﬁxed 15-min

clock times while the excessive and MSDP systems are free to move minute by

minute along the duration of an extreme event(s) searching for the most extreme

precipitation intensity. A close examination of the precipitation maxima table will

3 PRECIPITATION 819

also show that there are no 15, 30, or 45-min values published for the primary NWS

stations. These time periods were omitted in an effort to avoid confusion with the

values presented for the same time periods presented in the MSDP tables in the LCD

publication for the same station. There are 1-, 2-, 3-, 6-, 12-, and 24-h values

published for primary NWS stations in the precipitation maxima table of the

HPD; however, for continuity purposes, the maxima table precipitation values for

the primar y stations are ﬁxed ‘‘clock’’ hour times like the cooperative stations that

constitute the majority of stations in the HPD publication. Therefore, these 1-, 2-,

and 3-h values will seldom agree with the 60-, 120-, and 180-min values published

in the MSDP table of the LCD for that same primary NWS station.

4 SNOW

Given the numer ous problems in measuring snow accurately and uniformly

(Doesken and Judson, The Snow Booklet, 1997) it is no surprise that the quality

of snow data in the climate databases are not always what we would like to see. This

is true not only of data from the cooperative network stations but unfortunately also

from some of the primary NWS stations where, prior to the installation of ASOS

(which does not measure snowfall or snow depth) in the mid-1990s, there were

professionally trained personnel making the observations.

Obviously, no after-the-fact assurance review of snow data, or indeed any meteor-

ological data, can know exactly what occurred at each reporting site; however, with

other accompanying meteorological data or information from that site one can make

some reasonable assessments of the validity of the snow reports. The scope of this

chapter does not allow a full discussion of all the observational problems encoun-

tered in the measurement and quality assurance of snow, but some of the major

difﬁculties encountered will be listed. These will be grouped by (1) observations

from the cooperative network observers and (2) those from the NWS primary station

network.

Snow observations from cooperative stations presented the greatest quality assur-

ance challenges not only because of a wide variety in the knowledge and training of

each observer, but also because of limited supporting meteorological data from the

same station. For example, at least 2000 of the average 6800 published cooperative

stations report precipitation only (no temperatures), and if an observer at any one of

those 2000 sites reports liquid precipitation but no snowfall or snow depth, it is

obviously difﬁcult to know whether to assign a ‘‘missing’’ or a ‘‘0’’ to the database

and publication.

Such determin ations are not too difﬁcult in North Dakota in January or even in

Alabama during the blizzard of March, 1993. Unfortunately, these two cases are not

representative of most stations during the transitional months of the snow season or

for stations located at the geographic margins of a snowstorm.

For those cooperative stations that do report temperatures in addition to precipita-

tion, a ‘‘snow–no snow’’ determination is greatly improved. Automated checks can

820 DATA VALIDITY IN THE NATIONAL ARCHIVE

easily ﬁnd precipitation reports with both maximum and minimum temperatures

below freezing and ﬂag any that do not also include frozen precipitation.

Once the validity of a frozen precipitation event has been established, then the

next challenge is to conﬁrm the magnitude of the reported amount. This thought may

come as a surprise to some, when one realizes that a signiﬁcant portion of the U.S.

population at large (which includes weather observers) does not clearly understand

the proper placement of decimals in writing a bank check (thus the required worded

expression of the numerical value) or reporting precipitation on a weather form.

Liquid precipitation and=or melted snow is reported in the United States to inches

and hundredths (0.00); snowfall to inches and tenths (0.0), and snow depth to whole

inches (0).

A survey (by this writer) of precipitation (including snowfall and snow depth)

reports during the March, 1993, snowstorm showed that only about 25% of the

reports were correct as received by NCDC. Some had only liquid (meltwater)

amounts reported, while most reported snowfall or snow depth with no meltwater.

Combine the above reports with snow depth written in the snowfall column or

snowfall (inches and tenths) written in the snow depth column (whole inches) and

a 24.5-in. snowfall would become a 245-in. snow depth. Several reports had the

snowfall reported in feet, which uncorrected would have a 3-ft (3

) snowfall recorded

as 0.3 in. in the database and publication. Despite these numerous observational

problems, NCDC’s quality assurance specialists are quite adept in extracting the

correct information from the manuscript reports. Much of that ability came from a

knowledge of how each station entered their data in the past as well as an areal

station-to-station check for magnitude errors.

The automated checks of cooperative precipitation and snow observations at

NCDC is very intensive, and the major reason this could be accomplished was

because the observation of each precipitation element (i.e., meltwater, snowfall,

and snow depth) is made at the same time each day at any given station. Unfortu-

nately, the same cannot be said for the primary NWS stations because their obser-

ving practices require a midnight-to-midnight [local standard time (LST)] period for

precipitation=meltwater and snowfall, but a 1200 UTC time (7 a.m. EST, 6 a.m.

CST, 5 a.m. MST, and 4 a.m. PST, etc.) for the snow depth observation, and if the

snow depth on the ground was 2 in. at 1800 UTC time for the measurement of the

meltwater equivalent of the snow on the g round. These observational time differ-

ences caused automated comparison checks to falsely ﬂag more good data than bad,

and, as a result, the automated checks were turned off. As one might expect, data

errors passed on through to the archives and publications (Schmidlin, 1990).

The presence of these errors (mostly meltwater equivalent of snow on the ground)

in the database was not tota lly due to NCDC oversight. There has been a long-

standing verbal understanding between NCDC quality assurance specialists and

NWS observers that NCDC would not change any observed values without

agreement from the observing site in question. Even when presented with signiﬁcant

errors, some observers refused to agree to any correction. One obser ver from a

northern state admitted to this writer that his station only made meltwater equivalent

observations once a week instead of the required once daily time period. Following

4 SNOW 821

are two other examples of obvious water equivalent errors that were recorded at two

primary NWS stations during the March, 1993, blizzard and still exist in the data-

base and publications to this day: A station in the southeast United States recorded

1.85 in. of liquid precipitation for the 18.2 in. of snowfall that fell during the storm;

however, the station also reported 3.8 in. of liquid meltwater equivalent for the 18 in.

of snow on the ground. Note there was no snow on the ground prior to the storm. The

second station was in the northeast and reported 6.3 in. of liquid meltwater, equiva-

lent for 23.2 in. of new snow depth when most other stations in the area reported 1.5

to 2.0 in. of liquid equivalent for similar depths of new snow. Clearly, these two cases

were in error.

5 WIND

Throughout the history of the U.S. Weather Bureau and the National Weather

Service, the basic inst rumentation used for the determination of wind speed has

been a rotating cup anemometer (History of Weather Bureau Wind Measurements

1963). Despite this common design, there have been and continue to be many other

factors that inﬂuence the values of wind speed that are recorded in the publications

and databases. These factors include cup design, time constants=averaging periods,

units of measure, and sensor exposure.

The ﬁrst real cup anemometer (Histor y of Weather Bureau Wind Measurements

1963) was introduced by Robinson in 1846, but it was not until the mid-1870s that

the U.S. Army Signal Service (prede cessor to the U.S. Weather Bureau, which began

in 1890 under the U.S . Department of Agriculture) began using these cup anem-

ometers to record wind speed data. These early anemometers were constructed with

four cups that were hemispherical in shape (i.e., shaped like half of a ball) and had

smooth edges. Over the early decades of the twentieth century scientists learned that

three cups shaped like cones and possessing beaded (rounded) edges on the wind-

ward face of the cups dramatically improved the accuracy of cup-derived wind

measurements. Just to give the reader some sense of the inaccuracy of the four-

cup anemometer speeds, they were about 15% too high at 20 mph, 20% too high at

50 mph, and 25% at 100 mph. In comparison the three-cup speeds were <1% too

high at 20 mph, 2% too high at 50 mph, and <5% at 100 mph.

The above information becomes more important when one learns that the vast

majority of the wind speeds from the four-cup and early three-cup anemometers

were not corrected prior to entr y on the ofﬁcial observation forms until Januar y 1,

1932. Beginning on that date speeds were correcte d prior to entry on the forms.

These corrections continued until the 1950s when the accuracy of the ofﬁcial three-

cup anemometer (Model F-420C) became sufﬁciently accurat e that corrections were

no longer needed.

The three-cup, conical-shaped, beaded-edge de sign of the F-420C was carried

through to the advent of the Automated Surface Observing System (ASOS) in

the 1990s. However, the method of measuring the rotational speed of the

822 DATA VALIDITY IN THE NATIONAL ARCHIVE

cups was changed from an analog method to a digital. We will revisit this important

issue later.

Despite the fact that electricity was used as far back as the 1870s to assist in the

recording of wind speed and direction, the technology of the time still did not

provide the instantaneous measurement of wind speed. At that time wind speeds

were determined by counting (manually and=or graphically) the number of miles

or tenths of miles per unit of time of the anemometer as indicated by the number

of marks drawn on an autographic chart or the number of times an indicator light

was illuminated and counted by the weather obser ver. This type of measurem ent is

essentially the same as reading the miles or tenths of miles off the odometer of an

automobile. If one measures the time required to travel a mile, he or she will then

be able to compute the average speed for that mile. However, there is no way to

know what the maximum speed was during the mile. Therefore, there was no way

to obtain a maximum ‘‘instantaneous speed’’ without a continuous direct read out

of the speed (i.e., a speedometer in the above analogy). It was not until the 1950s

that instantaneous ‘‘peak gust’’ wind values began to be recorded by the military

and not until the 1960s to 1970s for the U.S. Weather Bureau. These direct reading

anemometers used a small electric generator (also known as a magneto) that

produced an electric current that is linearly related to the rotational speed of the

cups of the anemometer and thus the wind speed. The generated electrical

current was displayed on a wind dial or strip chart recorder that was located in

the weather ofﬁce.

Operational requirements of the aviation community were inﬂuential in bringing

about the measurement of these shorter duration wind speeds, and the resulting

recorded values were of a great value to the engineering community for determin-

ing building and construction standards as well as the analysis of post-storm-

related damage or structural failure. Prior to the advent and use of these direct

reading (speedometer) type of anemometers the shortest duration (odometer) wind

speed value that was recorded and published was the ‘‘fastest mile.’’ The observer

usually derived these values at the end of the observational day by ﬁnding the two

closest mile marks on the single or triple register autographic recorder and then

enter that value on the observation form. It has probably already occurred to the

reader that a ‘‘mile’’ is a distance and not a duration. Obviously, the time that it

takes the wind or a car (to continue our analogy above) to go 1 mile will vary

greatly with the speed. (For an average speed of 30 mph, 1 mile will take 2 min; at

60 mph 1 min; and at 120 mph only 30 s.) The extreme wind databases contain not

only these fastest mile values but also ‘‘fastest one-minute’’ values. These one-

minute values were introduced randomly in time during the 1970s and 1980s at the

National Weather Service stations as the old triple-register=multiregister recorders

broke down. Whenever these instruments ceased operating, there was no longer

any standard method by which the fastest mile could be observed or record ed,

therefore the NWS instructed the weather observers to observe the wind dial

(speedometer) for 1 min and record the value both for the regular hourly observa-

tion as well as the ‘‘fastest observed one-minute’’ for the day. Assuming an

observer could accurately integrate the movement of the wind speed dial for a

5 WIND 823

1-min period, the derived values would NOT be comparable (for climatic or

engineering use) to the earlier discussed fastest mile values. Remember that

1 min is a unit of time while 1 mile is a unit of distance and the only speed

where the two values are equal is 60 mph. Let’s go back to our earlier example of

30, 60, and 120 mph. A true ‘‘one-minute’’ wind of 30 mph is one-half the time of

a fastest mile of 30 mph, an equal time at 60 mph, and twice the time of a fastest

mile of 1 20 mph. One other reason why the fastest observed one-minute and

fastest mile values are not comparable without statistical adjustment is the fact

that the fastest mile value was taken from a record er chart that would include

every mile of wind that passed the anemometer during the day while the fastest

observed one-minute value only represented the observed wind speed for one

minute out of each hour. Obviously, the true peak one-minute wind would likely

not be recorded under these circumstances.

During the 1970s and 1980s the NWS began installing strip chart ‘‘gust recor-

ders’’ that constantly recorded the wind speed and allowed the observer to see the

near instantaneous speed of the wind. This made it easier for the observer to make a

more accurat e estimation of a one-minute wind speed but was still difﬁcult during

periods of strong or gusty winds.

However, these gust recorders did provide the accurate measurement of wind

‘‘gusts,’’ i.e., short duration wind speeds of 1 to 3 s depending upon the speed and

gustiness of the wind. As was previously mentioned, these shor t duration values were

important for aviation, structural design, and forensic use. For instance, the maximum

dynamic pressure of the wind upon an average size home can occur in as little as 2 to

3 s. Smaller structures or objects can be affected by even shorter duration gusts.

Starting in 1992 the NWS and later the Federal Aviation Admini stration (FAA )

and U.S. Military Services began installing automated weather observing platforms

(ASOS=AWOS). Even though most of the platforms use a three-cup anemometer

that is nearly identical to the older F-420 series used by the NWS, the internal

workings of the anemometer are different in that a ‘‘light-chopper’’ device sends

digital pulses to the computer, which samples the wind speed every second and

builds a temporary storage of ﬁve (1 s) wind speed values and then computes a 5-s

average that is stored for the observational day to be transmitted in real-time reports

as well as the now ‘‘peak 5-s’’ wind for the day that is passed on to the climate

databases and publications.

A 5-s gust differs signiﬁcantly from the 1- to 3-s gusts that were recorded on the old

analog strip chart gust recorders when you realize that the 5-s digital value is trun-

cated. That is because the 5-s sum and resulting computer average is not recomputed

with each new 1-s value. Rather the ASOS wind algorithm ingests ﬁve values and

computes an average and then repeats the process 5 s later. The net effect of this logic

dampens the magnitude of a true 5-s wind gust if computed with a running 1-s update.

Unfortunately, for data users the difference between the old analog gust values

and the newer digital ASOS records were further complicated by the discovery that

the ﬁrst 3 or 4 years of ASOS-derived 5-s gust values were actually generated by the

manufacturer’s ‘‘test’’ algorithm that produced wind speed values that were lower

than the correct running 5-s means. The erroneously determined values remain in the

824 DATA VALIDITY IN THE NATIONAL ARCHIVE

ofﬁcial databases and publications without adjustment or documentation as to when

the correct algorithm was installed.

In addition to the previously discussed changes in wind speed observations, there

is one other signiﬁcant change that coincided with the implementation of ASOS and

that was a doubling of the averaging period from 1 to 2 min for the regular hourly

wind speed observations.

The ﬁnal issue to be discussed here in relation to wind speed observations is not

as signiﬁcant as those previously discussed but nevertheless may show up in the

reader’s research or analysis. This issue deals with the units of measure.

Historically, wind speed measurements in the United States have been in miles

per hour (mph); however, with the increasing inﬂuence of the military services on

civilian weather observations during and after World War II, the knot (nautical mile

of 6076 ft or 1.1508 statute miles) was eventually adopted by the U.S. Weather

Bureau as the standard unit for wind speed in the mid-1950s. However, for public

consumption, miles per hour were still used both in terms of nonmarine forecasts as

well as historic publications like the Local Climatological Data (LCD) published by

the NCDC. The NCDC digital databases and publications contain wind speed values

in units of mph and=or knots.

ASOS wind observations are also recorded and disseminated to the aviation

community in knots, but disseminated to the general public and published by

NCDC in the daily summaries of the LCD in mph.

The reason for raising this issue is that in the arithmetic conversion

(1.1508 knots) of a whole knot value to a whole mph value causes certain mph

values to be omitted. For example, 3 knots 1.1508 ¼3.45 mph when rounded

becomes 3 mph while 4 knots 1.1508 ¼4.60 mph rounded becomes 5 mph and

thus the value of 4 mph never shows up.

The same is true for the mph values of 11, 19, 27, 34, 42, etc., thus undermining

wind speed frequency distributions of 1 mph classes.

6 WEATHER

Changes in equipment and observational practices are not restricted to those

discussed above; rather, this is probably just the tip of the proverbial iceberg. The

transition from human to automated determination of sky cover and visibility has

brought its own set of discontinuities. In terms of sky cover the human observer

viewed the whole sky while the automated (ASOS=AWOS) systems use a ﬁxed laser

beam and depend on the sky=clouds to move past them to be able to integrate a sky

condition. This integration process works well most of the time, but still is limited to

a ceiling height of 12,000 ft above ground level (agl) and thus does not indicate the

presence of any high-level altocumulus, altostratus, or cirrus clouds. As for visibility,

it is determined from an approximately 3-ft baseline at one location and is limited to

a maximum reported distance of 10 miles. Therefore, any historic visibility studies

that involved values greater than 10 miles would end at the time ASOS was

commissioned at each observation site.

6 WEATHER 825