SURFRAD-A National Surface

Radiation Budget Network

for Atmospheric Research
John A. Augustine,* John J. DeLuisi,* Charles N. Long+

ABSTRACT

A surface radiation budget observing network (SURFRAD) has been established for the United States to support
satellite retrieval validation, modeling, and climate, hydrology, and weather research. The primary measurements are
the downwelling and upwelling components of broadband solar and thermal infrared irradiance. A hallmark of the net-
work is the measurement and computation of ancillary parameters important to the transmission of radiation. SURFRAD
commenced operation in 1995. Presently, it is made up of six stations in diverse climates, including the moist subtropi-
cal environment of the U.S. southeast, the cool and dry northern plains, and the hot and arid desert southwest. Network
operation involves a rigorous regimen of frequent calibration, quality assurance, and data quality control. An efficient
supporting infrastructure has been created to gather, check, and disseminate the basic data expeditiously. Quality con-
trolled daily processed data files from each station are usually available via the Internet within a day of real time. Data
from SURFRAD have been used to validate measurements from NASA's Earth Observing System series of satellites,
satellite-based retrievals of surface erythematogenic radiation, the national ultraviolet index, and real-time National En-
vironmental Satellite, Data, and Information Service (NESDIS) products. It has also been used for carbon sequestration
studies, to check radiative transfer codes in various physical models, for basic research and instruction at universities,
climate research, and for many other applications. Two stations now have atmospheric energy flux and soil heat flux
instrumentation, making them full surface energy balance sites. It is hoped that eventually all SURFRAD stations will
have this capability.

1 • Introduction site selection, data handling methods, instrument cali-

bration philosophy, and operational procedures en-
The National Oceanic and Atmospheric Admin- sures a continuous research-quality dataset of the
istration's (NOAA's) Surface Radiation budget net- highest caliber. The varied uses of SURFRAD data
work (SURFRAD) is the first of its kind to operate include validation of radiation transfer codes and sur-
across the United States. The network began in 1995 face physics packages of weather, hydrology, and cli-
with four stations and expanded to six in 1998 (Fig. 1). mate models; assessment of satellite-based retrievals
Its mission is to provide the climate research, weather of surface radiation parameters; monitoring trends;
forecasting, satellite, and educational communities instruction; and basic research. A key to the appeal of
with continuous, accurate, high quality surface radia- SURFRAD data for research is its suite of supplemen-
tion budget measurements for different climates of the tary measurements and calculated quantities that af-
United States. Quality assurance in the station design, fect the transmission of radiation to and from the
surface of the earth. These aid in the interpretation of
the measurements and help improve the understand-
*NOAA/Air Resources Laboratory, Surface Radiation Research ing of the effects of clouds, aerosols, and water vapor
Branch, Boulder, Colorado. on the surface radiation budget.
+Atmospheric Radiation Measurement Program, Pacific North- Prior to SURFRAD, the last substantial effort by
west National Laboratory, Richland, Washington. NOAA to monitor solar radiation operationally over
Corresponding author address: John A. Augustine, NOAA/Air the United States was the solar radiation (SOLRAD)
Resources Laboratory, Surface Radiation Research Branch, 325
Broadway, Boulder, CO 80303. network. It began in the middle 1970s and consisted
E-mail: augustin@srrb.noaa.gov of global and normal incident solar radiation measure-
In final form 21 April 2000 ments at 30 National Weather Service (NWS) stations.

Bulletin of the American Meteorological Society 2339

With its demise from neglect in the early 1990s, the This article describes the SURFRAD network, focus-
satellite, climate, and modeling communities were sing on its rationale, design, instruments, operating
without suitable surface radiation measurements over procedures, data management, and plans.
the United States for use in validation activities and
research. A notable consequence was that scientists at
the National Aeronautics and Space Administration 2. Rationale
(NASA) were forced to use European network data to
validate the first attempts at retrieving the surface ra- In 1990, the Committee on Earth and the Environ-
diation budget and other parameters from satellite data mental Sciences (CEES) stated that the understand-
(Darnell et al. 1992). Given that stark reality, and the ing of the effects of clouds in climate and hydrologic
perception that more comprehensive high quality sur- systems was the highest priority of the nascent U.S.
face radiation measurements were needed over the Global Change Research Program (USGCRP) (CEES
United States, the NOAA Air Resources Laboratory 1990). This committee's hierarchy of priorities sup-
(ARL) began an effort in 1993 to reclaim the plied the impetus for the Department of Energy's
SOLRAD network by creating the Integrated Surface (DOE) Atmospheric Radiation Measurements (ARM)
Irradiance Study (ISIS; Hicks et al. 1996). The plan program (Stokes and Schwartz 1994), in recognition
was to refurbish 10 of the original stations to provide that in addition to precipitation and the release of la-
some continuity to the historical data record, and to tent heat, the role of clouds in weather, climate, and
add broadband ultraviolet-B (UVB) measurements. hydrology is gauged by their radiative properties.
Another component of the ISIS strategy was More recently, the National Academy of Sciences
SURFRAD, which was endorsed by NOAA's Office Committee on Global Change Research (CGCR) re-
of Global Programs (OGP) in support of the Global affirmed the USGCRP's need for long-term climate-
Energy and Water Cycle Experiment's (GEWEX) based in situ observations of fundamental external
Continental-Scale International Project (GCIP). forcing parameters, including surface radiation (CGCR
SURFRAD station locations were to be carefully cho- 1998). Their key recommendations emphasized more
sen for their suitability to satellite retrieval validation balance between satellite and terrestrial observations
and to benefit GCIP hydrometeorological research. and a crucial need for innovative well- and cross-
With that, and an expressed need by the World Cli- calibrated observing systems. The SURFRAD and
mate Research Program for a U.S. component of a glo- ARM Southern Great Plains (SGP) networks (Fig. 1)
bal ground-based Baseline Surface Radiation Network directly address these needs and complement each
(BSRN; WMO 1991), SURFRAD was established. other well. While both support satellite validation, the
ARM SGP network clusters stations
over a mesoscale area to support process-
type research and climate model valida-
tion, and SURFRAD extends the same
measurements to different climates of the
United States. Ancillary measurements
of cloud fraction and cloud opacity, in
conjunction with the radiation measure-
ments, make SURFRAD's data ideal for
studies of the effects of clouds on the sur-
face radiation budget, and for the valida-
tion of satellite retrieval algorithms. The
recent addition of surface energy flux,
soil temperature profiles, and C0 flux
2

instrumentation at some stations makes

the network ideal for the validation of the
surface physics of weather and climate
FIG. 1. The SURFRAD surface radiation budget network. The rectangular area models, and well suited for research.
over Kansas and Oklahoma represents the location of the Department of Energy
Atmospheric Radiation Measurements program's Southern Great Plains mesoscale Over its first five years of operation,
array of about 20 surface radiation budget stations. SURFRAD has built a diverse clientele,
2340 Vol. 81, No. 10, October 2000
including scientists at NOAA's National Centers for guidelines stipulated that stations should not be located
Environmental Prediction (NCEP), NASA, other gov- near concentrated sources of anthropogenic aerosol
ernment agencies, universities, and many others who emissions, nor should they be located close to large
use the surface radiation budget data for research and bodies of water. The latter was a concern because cloud
education. The SURFRAD network has been named patterns associated with the land-water difference might
the primary validation network over the United States not be representative of the general area. Hydrologists
for the Clouds and the Earth's Radiant Energy Sys- from Princeton University and the NOAA Office of
tem (CERES) instrument aboard the NASA Earth Ob- Hydrology requested that some of the initial stations
serving System (EOS) satellites TRMM (Tropical should be located within well-monitored drainage
Rainfall Mapping Mission), Terra, and Aqua. Its data basins to aid hydrological modeling and other GCIP
have been used extensively in the USGS Mississippi research. A final condition was to avoid central Kansas
Basin Carbon Project, which studies processes affect- and Oklahoma because of redundancy with the Depart-
ing carbon sequestration in terrestrial ecosystems. Spe- ment of Energy's (DOE's) ARM SGP site.
cifically, the surface radiation budget data are applied
in the development of relationships between climate b. Stations
variables and soil moisture in models of carbon flux, The SURFRAD network is shown in Fig. 1, and
and they have used SURFRAD data in modeling stud- Table 1 lists pertinent information about the stations.
ies that attempt to predict changes in the net primary The three initial stations at Bondville, Illinois, Fort
product as the climate changes. SURFRAD is also the Peck, Montana, and Goodwin Creek, Mississippi,
primary validation source for real-time surface radia- were installed in 1994. A fourth was added in 1995 at
tion fluxes derived by NESDIS for the Geostationary Table Mountain near Boulder, Colorado. All four were
Operational Environmental Satellite (GOES) GCIP/ located within the Mississippi River drainage basin in
Surface Radiation Budget project. SURFRAD data support of the GCIP experiment. Goodwin Creek was
have been used to study the diurnal, seasonal, and in- selected with an emphasis on hydrology; it is within
terannual variations of surface radiation, and the ef- the Goodwin Creek Experimental Watershed and is
fects of sky conditions on surface radiation (Jin and hosted by the U.S. Department of Agriculture's (USDA)
Dickinson 1999). The network has likewise developed National Sedimentation Laboratory. Bondville is in a
an international stature. Since access statistics from the flat agricultural region and is collocated with several
network's Web site commenced in late January of 1999, other research stations on a large site managed by the
the SURFRAD page has served requests from over 60 Illinois State Water Survey. The station at Table
countries, with several institutions registering hun- Mountain is on a mesa approximately 15 km north of
dreds of contacts per month. Boulder, Colorado, and a few kilometers east of the
foothills of the Rocky Mountains.
The Fort Peck station is on the Fort Peck Tribes
3. The SURFRAD network Reservation in northeastern Montana, and is attended
by their Environmental Office. Its selection satisfied
a. Site selection the needs of hydrologists, satellite scientists, and cli-
The site selection process for SURFRAD was a col- matologists. The station is located along the Poplar
laborative effort among NOAA, NASA, and university River that is well monitored with stream gauges from
scientists. Station locations were chosen to represent the Canadian border to the point where it flows into
diverse climates of the United States. With attention to the Missouri River. An important consideration for
satellite validation, consideration was given to locations hydrologic research is that the Poplar River has no
where the landform and vegetation were homogeneous artificial diversions, that is, it is not used for irrigation
over an extended region so that the station measure- nor for a municipal water supply. Another motive for
ments would be representative of a "pixel-sized" area. its selection is that northeastern Montana has a high
Satellites from which surface radiation parameters are interannual variation of snow cover, which Karl et al.
computed (e.g., AVHRR, NASA EOS, GOES) have (1993) reports may have an important bearing on re-
spatial resolutions at nadir that range from 1.1 km to gional climate change. This characteristic is evident
20 km. To account for the largest of these, the land- in the first 3 yr of monthly mean albedo from Fort Peck
form and vegetation of locations being considered was (Fig. 2). Note the high albedo for five months during
surveyed for uniformity over a 10-km radius. Other the winter of 1996-97 (indicating the persistent snow

Bulletin of the American Meteorological Society 2339

 State College, Pennsylvania, and is hosted and at-
tended by The Pennsylvania State University's Depart-
ment of Meteorology.
The land form at both Fort Peck and Bondville is
flat prairie with native grasses and very few trees. That
at Goodwin Creek is rolling hills covered with pas-
ture grass and sparsely distributed deciduous trees. At
the Penn State station, the surface viewed by the up-
welling radiometers is approximately three-fourths
grass and one-fourth crops (southwest corner). The
surface at Table Mountain is sandy with a mix of ex-
posed rocks, sparse grasses, desert shrubs, and small
cactus. The character of the underlying flora there
FIG. 2. Mean monthly surface albedo for the Fort Peck changes seasonally, that is, it is usually green in the
SURFRAD station from 1995 through 1998. High values (> 0.5) late spring and early summer, and browns signifi-
indicate persistent snow cover during the month.
cantly by midsummer. At Desert Rock the surface is
made up of fine rock and scattered desert shrubs
cover) compared to the large variations during the (mainly creosote bush, see Fig. 3). The surface is fairly
winters before and after. stable, having tightly bound soil and practically no
Two stations were added in 1998. An ISIS station at seasonal change in vegetation. Generally the stations
Desert Rock, Nevada, was converted to a SURFRAD are not enclosed, however, barbed wire fences were
station in March of that year (Fig. 3). It is located near necessary at Goodwin Creek and Fort Peck to prevent
the southern boundary of the Department of Energy's livestock from damaging the stations. Pictures of all
Nevada Test Site, and serves the need for sampling a of the stations may be found on the World Wide Web
hot arid climate. It is also the only station to be collo- at http://www.srrb.noaa.gov/surfrad/surfpage.htm.
cated with an operational rawinsonde station (DRA).
The SURFRAD station in central Pennsylvania (Penn
State; Fig. 4) was installed in June of 1998, and ful- 4. Station design
fills the desire of the original scientific advisory board
for sampling the northeastern U.S. environment. It is a. Arrangement of the stations
located in a broad Appalachian valley on an agricul- The three primary components of a SURFRAD
tural research farm approximately 10 km southwest of station are a main platform, a 10-m tower, and a solar

TABLE 1. Station location information.

Annual
rainfall Elevation U.S. Station
Station name Latitude, longitude (cm) (m) Nearest city state Start date ID
Bondville 40.06°N, 88.37°W 94 213 Bondville IL 1 Jan 1995 BON
Fort Peck 48.31°N, 105.10°W 30 634 Poplar MT 28 Jan 1995 FPK
Goodwin Creek 34.25°N, 89.87°W 145 98 Batesville MS 1 Jan 1995 GWN
Table Mountain 40.125°N, 105.237°W 46 1689 Boulder CO 19 June 1995 TBL
Desert Rock 36.626°N, 116.018°W 18 1007 Mercury NV 16 Mar 1998 DRA
Penn State 40.72°N, 77.93°W 99 376 Pine Grove Mills PA 29 June 1998 PSU

2340 Vol. 81, N o . 10, October 2000

tracker stand. The main platform is a fiberglass grat-
affects the earth and atmosphere. Primary measure-
ing (0.3 m x 3 m) that is elevated 1.5-2 m above ments are the up welling and downwelling components
ground level by two steel posts (Fig. 4). Weatherproof
of broadband solar and thermal infrared irradiance.
boxes that house the power distribution system, data Downwelling global solar and thermal infrared are
logging assembly, and communications equipment are measured on the main platform by an upward-facing
fastened to these posts. Within 10 m of the main plat-
pyranometer and an upward-facing pyrgeometer, re-
form is a separate stand that supports an automatic spectively. During the annual station visits in 2000,
solar tracker. The 10-m tower is generally within the pyrgeometer on the main platform will be moved
100 m of the main platform and to the north. An ex- to the solar tracker so that its dome will be shaded, fol-
ception to this is Desert Rock, where local conditions
lowing suggestions by the BSRN (McArthur 1998).
forced the placement of the tower to the south of the main
To measure upwelling irradiance, a pyranometer and
platform, but far enough away to not interfere with the
a pyrgeometer are mounted facing downward within
downwelling measurements. The Penn State station's closed-end cylindrical sun shields at either end of a
main platform is the tallest of all of the stations (~2 m).
2.4-m cross arm near the top of the tower.
It was built that way to preclude interference from tall Although upwelling and downwelling solar and
crops that may be grown on the research farm. infrared measurements are all that are necessary for the
surface radiation budget, other measurements are made
b. Radiation measurements to bolster the network's value to research. The direct
Radiation measurements at a SURFRAD station and diffuse components of downwelling solar irradi-
cover the range of the electromagnetic spectrum that ance are measured independently by a normal-

FIG. 3. The Desert Rock SURFRAD station as viewed from FIG. 4. The SURFRAD station at Penn State as viewed from
the north. In the foreground is the solar tracker; on the right is the the southeast. The main platform is in the foreground. Directly
Total Sky Imager (TSI); on the left is the main platform; and in behind the main platform is the solar tracker stand, and in the
the background is the 10-m tower. The surface at Desert Rock is background stands the 10-m tower. The scaffolding in the far
made up of small rocks; the sparse vegetation visible in the pic- background is not part of the SURFRAD station.
ture is primarily creosote bush.

Bulletin of the American Meteorological Society 2339

incidence pyrheliometer (NIP) and a shaded pyranom- trum that is active in photosynthesis (400-700 nm).
eter, respectively. The sum of these components is con- For aerosol optical depth, each station either has, or
sidered to be the best measure of downwelling global has local access to, a Multi-Filter Rotating Shadow-
solar irradiance because at large solar zenith angles, band Radiometer (MFRSR) (Harrison et al. 1994). The
when the sun is not obstructed by clouds, the global MFRSR alternately measures global and diffuse so-
solar measurement may be degraded by cosine errors. 1 lar radiation in one broadband solar channel and six
Other radiometers at SURFRAD stations monitor narrow bands of the solar spectrum. The narrowband
wavebands of special interest. A broadband ultravio- measurements are useful for recovering aerosol opti-
let radiometer measures the degree of ery thematogenic cal depth, column ozone, column water vapor, and may
UVB irradiance (280-320 nm) that reaches the surface, be manipulated to estimate the size distribution of
that is, that portion of the UVB irradiance that is harm- aerosols in the column. The Quantum sensor and
ful to human skin (McKinlay and Diffey 1987). A MFRSR also provide a quality check on the primary
Quantum sensor monitors that part of the solar spec- downwelling solar measurements.
Table 2 lists the salient characteristics of radiom-
eters used at SURFRAD stations. Spectral responses of
the instruments are a complicated function of the type
'Cosine errors associated with a global solar measurement are of detector and the transmission characteristics of the
caused by differential reflection of the solar beam by the horizontal domes, diffuser disks, and filters inherent to the instru-
sensor of the pyranometer as a function of the solar zenith angle.
Ideally, the sensor should absorb all of the incident radiation, but ments. To provide a definitive spectral responsivity of
in actuality it does not, especially when the sun is low (large so- each radiometer is beyond the scope of this article,
lar zenith angles). however, the nominal waveband that each is respon-

TABLE 2. SURFRAD radiometer characteristics.

Waveband or Dome/diffuser SURFRAD

Instrument Manufacturer wavelength (nm) and filters Detector measurements
Pyranometer Spectrolab and 300-2700 two WG295 thermopile global, diffuse,
Eppley Schott filter and up welling
domes solar irradiance
Pyrgeometer Eppley 4000-50 000 silicon dome thermopile downwelling and
coated with upwelling
interference infrared irradiance
filter
Pyrheliometer Eppley 200-4200 1-mm infrasil thermopile direct solar
quartz plate irradiance
UVB radiometer Yankee Environmental 280-320 Schott WG280 GaAsP global
Systems UV-transmitting photodiode ultraviolet B
dome, and irradiance
UG11 filter
Quantum sensor LI-COR 400-700 acrylic diffuser silicone global
disk, interference photodiode photosynthetic-
filters ally active
irradiance (PAR)
MFRSR Yankee Environmental 415, 500, 615, spectralon silicone global and diffuse
Systems 673, 870, 940 diffuser disk, photodiode spectral solar
interference irradiance
filters

2346 Vol. 8 7, N o . 7 0 , October 2000

sive to, as reported by the manufacturers, is given in welling global solar traces from an unventilated and
Table 2. ventilated pyranometer are contrasted on a clear-sky
morning following a nighttime snowfall. In this case,
c. Meteorological measurements the ventilated pyranometer is part of the SURFRAD
Atmospheric pressure, air temperature, relative suite at the Table Mountain station. Although the
humidity (RH), and wind are monitored at SURFRAD morning sky was relatively cloud free, snow on the
stations. All but surface pressure are measured near the unventilated pyranometer caused erroneous measure-
top of the 10-m tower. A lOOO-Ohm precision resis- ments until about 1700 UTC (1000 AM, LST),
tance thermistor is used for air temperature, and RH whereas the ventilated SURFRAD pyranometer mea-
is measured with a capacitive humidity sensor. These surements show no detrimental effects. Anomalous
sensors are combined into one probe that is housed in scattering by the snow on and around the dome of the
a gilled radiation shield. Station pressure is measured unventilated pyranometer caused its measurements to
with an analog silicon capacitive barometer that is be initially high by about 50 W m~ , then low by about
2

mounted in the data logger enclosure under the main 50 W nr , and then high again by > 200 W nr . After
2 2

platform. the snow melted, traces from the two pyranometers

Accuracy of the temperature sensor is ±0.5°C; that mirror each other with only slight discrepancies that
of the relative humidity sensor is ±2%-3% below 90% may be attributable to calibration differences, stabi-
RH, and ±6% above 90% RH. The wind monitor is a lizing effects of the ventilator, or both. The UVB ra-
propeller anemometer/vane combination. After its diometer and MFRSR are thermally stabilized at a
annual refurbishing, which is conducted according to high temperature (-45 °C). The primary purpose of
the manufacturer's recommendations by the Atmo- this is to stabilize the characteristics of the filters used
spheric Turbulence and Diffusion Division of ARL, in these instruments, but the warm surface also pre-
the resulting accuracy is ±2% for wind speed and ±5° cludes the formation of ice and dew on the instru-
for wind direction. The stated accuracy of the barom- ments, and minimizes residence time of rain and snow.
eter is ±4 mb between -20° and 0°C, and ±2 mb be- The tower radiometers, Quantum sensor, and pyrhe-
tween 0° and 40°C. liometer currently have no protection from the effects
of moisture. However, by virtue of their orientation
and the sheltering provided by the sun shields, the
5. Quality assurance tower radiometers are rarely adversely affected by the
weather.
a. Quality assurance in the station design The cross arm on which the tower radiometers are
Preventive measures in the station design help to mounted is oriented north-south. The pyranometer is
ensure high quality measurements and minimum at the south end to minimize the effect of the tower
down time. Because of local considerations, each
station's arrangement is unique; however, all of the
layouts ensure that the station structures or the instru-
ments themselves do not interfere with the measure-
ments. The primary factor preventing self interference
is that the three platforms are oriented and aligned
north-south. This assures that when the sun is near the
horizon, the long shadows of the structures and instru-
ments are cast normal to the orientation and alignment
of the platforms. Also, to prevent inadvertent shad-
owing, instruments are mounted on the main platform
such that their sensors are approximately at the same
height.
The pyranometers and pyrgeometers on the main
platform and solar tracker are ventilated for thermal
stability and to reduce the detrimental effects of dew, FIG. 5. Time series of global solar irradiance from a ventilated
frost, and precipitation. The importance of dome ven- (solid) and an unventilated (dashed) pyranometer on a day fol-
tilation is demonstrated in Fig. 5 where daily down- lowing a nighttime snowfall.

Bulletin of the American Meteorological Society 2339

shadow on the measured upwelling solar radiation. change, returning instruments are also operated along-
The pyrgeometer is at the north end of the cross arm. side the reference instruments, completing a pseudo-
The cylindrical sun shields that protect the tower ra- overlap of the two monitoring sets. These comparisons
diometers are made of aluminum that is powder coated contain valuable information that can be used to ex-
on the outside with a glossy white paint to reflect sun-plain discrepancies between monitoring epochs at a
light. Access ports for the instrument cables are cut particular station. However, such differences should
into the north-facing sides of the shields, eliminating be minimized owing to SURFRAD's frequent calibra-
the possibility of direct sunlight illuminating the in- tion policy.
struments. The way the radiometers are mounted The precision of the network data may be quanti-
within these shields also prevents unwanted signals fied using the pseudo-overlap data. To do this, the
from reaching the sensors. The pyranometer's sensor overlap data was compiled for all instrument ex-
is recessed slightly within the cylindrical shield's changes that have taken place. The pyranometers were
opening so that the beam of the rising or setting sun found to be quite stable. The mean absolute difference
will not be detected, and the interior of the shield is between the returning and replacement instruments
painted flat black to absorb stray visible light. The was 1.2%, with a standard deviation of 0.9%. The
sensor of the tower pyrgeometer protrudes slightly Quantum sensors were found to be less stable, with a
below the shield opening, thus preventing any infra- mean absolute difference of 4.7% and a standard de-
red emission from the shield itself from being sensed viation of 3.6%. The behavior of the latter was ex-
by the thermopile. The interior of the pyrgeometer's pected because the Quantum sensor is a filtered
protective cylindrical shield is left unpainted because instrument that is not thermally stabilized (as are the
bare aluminum has less of an emissivity than paint in UVB radiometer and MFRSR). For the UVB instruments
the infrared. and pyrgeometers, the pseudo-overlap information is
All stations use commercial power and telephone used to calibrate those instruments, and thus cannot
services. The electrical power is conditioned by a be used to quantify their precision in the network.
uninterruptable power supply (UPS) with battery The manufacturer of the humidity chip recom-
backup that will sustain the entire system for several mends that it should be replaced biennially; however,
hours if the power fails. Power lines throughout the to ensure high quality measurements, it is replaced
station are run separately from the data cables to pre- each year. Before conversion to the temperature and
vent radio frequency (RF) noise from corrupting the relative humidity probe now deployed, a lesser qual-
instrument signals. A lightning rod at the top of the ity sensor was used. That RH chip was designed to be
tower is connected to heavy-gauge braided copper wire replaced annually, but in very humid environments
that is firmly secured to a 2.4-m copper grounding rod such as Goodwin Creek, its sensitivity usually waned
that is driven vertically into the ground. The solar within six months. Conversion to the more robust tem-
tracker and main platform are similarly grounded. perature and RH probe took place in May 1996 at
Signal cables have been standardized to shielded 20- Table Mountain, in October 1996 at Fort Peck, and in
gauge wire. All cables that run from the main platform June 1998 at Goodwin Creek and Bondville. The Penn
to the solar tracker and tower pass through a buried State and Desert Rock SURFRAD stations began their
liquid-tight conduit to protect them from UV exposure, tenure with the more robust probe. Before deployment
heat, moisture, and wildlife. of the wind monitors, the bearings are replaced and the
electronics are checked following the manufacturer's
b. Annual cycling of instruments recommendations. The temperature and pressure sen-
All radiometers at SURFRAD stations are ex- sors are not serviced annually, but they are continu-
changed annually. An integral part of this procedure ally monitored and replaced if necessary.
is the documentation of differences between the re-
placement instruments and those being replaced. c. Frequent calibration of the reference
Before an exchange, the designated replacement in- instruments
struments are operated for at least five days alongside The reference instruments are calibrated annually.
sets of reference instruments at ARL's Table Moun- The three reference pyrgeometers are sent to the World
tain Field Calibration and Test Facility. This also Radiation Centre (WRC) at Davos, Switzerland, and
serves as an opportunity to identify troublesome in- the three reference pyranometers are calibrated at the
struments before they are deployed. After the ex- National Renewable Energy Laboratory (NREL) in

2340 Vol. 81, No. 10, October 2000

Golden, Colorado. Both of these centers are recog- tional Institute for Standards and Technology (NIST).
nized by the World Meteorological Organization The stated uncertainty of those calibrations is ±5%.
(WMO). The three reference UVB instruments are main-
The method employed at NREL to calibrate tained by ARL's Central UV Calibration Facility
pyranometers and pyrheliometers is the Broadband (CUCF), which was developed in close cooperation
Outdoor Radiometer Calibration (BORCAL) (NREL with the NIST in Gaithersburg, Maryland. The CUCF
2000), which uses the sun as the source. In this pro- calibrates the UVB reference instruments as a function
cedure, direct and diffuse components of solar radia- of solar zenith angle by comparing their measure-
tion are measured independently using an absolute ments to the output of a state-of-the-art UV spectro-
cavity radiometer and a shaded pyranometer, respec- radiometer (Lantz et al. 1999). Lantz et al. (1999)
tively. The calibration of the cavity radiometer used report that the uncertainty of this calibration transfer
at NREL is traceable to the World Radiation Refer- process is ±4%.
ence (WRR) standard at the WRC in Davos, Switzer-
land. Cavity radiometer data is used to directly d. Frequent calibration of the monitoring
calibrate pyrheliometers. The combination of the cav- instruments
ity radiometer data and diffuse solar measurements is No instrument is deployed in the network without
used to calibrate the pyranometers. Only cloudless a fresh calibration. All pyranometers and pyrheliom-
periods with small variations in the solar signal are eters removed from the stations during the annual ex-
used in the BORCAL analysis, thus minimizing the changes are sent to NREL for BORCAL calibration.
contribution of the diffuse component, which has the The returning Quantum sensors are sent to the factory
highest uncertainty. Although measurements are taken for calibration before they are used again.
throughout the day, only those made between 45° and Results of the FIRE II experiment's Infrared Ra-
55° solar zenith angles are used to fix the reported sen- diometer Intercomparison at Coffeyville, Kansas
sitivity factors. Those times are used because the co-
2
(DeLuisi et al. 1992), suggest that the manufacturer-
sine errors of pyranometers are typically near zero in supplied calibrations for pyrgeometers are acceptable
that range of solar positions. Reported uncertainties to within ±10 W nr . For SURFRAD, a novel technique
2

from the BORCAL procedure generally range from was developed at ARL to transfer the mean calibra-
±2% to ±5% for pyranometers and ±2% to ±3% for tion of the reference pyrgeometers to the monitoring
pyrheliometers. pyrgeometers. Test and reference instruments are op-
There is yet no world reference blackbody calibra- erated together outdoors for a week to 10 days. Periods
tion device for pyrgeometers. A first step in isolating of stable signals common to both sets of instruments
a device or technique that could be used as a world are objectively isolated for daytime and nighttime
reference was the BSRN Pyrgeometer Round Robin periods. Calibration equations, reported in Albrecht
Calibration experiment (Philipona et al. 1998) where and Cox (1977), are written for each of those two pe-
11 blackbody infrared calibration devices were blindly riods using mean signals from the test instruments, and
tested by calibrating the same five pyrgeometers. Six setting both equations equal to the mean irradiance of
of the 11 showed little difference in their calibrations the reference instruments. The two equations are then
of the five instruments. One of those six, that at the solved for the two constants in the Albrecht and Cox
WRC, is used to calibrate the three SURFRAD refer- (1977) formulation. This calibration transfer technique
ence pyrgeometers. The stated accuracy of their cali- appears to improve the test instruments' absolute cali-
brations is ±4 W m . 2 brations such that typically > 99% of the test data are
The three reference Quantum instruments are sent within ±5 W n r of the reference values. Adding the
2

annually to the factory where they were manufactured reported uncertainty of the reference instruments' cali-
to be calibrated with a standard light source that has brations yields an uncertainty of ±9 W n r for the
2

been referenced against a lamp certified by the Na- monitoring pyrgeometers.

The UVB reference instruments are used to cali-
brate the field units before they are deployed. To do
Both sensitivity factors and calibration factors are numbers that this, a test instrument is run alongside the three refer-
relate electrical signals from the radiometers to physical units. The ence instruments for at least a week. A scale factor,
2

semantic difference is that calibration factors are multipliers and defined as the ratio of the test instrument's daily inte-
sensitivity factors are divisors. gral of UVB irradiance to the mean daily integrated

Bulletin of the American Meteorological Society 2339

irradiance of the three references, is computed for each the location and time that the scale factors are applied
day that the instruments are operated together. The operationally to SURFRAD data, the references' cali-
representative scale factor that transfers the references' bration values valid for 300 Dobson units are used.
mean calibration to the field instrument is simply the Therefore, the UVB irradiance reported in the dissemi-
mean of the daily scale factors. The variation in the nated SURFRAD data are valid for an ozone level of
daily scale factors (used to compute the mean) is usu- 300 Dobson units. However, a family of curves is
ally less than ±1%. Adding to this the ±4% error of available to users of the data to scale the reported UVB
the standards' calibration yields ±5% for the monitor- irradiance to a value appropriate to the actual total
ing UVB instruments. ozone amount at the time of the measurement (see
A problem with UVB measurements is that the Fig. 10 of Lantz et al. 1999).
level of erythematogenic UV irradiance at the surface Data quality associated with the instruments used
is dependent on the amount of total ozone present at in SURFRAD is summarized in Table 3. Information
the time of the measurement (Lantz et al. 1999). For about the instruments including their calibration his-
times when the UVB standards are calibrated, the to- tories, status, etc., are maintained in a database. Also,
tal ozone amounts are known from local measure- a history of instrument suites deployed at SURFRAD
ments. A radiative transfer model is used to scale the stations and their associated calibration values for all
standards' erythemal calibration factors to varying monitoring epochs are available in the SURFRAD sec-
ozone amounts. Because total ozone is not known for tion of the World Wide Web site at www.srrb.noaa.gov.

TABLE 3. Data quality of SURFRAD monitoring instruments.

Instrument Manufacturer Model Measurements Accuracy

Pyranometer Spectrolab SR-75 global, diffuse, ±2% to ±5%

and Eppley (Spectrolab) upwelling solar
PSP (Eppley) irradiance
Pyrgeometer Eppley PIR downwelling and ±9 Win- 2

upwelling infrared
irradiance
Pyrheliometer Eppley NIP direct solar irradiance ±2% to ±3%
UVB radiometer Yankee Environmental UVB-1 global ultraviolet B ±5%
Systems erythemal irradiance
Quantum sensor LI-COR Quantum global ±5%
photosynthetically
active irradiance
Temperature Vaisala CS500 air temperature ±0.5°C
and relative
humidity probe relative humidity ±2%-3% below 90%
RH, and ±6% above
90% RH
Wind monitor R.M.Young 05103 wind speed ±2% for wind
and direction speed
±5° for wind
direction
Barometer Vaisala PTB101B station pressure ±4 mb (-20°-0°C)
±2 mb (0°-40°C)

2340 Vol. 81, No. 10, October 2000

e. Onsite maintenance 6. Data
Trained technicians at the stations provide routine
maintenance on at least a weekly basis. Because the a. Sampling and recording
stations are automated, routine maintenance involves The sampling rate for all instruments, except the
only cleaning the domes and diffuser disks, and a cur- MFRSR, is 1 s with an averaging period of 3 min.
sory inspection of the station. Each station also has Standard deviations of the 1-s samples are also re-
a local technical contact who is called to address ma- corded. Only the basic signals from the instruments
jor problems. The annual visits by ARL personnel to are logged, that is, voltages, resistances, etc. These
exchange instruments also serve as an opportunity data are downloaded daily from each station and saved
to conduct maintenance, implement improvements, in that form. Conversion to physical units and data
train technicians, and discuss issues with the station quality control are done in postprocessing. For UVB,
hosts. erythematogenic-weighted irradiance is computed.
Conveniently, this is the most accurate measurement
f Data recovery efficiency from the UVB instrument used because its spectral
A myriad of problems can affect an operational response mimics that of human skin to ultraviolet ra-
measurement network. Among the most common are diation. The MFRSR has its own logger that spot
equipment failure, communication problems, and samples at 15-s intervals and records 2-min averages.
weather. Although some events—for example, light-
ning strikes—cannot be avoided, quality assurance in b. Daily processing and file structure
the design and operation of the network has reduced An infrastructure for data collection and process-
the probability of failure. For SURFRAD, the use of ing has been developed to ensure a timely release of a
mutually compatible and proven equipment, care ap- quality controlled product, usually within a day of real
plied in the design of the network, and other preven- time. Every day, just after 0000 UTC, each station is
tative measures have resulted in a high good-data polled by a computer to retrieve the previous day's
recovery efficiency. Over the first two years of opera- data. If a call fails, retries are regularly scheduled un-
tion, recovery rates for acceptable radiometer data at til the data are successfully retrieved. Nine hours af-
Bondville, Fort Peck, Goodwin Creek, and Table ter the initial call, another automated program converts
Mountain were 95%, 91%, 97%, and 98%, respec- the raw data to physical units, subjects the data to
tively. In 1999 and 1998 the recovery rates for most simple physical-limit data quality control, and pro-
variables improved to between 98% and 99%. The duces diagnostic plots.
exception was for the instruments mounted on the Processed daily files for each station contain data
solar trackers at Bondville and Fort Peck, for which from 0000 to 2357 UTC. All of the data files are writ-
data collection rates fell to between 76% and 93%. ten as text and have the same format, that is, two header
This slippage was associated with aging solar track- records followed by 24 h of 3-min data. Each line rep-
ing systems. To alleviate this problem, new solar resents a 3-min period, and contains the date and time
trackers with a more robust design were installed in in various forms, the solar zenith angle for the mid-
late 1999 and early 2000. The data collection system point of the 3-min averaging period, followed by the
at all stations was also improved during that year. measured and computed variables. The time label of
More efficient and simpler data logging assemblies each line represents the ending time of the 3-min pe-
with greater data storage capacity and new commu- riod. Following each individual data entry is an inte-
nication equipment were installed. Prior to this, less ger quality control parameter; if it is greater than 0,
expensive third-party modems were used that had rela- then the preceding value is questionable.
tively short field lifetimes and lower connection The final audit given to the processed data is an
success rates. The new data loggers can store over "eye check" using an interactive graphical display, as
10 days of data before the oldest data are overwritten, recommended by BSRN. There, questionable values
thus providing a nine-day buffer to address any prob- that escaped the automated data quality control may
lems that may interfere with the normal schedule. be flagged by the network manager. These modifica-
Since the new equipment was installed in 1999, the tions are recorded both in a digitalfileand on hard copy.
data recovery rate has increased to over 99% for nearly Upon acceptance by the manager, processed daily data
all parameters. files for each station are released to anonymous ftp
directories: ftp://ftp.srrb.noaa.gov/pub/data/surfrad/.

Bulletin of the American Meteorological Society 2339

 The SURFRAD section of the SRRB Web page ing assigned bad QC flags. The station will be brought
(http://www.srrb.noaa.gov) is configured to facilitate back to normal as soon as possible.
free access to the basic data and includes convenient
interactive plotting and data downloading features. d. Sample data
Access to anonymous ftp directories is also available Figure 6 shows various 24-h time series from
from the main SRRB Web page. A "README" text Goodwin Creek for 1 December 1999. The upper-left
document that may be downloaded contains a descrip- panel (a) displays the four solar measurements: glo-
tion of the file structure and sample software to read bal, upwelling, direct, and diffuse. Measuring the di-
the data. SURFRAD data are also routinely transmit- rect and diffuse components of downwelling solar
ted to the National Climate Data Center for archival. radiation is useful for research, and the redundancy
provides a quality check on the downwelling global
c. Reprocessing capabilities and error reporting solar measurement. The direct solar component exhib-
Efficient methods for reprocessing SURFRAD its the largest values because the pyrheliometer sen-
data have been developed. Saving the station data in sor is always normal to the sun's beam. It is reported
their most basic form makes this possible. To avoid in the data file as it is measured; that is, it is not cor-
discrepancies with the original processed data, the re- rected to a horizontal surface. Because the global and
processing software mimics that used in the daily pro- diffuse solar measurements are made with a horizon-
cessing. It accesses the same programs and calibration tal sensor, before summing the direct and diffuse, the
files as those used operationally, thus ensuring that the direct component must first be corrected to that por-
irradiances computed for the reprocessed files are the tion of the sun's beam that would be measured on a
same as the originals. After the reprocessed files are horizontal surface. This is done by multiplying the
produced, modifications that were introduced by the direct normal measurement by the cosine of the solar
network manager during the daily "eye checks" are zenith angle, which is reported in the data file for each
reapplied by software that accesses the file of histori- 3-min period.
cal modifications. At this point, the new data files are These solar data indicate that during the morning
exactly the same as those that had previously been of 1 December 1999, Goodwin Creek was relatively
released, except of course for inclusion of the changes free of clouds, as is evident by the large direct solar
that necessitated the reprocessing. A list of algorithm values that rise to nearly 1000 W n r within a few
2

improvements and problems that have necessitated re- hours after sunrise. The very low diffuse solar values
processing of network data are given under the head- during that period are also indicative of cloud-free
ing "Problems" on the SURFRAD Web page. An skies. In the afternoon, sharp decreases in the direct
example of an actual notification is given below: component indicate that clouds periodically obscure
the sun's beam. Note that when the direct component
Station(s): Bondville decreases, the diffuse component increases proportion-
Parameter(s): upwelling PSP, UVB, tem- ally to account for the increased scattering by the
perature, RH, and station clouds.
pressure Figure 6b shows the two components of thermal
Date(s): 4 June-22 July 1999 infrared irradiance throughout the day. Upwelling in-
Description of problem: 11 June 1999 frared irradiance should always be greater than the
downwelling because the pyrgeometer at the top of the
All data after the lightning strike on 4 June-7 June tower faces the surface that is warmed by solar radia-
were lost. By shifting connections on the data logger tion during the day. Consequently, the upwelling in-
from blown channels to good ones, we were able to frared radiation trace should resemble the diurnal
record data from all of the instruments. At that point variation of the surface skin temperature. Under clear
it was determined that the pyranometer on the tower or mostly clear skies, the downwelling infrared irra-
(measuring upwelling solar radiation) was damaged, diance is relatively flat because the pyrgeometer on the
and the barometer and temperature and relative humid- main platform views the cold target of space. When
ity probe may have been destroyed. The UVB radiom- clouds intervene, the two infrared signals converge
eter is now not heating properly, however the data are because the cloud base is a warmer target than space,
being recorded. Since the instrument may not be at a causing the downwelling infrared signal to increase. An
stable temperature, the UVB irradiance values are be- example of this is seen in the first few hours of the day.

2340 Vol. 81, No. 10, October 2000

 Because of the paucity of its signal,
the UVB irradiance in Fig. 6c is plotted
in units of mW nr . Note that UVB is not
2

as responsive to clouds as are the solar

and infrared measurements. This is be-
cause, by virtue of its very short wave-
length, and thus its predisposition to
Rayleigh scattering, a large part of the
UVB signal is diffuse.
Figure 6d shows the net radiation
computed from the four surface radiation
budget measurements. These parameters
are also included in the data files. Net ra-
diation is defined as downwelling minus
upwelling irradiance. Net solar should
always be zero at night and positive dur-
ing the day. In reality the pyranometers
usually show a small negative signal at
night, which is caused by the domes and
thermopile cooling to space. This erro-
neous signal is not removed in the
SURFRAD data files. Net infrared FIG. 6. Sample data from the Goodwin Creek SURFRAD station for 1 Dec
should always be zero or negative be- 1999. Included are (a) the four solar measurements, (b) upwelling and downwelling
cause the upwelling component is al- infrared irradiance, (c) erythematogenic UVB irradiance, and (d) net radiation.
ways greater than the downwelling All measurements are in W nr , except the UVB, which is in mW nr .
2 2

component, except for some very un-

usual circumstances. Total net radiation is defined as arrangements were made with their operators for ARL
the sum of the net solar and net infrared, therefore it to access the raw data. Absolute calibrations of the
should be zero or negative at night, and it is usually MFRSR channels are not necessary because aerosol
positive during the day. optical depth is a relative measurement obtained from
the slopes of the output from the various spectral chan-
7. Value-added
features
TABLE 4. Value-added features at S U R F R A D stations.
An ongoing priority is to
make SURFRAD data more fa- Parameter(s) Instrument/calculation Availability Data accessibility
vorable for research by measur- Aerosol optical MFRSR All stations Request
ing or computing parameters depth
that affect the transmission of ra-
diation or reflect the response of Hemispheric Total Sky Imager All stations WWW ftp, request
the local environment to the sur- cloud fraction
face radiation budget. These Rawinsonde-type Interpolated, All stations WWW, request
value-added features are de- profiles except for DRA
scribed below and are summa-
rized in Table 4. Latent, sensible, Sonic anemometer, FPK, BON Request
Three stations have an and C0 2 Infrared hygrometer
MFRSR for aerosol optical depth atmospheric fluxes
measurements. For stations that Soil temperature Thermocouples FPK, BON Request
do not have an MFRSR, nearby profile
instruments were identified and

Bulletin of the American Meteorological Society 2339

nels plotted as a function of air mass. As of this It is desirable to have profiles of temperature and
writing, an acceptable method for analyzing moisture at SURFRAD stations to initiate radiative
MFRSR data has not been developed, and the op- transfer models. However, because of strict selection
tical depth analyses that have been done are con- criteria, most SURFRAD stations are located far from
sidered provisional. rawinsonde sites. The exception is Desert Rock, which
To address the issue of cloud effects, a total sky is at an operational sounding station. Software has
imaging (TSI) system was developed at ARL and is been developed to interpolate soundings from the
now being commercially produced. During 1999, NWS operational rawinsonde network to SURFRAD
TSIs were installed at all SURFRAD stations except station locations. Sounding parameters, including tem-
Fort Peck. That for Fort Peck will be installed in 2000. perature, dewpoint temperature, geopotential height,
Each minute the TSI samples the hemispheric sky and wind, are first interpolated vertically to common
and processes the image for cloud fraction and rudi- pressure levels, and then horizontally at those levels
mentary cloud properties. This information will aid to the locations of the SURFRAD stations. Balloon
in the interpretation of the surface radiation measure- drift is accounted for and special steps have been taken
ments during the daytime, and also offer ground truth to stabilize the interpolations near the surface. The
for estimates of cloud fraction made from satellite data method used for the horizontal interpolation is the
and numerical models. The SURFRAD TSIs are analytic approximation scheme described in Caracena
now in a break-in period where the proper settings (1987). Figure 7 shows a comparison of an interpo-
for each particular location need to be determined. lated sounding for Bondville and the nearest opera-
When these have been determined, the most recent tional sounding. Separate files for 0000 and 1200 UTC
2 weeks of hourly images and all daily 1-min cloud of each day, containing interpolated soundings for five
fraction files will be made available on the SRRB SURFRAD stations and the actual sounding for Desert
Web site. Rock, are available from the "Sound-
ings" link on the SURFRAD Web page
for every day from 1 January 1995 to the
present. Text files that explain the format
of the sounding files are also accessible.
In support of GCIP activities, a novel
state-of-the-art energy/carbon flux sys-
tem (Baldocchi and Meyers 1998) was
installed at the Fort Peck SURFRAD sta-
tion in the fall of 1999. This system also
measures soil temperature profiles to a
depth of 60 cm. Similar instrumentation
was installed near the Bondville station
in 1996, and will be installed at Goodwin
Creek in 2000. A goal is to eventually
add this capability at the other stations.
With this, SURFRAD will become a
valuable validation source for the surface
physics of NCEP's operational and ex-
perimental weather forecast models.
These platforms are considered experi-
mental and presently the data from them
are not routinely available.

FIG. 7. Soundings of temperature (°C, solid), dewpoint temperature (°C, g Collaborotion

dashed), and winds (full barb = 5 m s ), interpolated to the location of the Bondville
_1

SURFRAD station (thin lines), and the actual NWS sounding for Lincoln, IL (thick
lines), 100 km to the west of Bondville, for 0000 UTC 14 Apr 2000. The winds From its inception, SURFRAD was
from the Lincoln sounding are plotted with thick lines. a collaborative effort among universities
2340 Vol. 81, No. 10, October 2000
and government agencies, who all had input in the site
selection process, and right of refusal. Among the par-
ticipants were Princeton University, the Office of
Hydrology, the GEWEX Program Office, NESDIS,
and the NASA Langley Research Center. SURFRAD
has since developed several linkages with the USDA.
A USDA UV monitoring station is located at the Fort
Peck SURFRAD site. Through a special agreement
ARL accesses MFRSR data from USDA instruments
at Fort Peck and Bondville. The Goodwin Creek sta-
tion is sponsored by the USDA National Sedimenta-
tion Laboratory and is collocated with one of their
installations in the Goodwin Creek Experimental FIG. 8. Percentage difference between TOMS-based estimates
Watershed, where research quality measurements of of erythematogenic UVB flux at the surface and measurements
rainfall, soil temperature, and soil moisture are regu- at Goodwin Creek for totally clear-sky periods. The measured
larly made. UVB flux is averaged over 1-h periods centered on the satellite
SURFRAD data are now being used for a variety overpass time.
of validation activities. In 1997, the network was cho-
sen as a primary U.S. validation source for NASA's
EOS satellites. In fulfillment of this obligation, NOAA laborative effort among several interested groups who
provides reprocessed SURFRAD data, cloud-cover had a stake in the use of high quality surface radiation
information, interpolated soundings, and quantitative budget data, including OGP, NASA, NOAA's Office
estimates of cloud effects on the surface radiation bud- of Hydrology, university professors, and the GEWEX
get. SURFRAD data are also being used to validate Program Office. Consisting of six stations, it is the first
operational products such as weather forecasting mod- such network to continually measure the surface ra-
els, the NWS UV index, and NESDIS estimates of diation budget in various regions of the United States.
surface solar irradiance. SURFRAD nicely complements DOE's mesoscale
SURFRAD data have been used to check NASA's network of about 20 surface radiation budget stations
Total Ozone Mapping Satellite (TOMS) estimates of in Oklahoma and Kansas. Measurements include in-
erythematogenic UVB surface flux. Figure 8 shows dependent upwelling and downwelling components of
how SURFRAD data were useful in revealing that solar and thermal infrared irradiance, broadband
TOMS-based UVB estimates may be systematically erythematogenic UVB irradiance, photosynthetically
high. Data plotted in Fig. 8 are for times when the sat- active radiation, spectral solar, and basic meteorologi-
ellite passed directly over the Goodwin Creek station cal quantities. A theme of quality assurance guided the
during hemispheric cloud-free periods. The plotted design of the stations and the data processing system.
UVB data have been corrected for total ozone and The network is backed by a robust system of data re-
averaged over 1-h periods centered on the overpass trieval and distribution, redundant measurements, daily
time. Differences shown do not reflect a possible bias manual data checks, a rigorous instrument calibration
of an individual instrument because three different schedule, and efficient data processing and reprocess-
UVB radiometers were deployed during the period ing capabilities. Network instruments are cycled an-
depicted. nually and regularly calibrated to maintain accuracy
and precision. Data are released on a quasi-daily ba-
sis. Free access to the processed data is expedited by
9. Summary an ftp site and Web page from which daily data files
from individual stations may be downloaded. An in-
The SURFRAD network commenced in the teractive Web page also offers a feature where daily
middle 1990s. It arose from an expressed need by hy- time series of any parameter or group of parameters
drologists, modelers, the satellite community, and may be plotted. Likewise, there are Web pages that
GCIP scientists for a high quality, long-term surface allow the downloading and plotting of the interpolated
radiation budget network over the United States, which soundings, and the plotting of annual time series of
had been lacking. Planning for SURFRAD was a col- monthly means of measured or computed parameters.

Bulletin of the American Meteorological Society 2339

 For research, SURFRAD includes ancillary mea- to Tom Stoffel of NREL whose metrology group provides accu-
surements and calculations of parameters that affect rate and timely calibrations for all of the network's solar instru-
the transmission of radiation, such as cloud fraction, ments, giving its data the attributes of credibility, precision, and
quality assurance. Comments by James Angell, Detlef Matt, Chris
humidity, aerosol optical depth, and interpolated Cornwall, Ehrhard Raschke, Amy Stevermer, and two anonymous
radiosonde-type profiles. Two stations now have at- reviewers improved the quality of the manuscript.
mospheric and soil heat flux systems, making them
full surface energy budget sites; another is planned for
installation in 2000. A goal of the program is to add References
this capability to the remaining stations. With that,
SURFRAD will become more valuable for research, Albrecht, B., and S. K. Cox, 1977: Procedures for improving
and as a validation source for surface physics pack- pyrgeometer performance. J. Appl. Meteor., 16, 188-197.
ages of the newest numerical models. Baldocchi, D., and T. Meyers, 1998: On using ecophysiological,
The quality of SURFRAD data has been recog- micrometeorological and biogeochemical theory to evaluate
nized through its acceptance by large and reputable carbon dioxide, water vapor, and trace gas fluxes over veg-
scientific organizations. Regular users of SURFRAD etation: A perspective. Agric. Forest Meteor., 90, 1-25.
Caracena, F., 1987: Analytic approximation of discrete field
data include NASA, NESDIS, the National Centers for samples with weighted sums and the gridless computation of
Environmental Prediction, the National Weather Ser- field derivatives. J. Atmos. Sci., 44, 3753-3768.
vice UV index validation project, universities, class- Committee on Earth and Environmental Sciences, 1990: Our
room instructors, and many international scientists. Changing Planet: The FY 1991 Research Plan of the U.S.
With continued care in the handling of the instruments Global Change Research Program. U.S. Geological Survey,
and data, SURFRAD will become the U.S. baseline 250 pp.
Committee on Global Climate Research, 1998: Global Environ-
surface radiation reference for satellite systems and mental Change: Research Pathways for the Next Decade,
models that will eventually provide the datasets Overview. National Academic Press, 82 pp.
needed for breakthrough climate and weather research. Darnell, W. L., F. Staylor, S. K. Gupta, N. Ritchey, and A. C.
The question of how many stations are needed to sat- Wilber, 1992: Seasonal variations of surface radiation budget
isfy the research needs of atmospheric scientists is derived from International Satellite Cloud Climatology Project
data. J. Geophys. Res., 97, 15 741-15 760.
open at this time. Suggestions from users of the data DeLuisi, J. J., K. Dehne, R. Vogt, T. Konzelmann, and A. Ohmura,
are welcomed and will help with the evaluation of 1992: First results of the Baseline Surface Radiation Network
the network's progress, usefulness, and the need for (BSRN) broadband infrared radiometer intercomparison at
expansion. FIRE II. Proc. Int. Radiation Symp. 92, Tallinn, Estonia,
A. Deepak Publishing, 559-564.
Acknowledgments. Establishment of the SURFRAD network Harrison, L., J. Michalsky, and J. Berndt, 1994: Automated
was made possible by funding from NOAA's Office of Global multifilter rotating shadow-band radiometer: An instrument for
Programs, and its advocates there: Bill Murray, Rex Fleming, Arnie optical depth and radiation measurements. Appl. Opt., 33,
Gruber, and Mike Johnson. The network's success, quality, and 5118-5125.
efficient operation are enabled through the cooperation, enthusi- Hicks, B. B., J. J. DeLuisi, and D. R. Matt, 1996: The NOAA In-
asm, and generosity of several other agencies and people. Chris tegrated Surface Irradiance Study (ISIS)—A new surface ra-
Cornwall, Gary Hodges, Dennis Wellman, Scott Sandberg, Tony diation monitoring network. Bull. Amer. Meteor. Soc., 77,
Vergamini, Mike Rubes, and Dave Theisen are credited with pro- 2857-2864.
viding soundness in the station design, hardware, software, and Jin, M., and R. E. Dickinson, 1999: Interpolation of surface ra-
data retrieval system. Chris Cornwall, Mike Rubes, and Paul diation temperature measured from polar orbiting satellites to
Kooros deserve much of the credit for the Web page development. a diurnal cycle. Part 1: Without clouds. J. Geophys. Res., 104,
The USDA/ARS National Sedimentation Laboratory in Oxford, 2105-2116.
Mississippi, the Department of Meteorology of The Pennsylva- Karl, T. P., P. Y. Groisman, R. W. Knight, and R. R. Heim Jr.,
nia State University, and the Special Operations and Research 1993: Recent variations of snow cover and snowfall in North
Division of NOAA's Air Resources Laboratory host stations and America and their relation to precipitation and temperature
charge no fees. They also graciously provide utilities and techni- variations. J. Climate, 6, 1327-1344.
cal support for the stations. The existence of the Bondville and Lantz, K. O., P. Disterhoft, J. J. DeLuisi, E. Early, A. Thompson,
Fort Peck stations is a direct result of professional cooperation D. Bigelow, and J. Slusser, 1999: Methodology for deriving
from the Illinois State Water Survey and Fort Peck Tribes. Without clear-sky erythemal calibration factors for UV broadband radi-
the support of NASA, the total sky imagers, and the adaptation of ometers of the U.S. central UV calibration facility. J. Atmos.
the sounding interpolation algorithm to SURFRAD would not Oceanic Technol., 16, 1736-1752.
have been possible. The inventory of wind monitors for McArthur, L. J. B., 1998: Baseline Surface Radiation Network
SURFRAD is kept in top working order by Detlef Matt of (BSRN) Operations Manual (Version 1.0). World Climate
ARL/ATDD. The SURFRAD program owes a wealth of gratitude Research Programme, WMO/TD-No. 879, 69 pp.

2340 Vol. 81, No. 10, October 2000

McKinlay, A. F., and B. L. Diffey, 1987: A reference action spec- Radiation Research Laboratory, National Renewable Energy
trum for ultraviolet induced erythema in human skin. Human Laboratory, 1617 Cole Blvd., Golden, CO 80401-3393.]
Exposure to Ultraviolet Radiation: Risks and Regulations, Stokes, G. M., and S. E. Schwartz, 1994: The Atmospheric Ra-
W. F. Passchier, and B. F. M. Bosnjakovic, Eds., Elsevier, diation Measurement (ARM) program: Programmatic back-
83-87. ground and design of the cloud and radiation testbed. Bull.
Philipona, R., and Coauthors, 1998: The baseline surface radiation Amer. Meteor. Soc., 75, 1201-1221.
network pyrgeometer round-robin calibration experiment. J. WMO, 1991: Workshop on the implementation of the Baseline
Atmos. Oceanic Technol., 15, 687-696. Surface Radiation Network, WMO/TD-No. 406, 21 pp.
NREL, 2000: Broadband Outdoor Radiometer Calibration, 2000-
01. NREL Metrology Report, in press. [Available from Solar

^4

With the development of meteorological science and the continual refinement of the
technologies used in its practical application, the need to produce a new edition of the
International Meteorological Vocabulary (IMV) became evident (the original edition was
published in 1966). This volume is made up of a multilingual list of over 3500 terms
arranged in English alphabetical order, accompanied by definitions in each of the
languages (English, French, Russian, and Spanish) and an index for each language. This
new edition has been augmented with numerous concepts relating to new meteorological
knowledge, techniques, and concerns. It should help to standardize the terminology used in
this field, facilitate communication between specialists speaking different languages, and
aid translators in their work.
WMO No. 182, 784 pp., softbound, color-coded index, $95 (including
postage and handling). Please send prepaid orders to: WMO
Publications Center, American Meteorological Society, 45 Beacon St.,
Boston, MA 02108-3693. (Orders from U.S. and Canada only.)

International
Meteorological
Vocabulary
Bulletin of the American Meteorological Society 2339
 »*%urnals O n l i n e n o w «

N o w w i t h r i € : w f u n c t i o n a l i t y !
HHHBBBHHBHH
Now the complete contents of the A M S journals
and Bulletin articles are available on the Web!

All users can view titles and abstracts of articles from the AMS journals (starting with the January
1997 issues), while subscribers will have the added ability to view the complete articles delivered
in two forms—as HTML, optimized for on-screen viewing, and as Adobe Acrobat PDF files, for
printing a precise "reprint" of the article as it appeared in the journal. The full text of Bulletin articles
will be available free of charge to all users of the site.
• Flexible full-text and fielded search capability
• Journal contents available online prior to the print journal
Journal of the Atmospheric Sciences Weather and Forecasting
Journal of Climate Journal of Applied Meteorology
Journal of Atmospheric and Oceanic Journal of Physical Oceanography
Technology Bulletin of the American Meteorological
Monthly Weather Review Society
Journal of Hydrometeorology

Access the AMS Journals Online from the AMS Web site:
http://www.ametsoc.org/AMS

Subscriptions are available in the form of institutional network site licenses or individual
subscriptions. See the AMS Web site for details, or write to the American Meteorological Society,
45 Beacon St., Boston, MA 02108.

2340 Vol. 81, No. 10, October 2000