The Spectral Radiance Experiment


Project Description and Sample Results

Robert G. Ellingson
Department of Meteorology
University of Maryland
College Park, MD 20742

Warren J. Wiscombe

NASA Goddard Space Flight Center
Greenbelt, MD 20771

August 1995

(Final Revision March 1996)


The fundamental climatic role of radiative processes has spurred the development of increasingly sophisticated models of radiative transfer in the Earth-atmosphere system. Since the basic physics of radiative transfer is rather well known, this was thought to be an exercise in refinement. Therefore, it came as a great surprise when large differences (30 to 70 W m-2) were found among longwave infrared fluxes predicted by over 30 radiation models for identical atmospheres during the InterComparison of Radiation Codes used in Climate Models (ICRCCM) exercise in the mid 1980s. No amount of further calculation could explain these and other intermodel differences; thus, it became clear that what was needed was a set of accurate atmospheric spectral radiation data measured simultaneously with the important radiative properties of the atmosphere like temperature and humidity.

To obtain this dataset, the ICRCCM participants charged the authors to develop an experimental field program. So, we developed a program concept for SPECTRE (SPECTral Radiance Experiment), organized a team of scientists with expertise in atmospheric field spectroscopy, remote sensing, and radiative transfer, and secured funding from the Department of Energy and NASA. The goal of SPECTRE was to establish a reference standard against which to compare models, and also to drastically reduce the uncertainties in humidity, aerosol, etc., which radiation modelers had invoked in the past to excuse disagreements with observations. In order to avoid the high cost and sampling problems associated with aircraft, SPECTRE was designed to be a surface-based program.

The field portion of SPECTRE took place Nov 13 to Dec 7, 1991, in Coffeyville, Kansas, in conjunction with the FIRE Cirrus II field program, and most of the data have been calibrated to a usable form and will soon appear on a CD-ROM. This paper provides an overview of the data obtained; it also outlines the plans to use this data to further advance the ICRCCM goal of testing the verisimilitude of radiation parameterizations used in climate models.

1. Introduction

This article describes a comprehensive field program designed specifically to test longwave radiation models. It focuses on a particular piece of climate model physics and pays careful attention to measuring all variables necessary to close the problem. It demonstrates a strategy through which the worlds of modeling and field observation can be brought much closer together.

Our goal has been to make this article accessible to the general meteorological community. An irreducible minimum of radiation jargon is used, which is defined in Appendix A; look there for unfamiliar terms. Acronyms which are not needed to comprehend SPECTRE are defined in Appendix B rather than in the text.

2. Intercomparison of Radiation Codes used in Climate Models (ICRCCM): Background for SPECTRE

"Since the transfer of solar and longwave radiation is the prime physical process that drives the circulation of the atmosphere and its temperature structure, it is natural that an evaluation of the modeling of physical processes important to climate begin with radiation." (Luther, 1984)

This was the rationale for the InterComparison of Radiation Codes used in Climate Models (ICRCCM), sponsored by the U.S. Department of Energy (DOE), the World Meteorological Organization, and the International Radiation Commission. DOE's sponsorship of ICRCCM stemmed from its role as lead agency for investigating the CO2-climate problem (Riches, 1983), now more generally known as the greenhouse warming problem because other trace gases like ozone, CH4, N2O and halocarbons are also involved (Wang et al., 1992). Greenhouse warming is entirely forced by a radiative perturbation of a few W m-2 (Watts per square meter) or less, yet neither field measurements nor radiation model intercomparisons had ever achieved anywhere near this level of accuracy, at least not in the atmospheric and climate community.

Before ICRCCM, the assumption had been that radiative transfer physics and gaseous and cloud absorption and scattering were well understood. Indeed, many of the individual pieces of radiative transfer theory had been tested in the laboratory (even the Mie scattering by an individual water drop) and not found wanting. Thus, so it was thought, the models could not possibly disagree much. There was even a strong tendency to trust radiation models more than observations, partly because the usual observational data were inaccurate, poorly calibrated, and lacked spectral resolution. This complacent faith in models was shattered by the actual results of ICRCCM. ICRCCM began in 1984 with longwave clear-sky cases, because the longwave spectrum (4-50 mm wavelength, or 200-2500 cm-1 wavenumber) is where greenhouse forcing takes place, and because clear-sky cases are the simplest from a radiative transfer point of view. The participation was unprecedented: about 40 sets of results were contributed, representing almost all the world's major radiation modeling groups. These results revealed maximum inter-model disagreements ranging from 30 to 70 W m-2 (Luther, et al., 1988). These disagreements are substantial when compared both with typical longwave fluxes (200-400 W m-2) and with the 4 W m-2 change in flux at the tropopause which forces the doubled-CO2 problem (Ramanathan, 1979). The disagreements are worse for pure H2O and pure CO2 atmospheres, indicating that the smaller disagreements found in the all-gas cases (10-40 rather than 30-70 W m-2) are partly accidental, due to compensating errors.

Some of the most outlying model results were a result of poor treatment of the known physics and improper implementation of computational algorithms. Even simple things like numerical quadratures were sometimes poorly done. However, the participants had four years to fix their models, and withdraw or resubmit results. Probably physical and algorithmic deficiencies remain in some models, but the large disagreements quoted above, which remained even after this long weeding process, represent a good snapshot of the situation in 1988.

The disagreements found by ICRCCM are unacceptably large for climate studies. Two major components of the World Climate Research Program, TOGA (WCP-92, 1984) and ERBE (Barkstrom et al., 1989), have insisted upon radiative flux accuracies of 10 W m-2 or better at the bottom and top of the atmosphere, respectively. Based on the ICRCCM results, many radiation models could not meet such an accuracy goal in 1988 - and, in many cases, still cannot meet it because of the slow pace at which they have (if at all) corrected their shortcomings and improved their agreement with observations.

Line-by-line (LBL) modelers were especially active in trying to resolve their disagreements in the 1984-88 period (LBL models are the most faithful to radiative transfer theory because they make no spectral averaging approximations). Clough et al. (1992) describe one such exercise, which showed that the main points of disagreement were: (a) where to cut off absorption lines (the theoretical line shape functions extend to infinite wavenumber); and (b) what water vapor continuum formulation to use (see Tipping and Ma, 1995, for a review of this subject). When the LBL modelers used the same line cutoffs and the same continua, their results agreed to 1% or better (or 1-2 W m-2), indicating that the rest of their physics, their computational algorithms, and their absorption line databases were not an important source of disagreement. Even when using their own preferred line cutoffs and continua, the LBL models never differed by more than about 8 W m-2 (Clough et al., 1992).

With this tight agreement among LBL models as a beacon, many of the more approximate models made improvements and resubmitted results which agreed better with the LBL models (Ellingson et al., 1991). Nevertheless, the ICRCCM reluctantly concluded that no radiation model, not even an LBL model, could furnish an absolute standard by which to judge other models (Luther et al., 1988):

"Uncertainties in the physics of line wings and in the proper treatment of the continuum make it impossible for line-by-line models to provide an absolute reference ..."

The ICRCCM participants then carefully considered whether existing data sets could resolve the disagreements. They concluded in 1984 (Luther, 1984), and again at regular intervals right up through their 1988 meeting in Paris, that previous laboratory, satellite, aircraft and surface measurements were not suitable for a variety of reasons, including:

In sum, the ICRCCM participants had reached an impasse:

  1. their models disagreed at a level of major significance for climate, and
  2. there was no reference standard, either among models or in field data, that could resolve the disagreements to the accuracy needed.

Recognizing these facts, the ICRCCM group recommended (following their 1988 Paris workshop), and the International Radiation Commission and the World Climate Research Program endorsed, a second phase of ICRCCM to test high spectral resolution radiation models through comparison with observations taken in a dedicated field program.


The authors were formally charged by the ICRCCM participants to shape such a field program. The key ICRCCM recommendation was to obtain "accurate spectral radiances rather than [spectrally] integrated fluxes as a basis for evaluating model performance" (Luther, 1984). We began by carefully reviewing past field measurements of atmospheric longwave emission spectra to see what lessons had been learned. According to one of the pioneers (Bell, et al., 1960) :

" ... the general features of the sky's infrared radiance are qualitatively understood... but a quantitative prediction of the radiance of the sky would demand an intimate knowledge of the distribution of the absorbing and scattering species along the atmospheric path of observation, together with detailed information about the temperature and pressure distribution and many other parameters."

This concern for proper quantitative characterization of the atmosphere became a cornerstone of SPECTRE. Many other concerns raised in the pioneering studies were also considered very carefully in the design of SPECTRE, including: radiometric calibration errors, horizontal inhomogeneity, the assumption of statistical stationarity of the atmosphere during a spectral scan, emission by optical elements and gases in the optical path, and fogging of optical elements.

Next we reviewed the currently available capabilities in field spectrometry and remote profiling of temperature, humidity and aerosol from the surface, and concluded that the technology needed for SPECTRE not only existed but had successfully undergone several field trials.

At this point, the key features needed for SPECTRE to incisively test radiation models became clear:

(1) simultaneous and instantaneous profiles of temperature, humidity, aerosol and cloud

[Reason: Radiation responds to the instantaneous vertical state of the atmosphere, while radiosondes and in situ aircraft sensors measure time-lagged profiles along twisted curves which are neither temporally nor spatially coincident with the radiation observations.]

(2) spectral detail in the form of continuous spectra, not broadband measurements nor disjointed spectral bands

[Reason: This is the only way to assess radiation model differences, since basic radiation physics indicates that such differences are most likely associated with specific absorbing gases and thus confined to specific absorption bands, not spread across the entire spectrum.]

(3) zenith radiance (intensity) rather than flux measurements

[Reasons: Radiance, not flux, is the quantity actually predicted by radiative transfer theory; comparing radiance measurements to theory is thus much more natural and informative. Also, clear-sky zenith radiance contains only photons from directly overhead where the atmospheric state parameters are also being measured. We are still far from a radar-like capability to volumetrically measure all the atmospheric state parameters needed to make a 3D radiative transfer calculation.]

(4) redundant measurements of radiance - at least three

[Reason: Disagreements among two instruments are difficult to resolve; and three or more instruments give a better assessment of random error bars.]

(5) frequent and careful radiometric calibration in the field against known standards

[Reason: 1% changes in radiation energy are climatically significant. Observations accurate to 1% would deny models much of their accustomed tuning latitude and discriminate among theories that agree at the 10% level. The nominal goal was an accuracy of 1% of the maximum (over all spectral intervals) radiance; this referred to both relative (from one spectral interval to another) and absolute accuracy.]

For the range of conditions observed during SPECTRE, the maximum downward radiance at the surface was about 150 mW m-2 sr-1 (cm-1)-1. The accuracy goal of SPECTRE was 1% or this, or about 1 mW m-2 sr-1 (cm-1)-1; hereinafter, we will use the notation RU for this radiance unit.

SPECTRE always remained focused on its main goal: to sharply reduce the uncertainties in the specification of atmospheric state by which radiation models had, in the past, eluded incisive comparison with measurement. Based on features (1)-(5) above, the two primary strategies for achieving this goal were: first, drastically reduce reliance on radiosondes and use, to the maximum extent possible, vertical remote profiling technology for humidity, temperature, ozone, aerosol and cloud; and second, use several sophisticated research-class spectrometers with cryogenic cooling to make accurate and spectrally detailed zenith longwave radiance observations with high spectral resolution. Looking upward was especially important because:

SPECTRE emphasized relatively clear atmospheres because ICRCCM showed large disagreements between models already for the clear cases. Also, clear cases are the background against which all the more complicated cloud and aerosol cases are played out. Therefore it is important to gain a superb quantitative understanding of clear cases first.

We made several attempts to develop SPECTRE as a stand-alone program. This failed mainly because the prevailing philosophy gravitated toward large multidisciplinary field programs rather than small focused ones. Nevertheless, the funding agencies recognized the value of SPECTRE and ultimately recommended that it be attached to the large multidisciplinary cloud program FIRE (Cox, et al., 1987). In particular, it seemed appropriate and economical to attach SPECTRE to FIRE Cirrus II (see FSET, 1989), which was about to go into the field to Coffeyville, Kansas for a month in Nov-Dec 1991. This allowed SPECTRE to take advantage of FIRE's infrastructure (which furnished e.g. the SPECTRE trailer and electric power) and of its remarkable concentration of surface and aircraft in situ and remote sensors, including lidar, millimeter radar, wind profilers, radiosondes, radiometers, and aircraft cloud physics instruments. (The instruments and measurements of FIRE are described in the FIRE II Special Issue of Journal of the Atmospheric Sciences (1 Dec 1995).) This formidable instrument array, supplemented by that of SPECTRE, promised to measure atmospheric profiles well enough to satisfy SPECTRE's goals. Thus, SPECTRE became an augmentation of FIRE Cirrus II.

Because the FIRE location and time were chosen expressly to take advantage of a high climatological incidence of cirrus, it seemed unlikely that SPECTRE would find an abundance of perfectly clear cases. Aerosols and/or cirrus clouds would probably be present most of the time. (Actually, SPECTRE enjoyed many almost perfectly clear days.) But aerosols are only perturbations to the pure clear case. Aerosol particles are so small that aerosol optical depths are at least a factor of four smaller in the longwave than in the solar spectrum (see Plate 2.1 in Lacis and Mishchenko, 1995). Also, much of the longwave spectrum, even in the transparent "windows", has a gaseous optical depth much larger than that of a typical aerosol. So, an aerosol that seems significant to the eye (for example the Pinatubo aerosol that produced spectacular sunsets during SPECTRE) generally has almost no longwave effect. And fortunately the scattering and nonspherical ice crystal issues which so complexity solar-spectrum cirrus radiation modeling are much less important in the longwave, where cirrus act nearly like grey bodies (blackbodies with emissivity < 1). Furthermore, the FIRE lidars, millimeter radars, and aircraft provided much correlative information (e.g. Uttal et al., 1995) which should allow aerosols and cirrus to be prescribed accurately enough in models to (nearly) reach SPECTRE's high accuracy goals for longwave modeling.

The SPECTRE proposals subsequently served as the blueprint for the Dept. of Energy's Atmospheric Radiation Measurements (ARM) program (Stokes and Schwartz, 1994) which was inaugurated several months after plans for SPECTRE were finalized. ARM was launched so rapidly that it was able to provide assistance to several instrumentation groups participating in FIRE. ARM's goal is to advance the parameterization of clouds and radiation in climate models through a decade-long program of field observations and modeling, whereas SPECTRE was a short-term program directed at selected case studies. Nevertheless, the instruments deployed during SPECTRE served as prototypes for the ARM Southern Great Plains Central Facility in Oklahoma, and the data collected during SPECTRE will serve many ARM scientists as well as ICRCCM participants.

4. The SPECTRE team

SPECTRE was a tightly coordinated team effort funded by two umbrella proposals (one to DOE, one to NASA) assembled by the authors. Highly experienced subteams carried out the three main functions: spectrometer, remote, and in situ measurements. The members of these subteams are listed in Table 1.

a. Spectrometer Measurements

The spectrometers that were chosen for SPECTRE are listed in Table 2. These instruments are characterized by:

a large wavelength range (3-18 µm);
high spectral resolution (1 cm-1 or better);
cryogenic cooling of the detectors; and
routine blackbody calibration in the field.

The 3-18 µm wavelength range covers almost all the trace gas absorption bands important for greenhouse warming and encompasses around 60-70% of the longwave energy in midlatitude regions. Below 3 µm, there is virtually no longwave radiation. 18 µm is the cutoff of standard spectrometer detectors; anyway, at the surface, beyond about 21 µm, the midlatitude atmosphere is nearly a blackbody radiating at the local air temperature and thus holds little interest for testing longwave models. The 1 cm-1 spectral resolution represented a compromise between competing factors, including achieving high signal to noise and rapid spectral scans (which requires lower resolution) and resolving absorption bands well enough to really test models (which requires higher resolution). SIRIS was the only participating spectrometer that could resolve individual spectral lines, which are typically 0.1 cm-1 wide at the surface; therefore it played a unique role in SPECTRE.  Cryogenic cooling was necessary to reduce thermal noise in the detectors and achieve the high accuracies SPECTRE aimed for.

SPECTRE's spectrometers and their associated teams had considerable field experience in high-altitude balloons and aircraft. Spectrometer descriptions and results have been presented in the literature and at relevant meetings (e.g., Kunde et al., 1987; Brasunas et al., 1988; Murcray et al., 1984; Murcray 1984; Revercomb et al., 1988, Smith et al. ,1993; Smith et al. ,1995; Collard et al., 1995). The instruments listed in Table 2 were not originally designed for operations at the surface. Therefore, the pre-field period was used to make modifications and/or build new equipment for use in the SPECTRE field phase.  It was originally planned for the Wisconsin team to bring a laboratory version of their aircraft HIS instrument. However, ARM made it possible for them to develop and bring a prototype of the AERI, a ruggedized HIS specifically designed for autonomous surface operation. (The mature AERI is now operating continuously at the ARM Oklahoma facility.)

The AERI prototype was built around the same basic BOMEM interferometer system as the Denver team's spectrometer. However, the reduction of the interferograms to calibrated spectral radiances is performed differently by the two groups. During each 10-min scan sequence, the AERI calibrates against two internal temperature controlled black bodies (ambient and hot, plus a periodic view of a liquid nitrogen bath). Performance evaluation has included characterization of AERI radiometric response, signal to noise ratio, blackbody temperature monitoring accuracy, and calibration accuracy, with the goal of providing the best possible calibrated radiance measurements.

Since SPECTRE's goal was 1%-of-full-scale accuracy, or 1 RU, and since this is a considerable challenge even for the excellent spectrometers of the present day, the spectrometer subteam considered it important to have a "universal" blackbody calibration source in the field that could be regularly used for all the spectrometers. The SIRIS team constructed such a source and made it available during SPECTRE. However, in practice the SIRIS team were the main users of this blackbody; it proved more difficult than expected to build a universal blackbody that meets the needs of a variety of spectrometers. The goal was laudable, and is being pursued with vigor in ARM, but SPECTRE fell short of it.

Spectral scans by the SIRIS, the slowest of the spectrometers, take nearly 10 minutes. So 10 minutes became the "official SPECTRE spectrometer time interval", and all spectrometer groups were asked to produce 10-min average spectra centered on identical times. However, faster lower-resolution spectrometers like AERI, which completes a spectral scan in 30 s, were able to obtain a meaningful variance as well as a mean over 10 minutes (and in fact this has become a routine ARM data product).

b. Remote Sensing Measurements

Previous radiation studies have had to rely upon radiosonde or aircraft measurements of the profiles of temperature, humidity, cloud and aerosol.  There are times when the atmosphere is sufficiently stationary and horizontally homogeneous that such measurements are relevant to radiation. However, radiosondes and in situ aircraft sensors measure spatially-skewed, time-lagged profiles which are usually only loosely related to the radiation observations, because longwave radiation is a result of the instantaneous state of the atmosphere in the limited field of view of the radiometer. Radiosondes take an hour or more to complete their ascent, and in the process usually drift far out of the effective field of view of any radiometer, even a flux radiometer. Radiosonde and aircraft profiles of humidity tend to be particularly problematical, because water vapor may have spatial and temporal variability comparable to that of patchy cloudiness, especially in the lower troposphere (see our Fig. 5; also Melfi and Whiteman, 1985; Melfi et al., 1989; Ferrare et al., 1995).

But advances in surface profiling technology in the past decade now allow the near-instantaneous measurement of the vertical profiles of all radiatively-important variables to altitudes of 5-10 km. The profiling systems used in SPECTRE are listed in Table 3 and described briefly in the table caption.

Since humidity and temperature are the main variables controlling downward longwave radiation at the surface, special emphasis was placed on the Raman lidar (humidity) and RASS (temperature) profiling systems. Both of these systems have their Achilles' heels: the Raman lidar was calibrated each day against a different radiosonde (whose humidity shortcomings are well known) and was designed to work only at night; the RASS systems proved unexpectedly sensitive to high near-surface winds (which blew the acoustic pulses away before the radar could detect them) and radio frequency interference from human sources. Yet both offered significant advantages over radiosondes in terms of rapid vertical sampling.

The Raman lidar system worked only at night because its highly sensitive detectors were saturated by sunlight in the daytime. During SPECTRE the Raman team also tested a new ultraviolet Raman lidar, capable of daytime operation by virtue of using wavelengths where ozone absorbs sunlight, and eventually retrieved daytime humidity data up to 2-3 km (a similar system is now functioning at the ARM Oklahoma facility). SPECTRE adapted to the Raman lidar's limitations by operating on a noon-to-midnight schedule and, to get better humidity profiles during the daytime half, interleaving SPECTRE radiosonde launches with those of FIRE so that launch frequency was sometimes as rapid as 1.5 hr and rarely worse than 3 hr.

Three FIRE teams also brought lidars, chiefly for aerosol and cloud detection and retrieval, but the Raman lidar also furnished the standard aerosol and cloud backscatter information by simultaneously operating in Raman and normal-lidar modes.

c. In situ Measurements

The remote profilers are blind to variations below 100 to 300 m (their first "range gate") due to response time limitations in their electronics and other factors. Therefore, our original plan called for the use of a tethersonde for measuring profiles of temperature, water vapor and ozone in this region, and also for calibrating the Raman lidar at its lowest range gate. Shortly before the experiment, however, the FAA denied permission for flying the tethersonde as planned, even though there was virtually no night traffic in and out of the small rural airport at which FIRE and SPECTRE operated. To compensate, more frequent radiosonde launches were made by the Wallops team throughout the SPECTRE field phase.

Our sensitivity analysis of radiation model calculations (discussed below) showed relatively small sensitivity to variations in trace gases as compared with variations in water vapor and temperature. Nevertheless, for completeness the NOAA team obtained flask sample measurements of the concentration of trace gases (CO2, CH4, N2O, CO, Freons F11 and F12) near the surface. They also made Dobson spectrometer measurements of the total column ozone, and vertical profiles of the ozone concentration from the surface through the stratosphere using standard ozonesondes.

The NOAA team also conducted an intercomparison of downward longwave flux observations from over 20 different pyrgeometers for the World Climate Research Program at a site about 100 m distant from the main SPECTRE encampment. The SPECTRE data are proving valuable in estimating the absolute accuracy of these pyrgeometer observations.

5. Pre-field-phase sensitivity analysis

In spite of the fact that no perfectly correct radiation model could be identified, all models show roughly similar sensitivities to changes in their input variables, and so it still made sense to do sensitivity analyses prior to the field phase. In particular, it was crucial to see whether SPECTRE's profiles of temperature and water vapor would be accurate enough to test uncertainties in radiation models. So, we calculated the sensitivity of the zenith radiance at the surface to possible errors in the measurement of the radiatively important variables (temperature, moisture, ozone, CO2, N2O, CH4, Freons and aerosols) for clear-sky conditions. Since improved understanding of the water vapor continuum is particularly important, and since opaque spectral regions mainly depend on the near-surface air temperature, the sensitivity study concentrated on the semi-transparent portion of the spectrum from 700 to 1220 cm-1. The calculations were performed with the line-by-line model FASCODE2 (Smith et al., 1978) using the McClatchey et al. (1971) Midlatitude Summer profile of pressure, temperature, humidity and ozone. The line-by-line results were averaged over 1 cm-1 intervals to roughly simulate the Wisconsin and Denver spectrometer resolution.

The calculations show that, for the rural aerosol conditions typical of the SPECTRE site, by far the most significant sources of radiance uncertainty arise from errors in measured temperature and water vapor. The spectral radiance has a spectrally averaged rms error of:

(a) about 0.25 RU per degree Kelvin temperature random error, and
(b) about 0.15 RU per percent water vapor mixing ratio random error.

In general, systematic errors in temperature and water vapor profiles have about twice the effect of random errors. The results show that the expected 5% humidity errors and 0.5-1°K temperature errors (Table 3) cause spectrally varying zenith radiance errors that range between 1 and 3 RU in the regions of relatively weak absorption and about 0.5 RU in regions of strong line absorption. The far greater part of these radiance errors is caused by humidity errors, and the reduction of errors in humidity profiles is vital if really precise tests of longwave models are ever to be made. But even 5 RU radiance errors are smaller than the effects of using different continuum formulations in climate model radiation codes; thus the spectral radiance and profiling data will allow more stringent tests of the parameterizations of the water vapor continuum than heretofore possible.

6. SPECTRE field phase

The field phase of SPECTRE was carried out in conjunction with the surface based portion of FIRE Cirrus II at the Coffeyville, Kansas airport during the period from 12 November through 7 December 1991. The suite of SPECTRE instruments was located on an unused concrete airport runway on the northern edge of the airport between the active runways, about 1 km from normal airport operations and the FIRE surface sites. Figure 1 shows a diagram of the SPECTRE site. The site was put as far as logistically possible from the airport buildings and FIRE aircraft in order to minimize pollution contamination of the SPECTRE measurements. The vegetation surrounding the site was mainly in a dormant stage, although there were several winter wheat fields surrounding the airport, and leafed-out trees along a nearby river course. There were three other potentially important pollution sources nearby, including a PCB destruction facility adjacent to the southern boundary of the airport, and a refinery and power plant both located in Coffeyville approximately 4 mi from the SPECTRE site. Naturally, human thermal, gaseous, and aerosol emissions from all these possible sources, as well as similar anomalies caused by nearby natural sources, were a big a priori concern, but during the field phase there appeared to be many almost perfectly clear 10-min periods uncontaminated by any such emissions.

The three SPECTRE spectrometers were located within 10-20 m of one another, with their data systems housed in a common trailer. A fourth, independent spectrometer, the NOAA FIRS, was about 40 m from the others in a small self-contained trailer next to the NOAA passive microwave radiometer trailer. All the other SPECTRE components except the Dobson spectrophotometer (kept in the FIRE aircraft hangar) were located along the abandoned runway within 100 m of the spectrometer trailer. These included the Raman lidar, 404 and 50 MHz RASS units, SPECTRE radiosonde/ozonesonde launching unit, microwave radiometer (see Han et al. 1994), international pyrgeometer array, trace gas sampling unit, and an NCAR-supplied Portable Automated Meteorological Station (PAMS). The trailers, electricity and necessary supplies for SPECTRE components were furnished by NASA.

Other FIRE participants collected data that are of use in interpreting SPECTRE's spectra, especially in cloudy conditions. In addition to several FIRE aircraft that flew almost exclusively on cirrus days, FIRE radars, lidars, shortwave and longwave radiometers, and CLASS sonde-launching facilities were concentrated at two sites within 0.8 km of the FIRE hangar, which in turn was about 0.7 km from the SPECTRE site. Some FIRE data is available from the Langley DAAC (Baum and Barkstrom, 1993; and some of the most relevant FIRE data will also be included on the SPECTRE CD-ROM.

The days of operations of the various platforms and the typical schedule followed for the sonde launches during SPECTRE operations are shown in Figs. 2a and 2b, respectively. The spectrometers usually operated continuously between local noon and midnight, whereas the Raman lidar typically operated between sunset and sunrise. Dobson measurement were obtained near local noon when possible, and flask samples of trace gases were collected at least twice per day. Radiosonde soundings, either FIRE (CLASS) or SPECTRE, were made at least every 3 hr, and, during intensive FIRE periods, about every 1-1/2 hr. Because of their expense ($700), SPECTRE ozonesondes were typically launched only once or twice a day, and then only on selected days when clear-sky conditions were particularly good.

The field phase proceeded relatively smoothly. The first week was mostly lost to fixing unanticipated problems: the SIRIS spectrometer had a catalog of problems keeping it in the FIRE hangar for a week; the Raman lidar circuitry experienced massive radio frequency interference from the nearby 50 MHz RASS; the RASS's experienced continuing antenna and electronic problems. All spectrometers and the Raman lidar were performing up to expectations after about a week, however. Several thousand spectra of the 3 to 18 µm region were obtained by the spectrometers during the following three weeks. Both RASS's experienced continuing problems, however, the 404 MHz because of high surface winds and the 50 MHz because of electromagnetic interference and a flawed antenna; this forced greater reliance on radiosondes for temperature profiling. But during many SPECTRE observing periods, after the radiosonde balloon burst, the descending temperature profile agreed almost perfectly with the ascending profile (to within 1-2 K), alleviating our concerns about horizontal inhomogeneity in temperature.

Fortunately, observing conditions during SPECTRE spanned a wider than expected range of the conditions important for infrared radiation. There were an unusually large number (from a climatological point of view) of clear-sky periods as well as several periods with overcasts of cirrus, middle and low-level clouds. Some clear-sky periods followed rainstorms which left the atmosphere remarkably aerosol-free for many hours. The surface temperature and column water vapor ranged from about 0 to 23°C and 0.5 to 2 cm, respectively. (2 cm of water vapor is not large for the tropics, where values of 5-6 cm are common, but it was high for Kansas in November.)

A partial in-field intercalibration of the spectrometers, using the calibrated black body furnished by the SIRIS team, was also attempted. This proved more difficult than anticipated, however, due to the differing mechanical configurations of the spectrometers and their resulting inability to fully view the blackbody. The black body was taken to Denver and to Wisconsin to complete the instrument intercalibration. Overall, tests in the field and subsequent analysis at the home institutions showed that the various spectrometers collected useful information on all of the days shown in Fig. 2a. The spectrometer groups estimate that they nearly met the stringent SPECTRE accuracy goals: an absolute accuracy of 1-2 RU for the spectral region 550 to 3000 cm-1, and a relative within-spectrum accuracy (from one spectral interval to another) of several tenths of a percent. Spectral radiances measured by all three groups agree to within 1-2 RU, at least in the 800-1200 cm-1 window.  Therefore, these data should allow more incisive tests of longwave radiation models over a wider range of atmospheric variables than heretofore possible.

7. Examples of the data

All groups have now produced calibrated data sets for each of the observation days, with special attention to the "SPECTRE special-case days". These data are nearly ready for publication on the SPECTRE CD-ROM.

Examples of data from the spectrometers are shown in Figs. 3 and 4. Figure 3 shows the distribution of downwelling clear-sky radiance measured by the Wisconsin spectrometer on Dec 6, 1991; the ranges of the absorption bands of the major infrared-active gases are also indicated. Comparisons of measurements between the Wisconsin, Denver and NASA-Goddard spectrometers for a portion of the atmospheric window between 10 and 10.5 µm are shown in Fig. 4. The three spectrometers agree fairly well for this case, although there are many differences beyond SPECTRE's 1 RU accuracy goal which may be due to the different resolutions of the spectrometers and/or the different calibration procedures. The 0.06 cm-1 resolution of the SIRIS spectrometer gives a false impression of large disagreements, which would disappear if its spectra were degraded to the 0.5 cm-1 resolution of the other two spectrometers; we show it at its natural high resolution to illustrate how it not only allows the identification of individual lines, but also the weakly absorbing regions between strong lines, neither of which is possible with the Wisconsin and Denver spectrometers.

The Raman lidar data (Fig. 5 shows an example for Dec 5) have been reduced to yield vertical profiles of the tropospheric H2O mixing ratio and the tropospheric and stratospheric aerosol back-scatter at approximately 100 m resolution at ten minute intervals (i.e., roughly the time intervals for spectra). Some significant differences have been found between radiosonde and Raman humidity soundings, in spite of the fact that the Raman lidar is calibrated against a radiosonde at the beginning of its daily operational period. These differences are most likely due to the vertical profile of humidity being different from the profile measured along the twisted curve followed by the sonde. Our preliminary analysis indicates that spectra calculated with Raman humidities agree noticeably better with SPECTRE observations than spectra calculated with radiosonde humidities, as indeed one would expect - but this had never been proven before. This tends to confirm that humidity is much less horizontally homogeneous than temperature, especially at low levels where it is most important for surface radiation, and that radiation models and radiation measurements cannot be brought into agreement using radiosondes alone.

The RASS data were SPECTRE's greatest disappointment. There had been some hope that post-processing could remove the electronic noise from the 50 MHz RASS data, but this proved impossible. A rescue effort for the 404 MHz RASS data was more successful, although it was not possible to remove all the effect of the abnormally high near-surface winds, and so virtual temperature profiles were only recovered to about 1-1.5 km instead of the 2-3 km originally expected.

Data from the SPECTRE radiosondes, Dobson spectrometer, ozone-sondes and trace-gas flask samples have been reduced and quality checked.  Summary results from some of the trace gas measurements are shown in Figs. 6 and 7.  Briefly, rather large day-to-day variations of ozone were found, both total column amount and lower troposphere profile. Furthermore, the concentrations of CO2 and CH4 were found to vary over a relatively large range and were not constant with altitude below about 1.5 km. Measurements of these variations may be important in resolving discrepancies between calculations and SPECTRE observations in particular spectral regions.

Mt. Pinatubo aerosol was a dramatic presence during SPECTRE, including producing spectacular orange sunsets almost every evening. The Raman lidar easily detected this aerosol, as described by Ferrare et al. (1992). Since fresh volcanic aerosol particles are submicron in size, and therefore interact only very weakly with longwave radiation (see Lacis and Mishchenko, 1995), Pinatubo aerosol apparently had a negligible effect on the observed spectra. Preliminary radiation model calculations with the LOWTRAN7 (Kneizys et al., 1988) volcanic aerosol model show the maximum expected aerosol effect in the 8-12 micron window region to be about 0.5 RU. The Raman and other FIRE observations may provide sufficient information about Pinatubo aerosols to help identify this effect in the radiance spectra.

8. Preliminary comparisons of observations with calculations

As the first step for selecting data for distribution to ICRCCM participants, the SPECTRE team examined the variability of the observed spectra and observations of the surface temperature, column water vapor and cloudiness in order to select case study periods which represented the extremes of the range of variability encountered. Particular emphasis was given to selecting periods where the spectra had very low variance over at least an hour, indicating a nearly perfectly horizontally homogeneous atmosphere (as assumed by the models). One of the selected periods was a uniform cirrus overcast, the rest were clear: Cold-Moist, Cold-Dry, Warm-Moist, Warm-Dry and intermediate temperature and water vapor. Summary information for these periods is given in Table 4. Examples of spectra from the Wisconsin AERI spectrometer for the Cold-Moist, Cold-Dry, Warm-Moist and Cirrus cases are shown in Fig. 8. Note how the 8-12 micron window region tends upward toward a Planck blackbody curve as its opacity increases due to cirrus cloud or increasing water vapor.

We have made a number of comparisons of the observed spectra with line-by-line and narrow-band radiation models. Examples of differences between observed and calculated spectra for the Warm-Moist and Cold-Dry special case days are shown in Figs. 9(a,b).  The calculations were performed with the line-by-line model LBLRTM (Clough and Iacono, 1995) using near simultaneous Raman water vapor, radiosonde temperature and observed trace gas data with the 1992 HITRAN absorption line parameter compilation (Rothman et al., 1992) as input. In general, the line-by-line model captures most of the qualitative features in the observed clear-sky spectra. It may be premature to make firm conclusions concerning the spectral and absolute character of the differences because we have not examined all of the data, and the differences below 600 cm-1 may be partly due to increasing detector noise as the detector sensitivity rolls off toward zero at 520 cm-1. Nevertheless, assuming a radiance observation error of 1 RU or about 1 unit along the y-axis in Figs. 9(a,b), the differences in Figs. 9(a,b) between the observed and calculated spectra are significant and unexplained in the following spectral regions: the 550-620 cm-1 region (which plays a large role in polar radiative balance); the 700-800 cm-1 shoulder of the CO2 15 mm band (heavily used for satellite temperature and humidity profiling) and the 9.6 mm ozone band, 1000-1100 cm-1 (which is important for the longwave balance in the stratosphere).

Calculations not shown here indicate that the LBLRTM water vapor continuum absorption formulation (CKD2) gives generally better spectral agreement with SPECTRE observations than the popular continuum formulation of Roberts et al. (1976), still used in many radiation models (especially those in large climate models). However, this superiority is not uniformly evident: for example, it cannot be proven for the Cold-Dry case of Fig. 9(b), because the observational error of 1 to 1.5 RU is comparable to the differences shown in the 800-1000 cm-1 region.  As more spectrometer data are examined, their high relative accuracy will be used to check the spectral signatures of the differences and put error bars on them.

In addition to absolute values of radiance, climate modelers need to know the accuracy of the response of their radiation models to changes in atmospheric variables. Unfortunately, in the field it is not possible to change just one variable and thus observe the partial derivatives that modelers usually calculate. Nevertheless, differences between observed and calculated spectra obtained under different atmospheric conditions can be used to test the ability of models to calculate total derivatives. An example, shown in Fig. 9(c), is the observed-minus-calculated double-difference spectrum D(Obs - Cal), where D means (Warm-Moist minus Cold-Dry). The calculated double-differences fall well within ±10% of the observed spectral changes D(Obs), the upper curve in Fig. 9(c).  Such double-differences may also prove useful in tests of continuum formulations since they remove residual systematic errors in the radiance observations.

Climate modelers are also anxious to test the ability of their radiation codes to calculate the total downward flux. But for longwave flux observations, they have had to rely upon pyrgeometers, which when operated routinely (rather than painstakingly nursed) have an accuracy no better than about 5% or 10-20 W m-2 (see Dutton, 1993) and which are notoriously difficult to calibrate and maintain. Now, however, there is an attractive alternative: SPECTRE (and other similar) radiance observations can be integrated over the observed portion of the spectrum, then supplemented by model estimates of the angular variation over the sky hemisphere to yield good estimates of the longwave flux.

As a first approximation, radiances can be assumed isotropic except when the atmosphere is very clear and dry, and the approximation gets better and better the wetter and/or more cloudy the atmosphere becomes (except when clouds are very patchy). However, using models for guidance, it is possible to do better. Define I(µ) as the spectrally-integrated radiance, where µ = cos(q) and q is the zenith angle; then, if F denotes flux, define a "radiance-to-flux conversion factor"

L = F/[p I(0)] (1a)

which is unity if I is isotropic (independent of µ). Here, radiative flux F is given by


To first order, L is a function only of the temperature and moisture profiles. Then a detailed model may be used to calculate L for each sounding (an assimilated sounding, or an actual sounding nearby in space and time, will give satisfactory accuracy, since L has small variability), and the uncertainty in an individual flux estimate becomes, approximately,

d F = p L [ Iobs(0) - Ical(0) ] (2)

For opaque regions L = 1, whereas for the 550 to 3000 cm-1 region L fell between 1.1 and 1.2 for SPECTRE clear-sky cases. For the clear-sky data we have examined, [Iobs - Ical] is about 0.3 RU, spectrally averaged over the 550 to 3000 cm-1 interval, which makes d F about 3 W m-2 {p*1.2*0.3 RU*(3000 - 550 cm-1) }.

The SPECTRE radiance data did not cover the 0-550 cm-1 region. However, for the moisture conditions observed during SPECTRE, the radiation in the 0-500 cm-1 region is nearly a Planck blackbody function at the surface air temperature, and so the uncertainty in calculating the downward flux in this portion of the spectrum is relatively small. Thus, we can estimate the flux for homogeneous clear and cloudy conditions as follows: for the 0 to 550 cm-1 region, with model calculations; and for the 550 to 3000 cm-1 region, from the spectrometer observations with the conversion factor L. Note that this is similar to the approximation used to estimate fluxes from ERBE observations (Wielicki and Green, 1989), but here the L values are calculated for each sounding rather than assumed constant for the latitude region.

Note that L must be calculated for each sounding in order to achieve better than 5 W m-2 accuracy. Nevertheless, calculations with different sophisticated narrow band radiation models indicate that model-to-model differences in L for a given sounding lead to a flux uncertainty of at most 1 W m-2, despite the different models disagreeing at the 10 W m-2 level. Thus, the models are much better at estimating the angular variation of radiance than the absolute value of radiance.

Fig. 10 shows comparisons of fluxes estimated from AERI observed and LBLRTM calculated radiances using L values calculated with the MODTRAN3 (Berk et al., 1989 ) radiation model for 26 clear-sky periods during SPECTRE. The average agreement is about 2.5 W m-2, and the RMS differences fall within about 4 W m-2, or agreement at about the 1% level over a 100 W m-2 range of fluxes. In the past, differences smaller than 10 to 20 W m-2 between routine pyrgeometer observations and calculations could not be interpreted because of the large uncertainties associated with those instruments.

We have begun to compare fluxes estimated from the AERI with those measured at the SPECTRE site as part of the international pyrgeometer intercomparison project. A comparison of the AERI fluxes shown in Fig. 10 with those from one of the pyrgeometers (not shown) gives an RMS agreement of 4 W m-2. We interpret this good agreement to indicate that given great care, the accuracy of the pyrgeometers can be pushed to the 5 W m-2 level, or better. Additional comparisons must be made, however, to solidify this conclusion.

9. Validation of radiation codes in climate models

The major unfinished ICRCCM goal is the evaluation of the ability of radiation codes used in climate models to simulate observations.  We are contributing to this goal by having another round of model intercomparisons; however, this time the models are being intercompared with observations taken during SPECTRE as well as with each other.

The intercomparison is following the procedures used throughout ICRCCM. Observed atmospheric data needed for model input for the case studies noted in Table 4 (Warm-Moist, etc.) have been made available to ICRCCM investigators. Since each model calculates radiance in its own unique spectral intervals, each modeler has been requested to calculate radiance at their model's highest spectral resolution. We will then integrate the observed radiance data and line-by-line calculations to match the spectral intervals of the various models. Following the completion of all of the calculations, each participant will be supplied with the observations and line-by-line calculations appropriate for their model intervals.

The first priority for distribution are the clear-sky, low aerosol concentration, case studies , because these will allow the best tests of radiation models in the semi-transparent and window portions of the spectrum, the regions of greatest importance for radiative forcing by greenhouse gases and for water vapor feedback effects. As more cloudy-day data become available from the other FIRE participants, intercomparisons for cirrus and other homogeneous cloud cases will also be inaugurated.

The distribution of the case study data is being done mainly through ftp from an Internet site at the University of Maryland. Non-ICRCCM participants are also welcome to retrieve the input data and submit results. Alternate procedures will be adopted for those not having access to the Internet. Those wishing to participate in the current exercise should contact Ellingson (email:

The ICRCCM intercomparison activities represent only a small aspect of the potential use of the SPECTRE data, and so it is important that those data be archived and made available to the scientific community. Although the data for a few case studies may be easily transported via the Internet, the large volume of data collected during SPECTRE is more conveniently distributed on a CD-ROM. We are in the process of preparing such a CD-ROM containing both the spectral and ancillary data. All spectra will be put in a standard ASCII format to facilitate intercomparison with each other and with models.

10. Field program design: Lessons from SPECTRE

We found many advantages in SPECTRE's tight focus, centralized decision-making, informality, flexibility, and efficiency. This was a direct consequence of its team, umbrella-proposal, hierarchical organizational method. Before the mid-1970s, most field efforts were, like SPECTRE, the product of such small tightly-focused teams with leaders having real authority and power of the purse string. But since then, most field program dollars have gone to large, weakly-coupled collections of individualistic PI efforts. Both organizational methods have their advantages and disadvantages. For example, the modern method may unleash greater creativity and lead to more unanticipated discoveries, while the SPECTRE method makes it much easier to rapidly juggle resources as logic and necessity dictate and in particular to rapidly fund new and originally unanticipated needs.

Yet the modern method has become dominant not so much because its organizational method has been judged superior, but because the interconnectedness of Earth system components argues for field programs that study many components at the same time. In order to do that, the programs have longer and longer lists of goals and have therefore grown large and expensive - and wide dispersal of participation is a political necessity for expensive experiments. To achieve the best dispersal, a call for proposals is issued, then the best proposals are selected by peer-review (usually a different reviewer for every proposal). Program managers can fine-tune the selections to achieve a better balance, but are mostly bound by the reviews. The result is a group of PIs whose proposals, taken individually, represent good science, but which as a group may be duplicative and only weakly interacting, may not hang together as a tightly focused experiment, and may not lead to success measured by any criterion other than PI publications. The programs serve the PIs, but it is less clear that the PIs serve the program goals.

The two organizational methods are generally viewed as mutually exclusive, with SPECTRE-type organization possible and practical for experiments under $1M or so, but the collection-of-individualistic-PIs organization the only alternative for larger experiments. We suggest that it would be possible to merge the two methods, and capture some of the best features of both. All that would be necessary in the call for proposals would be to define the field program in terms of its necessary sub-elements, rather than the customary list of topics and questions. (This topic list usually falls far short of actually defining a field program; it rests on the hope that the submitted proposals will do so.) These "defined sub-elements" might be variables (like radiative fluxes), naturally related groups of variables (like cloud parameters), types of instrument (like lidar), types of instrument platforms (like aircraft), a data system, or models needed to assimilate or incorporate the experiment's data. Even something larger like SPECTRE could be a defined sub-element: for example, it could be the surface sub-element of a field program designed to test radiation models. The key requirement would be: only one proposal for each sub-element would be selected. This requirement would encourage teaming arrangements formed in an effort to have the best scientists and capabilities on a given proposal, and hence the greatest competitive advantage. The proposals would necessarily be larger, but this method would preserve the goal of dispersing the experiment widely while still providing tighter focus, better up-front definition, stronger coupling, and less duplication than the pure individualistic-PI method.

Another way to take advantage of the SPECTRE-type organizational method within a larger experiment is exemplified by the ARM Program, which grew out of SPECTRE. In ARM, many infrastructure measurements like radiosondes are permanently emplaced and taken care of by the program, allowing PI's to propose small incremental experiments organized like SPECTRE. Several such small, focused experiments have already taken place at and around the ARM Oklahoma facility. This makes the ARM facility something like an astronomical observatory, taking maximum advantage of PI creativity and at the same time allowing effective, SPECTRE-like organizational principles.


This project would not have been possible without the support and activities of many people. The SPECTRE team is particularly thankful for the faith and support of Mike Riches of DOE and Bob Schiffer of NASA, whose programs supported the project. The authors extend their warmest thanks to all the SPECTRE Team members, without whose long hours and devotion none of this would have been possible, including: Tom Ackerman, John Brasunas, Jim Churnside, Paul Delaney, John De Luisi, Ellsworth Dutton, Rich Ferrare, Bob Knuteson, Andy Korb, Virgil Kunde, Harvey Melfi, Prentiss Moore, Frank Murcray, Hank Revercomb, Jim Shaw, Suhung Shen, Bill Smith, Jack Snider, Dick Strauch, Jim Wendell, Sam West, and Dave Whiteman. Last but not least, we thank Tony Clough and two anonymous referees for their thorough reviews of this paper. This research was funded in part by the US Department of Energy under Grants DEFG05-90ER61075 and DEFG05-90ER60971 and by NASA.


Barkstrom, B. R., E. Harrison, G. Smith, R. Green, J. Kibler, R. Crest, and the ERBE Science Team, 1989: Earth Radiation Budget Experiment (ERBE) archival and April 1985 results. Bull. Amer. Meteor. Soc., 70, 1254-1262.

Baum, B. and B. Barkstrom, 1993: Design and Implementation of a prototype data system for earth radiation budget, cloud, aerosol, and chemistry data. Bull. Amer. Meteor. Soc., 74, 591-598.

Bell, E., L. Eisner, J. Young and R. Oetjen, 1960: Spectral radiance of sky and terrain at wavelengths between 1 and 20 microns. II. Sky measurements. J. Opt. Soc. Amer., 50, 1308-1320.

Berk, A., L. Bernstein, and D. Robertson, 1989: MODTRAN: A Moderate Resolution Model for LOWTRAN7. Final Report GL-TR-9-0122, Geophysics Laboratory, Hanscom AFB, MA, 38pp.

Brasunas, J., V. Kunde, and L. Herath, 1988: Cryogenic Fourier spectrometer for measuring trace species in the lower stratosphere. Appl. Opt., 27, 4964-4976.

Clough, S. A., F. Kneizys and R. Davies, 1989: Line shape and the water vapor continuum. Atmospheric Research, 23, 229-241.

_______, M. J. Iacono, and J-L. Moncet, 1992: Line-by-line calculations of atmospheric fluxes and cooling rates: Application to water vapor. J. Geophys. Res., 97, 15761-15785.

_______, and M. J. Iacono, 1995: Line-by-line calculations of atmospheric fluxes and cooling rates II: Application to carbon dioxide, ozone, methane, nitrous oxide, and the halocarbons. J. Geophys. Res., 100, 16519-16535.

Collard, A., S. Ackerman, W. L. Smith, X. Ma, H. Revercomb, R. Knuteson and S.-C. Lee, 1995: Cirrus cloud properties derived from high spectral resolution infrared spectrometry during FIRE II. Part III: Ground-based HIS results, J. Atmos. Sci. 52, 4264-4275.

Cox, S., D. McDougal, D. Randall, and R. Schiffer, 1987: FIRE - The First ISCCP Regional Experiment. Bull. Amer. Meteor. Soc., 68, 114-118.

Dutton, E. G., 1993: An extended comparison between LOWTRAN7 computed and observed broadband thermal irradiances: global extreme and intermediate surface conditions. J. Atmos. Ocean. Tech., 10, 326-336.

Ellingson, R. and Y. Fouquart, 1991: The intercomparison of radiation codes in climate models (ICRCCM): An overview. J. Geophys. Res., 96, 8926-8929.

_______, J. Ellis and S. Fels, 1991: The intercomparison of radiation codes in climate models (ICRCCM): Longwave results. J. Geophys. Res., 96, 8929-8953.

Ferrare, R., S. H. Melfi, D. Whiteman, and K. Evans, 1992: Raman lidar measurements of Pinatubo aerosols over southeastern Kansas during November-December 1991, Geophys. Res. Letters, 19, 1599-1602.

_______, S. Melfi, D. Whiteman, K. Evans, F. Schmidlin and D. Starr, 1995: A comparison of water vapor measurements made by Raman lidar and radiosondes, J. Atmos. Ocean. Tech. , 12, 1177-1195.

FIRE Cirrus II Special Issue, 1995: J. Atmos. Sci., 52, 4041-4385

FSET, 1989: Fire Phase II Research Plan: A Component of the U.S. Cloud Climate research Program, FIRE Project Office, NASA Langley Research Center, Hampton, VA.

Han, Y., J. Snider, E. Westwater, S. Melfi and R. Ferrare, 1994: Observations of water vapor by ground-based microwave radiometers and Raman lidar, J. Geophys. Res., 99, 18695-18702.

Kneizys, F. X., E. P. Shettle, L. W. Abreu, J. H. Chetwynd, G. P. Anderson, W. O. Gallery, J. E. A. Selby, and S. A. Clough, 1988: Users Guide to LOWTRAN 7. AFGL-TR-88-0177, Air Force Geophysics Laboratory, Hanscom AFB, MA, 137pp.

Kunde, V., J. Brasunas, B. Conrath, R. Hanel, J. Herman, D. Jennings, W. Maguire, D. Walser, J. Annen, M. Silverstein, M. Abbas, L. Herath, H. Buijs, J. Berube, and J. McKinnon, 1987: Infrared spectroscopy of the lower stratosphere with a balloon-borne cryogenic Fourier spectrometer, Appl. Opt. ,26, 545-553.

Lacis, A. A., and M. I. Mishchenko, 1995: Climate forcing, climate sensitivity, and climate response: A radiative modeling perspective on atmospheric aerosols. In Aerosol Forcing of Climate, R, J. Charlson and J. Heintzenberg, Eds., John Wiley & Sons Ltd., 11-42.

Luther, F., ed., 1984: The Intercomparison of Radiation Codes in Climate Models, World Meteorological Organization, World Climate Program Report WCP-93, Geneva, Switzerland.

_______, R. Ellingson, Y. Fouquart, S. Fels, N. Scott and W. Wiscombe, 1988: Intercomparison of Radiation Codes in Climate Models (ICRCCM): Longwave Clear Sky Results. Bull. Amer. Met. Soc., 69, 40-48.

May, P., R. Strauch, and K. Moran, 1988: The altitude coverage of temperature measurements using RASS with wind profiler radars. Geophys. Res. Lett., 15, 1381-1384.

McClatchey, R., R. Fenn, J. E. A. Selby, F. Volz, and J. Garing, 1971: Optical Properties of the Atmosphere. Rep. AFCL-71-0279, Air Force Cambridge Research Lab., Bedford, MA, 85pp.

Melfi, S. H. and D. Whiteman, 1985: Observations of lower atmospheric moisture structure and its evolution using a Raman lidar, Bull. Amer. Met. Soc., 66, 1288-1292.

_______, D. Whiteman, and R. Ferrarre, 1989: Observation of atmospheric fronts using Raman lidar moisture measurements, J. Appl. Met., 9, 789-806.

Murcray, D., 1984: Atmospheric transmission in the 750-2000 µm region. J. Quant. Spect. Rad. Trans., 32, 381-396.

Murcray, F. H., F. J. Murcray, D. Murcray, J. Pritchard, G. Vanasse and H. Sakai, 1984: Liquid nitrogen-cooled Fourier Spectrometer system for measuring atmospheric emission at high altitudes. J. Atmos. Ocean. Tech., 1, 351-357.

Ramanathan, V., M. Lian and R. Cess, 1979: Increased atmospheric CO2: Zonal and seasonal estimates of the effect on the radiation energy balance and surface temperature, J. Geophys. Res., 84, 4949-4958.

Revercomb, H., H. Buijs, H. Howell, D. LaPorte, W. L. Smith, and L. Sromovsky, 1988: Radiometric calibration of IR Fourier transform spectrometers: solution to a problem with the high-resolution interferometer sounder. Appl. Opt., 27, 3210-3218.

Riches, M., 1983: CO2 Climate Research Plan, Report DOE/ER-0186, U.S. Dept. of Energy, Washington, D.C.

Roberts, R., J. Selby and L. Biberman, 1976: Infrared continuum absorption by atmospheric water vapor in the 8-12 micron window. Appl. Opt., 15, 2085.

Rothman, L., R. Gamache, R. Tipping, C. Rinsland, M. A. H. Smith, D. Benner, V. Malathy Devi, J.-M. Flaud, C. Camy-Peyret, A. Perrin, A. Goldman, S. Massie, L. Bown, and R. Toth, 1992: The HITRAN molecular database: editions of 1991 and 1992. J. Quant. Spect. Rad. Trans., 48, 469-508.

Smith, H. J. P., D. Dube, M. Gardner, S. Clough, F. Kneizys, and L. Rothman, 1978: FASCODE - Fast Atmospheric Signature Code. Report AFGL-TR-78-0081, Air Force Geophysical Laboratory, Hanscom AFB, Massachusetts.

Smith, W. L., X. Ma, S. Ackerman, H. Revercomb, and R. Knuteson, 1993: Remote sensing cloud properties from high spectral resolution infrared observations. J. Atmos. Sci., 50, 1708-1720.

_______, H. Revercomb, R. Knuteson, F. Best, R. Dedecker, H. Howell and H. Woolf, 1995: Cirrus cloud properties derived from high spectral resolution infrared spectrometry during FIRE II. Part I: The High Resolution Interferometer Sounder (HIS) systems, J. Atmos. Sci. , 52, 4238-4245.

Stokes, G. M., and S. E. Schwartz, 1994: The Atmospheric Radiation Measurement (ARM) Program: Programmatic background and design of the cloud and radiation test bed., Bull. Amer. Meteor. Soc.., 75, 1201-1221.

Tipping, R. and Q. Ma, 1995: Theory of the water vapor continuum and validations, Atmos. Res., 36, 69-94

Uttal, T., E. Clothiaux, T. Ackerman, J. Intrieri and W. Eberhard, 1995: Cloud boundary statistics during FIRE II, J. Atmos. Sci., 52, 4276-4284.

Wang, W-C., J. London, I. Isaksen, K. Shine, R. Ellingson, and F. Taylor, 1992: Summary report of the IUGG-IAMAP Workshop MW5: Climatic Effects of Atmospheric Trace Constituents, Vienna, 19-20 August 1991. Bull. Amer. Meteor. Soc., 73, 801-804.

WCP-92, 1984: Report of the TOGA Workshop on Sea Surface Temperature and Net Surface Radiation, La Jolla, CA, USA, 28-30 March 1984, World Climate Program Report, Geneva, Switzerland.

Wielicki, B. A., and R. N. Green, 1989: Cloud identification for ERBE radiative flux retrieval. J. Appl. Meteor., 28, 1133-1146.

Table 1. Leaders for the various SPECTRE teams

Spectrometer Measurements
V. Kunde, J. Brasunas NASA Goddard Space Flight Center
D. Murcray Denver University
W. Smith, H. Revercomb University of Wisconsin

Remote Measurements
H. Melfi, D. Whiteman NASA Goddard Space Flight Center
R. Strauch NOAA Environmental Technology Lab

In situ Measurements
H. Bell, P. Moore, S. West NASA Wallops Facility
J. DeLuisi, J. Wendell NOAA Climate Monitoring and Diagnostics Lab

Table 2. Spectrometers used in SPECTRE. Each instrument is a Fourier transform interferometer.

SIRISV. Kunde NASA Goddard 0.06
700 - 2000
D. MurcrayDenver Univ. 1.0
550 - 2000
AERIW. Smith Univ. of Wisconsin 0.5
550 - 3000
HISW. Smith Univ. of Wisconsin 0.5
550 - 3000
FIRSJ. ChurnsideNOAA/ETL 0.5
550 - 3000

Table 3. Remote profiling instrumentation used in SPECTRE. Raman lidar uses the unique wavelength shift of photons scattered back by water molecules to determine their abundance (see Melfi and Whiteman, 1985; Melfi, et al., 1989). RASS (Radio Acoustic Sounding System) uses radar signals backscattered and Doppler shifted by acoustic waves to determine the sound speed and thereby the virtual temperature (see May et al., 1988).

Quantity Measured
Vertical Range
Vertical Resolution
Averaging Time
Raman lidarwater vapor mixing ratio 300 m to

7 km

100-150 m 2 min0.1-0.2 g/kg
or 5%
400 MHz RASSvirtual temperature 300 m to

3 km

150 m 10 min0.5 - 1°C
50 MHz RASSvirtual temperature 2 - 7 km150 m10 min 0.5 - 1°C

Table 4. - SPECTRE case study periods.




Surface Temperature (°C)
Precipitable Water
Warm-Moist18 - 19 Nov.
T and H2O
26 Nov.
Warm-Dry 30 Nov.
Cold-Dry03 - 04 Dec.
Cold-Moist05 Dec.
Cirrus05 Dec.

Appendix A: Terms

black body

anything radiating according to the Planck function; when referring to a calibration source, means a cavity maintained at as constant a temperature as possible which is designed to absorb every infrared photon entering it (often appears perfectly black to the eye as well); the SPECTRE black body allowed its temperature to be continuously varied

continuum absorption

an absorption smoothly varying with wavelength (in contrast to line absorption); in Earth's atmosphere, due almost entirely to water vapor; probably present throughout the spectrum but can only be observed in "window" regions where line absorption is weak; dominant current theory ascribes it to far wings of strong lines (Tipping and Ma, 1995)

absorption band

a group of related absorption lines (usually 1000s)


unit of wavenumber; (wavelength in mm) = 104/(wavenumber in cm-1)

line absorption

absorption of a photon due to transitions between different quantum mechanical states of rotation and/or vibration of a molecule;

line parameters

the characteristics of an absorption line; for radiative transfer, the most important are the line position, half-width, and strength (integrated area under the line) at standard temperature and pressure; the half-widths and strengths at other temperatures and pressures are deduced from theoretical (quantum-mechanical) considerations

line shape

lines in a spectrum look like spikes but actually have a small width and a monomodal shape due to molecular collisions in the troposphere and due to the Doppler effect in the stratosphere


refers to a radiation model with such high spectral resolution that it resolves individual absorption lines; requires over 100,000 spectral points to cover a major fraction of the longwave spectrum

optical depth

dimensionless measure of the amount of absorbing and/or scattering; equal to -ln(Tr) where
Tr = transmission of a beam of radiation between two points; varies with wavelength


Radiance Unit, equal to 1 mW m-2 sr-1 (cm-1 )-1; about 1% of the maximum or full-scale radiance observed in SPECTRE; the nominal accuracy goal of SPECTRE


inverse wavelength (see cm-1)

Appendix B: Acronyms


Atmospheric Emitted Radiance Interferometer; Univ. of Wisconsin instrument whose prototype was used in SPECTRE and which is now permanently deployed at the Oklahoma ARM facility


Atmospheric Radiation Measurements; a Dept. of Energy program to improve climate models in the areas of radiation and clouds


Canadian company which manufactures a variety of popular laboratory interferometers


Crosschain Loran Atmospheric Sounding System; trailer-based mobile radiosonde launching system used by NCAR


Earth Radiation Budget Experiment; a three-satellite system to study the radiative energy budget of Earth in the 1985-1990 period


Federal Aviation Administration


Fast Atmospheric Signature CODE; first widely-available line-by-line radiative transfer model; developed at Air Force Geophysics Lab in early 1980s


First ISCCP Regional Experiment; a large multidisciplinary cloud field program concentrating on marine stratocumulus and cirrus


Fourier InfraRed Spectrometer; a mobile trailer-based interferometer developed by NOAA laboratories in Boulder and patterned after the AERI


High-resolution Interferometer Sounder; a rugged self-calibrating aircraft and surface interferometer; first high-resolution spectrometer used in field programs in mid- to late 1980s


InterComparison of Radiation Codes in Climate Models


InfraRed Interferometer Sounder; an Earth-viewing spectrometer flown on the Nimbus-4 satellite in 1974; scientists are still interested in this data because no other remotely comparable spectrometer has since been flown to view the Earth


International Satellite Cloud Climatology Project


Line By Line Radiative Transfer Model; successor to FASCODE, also publicly available but now through ARM program rather than Air Force Philips Lab


National Center for Atmospheric Research, Boulder, Colorado


Stratospheric InfraRed Interferometer Spectrometer; a rugged interferometer designed originally for stratospheric balloon flight in limb-scanning mode (to identify trace gas species)


Tropical Ocean/Global Atmosphere; a decade-long program punctuated by several intensive field experiments

Figure Captions

Fig. 1. Relative positions of equipment at the SPECTRE site.

Fig. 2a. SPECTRE observation days during FIRE Cirrus II. The shaded boxes indicate days on which the instruments were operating.

Fig. 2b. Nominal radiosonde launch schedule during SPECTRE observation periods. Solid lines represent times when sondes were always launched, whereas dotted lines correspond to times when sondes were launched during FIRE IOPs.

Fig. 3. Spectrum of zenith atmospheric emission measured by the Wisconsin spectrometer at 2000 UT on 6 December 1991. Note that the detector sensitivity decreases strongly below 550 cm-1 and thus the data become increasingly noisy there.

Fig. 4. Intercomparison of near-simultaneous spectra observed by the Wisconsin, Denver and NASA-Goddard spectrometers for the 10.0 to 10.5 µm (1000 to 950 cm-1) region near 2000 UT on 06 December 1991. Note that the Wisconsin spectrometer has twice the resolution of the Denver instrument, whereas the NASA-Goddard spectrometer has about 10 times the resolution of the Wisconsin instrument.

Fig. 5 Raman lidar data for Dec 5, 1991 (the Cold-Moist case). Note the interference from cirrus clouds above 8 km after about 0500 UT.

Fig. 6. Examples of ozone soundings: Nov 18 and Nov 25. This is typical of the considerable day-to-day variation in the ozone profile, esp. in the troposphere, although these large variations have only a 1-2% effect on the total longwave flux at the surface (but a larger effect in the ozone absorption bands).

Fig. 7. Daily averaged surface concentrations of CO2 and CH4 during SPECTRE, based on flask samples analyzed by the NOAA labs in Boulder, Colorado, using standard procedures adopted for their own international network.

Fig. 8. Examples of Wisconsin spectra from the case study periods; (a) Warm-Moist and Cold-Dry; (b) Cirrus and Cold-Moist.

Fig. 9. Observed (Wisconsin) minus calculated (LBLRTM) radiance for the (a) Warm-Moist and (b) Cold-Dry cases. Panel (c) shows the difference between the Warm-Moist and Cold-Dry spectra relative to the difference between (a) and (b). Note that the units are the same in each panel.

Fig. 10. Comparison of fluxes estimated from AERI and LBLRTM spectrally integrated radiances from SPECTRE clear-sky periods. Both the AERI and LBLRTM radiances were converted to flux using L values calculated with the MODTRAN3 radiation model using Raman water vapor and radiosonde temperature profiles measured near the time of the radiance observations as input.