Hyperspectral Atmospheric Sounding Validation

203FEBRUARY 2008AMERICAN METEOROLOGICAL SOCIETY |

EAQUATE An International Experiment For Hyperspectral Atmospheric Sounding Validation

BY J. P. TAYLOR, W. L. SMITH, V. CUOMO, A. M. LARAR, D. K. ZHOU, C. SERIO, T. MAESTRI, R. RIZZI, S. NEWMAN, P. ANTONELLI, S. MANGO, P. DI GIROLAMO, F. ESPOSITO, G. GRIECO, D. SUMMA, R. RESTIERI, G. MASIELLO, F. ROMANO, G. PAPPALARDO, G. PAVESE, L. MONA, A. AMODEO, AND G. PISANI

The validation of advanced infrared sounding satellites requires a

diverse set of coordinated synergistic observations utilizing ground-

based and airborne instrumentation.

T he European Aqua Thermodynamic Experi-

ment (EQUATE) was held in September 2004

in both Italy and the United Kingdom to vali-

date data from the Atmospheric Infrared Sounder

(AIRS) instrument on the Earth Observing System

(EOS) Aqua satellite. It also aimed to demonstrate

how combinations of ground-based and airborne

systems are useful for validating hyperspectral sat-

ellite sounding observations from satellites due for

launch through this decade. The era of hyperspectral

atmospheric sounding began with the launch of the

AIRS experiment of the EOS Aqua satellite during

May 2002. Following AIRS, the Infrared Atmospheric

Sounding Interferometer (IASI) was launched

aboard the Meteorological Operational (METOP)

polar-orbiting satellite in October 2006, and the

Cross-Track Infrared Sounder (CrIS) instrument is

to be orbited aboard the National Polar-Orbiting

Environmental Satellite System (NPOESS) Prepatory

Project (NPP) operational series of polar orbiters in

late 2009. Operational geostationary satellite imaging

hyperspectral sounding instruments, modeled after

the experimental Geosynchronous Imaging Fourier

Transform Spectrometer (GIFTS), are being planned

for implementation during the next decade.

The focus of this initial experiment was placed

on the validation of the AIRS instrument on the

EOS Aqua satellite. During EAQUATE, the Proteus

aircraft, carrying five separate remote sensing

instruments [the NPOESS Airborne Sounder

Testbed-Interferometer (NAST-I), NAST-Microwave

(NAST-M), Scanning High-Resolution Infrared

Sounder (S-HIS), Far-Infrared Sensor for Cirrus

(FIRSC), and Microscale Measurement of Atmospheric

Pollution Sensor (micro-MAPS)], was stationed in

Naples, Italy, from 4 to 11 September and in Cranfield

United Kingdom, from 11 to 19 September 2004.

During the Italian portion of the campaign, Proteus

underflew Aqua in coordination with ground-based

remote sensing measurements, including several

Raman lidar water vapor and temperature profilers

and radiosondes, provided by


FIG. 1. The Proteus aircraft showing instrumentation f itted during EAQUATE.

204 FEBRUARY 2008|

the Istituto di Metodologie per

l’Analisi Ambientale (IMAA),

the Dipartimento di Ingegneria

e Fisica dell’Ambiente (DIFA),

the University of Basilicata

in Potenza, and the Univer-

sity of Naples (UNINA), in

Napoli, Italy. During the U.K.

portion of the campaign, the

Proteus underflights of Aqua

were coordinated with the

U.K. Facility for Airborne

Atmospheric Measurements

(FAAM) BAel46-301 aircraft,

which flew a large payload of

in situ measurement instru-

ments, including dropsondes

and remote-sensing instru-

ments [e.g., the Airborne Research Interferometer

Evaluation System (ARIES) interferometer spec-

trometer], useful for validating the Aqua satellite

observations. A brief description of the instrumen-

tation used in the EAQUATE campaign is given in

“Instrumentation,” followed by a description of some

of the results from “The Italian phase” and “The U.K.

phase.” The “Conclusions” discuss the merits of cali-

bration and validation of satellite instruments and the

required synergy of different observing systems.

The results presented in this paper often rely on

complex retrieval theory. A good review of a range

of the retrieval techniques used here is presented in

Zhou et al. (2006) and Amato et al. (2002).

INSTRUMENTATION. Full details of the key

instrumentation used in the EAQUATE campaign

is given in Table 1.

Proteus (IPO). The Proteus aircraft (Fig. 1), sponsored

by the Department of Defense (DoD)–National

Oceanic and Atmospheric Administration (NOAA)–

National Aeronautics and Space Administration

(NASA) Integrated Program Office (IPO) for the

NPOESS carried five separate radiometers. The

Proteus generally maintained a f light altitude in

the range of 15–17 km, depending upon f light

duration, when underflying the Aqua satellite and

overf lying the IMAA/DIFA ground sites and the

FAAM BAe146-301 aircraft measurements.

FAAM BAe146-301 (Met Off ice and NERC). The

FAAM BAe146-301 (Fig. 2) is jointly funded by the

Met Office and the Natural Environment Research

Council (NERC). Capable of operating between

15 m and 10.5 km, the aircraft can carry a scientific

payload of 4,000 kg. For the EAQUATE campaign

the scientific payload consisted of a combination of

spectrometers for measuring the radiation from the

visible to the microwave region of the electromagnetic

spectrum plus an array of instrumentation to char-

acterize the troposphere both in terms of thermody-

AFFILIATIONS: TAYLOR AND NEWMAN—Met Office, Exeter, Devon, United Kingdom; SMITH—Hampton University, Hampton, Virginia, and University of Wisconsin—Madison, Madison, Wisconsin; CUOMO, ROMANO, PAPPALARDO, PAVESE, MONA, AND AMODEO—Istituto di Metodologie per l’Analisi Ambientale, CNR, Tito Scalo, Italy; LARAR AND ZHOU—NASA Langley Research Center, Hampton, Virginia; SERIO, DI GIROLAMO, ESPOSITO, GRIECO, SUMMA, RESTIERI, AND MASIELLO—Dipartimento di Ingegneria e Fisica dell’Ambiente, University of Basilicata, Potenza, Italy; MAESTRI AND RIZZI—Physics Department, Alma Mater Studiorum, University of Bologna, Bologna, Italy; ANTONELLI—Mediterranean Agency for Remote Sensing, Benevento, Italy; MANGO—NPOESS Integrated Program Office, Silver Spring, Maryland; PISANI—Consorzio Nazionale

Interuniversitario per le Scienze fisiche della Materia, University of Naples, Napoli, ItalyCORRESPONDING AUTHOR: Dr. Jonathan P. Taylor, Met Office, FitzRoy Road, Exeter, Devon, EX1 3PB, United KingdomE-mail: [email protected]

The abstract for this article can be found in this issue, following the table of contents.DOI:10.1175/BAMS-89-2-203

In final form 6 July 2007©2008 American Meteorological Society

https://www.researchgate.net/publication/222527960_The_s-IASI_code_for_the_calculation_of_infrared_atmospheric_radiance_and_its_derivatives?el=1_x_8&enrichId=rgreq-21c24320-86cc-47dc-8770-9d5ad0a0ab0d&enrichSource=Y292ZXJQYWdlOzIzMDc2NTkwNjtBUzoxMDQyNDMzMjMyMTE3NzlAMTQwMTg2NDk1MzM3Nw==

FIG. 2. The FAAM BAe146-301 aircraft, showing instrumentation.


namical variables like temperature and water vapor,

and also particulates in the form of aerosols and

cloud particles. The main focus of the work during

EAQUATE was cloud-free atmospheres, but one

flight in cirrus clouds was conducted under an Aqua

overpass. During the EAQUATE flights dropsondes

were launched to give vertical profiles of temperature

and water vapor. During satellite overpasses up to

14 sondes were launched in quick succession, giving

high levels of spatial detail. In addition to the ther-

modynamic measurements, the BAe146-301 made

continuous measurements of carbon monoxide and

ozone concentrations and also measured aerosol and

cloud particle size, shape, and concentration.

Ground based (IMAA/DIFA/UNINA). A wide range of

instrumentation was used during EAQUATE mea-

surement campaign: ground based, airborne, and

satelliteborne.

During the Italian phase of the experiment, the

ground-based instrumentation was located at three

different sites located in southern Italy. These systems

were operated continuously throughout the Proteus

airborne measurement period, which included the

Aqua overpass times.

Aqua AIRS (NASA). The AIRS instrument flying aboard

the EOS Aqua polar-orbiting satellite is the first space-

borne spectrometer designed to meet the 1-K/1-km

sounding accuracy objective of future operational satel-

lite sounders by measuring the infrared spectrum quasi

continuously from 3.7 to 15.4 μm, with high spectral

resolution (v/δv = 1200/1). The sensitivity requirements,

expressed as noise-equivalent differential temperature

(NEDT), referred to a 250-K target temperature and

ranges from 0.1 K in the 4.2-μm

lower-tropospheric sounding

wavelengths to 0.5 K in the

15-μm upper-tropospheric and

stratospheric sounding spectral

region. The AIRS instrument

provides spectral coverage in

the 3.74–4.61-, 6.20–8.22-, and

8.8–15.4-μm infrared wave-

bands at a nominal spectral

resolution of v/δv = 1200, with

2378 IR spectral samples and

four visible/near-infrared

(VIS/NIR) channels between

0.41 and 0.94 μm. Cross-track

spatial coverage with ~16-km

linear resolution, depending

on scan angle, and views of

cold space and hot calibration targets are provided by

a 360° rotation of the scan mirror every 2.67 s. The

AIRS radiance measurement accuracy and its ability

to achieve the 1-K/1-km sounding accuracy objective

are to be validated using the EAQUATE surface and

airborne datasets.

THE ITALIAN PHASE. The Italian phase of

the EAQUATE measurement campaign took place

between 6 and 10 September 2004.

Figure 3 shows the times for all measurements;

note that two f lights of the Proteus aircraft were

coincident with Aqua overpasses.

The Proteus aircraft f lew legs at high altitudes

along the line of the Aqua overpasses at 0101 UTC

8 September and 0056 UTC 10 September (Fig 4).

The numbers of overpasses of the ground sites by the

Proteus aircraft are given in the figures.

NAST-I, AIRS retrieval validation. The NAST-I instru-

ment on Proteus covers the entire spectral range

of AIRS at greater resolution, allowing the direct

comparison of radiances and level-2 products, like

temperature and water vapor. During the night of

9–10 September 2004 the Proteus aircraft flew a flight

track over southern Italy that was coincident with

the AIRS overpass at 0056 UTC 10 September. The

aircraft track passed over the Potenza site, allowing

for a direct comparison of the satellite, aircraft,

and ground-based profile information (Fig. 4). The

NAST team of scientists has developed an inversion

scheme. Detailed NAST-I physically based empirical

orthogonal function (EOF) regression and simulta-

neous matrix inversion (i.e., nonlinear multivariable

physical iteration) can be found in Zhou et al. (2002).

TABLE 1. Key instruments used during the EAQUATE campaign.

Sensor Sensor description Products Location Reference

NPOESS Airborne Sounding Testbed-Interferometer (NAST-I)

Spectral range of 3.6–16.1 μm, spectral resolution (0.25 cm–1); FOV: size of 130 m km–1 of aircraft altitude (e.g., from the Proteus altitude of 16 km, 2.0-km spatial resolution is achieved)

Profiles of temperature and water vapor, a vertical resolution of 1–2 km, surface temperature and surface emissivity, direct radiance validation of satellite instruments; scans across track

Northrop-Grumman Proteus Aircraft

Smith et al. (2005); Zhou et al. (2002)

NPOESS Airborne Sounding Testbed-Microwave (NAST-M)

29 spectral channels in the 50–425-GHz range (i.e., eight channels in the 50–56-GHz oxygen band, seven channels on the 118-GHz oxygen line, seven channels on the 183-GHz water vapor line, and seven channels on the 425 oxygen line); ground resolution of 130 m km–1 of flight altitude

Temperature and water vapor sounding information through nonprecipitating clouds and maps out the spatial distribution of precipi-tating cells in the atmosphere


Blackwell et al. (2001)

Scanning High resolution Interferometer Sounder (S-HIS)

Spectral range 3.0–17.0-μm resolution 0.5 cm–1 spatial resolution 100 m km–1 of flight altitude views upward and downward

Profiles of temperature and water vapor and direct radiance validation of NAST-I and satellites.


Tobin et al. (2006)

Far Infrared Spectrometer for Cirrus (FIRSC)

Band-1 upwelling radiation in the 10–47 cm–1 (300–1400 GHz) band-2 upwelling radiation in the 80–135 cm–1 (75–135 μm) spectral resolution of 0.1 cm–1

Observe the microphysical properties of cirrus clouds

Vanek et al. (2001)

Microscale Measurement of Atmospheric Pollution Sensor (MicroMAPS)

Nadir-viewing gas filter radiometer 4.67-μm band of carbon monoxide; spatial resolution of 52 m km–1 of flight altitude

Radiometric measurement of carbon monoxide


Online at www.resonance.on.ca/mm aps.html

ARIES Thermal Infrared Interferometer

Bomem FTS 3.3- to 16-μm region with 0.5 cm–1 spectral resolution with both upward and downward pointing

Radiative transfer model validation, temperature profiling and profiling of various species (e.g., water vapor, ozone), cloud and surface emissivity studies

On board FAAM BAe146-301 Atmospheric Research Aircraft, operating between 15 m and 10.5 km

Fiedler and Newman (2005); Taylor et al. (2003)

MARSS Microwave Radiometer

Operates at AMSU-B channels 16–20 (89–183 GHz) and scans both upward and downward (+/–40° along track)

Satellite intercomparisons, radiative transfer model validation, temperature and humidity profiling, column liquid water retrievals, precipitation, ice cloud and surface emissivity studies


McGrath and Hewison (2001)

Deimos Microwave Radiometer

Four channels (each of two orthogonal polarizations at 23 and 50 GHz); upward or downward viewing

Satellite intercomparisons, radiative transfer model

validation, surface emissivity studies


McGrath and Hewison (2001)

The Tropospheric Airborne Fourier Transform Spectrometer (TAFTS) (owned by Imperial College, United Kingdom)

12.5–125-μm region with 0.1 cm–1 spectral resolution; upward and downward views

Study of upper-tropospheric and lower-stratospheric water vapor and cirrus clouds


Canas et al. (1997)

206 FEBRUARY 2008|

Atmospheric State Parameters

Temperature, water vapor, ozone, carbon monoxide, liquid water, ice water, particle size and shape, aerosol size scattering and chemistry, etc; dropsondes.

Wide range of measurements made at frequencies between 64 and 1 Hz

On board FAAM BAe146-301 Atmospheric Research Aircraft, operating between 15 and 10.5 m

Online at www.faam.ac.uk

Raman lidar*3 Transmitter: Nd:YAG 355 nm, 532 nm, prf 20 Hz

Vertical profiles, both in daytime and nighttime, of aerosol backscatter and extinction coefficients, water vapor mixing ratio, atmo-spheric temperature

1: Potenza laser and Basilicata 48°38 4́5.23˝N, 15°48´29.32˝E, 770 m ASL; 2: UNINA Napoli 40°50´18˝N, 14°10´59˝E, 118 m ASL; 3: The Istituto di Metodologie per l’Analisi Ambientale (IMAA) in Tito Scalo (40°36 4́.32˝N, 15°43´25.32˝E, 760 m ASL)

Di Girolamo et Bösenberg (2001); Matthias (2004); Böckmann (2004); Pappalardo et al. (2004)

Raman lidar Nd: YAG laser with third har-monic generator (355 nm) prf 100 Hz

Water vapor mixing ratio profiles from about 60 m above lidar station up to the UTLS

IMAA in Tito Scalo (40°36´4.32˝N, 15°43´25.32˝N, 760 m ASL)

Cornacchia et al. (2004)

Infrared Interferometer ABB MR-100 spectral range 500–5000 cm–1 resolution of 0.5 cm–1

Water vapor, temperature profiles, and information on cloud properties

IMAA in Tito Scalo (40°36´4.32˝N, 15°43´25.32˝E, 760 m ASL)

REFIR/BB (Radiation Explorer in Far Infrared/Bread Board)

FTS measuring over the range 100–1050 cm–1 with a resolution of 0.5 cm–1

Study of the water vapor rotational bands


Radiosonde Vaisala-type RS90 and RS92 sondes

Profiles of pressure, temper-ature, and relative humidity


Microwave radiometer 12 channels Profiles of temperature and water vapor from surface to 10 km and low-level cloud liquid water content


Ceilometer Vaisala cloud-base detector Cloud base up to 7.5 km, 15-km vertical resolution; five profiles a minute



FIG. 3. Chart showing times of measurements for all instru-ments involved in the Italian phase of EAQUATE.

FIG. 4. The four Proteus flight tracks: (a) 6, (b) 7, (c) 8, and (d) 9 Sep 2004. The black circles indicate the positions of Napoli, Tito Scalo, and Potenza, showing that the Proteus overflew each of these sites on every flight.

208 FEBRUARY 2008|

NAST-I measurements and its inversion algorithm

have been validated through numerous dedicated

field experiments (e.g., Smith et al. 2005). Processing

both AIRS and NAST-I data through this inversion

scheme allows the intercomparison of thermody-

namic parameters.

AIRS original single field-of-view (FOV) radiances

are put through the NAST team inversion scheme to

produce AIRS retrievals to compare with the NAST-I

retrievals degraded to AIRS full spatial resolution

(i.e., AIRS single FOV). The same inversion algorithm

is applied to both AIRS and NAST-I data to minimize

the impact of inversion algorithm differences on the

retrieval products. However, it should be noted that

the forward radiative transfer models used differ

in that the Stand-Alone AIRS Radiative Transfer

Algorithm (SARTA) (Strow et al. 2003) is used

for the AIRS retrieval while the optimal spectral

sampling (OSS) fast molecular radiative transfer

model (Moncet et al. 2003; Liu et al. 2003) is used for

the NAST-I retrieval. These retrievals can then be

compared directly with the measurements of dedi-

cated radiosonde and Raman lidar (e.g., Cuomo et al.

2004; Di Girolamo et al. 2004, 2006). Under cloud-

free conditions, the retrieved effective surface skin

temperature is equivalent to the skin temperature.

Figure 5 shows the skin temperature retrieved from

AIRS and NAST-I, with a full spatial resolution of

~16 and ~2 km, respectively. Profiles of temperature

and moisture retrieved from AIRS and NAST-I are

validated with dedicated radiosonde and Raman lidar

observations during the EAQUATE. In the figure, the

lines with arrows indicate the Aqua and Proteus flight

directions with associated time. The open cycles rep-

resent AIRS single FOV within NAST-I ground track

swath width. All NAST-I retrievals from the NAST-I

single FOV falling into the AIRS FOV are averaged

for comparison with the AIRS retrieval.

The mean profile of the section shown in Fig. 5

(indicated with the open cycles) is plotted in Fig. 6

with a Vaisala-type radiosonde and two Raman lidar

observations from Potenza. One Raman lidar station

(labeled lidar 1) is located at Potenza (40°39Ń,

15°48É; 730 m above sea level), while another Raman

lidar station (labeled lidar 2) and a radiosonde

station are located at Tito Scalo, Potenza (40°36Ń,

FIG. 5. AIRS- and NAST-I retrieved effective surface skin temperature (K) during the night of 9–10 Sep 2004.

FIG. 6. Retrievals of (top) temperature and (bottom) relative humidity using AIRS and NAST-I data processed through the NAST team inversion scheme. Radiosonde and Raman lidar profiles from Potenza (Ptz) are plotted as references.


15°44´E; 760 m above sea level). A

dedicated radiosonde was launched

while Proteus was passing over

Potenza. The Raman lidar data were

extensively processed (i.e., a 10-min

integration time and variable vertical

averaging) in order to reach a higher

altitude for retrieval validation as

Aqua and Proteus were passing over.

The Raman lidar data plotted in

Fig. 6 are the mean profiles of Aqua

(0056 UTC) and Proteus (0037 UTC)

overpasses. The retrievals from both

AIRS and NAST-I compare favor-

ably to Raman lidar and radiosonde

obser vat ions . However, a few

noticeable features are included,

such as follows: 1) there are integra-

tion time differences between AIRS

and NAST-I; in other words, the

Aqua satellite passed over quickly

while the Proteus’s f l ight from

south to north took more than

1 h, 2) the averaged profile-reduced

vertical resolution of a single FOV

retrieval has a lower vertical reso-

lution (approximately 2–3 km),

while radiosonde and Raman lidar

observations have a much higher

vertical resolution (15–500 m, de-

pending on altitude), 3) a dry bias

in the humidity measurement from a

Vaisala-type radiosonde at altitudes

above ~8 km has been noticed here

as well as in other validations, and

4) the fine vertical structures (i.e.,

resolution) of the retrieved profiles

are partially due to the instrumen-

tation characteristics, such as the

spectral resolution and instrument

noise, which could cause a difference

in the profile. Overall, the retriev-

als from the two different sounders

compare favorably with each other.

The AIRS–NAST-retrieved tempera-

FIG. 7. Temperature retrieval over tree-covered land near Tito Scalo on 10 Sep. First guess shown as black solid line with radiosonde shown as blue curve. The results of the physical retrieval conducted on NAST-I data using the 2- and 3-band scheme are shown in red circles and green crosses (respectively).

FIG. 8. Profiles of water vapor mixing ratio. Key to lines and symbols as per Fig. 7.

FIG. 9. Schematic of aircraft operations during the U.K. phase of EAQUATE.

210 FEBRUARY 2008|

ture and moisture profile differences, plotted in the

right-hand panels of Fig. 6, are within the uncertainty

levels for these parameters. This implies that both

instruments are well calibrated and that the radiative

transfer models are equivalently accurate.

NAST- I retr ieval over land. One of the NAST-I

footprints at 0032 UTC 10 September was over a

small tree-covered hill close to the IMAA/Consiglio

Nazionale delle Ricerche (CNR) Institute. The

emissivity of the surface, which was covered by

conifer trees, was determined by fitting a set of 16

clear-sky spectra recorded by NAST-I (Masiello et al.

2006). A physical retrieval scheme using the σ IASI

code (Amato et al. 2002; Carissimo et al. 2005) has

been used to retrieve temperature (Fig. 7) and water

vapor profiles (Fig. 8). The physical retrieval scheme

can be run using either two or three spectral bands.

The two–spectral band option uses bands between

670–830 and 1100–1200 cm–1. The three-band op-

tion uses an additional band of 1450–1600 cm–1. The

three-band option includes a portion of the water

vapor vibrational band at 6.7 μm. The inclusion of

this band had a negligible impact on the tempera-

ture retrieval and yielded a modest impact upon the

retrieved water vapor in

the lower atmosphere. In

principle, especially for

water vapor, one would

have expected a larger im-

pact by the inclusion of

band three. The fact here

is that the NAST-I band 3

has a relatively larger noise

in comparison to that of

bands 1 and 2; therefore,

the related radiances are

in part filtered out because

of the large variance weight

they had been assigned in

the observational covari-

ance matrix.

Although illustrative, the

retrieval example shown in

Figs. 7 and 8 gives evidence

of the vertical resolution

https://www.researchgate.net/publication/222527960_The_s-IASI_code_for_the_calculation_of_infrared_atmospheric_radiance_and_its_derivatives?el=1_x_8&enrichId=rgreq-21c24320-86cc-47dc-8770-9d5ad0a0ab0d&enrichSource=Y292ZXJQYWdlOzIzMDc2NTkwNjtBUzoxMDQyNDMzMjMyMTE3NzlAMTQwMTg2NDk1MzM3Nw==

FIG. 10. The (top left) sub–satellite track, (bottom right) BAe146-301 track, and (bottom left) BAe146-301 alti-tude time series overlaid on the Aqua MODIS image for 1320 UTC 14 Sep 2004. The portion of the sub–satellite track studied during this flight is shown as the red line.


improvement that can be reached with hyperspectral

sounding in the infrared. The fine structure in the water

vapor retrieval in between 600 and 800 mbar is nicely

reproduced, which leads us to conclude that the NAST-I

hyperspectral observation is capable of retrieving water

vapor with a vertical resolution of 1–2 km in the lower

troposphere. It is also worth noting that the inversion in

Figs. 7 and 8 has been obtained with a research-oriented

inversion scheme, in which the retrieval can be con-

strained by the degrees of freedom (Serio et al. 2007).

For the analysis at hand, the final degrees of freedom

of the H2O retrieval was 7.2. It can be shown that below

6.5 degrees of freedom the vertical resolution for water

vapor rapidly decrease to levels for which structures of a

length scale of the order of 1–2 km are no more resolved.

The performance of the retrieved temperature is not as

good in the boundary layer where the impacts of the

spectrally variable and nonunit emissivity is a contribut-

ing factor. In the layers below 800 mb the temperature

differences are still around 1 K, while the mixing ratio

differences are 0.1 g kg–1 at the surface (using the two-

band scheme).

THE U.K. PHASE. During the U.K. phase of

EAQUATE, the Proteus aircraft worked in close

col laboration with the BAe146-301, studying

the atmosphere during two AIRS overpass days.

Flights were f lown on 14 and 18 September 2004.

During this phase of EAQUATE the aim was to

validate AIRS in the remote marine environment

under cloud-free conditions by stacking the AIRS

satellite, Proteus aircraft, and BAe146-301 aircraft

vertically over each other during the overpass, as

indicated in Fig. 9. The Proteus aircraft was operat-

ing the NAST-I and S-HIS interferometers, while

the BAe146-301 made measurements of the in situ

state of the troposphere and the surface; dropsondes

were also launched to more finely characterize the

atmospheric column.

The conditions on 14 September 2004 are clearly

shown in the Moderate Resolution Imaging Spectro-

radiometer (MODIS) image in Fig. 10. After transiting

to the operating area, the BAe146-301 conducted a

profile descent to an altitude of 15 m above the sea

surface in the area of interest and then flew a straight

FIG. 11. Averaged nadir spectra at their original as-measured spectral resolutions for NAST-I (red), S-HIS (green), and AIRS (cyan) for the intercomparison region on 14 Sep 2004.

212 FEBRUARY 2008|

and level run at 30 m along the line of the subsatellite

track to measure the sea surface temperature. During

this low-level run of the BAe146-301 the Proteus

aircraft f lew at an altitude of around 14 km along

the same track in the same direction. The ground

speed of the BAe146-301 and Proteus were perfectly

matched and both aircraft remained stacked above

each other for this run. Having characterized the sea

surface, the BAe146-301 profiled to 8.5 km and flew

backward and forward along the subsatellite track,

launching dropsondes en route. Having burned a bit

more fuel, the BAe146 was then able to climb to an

altitude of 10.6 km, continuing along the subsatellite

track. During all of these high-level runs the BAe146

made measurements with the ARIES interferometer

and its other radiometers. Finally, the BAe146-301

conducted a long, descending profile back to the sur-

face to get a final look at the tropospheric structure of

temperature, water vapor, ozone, carbon monoxide,

and clouds.

The evaluation of satellite data can be conducted in

terms of retrieval performance, but also, importantly,

in terms of direct radiance validation. The Proteus

aircraft flies at an altitude that puts it above most of

the atmosphere, and hence it is ideally suited for the

important job of direct radiance validation using the

NAST-I and S-HIS instruments. NAST-I and S-HIS

have been compared on many occasions before and

their calibration is well characterized. During the

EAQUATE campaign the BAe146 and Proteus on oc-

casion both operated at 8.5 km, allowing for a direct

comparison of NAST-I, S-HIS, and ARIES, these data

are not shown here for brevity.

NAST-I, S-HIS, and AIRS radiance intercomparison. Only

observations from aircraft platforms can provide the

proper spatial and temporal coincidence needed to

best validate the radiometric performance of satellite-

based measurements like Aqua. The high spectral and

spatial resolution aircraft sensor radiance data can

be spectrally and spatially convolved, respectively, to

simulate what should be measured by the spatially and

temporally coincident satellite observations during

overpass events. The much higher spatial resolution

of the aircraft sensor data can play an important role

in validating satellite-derived data products under the

conditions of variable surface and atmospheric radi-

ance (e.g., resulting from clouds) within the satellite

sensor footprint.

The 11-μm window band (MB31) from the MODIS

aboard Aqua has been used to determine the most

spatially uniform scene of coincidence between AIRS

IFOVs and the NAST-I and S-HIS instruments on

the Proteus aircraft for the flight on 14 September

2004.

Figure 11 shows a comparison of the resulting

averaged nadir spectra at their original as-measured

spectral resolutions for NAST-I, S-HIS, and AIRS.

Such a first-order comparison covering the entire

spectral extent effectively demonstrates the desired

overlapping nature of these spectra, along with some

expected differences. Most notably, spectral structure

is better resolved with the higher spectral resolution

NAST-I; the AIRS grating has gaps in coverage over

the spectral extent; and spacecraft versus aircraft

altitude differences cause upwelling radiance observ-

able by AIRS to deviate from that seen by NAST-I

and S-HIS in select spectral regions (e.g., O3 and

CO2 bands). Figure 12 shows this same comparison

of infrared spectral radiance in more detail. This

intercomparison serves to exemplify the high degree

of radiometric (and spectral/spatial) fidelity of these

airborne Fourier transform spec-

trometer (FTS) systems (with

NAST-I and S-HIS differing, on

average, by 0.08 K) and their

utility for direct radiance valida-

tion of satellite sensors (with the

mean difference between NAST-I

and AIRS having a magnitude

< 0.15 K). While such results do

vary slightly in magnitude when

examining different underflights

resulting from, in large part, the

degree of spatial/temporal collo-

cation combined with the amount

of geophysical field nonunifor-

mity, a similar relationship has

typically been observed in other

FIG. 12. Spectral radiance intercomparison for a 100 cm–1 interval in the infrared window region: (a) for NAST-I reduced to S-HIS spectral resolution, S-HIS as measured, AIRS as measured, and NAST-I, with the AIRS spec-tral response functions applied; and for (b) difference plots with S-HIS and AIRS subtracted from the spectrally equivalent versions of the NAST-I measurement shown in (a).


cases analyzed. One consistent conclusion reached

among the different cases examined regards the im-

portance of such airborne measurements for satellite

sensor calibration/validation; that is, while radiance

simulations using available ground truth (radio-

sondes, NWP analysis fields, etc.) are an important

contributor to validation, better agreement is almost

always achieved upon infusing directly measured ra-

diance data from these very well-calibrated airborne

FTS sensors; this is particularly important for non-

uniform and/or rapidly varying scenes, those of most

meteorological interest for research and operational

usage of such satellite sensor systems.

Spatial variability of water vapor and the importance of in situ measurements. The flight of 18 September

2004 saw the BAe146-301 and Proteus aircraft flying

together under an AIRS overpass that occurred at

1254 UTC. The Proteus flew at a pressure of 150 hPa,

while the BAe146 operated at around 350 hPa. While

the Proteus aircraft flew a series of runs up and down

a track measuring the upwelling radiances with its

range of spectrometers, the BAe146 flew along the

same track launching a series of nine dropsondes to

characterize the temperature and water vapor struc-

ture of the troposphere. This dataset is being used

to evaluate the performance of the NAST-I retrieval

scheme, which is in turn being used to validate the

AIRS instrument retrieval. In addition to this valida-

tion study the dataset allows a study of the horizontal

variability of the thermodynamic structures in the

atmosphere, which are important to understand when

considering the calibration/validation of satellite

instruments. Figure 13 shows the positions of NAST-I

nadir view retrievals of temperature and water vapor

between 1221:00 and 1308:53 UTC (a total of 208

retrievals), where the data have been filtered to

remove points with aircraft roll in excess of 10°.

The times of the dropsondes were 1252, 1255,

1258, 1301, 1303, 1312, 1315, 1319, and 1350 UTC;

FIG. 13. The position of the NAST-I nadir retrievals (black line) and the launch and splashdown positions of the nine dropsondes (red). Launch points are those to the west.

FIG. 14. Contour plot of the average difference in relative humidity from NAST-I retrievals as a function of pressure and horizontal spatial scale.

214 FEBRUARY 2008|

thus, are all within 30 min of the NAST-I retrievals,

with many of them being significantly closer in

time. To give an idea of scale, the horizontal distance

from the launch to splashdown position of the most

northern dropsonde shown in Fig. 13 is 10.97 km.

Understanding spatial variability in water vapor and

temperature fields is critical to the interpretation of

remote-sensing data and any data used in a valida-

tion exercise; here, NAST-I retrievals will be used for

this purpose. For each pressure level of the retrievals

the absolute value of the relative humidity difference

between adjacent retrieval pixels (1.75 km) has been

computed and the average has been taken. There were

208 retrievals in all, so for one pixel difference 207

humidity differences are available for each pressure

level. This exercise was

carried out again looking

at the differences for all

possible 2-pixel differences

(equivalent to spatial scales

of around 3.5 km) through

to a l l possible 50-pixel

differences (equivalent to

spatial scales of around

105 km). Figure 14 shows,

as a contour plot, the aver-

age difference in relative

humidity, evaluated from

NAST-I retrievals, as a

function of pressure and

horizontal scale. For high

levels in the atmosphere

(above 700 hPa) the vari-

ability of water vapor is

low, and hence for all scales

evaluated up to 105 km

the relative humidity difference does not vary from

one sounding to another above about 3% (absolute

RH). At lower levels around 800 hPa the humidity

differences grow to nearer 6% for distances in excess

of 40 km. In the boundary layer the story is totally

different; the weather on this day consisted of a high

pressure system with dry air aloft and a moist bound-

ary layer between the surface and around 875 hPa.

The data from the retrievals in the boundary layer

show that the variability in relative humidity grows

with horizontal distance. It is hoped that IR sounders

like AIRS and future sounders like IASI and CrIS

will be able to allow retrieval of relative humidity

to within 10% over a 1-km-thick layer. This analysis

suggests that to evaluate a retrieval to this level of

accuracy will require in situ truth measurements

from within ~10 km of the retrieval. Clearly, the

horizontal scales of importance will differ from day

to day with the prevailing meteorological conditions

and also with the horizontal averaging of the remote

sensing instrument. However, this analysis is indica-

tive of the importance of gaining high-quality and

spatially correlated in situ data, as were available

during the EAQUATE campaign.

Figure 15 shows the nine dropsonde profiles of

temperature (red lines), the AIRS retrievals from the

six pixels in the geographic region of Fig. 13 (blue

lines), and the 208 retrievals of temperature from

NAST-I (black crosses). The AIRS retrievals are

version-3 level-2 product 3 × 3 FOV averages with a

spatial scale of 45 km. Figure 16 shows the profiles

of relative humidity in the same way. Several features

FIG. 16. Profiles of relative humidity from the nine dropsondes (red), the six AIRS pixels in the region (blue), and the 208 NAST-I retriev-als (black crosses).

FIG. 15. Profiles of temperature from the nine dropsondes (red), the six AIRS pixels in the region (blue), and the 208 NAST-I retrievals (black crosses).


are apparent: the NAST-I retrievals of temperature

differ from the dropsondes the most around the 750–

800-hPa area, which is where the relative humidity is

changing most rapidly. Furthermore, the NAST-I and

AIRS retrievals of relative humidity are generally too

moist in the boundary layer.

Figure 17 shows the difference in temperature

retrievals while Fig. 18 shows the difference in relative

humidity. The differences are defined as the retrieval

minus the sonde. Prior to computing the difference

the profile of each dropsonde was resampled to the

respective pressure levels of the two retrievals (AIRS

and NAST-I). This was done by taking the average

of the sonde data over the respective

model layers. The NAST-I retrieval

product is defined at 32 levels in the

troposphere below the altitude of the

BAe146 (which launched the drop-

sondes), whereas the AIRS retrieval

is only defined at seven levels in this

region. The levels at which the AIRS

profile is defined are coarse, and

therefore the comparison of “aver-

age” relative humidity across a sharp

inversion that straddles two layers is

always going to be difficult.

Considering first the temperature

differences, the difference between

the average of the six AIRS retrievals

and the average of the nine drop-

sondes (dashed line) shows that the

AIRS retrievals are generally within

+/–1 K, except around 500 hPa where

the dropsondes clearly show signifi-

cant variation. The RMS difference

computed from all AIRS-minus-

sonde combinations (54 in total) is

shown as a dashed–dotted line, and

this shows that the RMS difference is

within 1 K, except around 500 hPa.

The differences between the NAST-I

and sonde profiles are presented at a

higher vertical resolution. The solid

line is the difference between the

average difference of all NAST-I

retrievals and the average of all the

sondes, while the dotted line is the

average difference of the NAST-I

retrieval and its closest sonde. The

agreement between NAST-I and

the sonde is generally within 1 K;

the exception is around 800 hPa,

where the difference is around 1.8 K;

interestingly, this occurs in a region of maximum

change in relative humidity.

The relative humidity difference (Fig. 18) is pre-

sented with the same line styles as the temperature

difference figure. The AIRS retrievals are within

10% of the dropsonde, except for the single point

at 850 hPa, which corresponds to the top of a sharp

inversion and is indicative of the inability to resolve

such sharp features. The RMS difference is also

within 10%, with the exception of the point at the

boundary transition. The NAST-I retrievals of relative

humidity are in better agreement with the dropsonde

data, but once again perform worst at the boundary

FIG. 18. Profile of relative humidity difference (retrieval minus sonde). The solid line is the difference between the average of all NAST-I retrievals and the average of all nine dropsondes. The dotted line is the average difference between sonde and NAST-I retrieval where each retrieval has been differenced from its closest sonde in distance. The dashed line is the difference between the average of all six AIRS retrievals and the average of all sondes. The dashed dotted line is the RMS difference between AIRS and the sondes.

FIG. 17. Profile of temperature difference (retrieval minus sonde). The solid line is the difference between the average of all NAST-I retrievals and the average of all nine dropsondes. The dotted line is the average difference between sonde and the NAST-I retrieval where each retrieval has been differenced from its closest sonde in distance. The dashed line is the differ-ence between the average of all six AIRS retrievals and the average of all sondes. The dashed dotted line is the RMS difference between AIRS and the sondes.

216 FEBRUARY 2008|

layer transition. Although only a single case study,

these results clearly show the ability of IR sounders

like AIRS and NAST-I to retrieve profile information

of temperature and relative humidity at the 1 K and

10% RH level.

CONCLUSIONS. The EAQUATE campaign

brought together an extensive range of airborne and

ground-based observing systems to provide valida-

tion of AIRS level 1 and 2 products. The campaign

focused on the provision of high-quality radiance

spectra from the NAST-I and S-HIS interferometers

on the Proteus aircraft, backed up with detailed

measurements of the tropospheric temperature and

water vapor structure. During the Italian phase of the

campaign a network of ground-based lidar systems

provided the detail of the tropospheric water vapor

structure at high temporal resolution. During the

U.K. phase of the campaign detailed measurements

of the atmosphere over the marine environment were

provided by the extensively equipped FAAM BAe146-

301 aircraft; a critical element of this was the use of

multiple dropsondes.

The radiance performance of satellite instruments

like AIRS can be evaluated through direct compari-

son with higher spectral resolution instruments like

NAST-I and S-HIS, which are resampled spectrally

and spatially to the AIRS specification. Analysis here

shows excellent agreement between the instruments,

with the mean difference between NAST-I and AIRS

being <0.15 K. During the U.K. phase of the campaign

the airborne interferometers on the BAe146-301 and

Proteus were directly compared during a wingtip-to-

wingtip intercomparison run, which gives confidence

in the radiometric performance of the airborne

interferometers.

Working from this position of confidence in

the direct radiometric performance of the various

instruments, the data gathered during EAQUATE

have been used to evaluate the range of level-2

products derived from AIRS. This has been done

through direct comparison with similar level-2 prod-

ucts from the airborne interferometers (NAST-I and

S-HIS) and through comparison with lidar and drop-

sonde profiles. Comparison of retrievals during the

Italian and U.K. phases of EAQUATE have shown that

the 1-K and 10% accuracy required for temperature

and humidity is mostly being met by AIRS.

In conducting this type of analysis there are many

issues associated with the representativeness of the

various data types that should be considered. Direct

radiance validation is fairly straightforward provided

that the instrument artifacts, like instrument line

shape, are accounted for correctly. Comparison of

retrievals is more complex because of the fact that a

retrieved product is reported at pressure levels (which

are a function of the forward radiative transfer model

used by the researcher), which by definition represent

intercomparison in the framework of the EARLINET

project. 2. Aerosol backscatter algorithms. Appl.

Opt., 43, 977–989.

Bösenberg, J., and Coauthors, 2001: EARLINET:

A European Aerosol Research Lidar Network.

Advances in Laser Remote Sensing, A. Dabas, C. Loth,

and J. Pelon, Eds., École Polytechnique, 155–158.

Canas, A. A., J. E. Murray, and J. E. Harries, 1997:

The tropospheric airborne fourier transform

spectrometer. SPIE, 3220, 91–102.

Carissimo, A., I. De Feis, and C. Serio, 2005: The

physical retrieval methodology for IASI: The δ-IASI

code. Environ. Modell. Software, 20, 1111–1126.

Cornacchia, C., A. Amodeo, L. Mona, M. Pandolfi,

G. Pappalardo, and V. Cuomo, 2004: The IMAA

Raman lidar system for water vapor measurements.

Proc., 22nd Int. Laser Radar Conf., Matera, Italy,

ICLAS, ESA SP-561, 107–110.

Cuomo, V., A. Amodeo, C. Cornacchia, L. Mona,

M. Pandolfi, and G. Pappalardo, 2004: Envisat

validation campaign at IMAA-CNR. Proc. Second

Validation Workshop on the Atmospheric Chemistry

Validation of ENVISAT (ACVE-2), Frascati, Italy,

ESA, CD-ROM, ESA SP-562.

Di Girolamo, P., R. Marchese, D. N. Whiteman, and B. B.

Demoz, 2004: Rotational Raman lidar measurements

of atmospheric temperature in the UV. Geophys. Res.

Lett., 31, L01106, doi:10.1029/2003GL018342.

—, A. Behrendt, and V. Wulfmeyer, 2006: Spaceborne

profiling of atmospheric temperature and particle

extinction with pure rotational raman lidar and of

relative humidity in combination with differential

absorption lidar: Performance simulations. Appl.

Opt., 45, 2474–2494.

Fiedler, L., and S. M. Newman, 2005: Correction of

detector nonlinearity in Fourier transform spectros-

copy with a low temperature blackbody. Appl. Opt.,

44, 5332–5340.

Liu, X., J.-L. Moncet, D. K. Zhou, and W. L. Smith,

2003: A fast and accurate forward model for NAST-I

instrument. Optical Remote Sensing, OSA Technical

Digest, OMB2, 16.

Masiello, G, Greco, G., Serio, C., and Cuomo, V, 2006:

Canopy emissivity characterization from hyper-

spectral infrared observations. Preprints, 12th Conf.

Atmospheric Radiation, Madison, WI, Amer. Meteor.

Soc., 5.2.

Matthias, V., and Coauthors, 2004: Aerosol lidar

intercomparison in the framework of the EARLINET

project. 1. Instruments. Appl. Opt., 43, 961–976.

McGrath, A., and T. Hewison, 2001: Measuring the accuracy

of MARSS—An Airborne Microwave Radiometer. J.

Atmos. Oceanic Technol., 18, 2003–2012.


information over a finite pressure interval arising

from the weighting function of the spectral channels.

In contrast, dropsonde or radiosonde data are point

measurements. Great care needs to be used therefore

in describing differences between disparate types of

measurement, and the reader needs to be aware of

these issues when drawing conclusions.

The EAQUATE campaign has shown the critical

importance of coordinated measurements by a range

of observing systems. Over the marine environ-

ment coordinated operations between high-altitude

and tropospheric aircraft are critical in satellite

calibration/validation. In particular, because of the

high variability of water vapor in the atmosphere, a

high density of collocated dropsondes within 10 km

of the retrieval footprint are required for a useful

assessment to be made of the retrieval performance.

IASI was launched in October 2006 on the

METOP platform, and this will be followed by CrIS

on the NPP/NPOESS platform in 2009. Calibration

and validation of these instruments will be vital in

order that the maximum benefit can be realized for

numerical weather prediction. A major calibration/

validation of IASI the Joint AirBorne IASI Validation

Experiment (JAIVEX) took place in April–May

2007 using the FAAM BAe146-301 aircraft and the

NASA WB-57 aircraft carrying the same range of

instruments. This campaign included measurements

over the remote marine environment and over the

Atmospheric Radiation Measurement Great Plains

site in Oklahoma. Lessons learned during EAQUATE

will allow for the optimization of data gathering and

analysis.

ACKNOWLEDGMENTS. The FAAM BAe146-301 is

jointly funded by the Met Office and the Natural Environ-

ment Research Council. The NAST-I program is supported

by the NPOESS IPO, NASA Headquarters, and NASA

Langley Research Center.

REFERENCESAmato, U., G. Masiello, C. Serio, and M. Viaggiano,

2002: The σ-IASI code for the calculation of infrared

atmospheric radiance and its derivatives. Environ.

Modell. Software, 17, 651–667.

Blackwell, W. J., J. W. Barrett, F. W. Chen, R. V. Leslie,

P. W. Rosenkranz, M. J. Schwartz, and D. H. Staelin,

2001: NPOESS Aircraft Sounder Testbed-Microwave

(NAST-M): Instrument description and initial

flight results. IEEE Trans. Geosci. Remote Sens., 31,

2444–2453.

Böckmann, C., and Coauthors, 2004: Aerosol lidar

218 FEBRUARY 2008|

Moncet, J. L ., and Coauthors, 2003: Algorithm

Theoretical Basis Document (ATBD) for the Cross

Track Infrared Sounder (CrIS) environmental data

records (EDR), V1.2.3. AER Document P882-TR-E-

1.2.3-ATBD-03-01, 157 pp. [Available online at eic.

ipo.noaa.gov/IPOarchive/SCI/atbd/cris_atbd_03_

09_01.pdf.]

Pappalardo, G., and Coauthors, 2004: Aerosol lidar in-

tercomparison in the framework of the EARLINET

project. 3. Raman lidar algorithm for aerosol

extinction, backscatter, and lidar ratio. Appl. Opt.,

43, 5370–5385.

Serio, C., G. Grieco, and G. Masiello, 2007: Capability

of high spectral resolution observations in the in-

frared to detect water vapor structures. Technical

Digest Fourier Transform Spectroscopy (FTS)

Hyperspectral Imaging and Sounding of the

Environment (HISE), Optical Society of America,

CD-ROM, HThB2.

Smith, W. L., D. K. Zhou, A. M. Larar, S. A. Mango,

H. B. Howell, R. O. Knuteson, H. E. Revercomb,

and W. L. Smith Jr., 2005: The NPOESS Airborne

Sounding Testbed Interferometer—Remotely sensed

surface and atmospheric conditions during CLAMS.

J. Atmos. Sci., 62, 1117–1133.

Strow, L. L., S. E. Hannon, M. Weiler, and K. Overoye,

2003: Prelaunch and in-flight radiometric calibration

of the Atmospheric Infrared Sounder (AIRS). IEEE

Trans. Geosci. Remote Sens., 41, 274–286.

Taylor, J. P., S. M. Newman, T. J. Hewison, and

A. McGrath, 2003: Water vapour line and continuum

absorption in the thermal infrared—Reconciling

models and observations. Quart. J. Roy. Meteor. Soc.,

129, 2949–2969.

Tobin, D. C., H. E. Revercomb, C. C. Moeller, and T. S.

Pagano, 2006: Use of Atmospheric Infrared Sounder

high-spectral resolution spectra to assess the calibra-

tion of Moderate resolution Imaging Spectroradi-

ometer on EOS Aqua. J. Geophys. Res., 111, D09S05,

doi:10.1029/2005JD006095.

Vanek, M. D., 2001: Far-infrared sensor for cirrus

(FIRSC): An aircraft-based Fourier-transform spec-

trometer to measure cloud radiance. Appl. Opt., 40,

2169–2176.

Zhou, D. K., and Coauthors, 2002: Thermodynamic

product retrieval methodology for NAST-I and

validation. Appl. Opt., 41, 6957–6967.

—, and Coauthors, 2006: AIRS retrieval validation

during the EAQUATE. Proc. SPIE, 6362, 636 224.

Hyperspectral Atmospheric Sounding Validation

Documents

Transcript of Hyperspectral Atmospheric Sounding Validation