Kim Y.J. (Ed.) Advanced Environmental Monitoring

Подождите немного. Документ загружается.

212 S. Buteau et al.

Fig. 16.3 Normalized fluorescence signatures extracted from the numerous acquisition per-

formed by SINBAHD at JBSDS Demo II trial, 2005

sensor- dependent processed signatures, which can be considered as representing, to

a different degree depending on the signal/noise (S/N) ratio, the specific LIF for

these materials as measured by SINBAHD.

The fog oil and yellow smoke signatures are the most easily discernable among

the ten clear signatures. In addition to having clear distinct spectral features, their

fluorescence signal levels were much higher than the others (high S/N ratio). This

implies that these interferants would be easily categorized by a classifier algorithm

but could however, interfere with bioagent detection due to their relatively high

signal intensity. Signatures of the vegetative cells EH and irradiated Yp have quite

similar spectral features, which shows how difficult the discrimination may be

between different materials of the same type. The last six signatures show

discrepancies and differences between them to various degrees. The first interesting

fact that must be underlined from these results is the obvious spectral differences

between killed-Ba and BG, which are two species of Bacillus bacteria spores.

In contrast to the vegetative cells studied, the two Bacillus species can easily be

discriminated even though they are both bacteria spores. Note that the killing

processes utilised for both the Ba spores and the YP vegetative cell (killed vaccine

strain) probably had observable effects on their LIF signatures. Ovalbumin (OV)

and MS2, which are respectively toxin and virus simulants, have spectral features

much closer to killed-Ba than BG has. Discrimination between killed-Ba, MS2 and

OV may be challenging.

16 Bioaerosol Standoff Monitoring Using Intensified 213

16.6 Data Processing

Once the spectral signatures of all different materials have been obtained, the

MA can be performed on trial data acquired during the open-air crosswind

releases. The MA will evaluate the amplitudes of some pre-selected reference

signatures in order to obtain the best possible fit, in a least squares sense, of

the acquired spectra. Comparing the amplitudes obtained for different reference

spectra permits the classification of the detected signal, up to a certain point.

However, this is behind the scope of the present work. Depending on the

specificity of the collected spectra and the ‘goodness’ of the reference spectra,

the degree of confidence of this classification process may vary. In bio-medical

science, using the term ‘classification’ as compared to ‘identification’ convey

different meanings. The latter term implies that very specific microbiological and

biochemical techniques have been followed to obtain taxonomic genus and

species information.

The MA results obtained can then be converted into equivalent concentra-

tions to obtain the corresponding detection limit. Since SINBAHD did not

participate in the absolute measurement portion of the trial (sABT phase), its

MA results will have to be correlated to the calibrated concentration data

measured by the Dugway west desert reference LIDAR (WDL). This latter

process will hence facilitate the evaluation of SINBAHD concentration detection

limit for the different released materials. This process is limited in efficiency

due mainly to the spatial correlation between the WDL and SINBAHD collected

data. Indeed, they did not have the exact same line of sight through the target

cloud. In spite of that, valid linear correlation between the two systems should

be able to be obtained.

Figure 16.4 presents the WDL and SINBAHD MA results for a release of

20 g of killed Bacillus anthracis (BA) at about 1.2 km from sensors location on

June 13, 2005. The WDL concentration length (solid line) was obtained by

integrating the concentration over the region where the cloud appeared and

multiplying by the length of a bin. SINBAHD MA results presented in Fig. 16.4

include the background (triangles) and killed-Ba (dashed line) signature

amplitude. The concentration detection threshold, namely 4sigma, corresponds

to four times the standard deviation of the studied material signature amplitude

during the non-presence of this material. This 4sigma directly provides the

SINBAHD detection limit for that particular material at a given range. Once the

correlation between WDL and SINBAHD MA results is performed and the off-

signal standard deviation is calculated, the detection limit can then be

determined. In this particular case, for a cloud of 10 m along the LIDAR line of

sight at 1.2 km, the detection limit in cloud concentration would be of 17 kppl

(kppl = 1 × 10

particles per litre of air). One interesting point in this result is

the ability to detect the LIF signature (dashed line) at lower signal levels than

the background (full triangles). This result demonstrates the efficiency of the

MA to exploit the acquired data even at detected signal levels lower than the

background levels.

214 S. Buteau et al.

16.7 Conclusion

SINBAHD participation within the JBSDS Demo II field trial has provided further

opportunities to explore its detection potential and also to obtain new spectral

signatures in the pattern set (library). SINBAHD participated in 34 test releases for

crosswind open air trials. From these acquisitions, 19 signatures could be extracted,

nine did not have any fluorescence signal and one was lost due to a computer failure.

The five other acquisitions, from which no fluorescence signal could be visually

observed in real time, were processed with SINBAHD multivariate analysis. In

these five cases, a detection signal could be extracted but with a fairly low S/N ratio.

Signature differentiation was fairly obvious in most cases but not between the two

vegetative cells, irradiated-Yp and EH. Indeed, the killed vaccine strain of Yersinia

pestis (Yp) had about the same signature as Erwinia herbicola (EH), even though

the first had been irradiated but not the latter. Spectral differences between

irradiated-Ba and BG were noted (the two species of Bacillus bacteria spores). The

signature of OV and MS2 showed spectral features much closer to killed-Ba. The

background fluorescence was also properly characterized in order to optimise

the MA results. The background fluorescence signature has a fairly stable level over

an acquisition trial, and a quite stable spectral shape for different acquisition gate

widths. All the SINBAHD results, once processed by the MA, were correlated with

the data of DPG west desert LIDAR (WDL) referee system. From these correlations,

the 4sigma sensitivity could be evaluated for a given cloud depth and range.

−0.25

0.25

0.75

1.25

1.75

2.25

2.75

144 146 148 150 152 154 156 158 160

Time (minutes after midnight)

Concentration Length *10E+06

(ppl*m)

−500

500

1500

2500

3500

4500

5500

Normalized signature amplitude

(counts)

WDL referee data

SINBAHD: killed-BA

SINBAHD: background

off-signal std = 85

4sigma = 4*85ct *2550 Kppl*m/5100ct

detection limit = 170 Kppl*m

Fig. 16.4 WDL calibrated concentration (solid line) and SINBAHD (600 m gate width, around

10 m cloud depth) MA results for a release of killed-Ba at about 1.2 km (TDA015)

16 Bioaerosol Standoff Monitoring Using Intensified 215

Acknowledgements We are grateful to the Biological Standoff Detection Field Demo trial team

and to DPG personnel for their support throughout the trial. Special thanks to John Strawbridge,

JPM Bio Defence, to have given us the opportunity to attend this trial.

References

Bronson R. (1989), Matrix Operations, Schaum’s Outline Series, McGraw-Hill, New York.

Buteau S., Simard J.-R., and Roy G. (2005, October), Standoff detection of natural bioaerosol by

range-gated laser-induced fluorescence spectroscopy. (Paper presented at Optics East SPIE

conference—Chemical and Biological Standoff Detection III, #5995, Boston, USA).

Buteau S., Simard J.R., McFee J., Ho J., Lahaie P., Roy G., and Mathieu P. (2006), Joint Biological

Standoff Detection System Increment II:Field Demonstration—SINBAHD performances.

DRDC Valcartier TM2006-140, 83, UNCLASSIFED.

Chen C.L., Heglund D.L., Ray M.D., Harder D., Dobert R., Leung K.P., Wu M.,and Sedlacek III

A.J. (1997), Application of Resonance Raman Lidar for Chemical Species Identification, SPIE

Vol. 3065, pp. 279–285.

Condatore L.A.Jr., Guthrie R.B., Bradshaw B.J., Logan K.E., Lingvay L.S., Smith T.H.,

Kaffenberger T.S., Jezek B.W., Cannaliato V.J., Ginley W.J., and Hungate W.S. (1998,

October), U.S. Army Chemical and Biological Defense Command Counter-Proliferatioin

Long-Range Biological Standoff Detection System, (paper presented at the 4th Joint Workshop

on Standoff Detection for Chemical and Biological Defense).

Evans B.T.N., Roy G., and Ho J. (1989), The Detection and Mapping of Biological Simulants,

Part 2: Preliminary Lidar Results, DREV R-4480/89, UNCLASSIFIED.

Faris G.W., Copeland R.A., Mortelmans K., and Bronk B.V. (1997), Spectrally resolved absolute

fluorescence cross sections for bacillus spores, Appl. Opt., 36(4), 958–967.

Fell N.F.Jr., Pinnick R.G., Hill S.C., Videen G., Niles S., Chang R.K., Holler S., Pan Y., Bottiger

J.R. and Bronk B.V. (1998, November), Concentration, Size, and Excitation Power Effects on

Fluorescence from Microdroplets and Microparticles Containing Tryptophan and Bacteria,

(paper presented at the Conference on Air Monitoring and Detection of Chemical and

Biological Agents, Boston). SPIE Vol. 3533, pp. 52–63.

Gelbwachs J. and Birnbaum M. (1973), Fluorescence of atmospheric aerosols and lidar implica-

tions, Appl. Opt., 12(10), 2442–2447.

Grant W. B. (1977), Optical Bases for Remote Biological Aerosol Detection, Edgewood Arsenal

Contractor Report No. ED-CR-77025, Standford Research Institute.

Hairston P.P., Ho J.and Quant F.R. (1997), Design of an instrument for real-time detection of

bioaerosols using simultaneous measurement of particle aerodynamic size and intrinsic fluo-

rescence, J. Aerosol Sci., 28(3), 471–482.

Hargis P.J. Jr., Lang A.R., Schmitt R.L., Henson T.D., Daniels J.W., Jordan J.D., Shcroder K.L.

and Shokair I.R. (1998, October), Sandia Multispectral Airborne Lidar for UAV Deployment,

(paper presented at the 4th Joint Workshop on Standoff Detection for Chemical and Biological

Defense).

Ho J., Spence M., and Hairston P. (1999), Measurement of biological aerosol with a fluorescent

aerodynamic particle sizer (FLAPS): Correlation of optical data with biological data,

Aerobiologia, 15, 281–291.

Houston J.D. and Evans B.T.N. (1987), Monitoring Aerosols Optically, Photo. Spec., 93.

Hutt D.L., Bissonnette L.R., and Durand L. (1994), Multiple field of view lidar returns from

atmospheric aerosols, Appl. Opt., 33, 2338–2348.

Measures R.M. (1984), Laser Remote Sensing: Fundamentals and Applications, Wiley Interscience,

New York.

Press W.H., TeukolskyS.A., Vetterling W.T. and Flannery B.P. (1992), Numerical Recipes in C:

The Art of Scientific Computing, Second Edition, Cambridge University Press, Cambridge.

216 S. Buteau et al.

Roy G., Bissonnette L., Bastille C., and Vallée G. (1999), Retrieval of droplet-size density

distribution from multiple-field-of-view cross-polarized lidar signals: Theory and experimen-

tal validation, Appl. Opt., 38(24), 5202–5211.

Simard J.-R., Mathieu P., Theriault J.-M., Larochelle V., Roy G., McFee J.E., and Ho J. (1999).

Active Range-Gated Spectrometric Standoff Detection and Characterisation of Bioaerosols:

Preliminary Analysis, DREV TM1999–056, 34 pp, UNCLASSIFED.

Simard J.-R., Roy G., Mathieu P., Larochelle V., McFee J.E. and Ho J. (2003). Standoff Integrated

Bioaerosol Active Hyperspectral Detection (SINBAHD): Final report, DREV TR 2002–125,

118 pp, UNCLASSIFED.

Simard J.R., Roy G., Mathieu P., Larochelle V., McFee J.E. and Ho J. (2004), Standoff sensing of

bioaerosols using intensified range-gated spectral analysis of laser-induced fluorescence. IEEE

Trans. Geosci. Remote Sens., 42(4), 865–874.

Chapter 17

MODIS 500 × 500-m

Resolution Aerosol

Optical Thickness Retrieval and Its Application

for Air Quality Monitoring

Kwon H. Lee

, Dong H. Lee

, Young J. Kim

, and Jhoon Kim

Abstract Atmospheric aerosol optical thickness (AOT) was retrieved over urban

areas using moderate resolution imaging spectroradiometer (MODIS) 500-m

resolution calibrated radiance data. A modiﬁ ed Bremen Aerosol Retrieval (BAER)

algorithm was used to retrieve AOT over Seoul, Korea. Since the surface reﬂ ectance

of an urban area is typically brighter than that of vegetation or soil areas in the

visible wavelength region, the error associated with aerosol retrieval is higher.

Surface reﬂ ectance determination using the linear mixing model (LMM) produced

values that were smaller (∼0.02) than those obtained using the minimum reﬂ ectance

technique (MRT). This difference would lead to an overestimated AOT when using

the LMM approach. Retrieved AOT data using MRT and standard MODIS 10-km

aerosol products (MOD04) were compared to Rotating Shadow-band Radiometer

(RSR) observations at Yonsei University, Seoul, Korea. Regression analysis

showed that the root mean square error (RMSE) of MODIS AOT and MOD04

AOT was 0.05 in both cases. In addition, MODIS AOT data were compared

with ground-based particulate matter data (PM10) from air quality monitoring

networks of the National Institute of Environmental Research (NIER). MODIS

AOT data showed a relatively low correlation (r = 0.41) with surface PM10 mass

concentration data due to differences in ground-based and column-averaged data,

variability of terrain, and MODIS cloud mask. This result is similar to that of other

investigations. The application of ﬁ ne-resolution satellite data supports the feasibility

of local and urban-scale air quality monitoring.

Keywords: AOT, LMM, MODIS, MRT, PM10

Advanced Environmental Monitoring Research Center (ADEMRC), Gwangju Institute of Science

& Technology (GIST), 1 Oryong-dong, Buk-gu, Gwangju 500-712, Republic of Korea

Department of Atmospheric Sciences, Yonsei University, Shinchon-dong 134, Seodaemun-gu,

Seoul 120-749, Republic of Korea

*Corresponding author: Tel: + 82-62-970-3401, Fax: + 82-62-970-3404

217

Y.J. Kim and U. Platt (eds.), Advanced Environmental Monitoring,

217–230.

218 K.H. Lee et al.

17.1 Introduction

Satellite remote sensing has provided quantitative information on aerosols with an

accuracy comparable to that of surface measurements. The large-scale distribution

of aerosol concentration and characteristics, aerosol radiative forcing, and property

changes of clouds interacting with aerosols are among the important observations

that can be provided by satellite remote sensing. Due to the growing recognition of

the importance of aerosol properties for studies of climate and global change, it is

indeed fortunate that a number of very significant and greatly enhanced satellite

systems are being developed for launch in the next few years. These systems will

also provide information on the global distribution, including seasonal and interannual

variation, of aerosol sources (forest fires, desert dust and aerosol from the oxidation

of SO

emissions from industrial regions), aerosol loading and optical properties,

as well as direct and indirect radiative forcing.

Satellite measurements can also be inverted to yield information on aerosol

optical thickness (AOT), angular scattering properties, and size distribution.

Satellite measurements clearly have the advantage of being the only set of meas-

urements that provide a wide coverage. The disadvantage of the presently used

passive sensors is that they can only be applied under clear-sky conditions, and

their application over land is difficult. Furthermore, a number of additional

assumptions have to be made regarding atmospheric conditions. In spite of these

disadvantages, many researchers have studied global and local-scale aerosol

remote sensing using satellites. Techniques associated with satellite observation

in the field of atmospheric science are advancing rapidly (Kaufman et al. 1997a;

Gordon et al. 1997; King et al. 1999). Total ozone mapping spectrometer (TOMS)

measurements are designed to monitor ozone and provide valuable information

on aerosol geographical distributions (Herman et al. 1997). Such measurements

were used to detect an increase in biomass burning smoke in African savanna

regions during the 1990s (Hsu et al. 1999). The Sea-viewing Wide Field-of-view

Sensor (SeaWiFS), developed to provide ocean color data products for the study

of marine biogeochemical processes, produces an aerosol data product from its

atmospheric correction algorithm (Gordon and Wang 1994) through a separate

algorithm (von Hoyningen-Huene et al. 2003), and has been used to investigate

transport of Asian dust plumes (Lee et al. 2004). The launch of polarization and

directionality of the earth’s reflectance (POLDER) on ADEOS II adds greater

capabilities (Leroy et al. 1997). Extracting the full potential from the satellite data

requires correlative suborbital data to not only validate the satellite measurements

(Clark et al. 1997; Fraser et al. 1997; Vermote et al. 1997), but to supply information

not obtainable from space. The EOS/Terra and EOS/AQUA satellites and their

new instruments, such as the moderate resolution imaging spectroradiometer

(MODIS) and multi-angle imaging spectroradiometer (MISR), provide improved

measurement capabilities and unprecedented volumes of data regarding the

atmosphere, land, ocean, and radiative processes (Tanré et al. 1997; Wanner

et al. 1997).

17 Resolution Aerosol Optical Thickness Retrieval 219

AOT is a key property used in the characterization of atmospheric aerosols

from satellite data. The radiant energy obtained from an earth-observing satellite

sensor is converted into reflectance or apparent reflectance using information

about the atmospheric condition along the path between the ground and the satellite,

as well as the viewing geometry and solar geometry. Aerosols also block the view

for systems observing changes in land use. Basic aerosol retrieval techniques are

therefore employed to remove the aerosol signal, and can also be used to highlight

or enhance the aerosol-only signal. King et al. (1999) summarized the major

retrieval methods well.

Aerosol retrieval from remote sensing data has many important applications,

including a role in climate studies and the atmospheric correction of satellite

imagery. Remote sensing of atmospheric aerosol using MODIS satellite data

has been shown to be very useful in global/regional-scale aerosol monitoring.

Recently, satellite-retrieved AOT has also been used to estimate local/regional air

quality (Engel-Cox et al. 2004a,b; Hutchison 2003; Hutchison et al. 2004). Due to

the large spatial resolution of 10 × 10-km

for the MODIS standard aerosol

product (MOD04), AOT data have limitations for local/urban-scale air quality

monitoring applications. Its large spatial resolution is incapable of reflecting various

types of urban-scale (Seoul ∼ 600 km

) pollution with a complicated topography.

Therefore, a modified Bremen Aerosol Retrieval (BAER) algorithm (von

Hoyningen-Huene et al. 2003) as detailed by Lee et al. (2005) was used in this

study to retrieve AOT for fine resolutions of 500 × 500 m

over Seoul, Korea.

Figure 17.1 shows the study area.

Fig. 17.1 Map of the Korean peninsular and the Seoul metropolitan area. Symbols on the right

represent the 27 air quality monitoring stations (asterisks) and the radiometer measurement site

(large star)

220 K.H. Lee et al.

17.2 Methodology

MODIS is a multispectral (36 bands), multi-resolution (1 km, 500 m, 250 m)

sensor dedicated to the observation of the Earth. It is a key instrument aboard

the Terra (EOS AM) and Aqua (EOS PM) satellites. Terra’s orbit around the Earth

is timed so that it passes from north to south across the equator in the morning,

while Aqua passes south to north over the equator in the afternoon. Two bands are

imaged at a nominal resolution of 250 m at nadir, with five bands at 500 m and

the remaining 29 bands at 1 km. A ±55-degree scanning pattern at the EOS orbit

of 705 km achieves a 2,330-km swath and provides global coverage every 1–2

days. Global/regional-scale aerosol distribution information has been provided

by MODIS measurements since the launch of the satellites. However, the operational

MODIS AOT from the NASA algorithm is not appropriate for local/urban-scale

aerosol monitoring as it has a spatial resolution of 10 × 10-km

. Therefore, we

used TERRA/MODIS 500-m resolution data (0.47, 0.55, 0.66 µm) for aerosol

retrieval over an urban area for 2004 in this study. Figure 17.2 briefly describes the

aerosol retrieval processes using MODIS 500-m resolution data.

The aerosol retrieval algorithm for MODIS 500-m resolution data is a modified

version of BAER, which retrieves aerosol properties in greater detail and can be

applied to other sensors with relative ease. In order to retrieve AOT, a radiative trans-

fer equation was introduced to determine aerosol reflectance by decomposing the

top-of-atmosphere (TOA) reflectance into surface reflectance and Rayleigh path radi-

ance (Kaufman et al. 1997a). The TOA reflectance ρ

TOA

(θ

, θ

, φ) is expressed as

rqqfr qqftt qv

TOA S ATM S Aer Ray

Tot To

(,,) (,,, , ,(), )

()

00 0

⋅

ttS Surf S

Surf S Hem Tot

() (,)

(, ) ( ,)

qr qq

rqq t

⋅

−⋅

(1)

where q

is the solar zenith angle, q

the satellite zenith angle, and f the azimuth

angle. τ

Aer

, τ

Ray

and τ

Tot

represent aerosol, Rayleigh, and total optical thickness,

respectively. p(θ) is the phase function, ϖ

a single-scattering albedo, g the

asymmetry parameter, ρ

ATM

the atmospheric path reflectance, T

Tot

) the total

transmittance, ρ

Surf

, m

) the surface reflectance, and r

Hem

(τ

Tot

, g) the hemispheric

reflectance. The atmospheric contribution consists of the Rayleigh and aerosol

fractions. The Rayleigh path reflectance can be determined through the use of a

relevant wavelength method (Bucholtz 1995). The Rayleigh optical thickness

(τ

Ray

(λ)) can then be determined from the surface pressure p(z) in each pixel

through the following equation:

tl l

Ray

()

=⋅ ⋅

−+⋅+

⎛

⎝

⎜

⎞

⎠

⎟

(2)

where A, B, C and D are the constants for the total Rayleigh scattering cross-section

and total Rayleigh volume scattering coefficient for a standard atmosphere. p

is the

mean sea level surface pressure.

17 Resolution Aerosol Optical Thickness Retrieval 221

There exists an inherent problem in determining surface reflectance for aerosol

retrieval, since photons reflected by the surface reach the satellite after they transmit

through the atmosphere. However, aerosol retrieval over a bright surface such as

an urban area is problematic because of the large uncertainties concerning the

contribution of surface reflectance to satellite data. Although Kaufman et al. (1997c)

used the 2.1-um channel reflectance for estimation of surface reflectance at visible

wavelengths, it is mainly suitable over vegetation or dark soils. The original

BAER algorithm used a linear mixing model (LMM) between vegetation and soil

spectra for surface reflectance determination in a pixel. Surface reflectance

Surf

) can be calculated by portion of vegetation and bare soil spectra in the

following manner:

Fig. 17.2 Schematic diagram for aerosol retrieval in this study