At first glance, an examination of the bulk refractive indices for the 8–12 μm waveband of various bioaerosols suggests differentiation with respect to common background aerosols based upon the spectral characteristics of the absorption. The question of whether there is a spectral signature of bioaerosol clouds when those clouds are immersed in a typical atmosphere, including the boundary layer background aerosols, has been addressed in a simulation using the Weather and Atmospheric Visualization Effects for Simulation (WAVES) suite of codes. Using measured values of the refractive index for common bacterial spores, and their typical size distributions, the single-scattering, ensemble-averaged optical properties such as extinction/absorption coefficients, albedo, and the scattering phase function was computed for bioaerosol clouds at a resolution of 1 cm-1. WAVES was then used to calculate the radiative transfer for a finite sized cloud immersed in background. Results of this simulation indicate that, for a passive remote sensing measurement, it is unlikely that bioaerosol clouds can be identified from the spectral signature alone.
©2002 Optical Society of America
One of the most pressing scientific problems under scrutiny today is the detection and identification of biological aerosols, particularly harmful biological aerosols. A method by which to sense and detect harmful aerosols is the thrust of many research programs. Most harmful biological material is composed of material very similar to other biological and even non-biological material already in the environment. It is recognized that biological compounds have infrared spectra that indicate specific chemical structures common to biological compounds. Some progress has been made in the sensing and detection of chemicals using spectroscopic and imaging techniques –. These chemicals are often significantly different in composition from other materials normally found in the environment. This difference makes the sensing and detection of chemicals in the environment, a somewhat more tractable problem that the sensing and detection of biological aerosols in the environment. Since the imaging of chemical clouds has been suggested, modeled, and measured, imaging of biological aerosols has also been suggested . Modeling biological aerosol clouds is a natural extension of this research because imaging experiments can be expensive to conduct, imagers are expensive to build, and the release of biological aerosols in the natural environment is restricted.
This paper describes a computational experiment in which the scattering and absorption properties of several biological aerosols are modeled, and then these properties are integrated into overall radiative transfer and propagation model computations. The experiment compares the results of simulation of embedded biological clouds and clouds comprised of different water fogs for various background visibilities and ranges. The impact of the background, boundary–layer atmosphere on the passive spectra of the embedded biological and water fog aerosol clouds is demonstrated. The utility of using imaging spectrometers, operating in the Long–wave infrared (LWIR)(8–12 μm ) is then briefly discussed, based on the computational results of this paper.
1.1 Bulk properties
The bulk refractive index of water was generated from the routine REFWAT which contains a compilation of several measured databases for the temperature dependent refractive index of water. For our purposes, we generated the refractive index for water at a temperature of 305 K for the spectral region 850–1250 cm-1 (roughly, 8–12 μm) in 1.0 cm-1 steps. In this spectral region, the temperature dependence for the refractive indices is negligible. The experimentally determined refractive indices for the bioaerosols B. subtilis and M. luteus used in this study were provided by M. Milham, Edgewood . These indices were also given at a resolution of 1 cm-1. A plot of the real and imaginary components of the complex, bulk refractive indices m = n + ik can be seen in Figure 1. The real portion of the refractive index n has a nominal value of 1.50 for the bioaerosols with a ≾5% variation over the waveband while n for water varies almost linearly from 1.10 to 1.30 over the waveband. The imaginary portion of the refractive index k shows the most spectral differences between the bioaerosols and water. For water, the absorption in the selected spectral region is smoothly varying while both bioaerosols show a broad absorption resonance between 1000–1150cm-1.
1.2 Ensemble properties
The spectral characteristics of the bulk material will, in general, be significantly different than the spectral characteristics of small particles created by an encapsulation of the bulk material. The spectral characteristics are further changed when, for large number of particles, the scattering properties are defined by an ensemble average over a given aerosol size distribution. Atmospheric aerosols, including water fogs and biological aerosols, do not have a single size. Rather, they are found with a distribution of sizes, with each size exhibiting a slightly different wavelength dependence of the optical properties. The ensemble-averaged optical properties, determined by averaging over the size distribution, in general exhibits broader spectral features than may be found in the optical properties of single aerosol particles.
For the case of spherical homogeneous aerosols, the ensemble average optical coefficients are related to their respective optical cross-sections by a Fredholm I integral, namely
where 〈ci (k)〉 is the volumetric extinction, scattering, or absorption coefficient, σi (kr) is the corresponding single–particle cross section, and n(r) is the ensemble particle size distribution. The single–scattering, ensemble–averaged, spectral phase function is similarly defined as
where dσ scat/dΩ is the single–particle, differential scattering cross–section.
For the natural water fogs, we used the models presented by Shettle and Fenn. These models describe the aerosols as spherical particles with radius described by a Modified Gamma (MG) distribution given by
In Eq.’s 1 and 2, N is the number concentration of the aerosol, rc is the modal radius and the parameters α and γ describe the shape of the distribution. For both fog models used in this study, γ was set to 1. For the heavy advection fog model, the mode radius was 10 μm, with a number density of 20 particles/cm3, and α was set to 3. For the radiation fog, the mode radius was 2.0 μm, with a number density of 200 particles/cm3 and α was set to 6. Mie theory was used to describe the individual spherical particle scattering. The ensemble scattering properties were calculated using these size distributions. For the bioaerosols, we modeled the size distribution by a log-normal distribution given by
The same particle distribution was used for both the B. subtilis and M. luteus models; the modal radius was set to 0.5 μm with a width σ of 0.4 and the number concentration of 1 particle/cm3. This corresponds well with experimental values for the average size and typical size variations of bacterial spores . Again, as with the natural fog aerosols, the ensemble average scattering quantities were calculated using Mie theory. This is not entirely a correct procedure as the bioaerosols are in general not smooth, perfect spheres for which Mie theory is valid. However, in the 8–12 μm waveband, it is an accurate approximation as the particle size is much smaller than the typical wavelength in the band. The approximation is further supported by the nature of the ensemble average process which tends to smear out the morphologically-dependent resonances which may occur at specific sizes. In addition, for non–spherical particles, the random angular orientation of the particles will also tend to reduce the resonant structure.
Figure 2 show the results for the ensemble-averaged, single-scattering coefficients (extinction,scattering, and absorption) for the four aerosol types used in the simulation. For the heavy advection fog, Figure 2a, the scattering coefficients are essentially independent of wavenumber from 1000–1250 cm-1 and vary only slightly from 850–1000 cm-1. Considerably more wavenumber dependence can be seen in the scattering coefficients for the radiation fog, Figure 1b. For wavenumbers in the low end of our spectral range, the aerosol is dominated by absorption. Towards the high end of the region the radiation fog aerosol is dominated by scattering coefficients are essentially independent of wavenumber scattering.
The ensemble scattering phase functions for the four aerosol models used are shown in Figure 3. By direct observation of the surface plots we note that the phase functions for all four models are smoothly varying for the entire waveband. Heavy advection fog shows a strong forward scattering peak over the entire waveband. This is expected due to the relatively large size of the aerosol particles in comparison to the typical wavelength in the band. This effect is also seen to a lesser degree in the radiative fog aerosol. The bioaerosols phase function exhibit a nearly symmetric phase function about a scattering angle of which is expected for an aerosol with characteristic size much smaller than the wavelength.
2 3–Dimensional Radiative Transfer
2.1 Physical Setup
Figure 4 shows an overhead plan view of the physical volume used in this simulation. The simulation volume had physical extent of 20 km East–West, 5 km North–South, and 5 km vertical. We described the volume on a grid with spacing of 0.25 km. Spectrally, the calculation was performed from 850–1250 cm-1 in 5 cm-1 increments. We chose the size of the target aerosol clouds to represent what might typically be found in a man-made release. The target aerosol clouds were modeled as cubicle clouds with extent of 0.5km × 0.5km × 0.5km superimposed in the natural background. We define the background as the combination of the boundary-layer aerosol plus atmosphere. For this simulation, we used the Shettle and Fenn urban aerosol model coupled with the 1976 U.S. standard atmosphere model to describe the background. For computation speed, two separate target clouds were placed in the simulation volume. The clouds were placed a sufficient distance from each other that we could treat them as independent. The limiting path radiance, as a function of range from the target cloud, was computed using the general illumination values obtained using the Boundary Layer Illumination and Radiation Balance (BLIRB) model. BLIRB is the radiative transfer module in the U.S. Army Research Laboratories Weather and Atmospheric Visualization Effects for Simulations (WAVES) suite of environmental models . For this study, BLIRB solved for the radiative transport using a standard 8-stream, Discrete-Ordinates method. BLIRB calculates the global illumination which is then transformed to create local source terms for a line–of–sight (LOS) calculation to determine the limiting path radiance. The concentration of the target aerosol clouds were set in proportion to the background using parameter γ defined by
where the bracket denotes the average value over the waveband Γ. Thus, for γ ≈ 1, the average extinction of the aerosol cloud is significantly larger than that of the boundary–layer aerosol + atmosphere. For our selection of a 10km visibility (at 0.55μm) urban aerosol model for our boundary-layer aerosol together with the 1976 U.S. Standard Atmosphere model, we determined the band-averaged extinction coefficient for background to be 〈β e,background〉 = 0.1369km-1. For comparison, the advection and radiative fog models described previously correspond to a visibility of 0.14 km and 0.45 km and give corresponding values of γ = 0.987 and 0.946, respectively. Typically, number concentrations for biological aerosol releases vary appreciably depending on aerosolization technique. We believe a range up to 1000 particles/cm3 to be a reasonable estimate. For a 10km background visibility and cloud number concentration of 1000 particles/cm3, γ ≈ 0.5 for both bioaerosol models. The purpose for using this parameter for scaling the embedded aerosol clouds to background is to adjust for various background aerosol visibility conditions. Scaling in this manner more clearly shows when the path radiance is dominated by the background atmosphere (background–limited).
For each of the LOS’s depicted in Figure 4 through the cloud, as well as an LOS between the clouds to determine the background signature, the aerosol loading and position of the observer from the cloud was varied. For our purposes, we also used a relative position, denoted as R/R 0, where R 0 is the scale dimension of the embedded aerosol cloud. Although this study was considerably larger in detail, we will present the results for a small subset of the results. We will focus on a moderate, high, and extremely high aerosol loading for the cloud in a near and far–field arrangement.
Figure 5 shows the results for the limiting path radiance for the line–of–sight (LOS) through the background only. The contributions to the path radiance are from atmospheric molecular absorption lines (sharp peaks) and from a broader spectral feature arising from the molecular continuum and the boundary–layer aerosol. An important note to keep in mind is that a considerable constituent of the urban boundary–layer aerosol model is the continental aerosol which includes biological aerosols (soil–derived bacteria such as B. subtilis). Also, Figure 5 and Figure 6 show the results for a moderate cloud aerosol loading of γ = 0.5 for the cloud in near–field or far–field, respectively. One obvious observation is the very small differences in spectral path radiance with respect to range. This indicates that the spectral return is background limited. For this relatively moderate loading the small differences in returned spectra would probably not be detectable. In order to see more of the differences, much higher aerosol loading was necessary.
Figures 7 and 8 show the results for a relatively large aerosol loading of γ = 0.97 for near–field and far–field position of the clouds in the LOS, respectively. For this high of a loading with respect to background, we can see a considerable difference in the spectral path radiance. The differences between the cloud being in near vs. far–field is most noticeable in the suppression of the atmospheric molecular lines. The cloud for this case acts as a very bright diffuse source. When the clouds are in near–field, the aerosol clouds act as an absorber for the atmospheric emission that occurs from the portion of LOS on the opposite side of the cloud. The only background emission remaining is due solely to the background located in the near–field. For the far–field case, Figure 8, the intervening atmosphere will act as an absorber of the light emitted by the bright cloud. However, the question of whether there is sufficient differences between the different clouds for a classification of the cloud as “Bio–like” or natural is up for debate. The difficulty with classification is due to the relatively broad spectral features of bioaerosols which are very similar to background signatures. For an extremely high aerosol loading with respect to background, we can see this result more clearly.
Figures 9 and 10 show the results for a scaling factor of γ = 0.99. For this case, the aerosol cloud is essentially the only important contribution to the path radiance. For near–field, Figure 9, the cloud essentially absorbs all of the emitted radiation from the LOS outside of the cloud. Only the radiative fog has a significantly different spectral emission over this waveband. In far–field, Figure 10, the atmosphere between the cloud and observer acts as an absorbing medium. Also, the intervening atmosphere has significantly reduced the difference between the radiative fog cloud and the other bio-clouds and advection fog. For this extremely high aerosol loading, it may be possible to distinguish between “bio–like” clouds vs. natural clouds.
At first glance, it appears to be possible to detect and categorize bioaerosols in the LWIR from natural fogs by comparing their respective bulk optical properties. Unfortunately, these properties are not readily measurable in “real–life” conditions. Those optical properties of the aerosols which are available for passive remote sensing in the LWIR do not have as readily apparent spectral signature difference. Encapsulation of the bulk material to form particles, ensemble–averaging over the size distribution and most importantly embedding the aerosol cloud in the natural background considerable confuses the categorization between the aerosols. The dependence of the spectral signature on both the range of the observer from the cloud, as well as the dependence on the relative concentration indicate that categorization will be extremely difficult without the aid of additional information necessary to remove the background spectral signature.
References and links
1. D.F. Flanigan, “Hazardous Cloud Imaging: a new way of using the passive infrared,” Appl. Opt. 36, 7027–7036 (1997). [CrossRef]
2. D.F. Flanigan, “Hazardous Cloud Imaging: An In-Depth Study,” ERDEC-TR-416 (Clearinghouse for Federal Scientific and Technical Information, Cameron Station, VA., 1997).
5. M.L.G. Althouse and C. Chang, “Chemical vapor detection with a multispectral thermal imager,” Opt. Eng. 30, 1725–1733 (1991). [CrossRef]
6. L. Carr, L. Fletcher, P. Holland, J. Leonelli, D. McPherrin, and M. Althouse, “Characterization of filtered FLIR systems designed for chemical vapor detection and mapping,” in Infrared Imaging Systems: Design, Analysis, Modeling, and Testing, G.C. Holst, ed., Proc. SPIE Vol. 1309, 90–103 (1990). [CrossRef]
7. S.R. Horman, “Remote Identification of CW Agents by Spectral Techniques: Calculations of Cloud Emission in the Infrared,” NSWC–TR–3457, (Clearinghouse for Federal Scientific and Technical Information, Cameron Station, VA., 1976).
8. C.M. Gittins, L.G. Piper, W.T. Rawlins, W.J. Marinelli, J.O. Jensen, and A.N. Akinyemi, “Passive and Active Standoff Infrared Detection of Bio-Aerosols,” Field. Anal. Chem. Tech. 3, 274–282 (1999). [CrossRef]
9. D. Suhre and E. Villa, “Imaging spectroradiometer for the 8-12 micron region with a 2 cm-1 passband acousto-optic tunable filter,” Appl. Opt. 37, 2340–2345 (1998). [CrossRef]
10. D. Segelstein, ”The complex refractive index of water,” M.S. Thesis, University of Missouri- Kansas City (1981);P. Ray, ”Broadband complex refractive indices of ice and water,” Appl. Opt. 11, 1836–1844 (1972). [CrossRef] [PubMed]
11. private communication, for optical constants in the visible to short wave IR see P.S. Tuminello, E.T. Arakawa, B.N. Khare, J.M. Wrobel, M.R. Querry, and M.E. Milham, “Optical properties of Bacillus subtilis spores from 0.2 to 2.5 microns,” Appl. Opt. 36, 2818–2824 (1997). [CrossRef] [PubMed]
12. S. Yabushita and K. Wada, “The Infrared and Ultraviolet Absorptions of Micro-Organisms and their relation to the Hoyle-Wickamasinghe Hypothesis,” Astrophys. Space Sci. 110, 405–411 (1985). [CrossRef]
13. R.W. Fenn, E.P Shettle, and W.S. Hering, et.al., “Atmospheric optical properties and meteroo-logical conditions,” Atmos. Environ. 15, 1911–1918 (1981) [CrossRef]
14. E.P Shettle and R.W. Fenn, “Models for the aerosols of the lower atmosphere and the effects of humidity variations on their optical properties,” AFGL-TR-79-0214, (Clearinghouse for Federal Scientific and Technical Information, Cameron Station, VA., 1979).
15. K.P Gurton, D.A. Ligon, and R. Kvavilashvili, “Measured infrared spectral extinction for aerosolized Bacillus subtilis var. niger endospores from 3 to 13 μm,” Appl. Opt. 40, 4443–4448 (2001). [CrossRef]
16. D.A. Ligon, T.W. Chen, and J.B. Gillespie,“Determination of aerosol parameters from light-scattering data using an inverse Monte Carlo technique,” Appl. Opt. 35, 4297–4303 (1996). [CrossRef] [PubMed]
17. P.S. Gillespie, A.E. Wetmore, and D.A. Ligon, “Weather and Atmospheric Effects for Simulation: WAVES98 Suite Overview,” ARL-TR-1721-1, (Clearinghouse for Federal Scientific and Technical Information, Cameron Station, VA., 1998).