es403843t

Article pubs.acs.org/est

Night-Time Ground Hyperspectral Imaging for Urban-Scale Remote Sensing of Ambient PMModal Concentrations Retrieval Yael Etzion,† Tsafrir Kolatt,‡ Maxim Shoshany,† and David M. Broday*,† †

Civil and Environmental Engineering, TechnionIsrael Institute of Technology, Haifa 32000, Israel IIARThe Israeli Institute for Advanced Research, Rehovot 76100, Israel

‡

S Supporting Information *

ABSTRACT: Retrieval of aerosol loading in vertical atmospheric columns is a common product of satellite and ground instruments that measure spectral extinction of solar radiation throughout the entire atmosphere. Here we study ground hyperspectral imaging of artificial light sources as a complementary method for retrieving fine aerosol concentrations along quazi-horizontal ambient open paths. Previously, we reported hyperspectral measurements of the aerosol optical thickness in the 500−900 nm range over urban-scale distances (180 m to 4 km), measuring the extinction of radiation emitted from a halogen source. Here we confirm in a laboratory-setup the basic premise that different accumulation-size aerosols generate distinct hyperspectral signatures in this spectral range. Measured hyperspectral attenuation signatures of fine aerosols were comparable to calculated Mie scattering signatures, suggesting that modal aerosol concentrations can be retrieved. A genetic algorithm was adapted to estimate the aerosol modal concentrations from its hyperspectral extinction signature. Retrievals of aerosol concentrations from measured and synthetic hyperspectral signatures indicated a robust algorithm, with an expected retrieval error of 0.2−22% for typical ambient concentrations along an urban-scale open path. The retrieval accuracy was found to depend on the relative aerosol modal concentrations, especially when there is a substantial overlap between the modal spectral signatures.

■

regulate PM1 (PM of diameters 900 nm). The measured spectral signatures of the five different aerosol mixtures (M1− M5, Table 1, columns I−II) were found to be distinct and comparable to the signatures calculated based on the simultaneously measured bimodal size distribution of the aerosol mixtures (Figure 1). It is noteworthy that in some cases, discrepancies from ideal signatures of log-normal aerosol modes were observed, e.g., the experimental signature of aerosol M3 was found to be more intense than expected based on Mie scattering for λ < 600 nm (Figure 1). In part, such discrepancies could be expected for high particle concentrations that lead to coagulation into doublets and triplets and result in an extended (“smeared”) right tail. Namely, the PSD cannot be fully modeled as log-normal. Indeed, aerosol M4 (M5) exhibited a milder (steeper) extinction pattern than expected (Figure 1), suggesting an aerosol size distributions with extended right tail. Nonetheless, accounting for the measured right tail did not eliminate the discrepancy between measure and modeled τa, possibly since high concentrations can led to

coincidence errors in the OPC measurements thus adding to the underestimation bias, especially of the smaller particles. PM Retrieval. Laboratory Signatures. The objective error function in the 2-D modal concentration state-space, V (see SI), exhibited a consistent pattern of elliptical contours that are symmetrically nested around a minimum pointthe aerosol’s true modal concentrations (Figure 2). The common major axis of the elliptical contours represents a mild gradient, along which the objective function is less sensitive to variations in the modal concentrations, and therefore the convergence is harder. Indeed, retrieved modal concentrations for a given scenario by different GA runs were distributed along this axis, with their geometric center found close to the true value (Figure 2). The orientation of this axis indicates that the retrieval of the smaller mode is less sensitive than the retrieval of the larger mode. This behavior was consistent for numerous simulations of spectral response in the laboratory setup, covering different combinations of PSL modal concentrations (103−105 particles cm−3) and GA parameters (subset size, number of iterations, etc.; 1790

dx.doi.org/10.1021/es403843t | Environ. Sci. Technol. 2014, 48, 1787−1794

Environmental Science & Technology

Article

Table 2). In general, whereas retrieved concentrations of the larger mode (dm = 0.82 μm) were generally accurate, retrieved concentrations of the finer mode (dm = 0.5 μm) were overestimated when its true concentrations were lower than the concentrations of the larger mode by more than 1 order of magnitude (Table 2). These findings represent the more pronounced signature of the larger particles in the operational spectral range relative to the signature of the finer particles, and are consistent with Ferri et al.40 As expected, when exploring subranges within the 500−900 nm spectral range, we found that the best retrievals were obtained using the spectral subrange in which the calculated τa closely followed the measured τexp (Figure 1). For example, the best retrieval of M3 was obtained when the 600−850 nm spectral range was used (upper right plate in Figure 1). Considering the perfect sphericity and compositional uniformity of PSL particles, and the successful reconstruction of the spectral signals based on the retrieved concentrations (Figure 3), the relatively large error in the retrieval of M2 concentrations probably resulted from OPC measurement errors.

Table 3. Parameters of Synthetic Plausible Ambient Aerosol Scenarios (Adapted from Friedlander,8 Dubovik et al.37 and Jaenicke38) scenario

1 2 3 4 5 6 7 8

Figure 3. Hyperspectral signature of the PSL aerosol M2 (Table 1). Experimental signature (black jagged line) and calculated signatures based on Mie scattering of (a) the aerosol PSD that has been simultaneously measured by the OPC (black thick line), and (b) the GA retrieved size resolved concentrations (gray thin line).

size mode I

size mode II

type

dm (μm)

σ

dm (μm)

σ

water-soluble organics soot water-soluble organics water-soluble organics water-soluble organics water-soluble organics water-soluble organics water-soluble organics

0.12

1.84

dust

0.815

1.84

0.12 0.12

1.84 1.84

0.815 0.815

1.84 1.84

0.2

1.5

0.83

2.25

0.2

1.5

dust water-soluble organics water-soluble organics dust

0.83

2.25

0.36

1.51

0.83

2.25

0.36

1.51

water-soluble organics dust

0.83

2.25

0.36

1.51

water-soluble organics

0.64

2.17

type

Figure 4. Retrieval errors of eight ambient scenarios as specified in Table 3 (plate number) based on a 500−900 nm spectral range (circles) and a 300−1100 nm spectral range (triangles). Each scenario (plate) contains 12 different combinations of realistic modal concentrations.

Simulated Ambient Aerosols. Eight scenarios of ambient bimodal aerosol types with dm,i between 0.12 and 0.36 μm (finer mode) and 0.64−0.83 μm (larger mode), and σ between 1.5 and 2.25, namely wider modal PSDs than those that characterized the PSL aerosols, were simulated (Table 3). Every scenario included twelve combinations of realistic modal concentrations: between 250 and 104 particles cm−3 of the finer mode and up to 15 particles cm−3 of the larger mode. The signatures were calculated for an urban scale open path of 1 km. In general, GA retrievals of the synthetic scenarios reveal relative errors similar to those obtained during the retrievals of the PSL aerosols (Figure 4). Specifically, retrieval errors of 0.4− 12% and 0.14−16.6% were obtained for the finer and the larger modes, respectively (for a nonabsorbing particles). The retrieval improved for nominally and relatively high modal concentrations. Unlike in the laboratory measurements, retrievals of the larger mode were more sensitive to the modal concentrations ratio. This is due to the fact that the finer

mode outnumbered the larger mode and, therefore, had a significant contribution to the overall aerosol spectral signature. Scenarios that included an absorbing finer mode (soot) or poorly distinct size modes were less accurately retrieved. Specifically, the retrieval errors of the smaller and the larger size modes increased to 3.6−19.4% and 6−44%, respectively, and in the case of modes with very close dm,i the retrieval yielded 2−4 fold overestimated concentrations of the larger mode. In agreement with Lienert et al.,22 extending the retrieval spectral range to 300−1100 nm generally improved the retrieval of the larger mode concentration (Figure 4). To conclude, we demonstrated that retrieval of modal concentrations of aerosol particles in the accumulation size mode based on laboratory measured hyperspectral signatures in the visible-NIR range is feasible. In certain cases, however, the retrieval accuracy may be limited due to copresence of modes 1791



Article

lower wavelengths, a different light source may be used, e.g., xenon lamps are commonly used in active RS technologies29,30 and in optical particle counters.34 The experimentally limited spectral range we could use both weakened the signatures’ sensitivity to PM size and narrowed the retrievable range to 0.15 μm ≤ dp ≤ 1 μm (vertical dotted lines in Figure 5), as discussed previously by Etzion et al.32 Lienert et al.22 argued that a very accurate measurement of the spectral signature (∼0.5% RMS error) is necessary for an accurate PSD retrieval. Yet, such accuracy is clearly not realistic when τa ≤1, since offsets (systematic errors) and random errors of ∼5% are inherent in radiometric measurements and introduce uncertainty of Δτa(λ) = ± 0.01 − 0.02.43 Indeed, our measurements involved a maximum uncertainty of Δτa(λ) ≈ 0.013 − 0.021, revealing a RMS fit error of 2.3−6.6%. Similarly, based on AERONET radiometric measurements, Dubovik et al.44 demonstrated good retrieval of the PSD of water-soluble particles and of desert dust particles even when an uncertainly of Δτa(λ) = 0.02 was introduced (for τa(λ) ranging between 0.05 and 1). However, the accuracy of our retrievals depended on the relative aerosol modal concentrations. We showed that distinct modal extinction in the operational spectral range is essential for accurate retrieval of modal concentrations. Hence, for multimodal aerosols, Heintzenberg’s et al.15 criterion for determining the quality of modal parameters retrieval is not sufficient. It is noteworthy that previous studies that examined PSD retrievals accounted for a distinct aerosol size modes (e.g., a UF mode with dm,1 = 0.07−0.08 μm and a fine mode with dm,2 = 0.8−1 μm42), representing different types of ambient aerosols. However, the literature is deprived of studies that examined retrieval of aerosols that are characterized by coexisting accumulation modes of particles with different thermo-physical properties (representing different sources), i.e., of fine aerosol particles that could be described as externally mixed. In conclusion, urban-scale ground remote sensing is a promising approach for monitoring fine ambient PM at high spatiotemporal resolution. Current RS methods mostly estimate the AOT along vertical atmospheric columns during daytime, which is not easily related to ground level PM concentrations in vast regions of the world.14,45 However, size resolved aerosol concentrations near the ground and the physicochemical characteristics of these particles in terms of their refractive indices may be deduced by means of a similar analysis of spectral scattering and attenuation signatures through quasi horizontal atmospheric layers. Our previous work32 portrayed a general scheme for acquiring HS signatures of fine PM along urban-scale open paths using a portable HS camera system, with the differential extinction related to changes in aerosol concentrations relative to a reference level measured during clean conditions. In contrast to PM retrievals that are based on solar irradiance reflectance spectra (day time retrievals), a night time (dark scene) imaging was developed based on radiance of a halogen target. Measurement of the AOT of bimodal aerosols along a laboratory scale as well as a short ambient scale open paths demonstrated the feasibility to acquire characteristic HS signatures for various modal concentrations. Urban ambient conditions are characterized by a dominant contribution of the fine fraction to extinction in the visible-NIR spectral range, and by a relatively small contribution of the UF mode. The methodology presented in this study is expected to

with substantial overlapping contributions to the spectral signature, demonstrating the ill-posed nature of the problem.

■

DISCUSSION Retrieval of modal concentrations in the laboratory (open path of 30 cm) based on hyperspectral measurements was shown to be feasible for number concentrations >104 particles cm−3 per size mode. Hence, for an urban-scale open path that is 3−4 orders of magnitude larger than the laboratory chamber open path, HS measurements are expected to enable retrieval of particle concentrations as small as few particles cm−3. Synthetic simulations of both laboratory and ambient aerosols revealed that a broad spectral range has a clear advantage for the retrieval of aerosol concentrations. This result indicates the potential for a more accurate retrieval of modal concentrations, since a wider spectral range may capture distinct spectral signatures of the different aerosol size modes. Overall, the retrieval errors ranged between 0.2 and 22%, i.e., better than the estimated 15−25% accuracy of the optical scattering-based laboratory size distribution measurements of Fehsenfeld et al.41 and similar to the 0−20% errors in the airborne measurements of Flocas et al.42 as well as the experimental errors reported in Table 2. Spectral measurements in this study were restricted to the 500−900 nm range, beyond which the measurement noise was found to exceed 1% error. This range excludes part of the spectrum which may be useful for spectral analysis of fine PM and is indeed used by other RS platforms, e.g., the AERONET radiometers measure radiance between 340 and 1640 nm. Nonetheless, the variation in the ratio between the spectral extinction efficiencies at 500 and 900 nm confirms that our spectral data do carry retrievable information on the particle size distribution for 0.1 μm ≤ dp ≤ 1 μm (Figure 5), as was indeed observed experimentally. This particle size range corresponds to the fine aerosol fraction, which is the target “pollutant” in new regulations. It is noteworthy that the lower spectral limit, 500 nm, stemmed from the choice of the halogen illumination source. In order to expand the spectral range to

Figure 5. Ratio of marginal extinction efficiencies of retrievals in the 500−900 nm range (solid line), 600−850 nm range (+), and 500−650 nm range (circles). The vertical dotted lines correspond to the retrievable particle size range based on the measurable spectral range in this study. 1792



Article

comparison of satellite retrievals using different algorithms and instruments. Atmos. Res. 2007, 85, 372−394. (13) Hoff, R. M.; Christopher, S. A. Remote sensing of particulate pollution from space: Have we reached the promised land? J. Air Waste Manage. Assoc. 2009, 59, 645−675. (14) Bohren, C. F.; Huffman, D. R. Absorption and scattering of light by small particles; Wiley-VCH Verlag GmbH & Co.: Weinheim, 1983. (15) Heintzenberg, J.; Müller, H.; Quenzel, H.; Thomalla, E. Information content of optical data with respect to aerosol properties: Numerical studies with a randomized minimization-search-technique inversion algorithm. Appl. Opt. 1981, 20, 1308−1315. (16) Twomey, S. Introduction to the mathematics of inversion in remote sensing and indirect measurements; Elsevier Scientific Publishing Company: The Netherlands, 1977. (17) Amato, U.; Esposito, F.; Serio, C.; Pavese, G. Aerosol science and technology inverting high spectral resolution aerosol optical depth to determine the size distribution of atmospheric aerosol. Aerosol Sci. Technol. 1995, 23, 591−602. (18) Ramachandran, G.; Cooper, D. W. Size distribution data analysis and presentation. In Aerosol Measurement: Principles, Techniques, and Applications; Kulkarni, P., Baron, P. A., Willeke, K., Eds.; John Wiley & Sons Inc.: New York, 2011; pp 497−503. (19) Amato, U.; Carfora, M. F.; Cuomo, V.; Serio, C. Objective algorithms for the aerosol problem. Appl. Opt. 1995, 34, 5442−5452. (20) Goldberg, D. E. Genetic Algorithms in Search, Optimization, and Machine Learning; Addison Wesley Longman Inc.: Reading, MA, 1989; pp 60−87, 99−136. (21) Carroll, D. L. Chemical laser modeling with genetic algorithms. AIAA J. 1996, 34, 338−346. (22) Lienert, B. R.; Porter, J. N.; Sharma, S. K. Repetitive genetic inversion of optical extinction data. Appl. Opt. 2001, 40, 3476−3482. (23) Barkey, B.; Kim, H.; Paulson, S. E. Genetic algorithm retrieval of real refractive index from aerosol distributions that are lognormal. Aerosol Sci. Technol. 2010, 44, 1089−1095. (24) Kuzmanoski, M.; Box, M. A.; Box, G. P.; Schmid, B.; Wang, J.; Russell, P. B.; Jonsson, H. H.; Seinfeld, J. H. Aerosol properties computed from aircraft-based observations during the ACE-Asia campaign: 1. Aerosol size distributions retrieved from optical thickness measurements. Aerosol Sci. Technol. 2007, 41, 202−216. (25) Riefler, N.; Wriedt, T. Intercomparison of inversion algorithms for particle-sizing using Mie scattering. Part. Part. Syst. Charact. 2008, 25, 216−230. (26) Levy, R. C.; Leptoukh, G. G.; Kahn, R.; Zubko, V.; Gopalan, A.; Remer, L. A. A critical look at deriving monthly aerosol optical depth from satellite data. IEEE Trans. Geosci. Remote Sens. 2009, 47, 2942− 2959. (27) Johnson, R. S.; Zhang, J.; Hyer, E. J.; Miller, S. D.; Reid, J. S. Preliminary investigations toward nighttime aerosol optical depth retrievals from the VIIRS Day/Night Band. Atmos. Meas. Technol. 2013, 6, 1245−1255. (28) Barreto, A.; Cuevas, E.; Damiri, B.; Guirado, C.; Berkoff, T.; Berjón, A. J.; Hernández, Y.; Almansa, F.; Gil, M. A new method for nocturnal aerosol measurements with a lunar photometer prototype. Atmos. Meas. Technol. 2013, 6, 585−598. (29) Müller, T.; Müller, D.; Dubois, R. Particle extinction measured at ambient conditions with differential optical absorption spectroscopy. 1. System setup and characterization. Appl. Opt. 2005, 44, 1657− 1666. (30) Varma, R. M.; Venables, D. S.; Ruth, A.; Heitmann, U.; Schlosser, E.; Dixneuf, S. Long optical cavities for open-path monitoring of atmospheric trace gases and aerosol extinction. Appl. Opt. 2009, 48, B159−171. (31) Kyung, W. K.; Young, J. K.; Seung, J. O. Visibility impairment during yellow sand periods in the urban atmosphere of Kwangju, Korea. Atmos. Environ. 2001, 35, 5157−5167. (32) Etzion, Y.; Jarmer, T.; Kolatt, T.; Shoshany, M.; Broday, D. M. Night-time ground hyperspectral imaging for urban- scale remote sensing of ambient PM . I . Aerosol optical thickness acquisition. Aerosol Sci. Technol. 2012, 46, 1119−1128.

be applicable mainly for retrieving concentrations of the accumulation mode, with the exception of dust storm events where a massive fraction of coarse particles is present.

■

ASSOCIATED CONTENT

S Supporting Information *

Laboratory experimental setup (Figure S1); retrieval by genetic algorithm (GA); measured and fitted PSL size distributions (Figures S2 and S3); and additional references. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*Tel: +972-4-829-3468; fax: +972-4-822-8898; e-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The study was partially supported by the state of Lower-Saxony and the Volkswagen Foundation, Hannover, Germany, through the joint Lower Saxony−Israel Research Projects Program (Niedersachsen Project ZN2725), and by the Technion Center of Excellence in Exposure Science and Environmental Health (TCEEH).

■

REFERENCES

(1) Laden, F.; Schwartz, J.; Speizer, F. E.; Dockery, D. W. Reduction in fine particulate air pollution and mortality: Extended follow-up of the Harvard Six Cities study. Am. J. Resp. Crit. Care Med. 2006, 173, 667−672. (2) Pope, C. A.; Dockery, D. W. Health effects of fine particulate air pollution: lines that connect. J. Air Waste Manage. Assoc. 2006, 56, 709−742. (3) US EPA. Integrated Science Assessment for Particulate Matter (final report). EPA/600/R-08/139F.; Washington DC, 2009. (4) Whitby, K. T.; Sverdrup, G. M. California aerosols: Their physical and chemical characteristics. Adv. Environ. Sci. Technol. 1980, 8, 477− 525. (5) Morawska, L.; Thomas, S.; Jamriska, M.; Johnson, G. The modality of particle size distributions of environmental aerosols. Atmos. Environ. 1999, 33, 4401−4411. (6) Park, S.-U.; Kim, J.-W. Aerosol size distributions observed at the Seoul National University campus in Korea during the Asian dust and non-Asian dust periods. Atmos. Environ. 2006, 40, 1722−1730. (7) Morawska, L.; Keogh, D. U.; Thomas, S. B.; Mengersen, K. Modality in ambient particle size distributions and its potential as a basis for developing air quality regulation. Atmos. Environ. 2008, 42, 1617−1628. (8) Friedlander, S. K. Smoke, Dust and HazeFundamentals of Aerosol Dynamics, 2nd ed.; Oxford University Press, Inc.: New York, 2000. (9) Kumar, P.; Robins, A.; Vardoulakis, S.; Britter, R. A review of the characteristics of nanoparticles in the urban atmosphere and the prospects for developing regulatory controls. Atmos. Environ. 2010, 44 (39), 5035−5052. (10) Freiman, M.; Hirshel, N.; Broday, D. M. Urban-scale variability of ambient particulate matter attributes. Atmos. Environ. 2006, 40, 5670−5684. (11) Zhao, X.; Zhang, X.; Xu, X.; Xu, J.; Meng, W.; Pu, W. Seasonal and diurnal variations of ambient PM2.5 concentration in urban and rural environments in Beijing. Atmos. Environ. 2009, 43, 2893−2900. (12) Kokhanovsky, A. A.; Breon, F. M.; Cacciari, A.; Carboni, E.; Diner, D.; Di Nicolantonio, W.; Grainger, R. G.; Grey, W. M. F.; Höller, R.; Lee, K.-H.; Li, Z.; North, P. R. J.; Sayer, A. M.; Thomas, G. E.; von Hoyningen-Huene, W. Aerosol remote sensing over land: A 1793



Article

(33) Cooper, D. Methods of size distribution data analysis and presentation. In Aerosol MeasurementPrinciples, Techniques, and Applications; Baron, P. A., Willeke, K., Eds.; John Wiley & Sons: New York, 2001; pp 689−701. (34) Welas. PALAS Operating Manual−Aerosol Spectrometer Welas, 2000; Welas: Vienna, 2008. (35) Pinnick, R. G.; Hofmann, D. J. Efficiency of light-scattering aerosol particle counters. Appl. Opt. 1973, 12, 2593−2597. (36) Gebhart, J. Optical Direct-Reading Techniques: Light Intensity Systems. In Baron, P. A., Willeke, K., Eds. Aerosol Measurement: Principles, Techniques, and Applications; John Wiley & Sons: New York, 2001; pp 419−454. (37) Dubovik, O.; Holben, B.; Eck, T. F.; Smirnov, A.; Kaufman, Y. J.; King, M. D.; Tanré, D.; Slutsker, I. Variability of absorption and optical properties of key aerosol types observed in worldwide locations. J. Atmos. Sci. 2002, 59, 590−608. (38) Jaenicke, R. Tropospheric aerosols. In Aerosol-Cloud-Climate Interactions; Hobbs, P. V., Ed.; Academic Press: San Diego, CA, 1993; pp 1−31. (39) Taylor, J. R. An Introduction to Error Analysis. In The Study of Uncertainties in Physical Measurements, 2nd ed.; University Science Books: Sausalito, CA, 1997. (40) Ferri, F.; Bassini, A.; Paganini, E. Modified version of the Chahine algorithm to invert spectral extinction data for particle sizing. Appl. Opt. 1995, 34, 5829−5839. (41) Fehsenfeld, F.; Hastie, D.; Chow, J.; Solomon, P. Particle and Gas Measurements Chap. 5. In Particulate Matter Science for Policy Makers: A NARSTO Assessment; McMurry, P., Shepherd, M., Vickery, J., Eds.; Cambridge University Press: Cambridge, England, 2004; pp 159−190. (42) Flocas, H. A.; Assimakopoulos, V. D.; Helmis, C. G. An experimental study of aerosol distribution over a Mediterranean urban area. Sci. Total Environ. 2006, 367, 872−887. (43) Dubovik, O.; King, M. D. A flexible inversion algorithm for retrieval of aerosol optical properties from sun and sky radiance measurements. J. Geophys. Res. 2000, 105, 20673−20696. (44) Dubovik, O.; Smirnov, A.; Holben, B. N.; King, M. D.; Kaufman, Y. J.; Eck, T. F.; Slutsker, I. Accuracy assessments of aerosol optical properties retrieved from Aerosol Robotic Network (AERONET) sun and sky radiance measurements. J. Geophysical. Res. 2000, 105, 9791−9806. (45) Van Donkelaar, A.; Martin, R. V.; Brauer, M.; Kahn, R.; Levy, R. C.; Verduzco, C.; Villeneuve, P. J. Global estimates of ambient fine particulate matter concentrations from satellite-based aerosol optical depth: Development and application. Environ. Health Perspect. 2010, 118, 847−855.

1794


es403843t

Recommend Documents