^{1}

^{2}

^{3}

^{3}

^{3}

Observation of optical properties of atmospheric aerosols, especially their behavior near the surface level, is indispensable for better understanding of atmospheric environmental conditions. Concurrent observations of ground-based instruments and satellite-borne sensors are useful for attaining improved accuracy in the observation of relatively wide area. In the present paper, aerosol parameters in the lower troposphere are monitored using a plan position indicator (PPI) lidar, ground-sampling instruments (a nephelometer, an aethalometer, and optical particle counters), as well as a sunphotometer. The purpose of these observations is to retrieve the aerosol extinction coefficient (AEC) and aerosol optical thickness (AOT) simultaneously at the overpass time of Landsat-8 satellite. The PPI lidar, operated at 349 nm, provides nearly horizontal distribution of AEC in the lower part of the atmospheric boundary layer. For solving the lidar equation, the boundary condition and lidar ratio are determined from the data of ground sampling instruments. The value of AOT, on the other hand, is derived from sunphotometer, and used to analyze the visible band imagery of Landsat-8 satellite. The radiative transfer calculation is conducted using the MODTRAN code with the original aerosol type that has been determined from the ground sampling data coupled with the Mie scattering calculation. Reasonable agreement is found between the spatial distribution of AEC from the PPI lidar and that of AOT from the blue band (band 2) of Landsat-8. The influence of AOT on the values of apparent surface reflectance is also discussed.

Aerosol exerts impact on radiation balance of the Earth’s atmosphere by modifying cloud characteristics in addition to scattering and absorption of solar radiation [

Concurrent observations with ground-based instruments have been reported including MODerate resolution Imaging Spectroradiometer (MODIS) [

The atmospheric group of Center for Environmental Remote Sensing (CEReS), Chiba University, has developed and operated several lidar systems such as a multi-wavelength lidar system [

The following section of this paper is outlined as follows. In Section 2, the method and formulation are explained. In Section 3, the results are presented and discussed, followed by Section 4 as the conclusion of this paper.

The observations with ground-based instruments were carried out at Chiba University (35˚37'30'' and 140˚06'14''), located south-east of Tokyo metropolitan area (

3) March 20 2017, and 4) May 23 2017 are reported in the present paper, representing the aerosol distributions over the Kanto plain area under nearly cloud-free conditions.

The real-time data of aerosol particles in the atmospheric boundary layer are obtained using the following instruments routinely operated on the rooftop of the CEReS building (~30 and ~50 m above ground and sea level, respectively): a three-wavelengths (450, 550, and 700 nm) integrating nephelometer (TSI, Model 3563), a seven-wavelengths (370, 470, 520, 590, 660, 880, and 950 nm) aethalometer (Magee Scientific, AE-31), and optical particle counters (OPCs) that cover relatively small (0.08, 0.1, 0.2, 0.3, and 0.5 μm) and relatively large (0.3, 0.5, 1.0, 2.0, and 5.0 μm) particle sizes (Rion, KC-22B, KC-01D and KC-01E). The aerosol extinction coefficient, a_{ext}, can be calculated as a function of wavelength, l, as the sum of scattering and absorption coefficients:

α e x t ( λ ) = α s c a ( λ ) + α a b s ( λ ) , (1)

where α s c a and α a b s are the scattering and absorption coefficients derived from the nephelometer and aethalometer, respectively. The lidar ratio, S_{1}, is defined as the ratio between the extinction ( α e x t ) and backscattering (β) coefficients, and the following Mie scattering formulation can be used to calculate the value of S_{1}:

S 1 = α e x t β = ∫ 0 ∞ n ( model ) ( r ) σ e x t d ( log r ) ∫ 0 ∞ n ( model ) ( r ) ( d σ s c a d Ω ) θ = π d ( log r ) , (2)

where r is the particle radius,

n ( model ) ( r ) = d N ( r ) d ( log r ) = ∑ i k N i 2 π log 2 π log σ i exp [ − ( log r / r i ( m ) ) 2 2 ( log σ i ) 2 ] (3)

is the lognormal size distribution (k = 1 for mono-modal and k = 2 for bimodal), and α e x t and d s s c a / d Ω are the extinction cross-section and differential cross-section, respectively [

d σ s c a d Ω = σ s c a f ( cos θ ) , (4)

where f ( cos θ ) is the scattering phase function. The parameters that are used in the Mie calculation, namely, the mode concentration (N_{i}), mode radius ( r i ( m ) ), width ( log σ i ), and both the real and imaginary parts of refractive index are determined so as to reproduce the data of ground-sampling instruments (scattering, absorption, and particle size) [

The PPI lidar, installed at the same roof top of the CEReS building, is based on a diode-laser-pumped Nd:YLF laser operated at 349 nm with output energy of 60 μJ/pulse and 300 Hz pulse repetition rate. A 30-cm diameter Cassegrainian telescope coupled with a flat mirror tilted at ~47˚ is used to collect the backscattered signal, and a photo-multiplier tube (Hamamatsu, H10304-00) and a transient recorder (Licel, TR20-160) are employed for recording the signals. As shown in

The inversion analysis of PPI lidar signals is attained using the Fernald method [

α 1 ( R ) = − S 1 ( R ) S 2 α 1 ( R ) + S 1 ( R ) X ( R ) exp { 2 ∫ R R C [ S 1 ( R ′ ) X ( R ′ ) S 2 − 1 ] α 1 ( R ′ ) d R ′ } X ( R C ) α 1 ( R C ) S 1 ( R ) + α 2 ( R C ) S 2 ( R ) + ∫ R R C S 1 ( R ′ ) X ( R ′ ) 2 ∫ R R C [ S 1 ( R ′ ) X ( R ′ ) S 2 − 1 ] α 1 ( R ′ ) d R ′ (5)

where R is the range, X(R) = R^{2}P(R) is the range-corrected signal, R_{C} is the reference range, and the subscript 1 (2) refers to aerosol (air molecule) [_{1} is assumed for analyzing PPI data. From the general property of Rayleigh scattering of air molecules, S_{2} is assumed to be 8.52 sr [_{C} are adjusted so that the resulting AEC at R = 0 agrees with the value from sampling observations.

The values of AOT and Angstrom exponent are derived from the multi-band observation of a sunphotometer (Prede, PSF-100), routinely operated in CEReS. The measurement is made at the wavelengths of 368, 500, 678, and 778 nm. The Lambert-Beer law coupled with the Langley extrapolation method [

τ A ( λ ) = ln I 0 ( λ ) − ln I ( λ ) m ( θ ) − τ G ( λ ) − τ R ( λ ) . (6)

Here the variables τ A , τ G , and τ R represent the optical thickness due to aerosol, absorbing gas (ozone), and Rayleigh scattering, respectively; I_{0} is the extra-terrestrial intensity of solar radiation (dependent on the seasonal change of the Sun-Earth distance), I is the solar radiation intensity measured by the sunphotometer, and m is the air mass dependent on the solar zenith angle, q. The value of I_{0} can be determined by implementing the Langley extrapolation method on a clear-sky day with small aerosol loading [

Furthermore, the information on aerosol size distribution can be inferred from the wavelength dependence of AOT that can be expressed using Angstrom exponent [

p = − ln [ τ A ( λ 2 ) τ A ( λ 1 ) ] / ln ( λ 2 λ 1 ) , (7)

which can easily be extended for multi-wavelength fitting. The value of p is the order of unity. The value of Angstrom exponent becomes larger in the condition of the dominance of fine-mode aerosol, while it becomes smaller for the dominance of coarse-mode particles [

Here the satellite measurement of aerosol distribution is studied using images of the OLI sensor onboard the Landsat-8 satellite, since the data provide a fine resolution of 30 m [^{−2}∙sr^{−1}∙μm^{−1}), and apparent reflectance (dimensionless). Every satellite has its own empirical equation(s) for converting the DN values to radiance (L_{obs}) and apparent reflectance (ρ_{ap}) as a function of wavelength. For Landsat-8 OLI, the apparent reflectance is calculated using the procedure described elsewhere [

ρ a p ( λ ) = π d 2 E ( λ ) cos θ s × L o b s ( λ ) ,

L o b s ( λ ) = L M A X − L M I N Q C A L M A X − Q C A L M I N × ( D N − Q C A L M I N ) + L M I N . (8)

Here Q_{CALMAX} and Q_{CALMIN} are the maximum and minimum values of the quantized and calibrated pixel value in DN; L_{MAX} and L_{MIN} are the spectral radiance scaled to Q_{CALMAX} and Q_{CALMIN}, respectively. Besides, parameters d, E(λ), and θ_{s} stand for the Sun-Earth distance in astronomical unit (AU), solar irradiance at the top of atmosphere at 1 AU, and solar zenith angle, respectively.

The radiance components detected by a satellite sensor can be given schematically as shown in

Besides, scattering and absorption of air molecules and aerosol particles exert significant influence on the radiance value of each pixel. Therefore, radiative transfer calculation is indispensable for separating the atmospheric effects from the ground-reflected radiance in each satellite scene. In an early work, Chandrasekhar developed a radiative transfer equation where the scattering and absorption processes due to air molecule were considered [

Here we describe the basic features required for understanding the radiative transfer processes. As indicated in _{tot}(λ), can be separated into six different components as

L t o t ( λ ) = L g d ( ρ ) + L g i 1 ( ρ ) + L g i 2 ( ρ , ρ ¯ ) + L p s + L p m 1 + L p m 2 ( ρ ¯ ) . (9)

Here, L_{gd} and L_{gi} = L_{gi}_{1} + L_{gi}_{2} are the radiance components reflected from the surface directly and indirectly, respectively: only the reflectance of the target pixel, ρ = ρ(λ), is included in L_{gi}_{1}, while the reflectance averaged over adjacent pixels, ρ ¯ , is also considered for L_{gi}_{2}. The last three terms on the right-hand side of Equation (9) represent path radiance components: L_{ps} and L_{pm}_{1} are the components arising from single and multiple scattering, respectively, whereas the adjacent surface reflection is also considered in L_{pm}_{2}. It is noted that on the right-hand side of Equation (9), the λ dependence of each term has been omitted for the sake of simplicity. The radiance just after the surface scattering (which is assumed to be Lambertian) can be written as

L g ( λ ) = 1 π d 2 ρ ( λ ) E ( λ ) cos θ s T ( λ , θ s ) , (10)

where T ( λ , θ s ) is the atmospheric transmittance of the incoming solar irradiance [

L g d ( λ ) = L g ( λ ) T ( λ , 0 ) , (11)

where T(λ, 0) is the transmittance when the satellite observation is made toward the nadir direction, as is the case of Landsat-8 OLI. Detailed formulation and modification of the radiative transfer equation including the target reflectance (ρ), average reflectance ( ρ ¯ ), transmittance (T), and some other parameters can be found in a satellite guide and references therein [

The determination of AOT from each pixel of a satellite image can be carried out as follows. In order to implement the radiative transfer simulation on a satellite pixel, the aerosol model has to be specified with the value of AOT, usually in the form of τ A ( 550 ) = τ 550 , the AOT at 550 nm. Once the pixel reflectance (ρ) is known, the value of L_{tot}(λ) can be calculated with the simulation. Then, the value of τ 550 can be uniquely determined from the condition that L_{tot}(λ) is equal to the observed radiance, L_{obs}(λ), given in Equation (8). Practically, the implementation of this procedure is facilitated by constructing a lookup table, which summarizes the behavior of ρ_{ap} as a function of ρ and τ 550 . Also, it can be pointed out that by combining Equations (8)-(11), one obtains the following relation between the apparent reflectance and reflectance:

ρ a p ( λ ) = π d 2 E ( λ ) cos θ s L o b s ( λ ) = π d 2 E ( λ ) cos θ s L t o t ( λ ) = ρ ( λ ) T ( λ , θ s ) T ( λ , 0 ) + π d 2 E ( λ ) cos θ s ( L g + L p ) (12)

Here L_{g} and L_{p} stand for the ground reflectance (other than L_{gd}) and atmospheric scattering terms in Equation (9), respectively. When aerosol loading is limited and wavelength is not too short, both of L_{g} and L_{p} become small, representing some remaining contribution from the molecular (Rayleigh) scattering. Therefore, to a good approximation, we obtain

ρ a p ( λ ) ≅ ρ ( λ ) exp [ − ( 1 + m ) τ A ( λ ) ] , (13)

where m = (cosθ_{s})^{−1} is the air mass. This equation is useful for estimating the value of AOT. Among the nine bands of Landsat-8, band 2 centered at 482 nm is used for the present analysis. The radiance that represents band-i can be calculated as

L i = ∫ λ 1 λ 2 g i ( λ ) L t o t ( λ ) d λ / ∫ λ 1 λ 2 g i ( λ ) d λ , (14)

where g_{i}(λ) is the band response function covering the wavelength range from λ_{1} to λ_{2}.

The values of AEC, Angstrom exponent as well as lidar ratio are determined using the data of ground-based instruments at the time of Landsat-8 overpass. _{i} and σ_{i} in Equation (3). The real and imaginary parts of the complex refractive index (which are assumed to be independent of wavelength) are determined so as to reproduce the observed wavelength dependence of the scattering and absorption coefficients measured with the nephelometer and aethalometer. It is noted that before this analysis, the scattering data from the integrating nephelometer are subjected to the truncation error correction, which is related to the loss of contributions from relatively coarse particles, as well as the correction due to the evaporation of hygroscopic aerosols when the particles are introduced into the scattering volume of the instrument from ambient conditions with relative humidity higher than around 50% [_{A}(λ), which in turn have been retrieved from the sunphotometer observation using Equation (6).

The value of lidar ratio, S_{1}, can readily be obtained from the phase function information provided from the Mie scattering calculation. In the case of _{1} = 62.5 sr obtained for λ = 349 nm is employed for implementing the Fernald analysis given by Equation (5). The resulting parameters, namely, the size distribution parameters and refractive index characterize aerosol in the atmospheric boundary layer, are utilized also as input parameters in the radiative transfer calculation using MODTRAN.

Date | S_{1} (sr) | α_{ext}(km^{−1}) | p | RH (%) | T (˚C) |
---|---|---|---|---|---|

27 October 2016 | 74.10 | 0.075 | 1.422 | 45 | 24 |

31 January 2017 | 62.50 | 0.056 | 1.250 | 30 | 7 |

20 March 2017 | 48.90 | 0.330 | 1.750 | 26 | 15 |

23 May 2017 | 61.70 | 0.120 | 1.578 | 58 | 25 |

Date | DOY | SZA (deg) | SAA (deg) | AOT | Atm. Model | Aerosol Type | ||||
---|---|---|---|---|---|---|---|---|---|---|

N | r (μm) | logs | Re | Im | ||||||

27 Oct. 2016 | 301 | 51.42 | 158.88 | 0.087 | Mid Lat. Summer | 1.0000 | 0.0206 | 0.3017 | 1.5848 | 0.0558 |

31 Jan. 2017 | 31 | 58.14 | 152.51 | 0.079 | Mid Lat. Winter | 1.0000 | 0.0266 | 0.3242 | 1.6000 | 0.0373 |

20 Mar. 2017 | 79 | 41.98 | 144.46 | 0.476 | Mid Lat. Winter | 1.0000 | 0.0268 | 0.3323 | 1.5926 | 0.0231 |

23 May 2017 | 143 | 23.47 | 125.76 | 0.225 | Mid Lat. Summer | 1.0000 | 0.0246 | 0.3165 | 1.5867 | 0.0351 |

(width), Re (real part of refractive index), and Im (imaginary part of refractive index), resulting from the Mie scattering calculation (

_{1} = 452 nm and λ_{2} = 512 nm) is utilized for the construction of the lookup table.

_{ap}, changes as a function of the surface reflectance, ρ, for various values of AOT at 550 nm, τ_{550}. It is seen that the value of ρ_{ap} increases with ρ. Some deviations are noticeable for small values of ρ because of the influence of atmospheric scattering due to air molecules and aerosol particles. Nevertheless, since the relation between ρ_{ap} and ρ do not change significantly when τ_{550} is changed between 0 and 1, it is understood that the value of surface reflectance can be estimated from the apparent reflectance. _{550} for various values of ρ_{ap}. From this figure, it is seen that for a fixed value of τ_{ap}, the dependence of ρ on τ_{550} becomes noticeable for smaller values of ρ_{ap} (<0.15).

Figures 7(a)-(d) show the apparent reflectance (ρ_{ap}) distribution over the Kanto area obtained by applying Equation (8) to the four Landsat-8 band-2 images when the concurrent observations were performed. Since the AOT value and aerosol properties are known from the ground-based observations, considerations explained in _{ap}, are rather large, and particular enhancement is found in the urban area, e.g., around the coastal area of the Tokyo Bay. In _{ap} are rather suppressed, indicating the influence of decreased transmittance due to relatively large values of AOT (

_{ext}) derived by processing the PPI signal using Equation (5). In this calculation, the value of lidar ratio, S_{1}, listed in

_{482} = 0.06 − 0.08), except in the north part of Kanto area. The smallest condition is seen on (b) 31 January 2017, with a homogeneous distribution of AOT (τ_{482} = 0.05 − 0.07). The highest AOT (τ_{482} = 0.45 − 0.50) is seen on (c) 20 March 2017. In _{482} = 0.20 − 0.30) is found, which is more conspicuous than the other cases. It is noted that in

(τ_{482} = 0.476) while the smallest on (b) 31 January 2017 (τ_{482} = 0.079): these values have been used to convert the apparent reflectance (ρ_{ap}, _{obs} = L_{tot}, as explained in Section 2.4.

The main purpose of this novel monitoring technique is the comparison of AEC distribution derived from PPI lidar (_{ext}(z), can usually be described as a_{ext}(z) = a_{ext}(0) exp(−z/h_{a}), where h_{a} is a parameter called the aerosol scale height (h_{a} ~1 - 2 km), and a_{ext}(0) is the AEC directly measured with the PPI lidar. The value of AOT, on the other hand, can be obtained by integrating a_{ext}(z) over the whole troposphere, leading to the expression of τ_{A} ~ h_{a} a_{ext}(0). This indicates that AOT is nearly proportional to AEC, though the change in h_{a} may cause some deviations. In the four cases reported in the present paper, the estimated value of h_{a} is around 1 km, though the value is slightly large (−2 km) for the case of (d) 23 May 2017.

A novel monitoring technique for retrieving the spatial distribution of atmospheric aerosol optical properties has been proposed and demonstrated. The PPI lidar provides the distribution of AEC in the atmospheric boundary layer with the help of data from ground-sampling instruments (an integrating nephelometer, an aethalometer, and optical particle counters) to determine the lidar ratio in addition to the near-end boundary values of AEC. For this purpose, the Mie scattering calculation has been conducted. Besides, this procedure provides the parameters that define the aerosol type as needed for the MODTRAN computation. In the MODTRAN simulation of the satellite-observed radiance, the AOT data from the sunphotometer is also exploited for determining the reflectance distribution from a satellite image, which, in turn, is used to derive the spatial distribution of AOT for the band-2 (blue band) of Landsat OLI. Through this analysis, good consistency has been found between the spatial distribution of AEC from the PPI lidar and that of AOT from the satellite sensor. As a whole, the present work has demonstrated that aerosol characterization based on ground-based observations is useful for obtaining detailed insight into the horizontal distribution of near-surface aerosol, which also yields better constraint on the atmospheric correction of satellite remote sensing data. Future extension of the present method may involve more frequent observation (i.e., rapid scan) of the PPI lidar measurement with preferably higher laser power (i.e., wide range coverage) so that the temporal as well as spatial variation of near-surface aerosol can be monitored on an operational basis.

The first author (JA) would like to thank to Ministry of Research, Technology, and Higher Education Republic of Indonesia for supporting fellowship named Beasiswa Dikti.

The authors declare no conflicts of interest regarding the publication of this paper.

Aminuddin, J., Purbantoro, B., Lagrosas, N., Manago, N. and Kuze, H. (2018) Landsat-8 Satellite and Plan Position Indicator Lidar Observations for Retrieving Aerosol Optical Properties in the Lower Troposphere. Advances in Remote Sensing, 7, 183-202. https://doi.org/10.4236/ars.2018.73013