Aerosol Retrieval by Hyperspectral Remote Sensing

Contact Persons

Felix Seidel

Keywords

Aerosol Optical Depth (AOD), Microphysical Properties of atmospheric Particles, Radiative Transfer, MODTRAN, Spectral Decomposition

Abstract

Information about the optical properties of the atmosphere are important to measure physical quantities at the surface by an optical remote sensing instrument. The atmospheric particles (Aerosols) have a big effect on the atmospheric radiation and are changing over time and space. It is therefore important to retrieve aerosols directly from the remote sensing data, which is not trivial. New hyperspectral airborne instruments, such as APEX, combining the abilities to retrieve aerosols in the near-UV with a high spectral and spatial resolution to overcome traditional limitations and reduce the uncertainties.

The problem of separation between radiative contributions of surface and atmosphere in the spectrometer signal based on image pixels, especially over land areas, is being addressed in this work. Radiative transfer simulations, that show the non-uniqueness of the aerosol-optical effect with respect to optical particle properties and size distributions, define the look-up table.

This research leads to an operational Satellite and Airborne Hyperspectral Aerosol Retrieval Algorithm (SAHARA), which is able to map aerosol optical properties on a high spatial resolution. Applications of such a product are in regional climate studies and in the calibration and validation of air-quality dispersion models.
Methods

• Modeling signal sensitivity
  The sensitivity of the aerosol-optical effect to different boundary layer aerosol regimes and surface reflectances is assessed by means of the atmospheric radiative transfer code MODTRAN 4v3r1 [1]. Assumptions are Mie scattering and horizontal layering of the atmosphere. These calculations define the design of multi-dimensional look-up tables, which are employed for numerical inversion of the aerosol-optical effect to aerosol optical parameters.
• Satellite and Airborne Hyperspectral Aerosol Retrieval Algorithm (SAHARA)
  The aerosol retrieval algorithm is proposed to work on an iterative basis (see Fig. 2). A first step relies on known surface spectral references, which allow retrieving the aerosol model and AOD. The aerosol composition is assumed to remain homogeneous over a certain area and varies only in concentration. The next step trades in the known surface reflectance by the aerosol model and the algorithm can now be extended on dark dense vegetation targets. The retrieval itself is done by fitting the calculated to the measured radiance. The AOD and aerosol model, which provides the best fit with the RTM, are believed to represent the real aerosol feature. The fitting will be done in the near-UV/blue spectral region in order to maximize the aerosol induced signal and minimize the influences of uncertain RTM parameters, such as the surface reflectance for instance [2][3][4]. The use of a band ratio (classic two-channel approach [2][5][6]) enables to catch a distinct radiance feature at sensor between 385 nm and 410 nm. The ratio or slope is characteristic for the aerosol model and can therefore be used to approximate the aerosol composition (see Fig. 3).
• Aerosol parameter mapping with AVIRIS
  Airborne imaging spectrometry data in 224 spectral bands at wavelengths between 400 and 2500 nm and 10 nm resolution, and a spatial ground resolution of 20 m are provided by the AVIRIS sensor [7]. We use an image scene with strong topography and surface variation (Californian coast near Santa Monica), and expected spatial change of aerosol properties to develop methods for aerosol-optical parameter mapping over land, a problem that is yet to be solved in a general sense [2]. Data are geocoded and corrected for illumination effects by use of the PARGE [4] and ATCOR [5] software packages. Application of the band ratio method [2], which assumes a high correlation between reflectances at 2.1um and at 0.47um and 0.66um, respectively, for many low-albedo natural surfaces, yields an estimate of the aerosol-optical effect.

Results

• Sensitivity Study
  For the example shown in Figure 1, soot absorption and marked backscattering in the visible due to a high fine particle fraction in the two-mode case are discernible. Look-up tables for a numerical inversion of the aerosol-optical effect are calculated by means of these simulations.

Fig.1: Simulated aerosol-optical effect in VIS and IR spectrum for surface albedo 0.1.

• Satellite and Airborne Hyperspectral Aerosol Retrieval Algorithm (SAHARA)
  Figure 2 shows the flowchart of the proposed initial aerosol retrieval algorithm. It starts over a reference surface target to invert the AOD and the aerosol model, which is sequentially used to determine the AOD over a dark targets throughout a given scene. Figure 3 shows the relationship between the APEX Channel 1 (385 nm) apparent reflectance and the Channel ratio 1 / 3 (385 nm / 412 nm) for four aerosol models. The dots represent different at-sensor reflectance values due to different AOD. All four models were calculated with the same AOD under standard conditions.

Fig.2: Algorithm flowchart. (from [10])

Fig.3: Two-channel aerosol retrieval. (from [11])

• Aerosol parameter mapping with AVIRIS
  Figure 4 shows the change of aerosol optical thickness with topography, which is particularly well detected in the valley centered in the image. Pixels with too high or low albedo for the method to be applied are masked. Numerical inversion of this result provides the spatial variation of optical thickness and aerosol types.

Fig.4: Aerosol backscatter variation over complex topography.

References:

• [1] A. Berk, G.P. Anderson, P.K. Acharya, M.L. Hoke, J.H. Chetwynd, L.S. Bernsten, E.P. Shettle, M.W. Matthew, and S.M. Adler-Golden, 2003. MODTRAN4 Version 3 Revision 1 USERS MANUAL. AFRL Technical Report, Hanscom AFB, USA.
• [2] M.D. King, Y.J. Kaufman, D. Tanré, T. Nakajima, 1999. Remote Sensing of Tropospheric Aerosols from Space: Past, Present, and Future. Bull. Am. Meteor. Soc. 80, 2229-2259.
• [3] O. Torres, P. K. Bhartia, J. R. Herman, Z. Ahmad, J. Gleason, 1998. Derivation of aerosol properties from satellite measurements of backscattered ultraviolet radiation: Theoretical basis, J. Geophys. Res., 103, 17099-17110.
• [4] R. Höller, A. Higurashi, T. Nakajima, 2004. The GLI 380-nm channel Application for satellite remote sensing of tropospheric aerosol, Proc. EUMETSAT Meteorological Satellite Conference.
• [5] R. Santer, V. Carrre, P. Dubuisson, J.C. Roger, 1999. Atmospheric corrections over land for MERIS, Int. J. Remote Sensing, 20, no. 9, 1819-1840.
• [6] W.v. Hoyningen-Huene, M. Freitag, J.B. Burrows, 2003. Retrieval of aerosol optical thickness over land surfaces from top-of-atmosphere radiance, J. Geophys. Res., 108, 4260.
• [7] R.O. Green, M.L. Eastwood, C.M. Sarture, T.G. Chrien, 1998. Imaging spectroscopy and the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). Rem. Sens. Env. 65, 227-248.
• [8] D. Schläpfer, M. Schaepman, K.I. Itten, 1998. PARGE: parametric geocoding based on GCP-calibrated auxiliary data. SPIE Imaging Spectrometry, Vol. 3438, 334-344.
• [9] R. Richter, 1996. A spatially adaptive fast atmospheric correction algorithm. Int. J. Rem. Sens. 17(6), 1201-1214.
• [10] F. Seidel, J. Nieke, D. Schläpfer, K.I. Itten, 2006. Evaluation of near-UV/blue Aerosol Optical Thickness Retrieval from Airborne Hyperspectral Imagery. Accepted for Proceedings-IEEE IGARSS 2006, Denver.
• [11] F. Seidel, J. Nieke, D. Schläpfer, R. Höller, W.v. Hoyningen-Huene, K.I. Itten, 2005. Aerosol retrieval for APEX airborne imaging spectrometer: a preliminary analysis. in: K. Schäfer (Editor), Remote Sensing of Clouds and the Atmosphere X. Brugge, SPIE Vol 5979, 59791W.

Related Projects:

TROPOSAT - Retrieval of tropospheric aerosol properties from space (Johannes Keller, Paul-Scherrer-Institut)