A general two-stream algorithm for retrieving spectral surface albedo


Spectral surface albedo (SSA) has numerous applications in climate and environmental studies. Given that ground-based observations of SSA are very limited, space-borne remote sensing has been the primary means of acquiring SSA on large or global scales. To date, many satellite sensors measure reflectances in different spectral regions from which values of SSA are inferred. The inversion algorithms range from simple linear relationships to complex, full-fledged radiative transfer models. Often, algorithms were designed for application to a particular sensor or spectral band. In this study, we propose a more versatile parameterized algorithm that can be used for estimating SSA from satellite-measured spectral albedos at the top of the atmosphere (TOA). The algorithm was developed based on a three-layer atmospheric model. Monochromatic and band transmittances due to various absorbing species are parameterized. The reflectance and transmittance for direct and diffuse radiation in the second layer are determined by the generalized two-stream solutions. Except for the parameterization coefficients that vary with the bandpass and spectral response function of the satellite sensor, the framework of the inversion model is applicable to any sensor or spectral region. The model is tested by applying it to the results of detailed radiative transfer simulations for a wide range of conditions for various satellite sensors and spectral bands. The complexity and accuracy of the proposed model are intermediate relative to those of models currently in use. Résumé. L’albédo spectral de surface (SSA) a plusieurs applications dans les études climatiques ou environnementales. Étant donné que les observations de SSA au sol sont très limitées, la télédétection satellitaire constitue le moyen privilégié d’acquisition de SSA à grande échelle ou à l’échelle du globe. Aujourd’hui, plusieurs capteurs satellitaires mesurent les réflectances dans différentes bandes spectrales à partir desquelles la valeur de SSA est déduite. Les algorithmes d’inversion varient des simples relations linéaires aux modèles plus complexes de transfert radiatif. Souvent, les algorithmes ont été développés pour application à un capteur particulier ou une bande spectrale spécifique. Dans cette étude, nous proposons un algorithme de paramétrage plus versatile qui peut être utilisé pour l’estimation de SSA à partir de mesures d’albédo spectral acquises par satellite au sommet de l’atmosphère (TOA). L’algorithme a été développé à partir d’un modèle atmosphérique à trois couches. Les transmittances monochromatiques et intégrées sur la bande dues aux différentes composantes absorbantes de l’atmosphère sont paramétrées. La réflectance et la transmittance du rayonnement direct et diffus dans la seconde couche sont déterminées au moyen de solutions généralisées à deux flux. À l’exception des coefficients de paramétrisation qui varient en fonction de la largeur de la bande et de la fonction de réponse spectrale du capteur satellitaire, le cadre du modèle d’inversion est applicable à n’importe quel capteur ou région spectrale. Le modèle est testé en l’appliquant aux résultats de simulations détaillées de transfert radiatif pour une grande variété de conditions et pour divers capteurs satellitaires et diverses bandes spectrales. La complexité et la précision du modèle sont intermédiaires par rapport à ce que l’on utilise à l’heure actuelle. [Traduit par la Rédaction] Li et al. 399 Introduction The needs for spectral surface albedo (SSA) are multifold. First, the spectral variation of surface albedo is a unique signature of the target. Such a signature has been widely used in a variety of remote sensing applications. A chief example is the use of vegetation indices (VIs), which are often derived from measurements in two spectral bands, namely the visible (VIS) and near infrared (NIR) (Huete et al., 2002). The utilities of VIs in remote sensing are numerous, ranging from the retrieval of bio-geo-physical parameters (e.g., photosynthetically active radiation or PAR) (Li et al., 1997a) to the classification of land cover types (Cihlar et al., 1997a; 1997b). SSA is a fundamental input variable for inferring atmospheric parameters (cloud, aerosol, water vapour, etc.) from multispectral space-borne observations (Kaufman et al., 2002; King et al., 2003; Remer et al., 2005). It is because of inadequate knowledge of SSA that some remote sensing applications, such as aerosol optical thickness, have been limited to relatively uniform ocean surfaces (Nakajima and Higurashi, 1998; Mishchenko et al., 2003; Jeong and Li, 2005). Closure tests of radiative transfer calculations cannot be achieved unless SSA is known, especially under cloudy conditions due to multiple internal © 2005 CASI 391 Can. J. Remote Sensing, Vol. 31, No. 5, pp. 391–399, 2005 Received 31 January 2005. Accepted 2 August 2005. Z. Li.1 Department of Atmospheric and Oceanic Science, University of Maryland, College Park, MD 20742, USA. H.W. Barker. Meteorological Service of Canada, Downsview, ON, M3H 5T4, Canada. L. Moreau. Intermap Information Corporation, Nepean, ON K2E 1A2, Canada. 1Corresponding author (e-mail: zli@atmos.umd.edu). reflections between surface and atmosphere (Barker and Davies, 1989; Li et al., 2002). As general circulation models (GCMs) advance, there is an increasing demand for high-resolution SSA data from satellites (Dickerson et al., 1990). Not long ago, two-band GCMs were typical, but now GCMs with four or more bands are more popular (Barker et al., 2003). Modelling Earth’s climate and understanding the feedbacks between climate and land surface require a good knowledge of SSA on the global scale (Dickerson, 1983). A substantial amount of discrepancy in the simulation of Earth’s energy budget originates from large differences in surface albedo datasets used by various GCMs (Li et al., 1997b). It is highly desirable to derive an accurate, global, spectral, and broadband albedo climatology to facilitate climate change studies (Sellers, 1985). Satellite remote sensing remains a primary tool to meet the aforementioned requirements. Note that there are several steps to convert satellite-measured radiances to surface albedos, although an alternative approach was proposed recently to circumvent the steps by directly linking planetary reflectance with surface albedo following a hybrid modelling and statistical method (Liang, 2003). They include corrections of the spectral response function of the sensor (Trishchenko et al., 2002), the atmospheric effects (Vermote et al., 1997), conversion of reflectance as measured from a particular direction to an albedo defined over the entire hemisphere, or the commonly known bidirectional reflectance distribution function (Luo et al., 2005). If a broadband albedo is derived from narrow-band measurements, spectral conversion is required (Li and Trishchenko, 1999; Liang et al., 2005). Uncertainties are incurred in any of these steps. This study deals with one of the steps, namely, converting a planetary albedo into a surface albedo following atmospheric correction, assuming that the angular correction has been applied to convert the satellitemeasured reflectance to a planetary albedo. Although most BRDF models are for broadband radiances, a limited number of BRDF models have also been proposed (e.g., Chang et al. (2000) for the visible region and Ciren and Li (2001) for the ultraviolet region). At present, there exist some global broadband surface albedo datasets derived from various satellite sensors and systems such the Earth Radiation Budget Experiment (ERBE) (Li and Garand, 1994), the International Satellite Cloud Climatology Project (ISCCP) (Pinker and Laszlo, 1992), and the moderateresolution imaging spectroradiometer (MODIS) (Schaaf et al., 2002), in addition to many regional albedo estimates (e.g., Barker and Davies, 1989). Most notable is MODIS, which has 36 channels, 23 of which are located at solar wavelengths up to 4.0 μm. SSA is one of the MODIS products following atmospheric correction (Vermote et al., 1997) and bidirectional correction (Lucht et al., 2000; Li et al., 2001). The purpose of this study is to propose a satellite-based algorithm for retrieving SSA. The algorithm is intended for general application, instead of being tailored for a particular sensor. For a given satellite radiometer, the spectral filter function is required to derive specific coefficients in the parameterization schemes proposed in the paper. In terms of complexity, the proposed model is intermediate between the simple highly parameterized models (Chen and Ohring, 1984; Li and Garand, 1994) and more complicated radiative transfer models (Vermote et al., 1997), or a feed-forward neural network (Liang et al., 1999). The model developed here is more versatile and accurate relative to the simple parameterized models and at the same time involves less computation and fewer input parameters relative to the complicated models. Given that the parameters characterizing the state of the atmosphere contain many uncertainties, even in the era of Earth-observing systems (EOS), the model presented is of certain special utility. Model This model is based on a three-layer atmosphere–surface system as depicted in Figure 1. The top layer consists of ozone molecules that attenuate solar radiation due to absorption at some specific wavelengths. The radiative properties of this layer can thus be characterized by transmittance functions T1 and T1* for direct and diffuse solar irradiance, respectively. For monochromatic radiation, transmittance due to ozone absorption is determined by Beer’s law. The middle layer contains radiatively active constituents including air molecules, water vapour, CO2 and other gases, and aerosols. These agents have strong interactions via scattering and absorption. The bulk radiative properties of this layer include reflectance and transmittance for direct (R2 and T2) and diffuse (R2 * and T2*) sunlight. The third layer is the surface whose albedo (Rs) is to be derived. 392 © 2005 CASI Vol. 31, No. 5, October/octobre 2005 Figure 1. Schematic diagram of the radiative transfer in a threelayer atmosphere–surface system. For such a system, the following six equations govern the transfer of fluxes between three levels: E E T E E T E R E E E E T E R E E T 1 1

Extracted Key Phrases

5 Figures and Tables

Cite this paper

@inproceedings{Li2005AGT, title={A general two-stream algorithm for retrieving spectral surface albedo}, author={Zhanqing Li and Howard W. Barker and Louis Moreau}, year={2005} }