The Importance of Reflectance Terminology in Imaging Spectroscopy

Analysing databases, field and airborne spectrometer data, modelling studies and publications, it becomes obvious, that there is a lack of consistency in definitions of reflectance quantities throughout the imaging spectroscopy user community. One example is the term ‘BRDF’ assigned to significantly differing quantities, ranging from the bidirectional reflectance distribution function to hemispherical-conical reflectance factors. Our contribution summarizes basic reflectance nomenclature articles. Secondly we quantify differences of reflectance products, with special emphasis on wavelength specific effects, to stress the importance of adequate usage of reflectance definitions and quantities. Results from the comparison of directional-hemispherical reflectance versus bihemispherical reflectance and bidirectional reflectance factors versus hemispherical-directional reflectance factors are shown. We exemplify differences of these quantities using modelling results of a black spruce forest canopy and snow cover, as well as biome-specific MISR reflectance products of the year 2001. The actual differences in the reflectance products of a remotely sensed surface depend on the atmospheric conditions, the surroundings, topography, and the scattering properties of the surface itself. As these effects are highly wavelength dependent, the imaging spectroscopy community has to become more specific on the application and definition of reflectance quantities. As of today most delivered reflectance products from imaging spectrometers include the hemispherical illumination component, product algorithms based on resulting at surface reflectance data have to include the actual atmospheric conditions even for nadir view angles, e.g., in the form of a wavelength specific indication of the ratio of diffuse to direct illumination. The results urge the community to treat reflectance quantities with outmost care and consistency to reduce uncertainties of derived products. INTRODUCTION Imaging spectrometer data and products are constantly improved in quality. Optimization is usually performed at radiance level (enhanced calibration concepts, vicarious calibration, etc.) with uncertainties approaching 4% (i), at reflectance level (atmospheric correction) with uncertainties approaching 5% (ii), at product level (sophisticated integration of various sources, assimilation, etc.) with uncertainties approaching 10% (iii), but rarely on terminology, where uncertainties can still be much higher than 10%. The imaging spectroscopy community is developing an increased appreciation of the effects that are induced by the solar illumination and sensor viewing geometry on field, airborne and satellite data. The reflectance anisotropy of the Earth’s surfaces and the atmosphere contains unique information about their structure and the optical properties of the scattering elements. The underlying concept for the characterization of the anisotropy is the bidirectional reflectance distribution func© EARSeL and Warsaw University, Warsaw 2005. Proceedings of 4th EARSeL Workshop on Imaging Spectroscopy. New quality in environmental studies. Zagajewski B., Sobczak M., Wrzesień M., (eds) tion (BRDF). It describes the radiance reflected by a surface as a function of a parallel beam of incident light from a single direction into another direction of the hemisphere. Under natural conditions, i.e. for all ground based, airborne and spaceborne sensor measurements, the assumption of a single direction of the incident beam does not hold true. Natural light is composed of a direct part, as well as a diffuse component scattered by the atmosphere, and the surroundings of the observed target. The amount and spectral character of the diffuse light irradiating the observed surface is thus depending on the atmospheric conditions, as well as on the topography and the scattering properties of the surroundings. Without correction of this diffuse component, observed reflectance quantities depend on actual atmospheric conditions, especially in the Rayleigh scattering dominated wavelength region (400-800 nm), and are not limited to the desired intrinsic directional characteristics of the observed surface. Further, the (instantaneous) field of view ((I)FOV) of the instrument most often integrates over a large viewing angle and does not allow a single beam observation. Thus, imaging spectrometer measurements do not follow the protocol of directional reflectance quantities and resulting products can only be considered as rough approximations of the surface bidirectional reflectance, a fact that is often neglected. A physically based terminology, defining various reflectance quantities using the direction of illumination and observation, as well as their opening angle, was proposed by Nicodemus (iv), and updated by Martonchik (v). Further, recent advances originating from the Multi-angle Imaging SpectroRadiometer (MISR) science team have lead to a more uniform reflectance terminology. The operational MISR data products including different reflectance quantities are a major progress, and give users the opportunity to apply appropriate physical quantities for their investigations. Despite above-mentioned advancements, physical conditions of measurements and corresponding terminology of at-surface reflectance quantities are still very often neglected by the user community. The loose usage of the term ‘BRDF’ is one of the most striking examples. The community performing ground based multiangular at-surface reflectance measurements often calls acquired quantities BRDF or BRF data (e.g., (vi)). But the derivation of the bidirectional reflectance distribution function (BRDF) from measurements performed under ambient sky (i.e., hemispherical) illumination results in a considerable shape distortion of the resulting function with respect to the true BRDF in the visible and near-infrared when no correction for the diffuse part of the illumination is performed, even under clear sky conditions (vii). Thus the derived so-called BRDF databases do not only reflect intrinsic bidirectional reflectance properties of the observed surface, but also wavelength-dependent effects caused by the diffuse illumination component. This is especially true for diurnal multiangular observations with changing atmospheric conditions throughout the day. Consequently, Martonchik (viii) and Lyapustin (vii) have developed methods for an accurate atmospheric correction of measured hemispherical-directional reflectance data to enhance the experimental research of anisotropic surface reflectance. Given the confusion with and neglect of reflectance terminology as exemplified above, the aim of this paper is to summarize the basic nomenclature articles of Nicodemus (iv) and Martonchik (v) and make the updated definitions available to the imaging spectroscopy community. This overview helps to identify the correct definition for measured reflectance quantities and processed products, and to apply the appropriate quantity in physical as well as empirical approaches. The adequate use of reflectance data does not only require a precise, widely distributed and easy to use reflectance terminology, but also an in-depth understanding of spectrodirectional effects. Therefore, the second part of the paper is aiming in demonstrating the importance of adequate use of reflectance definitions and quantities. Uncertainties, introduced by neglecting the physical basis and the corresponding terminology, are exemplified through case studies, with special emphasis on the spectral and directional domain. This paper systematically highlights differences in at-surface reflectance quantities by their definition. We focus on the geometry of the opening angle of the illumination, i.e., directional and hemispherical extent. We quantitatively compare operational MISR directional and hemispherical reflectance products with respect to the corresponding optical depth and resulting wavelength-specific differences. Secondly, we perform a modeling exercise for forest and snow. Using a variation of