A forward image model for passive optical remote sensing of river bathymetry

By facilitating measurement of river channel morphology, remote sensing techniques could enable significant advances in our understanding of fluvial systems. To realize this potential, researchers must first gain confidence in image-derived river information, as well as an appreciation of its inherent limitations. This paper describes a forward image model (FIM) for examining the capabilities and constraints associated with passive optical remote sensing of liver bathymetry. Image data are simulated from the streambed up by first using information on depth and bottom reflectance to parameterize models of radiative transfer within the water column and atmosphere and then incorporating sensor technical specifications. This physics-based framework provides a means of assessing the potential for spectrally-based depth retrieval from a particular river of interest, given a sensor configuration. Forward image modeling of both a hypothetical meander bend and an actual gravel-bed river indicated that bathymetric accuracy and precision vary spatially as a function of channel morphology, with less reliable depth estimates in pools. A simpler, more computationally efficient analytical model highlighted additional controls on bathymetric uncertainty: optical depth and the ratio of the smallest detectable change in radiance to the bottom-reflected radiance. Application of the FIM to a complex, natural channel illustrated how the model can be used to quantify the effects of various sensor characteristics. Bathymetric accuracy was determined primarily by spatial resolution. due to mixed pixels along the banks and sub-pixel scale variations in depth, whereas depth retrieval precision depended oil the sensor's ability to resolve subtle changes in radiance. This flexible forward modeling approach thus allows the utility of image-derived river information to be evaluated in the context of specific investigations, leading to more efficient, more informed use of remote sensing methods across a range of fluvial environments. (C) 2009 Elsevier Inc. All rights reserved.