An interpretation of underwater LiDAR waveforms based on a modified Weibull probability distribution function

Abstract

One important research topic in marine optics and active remote sensing is the study of spatial distribution of inherent optical properties (IOPs) in 3-D and at very small scales (i.e., ten to hundredth of mm). The understanding of micro-patchiness has many useful implications in ecology [1], ocean color inversion based on satellite and airborne platforms [2], and improvement of underwater image processing [3]. Here a Fine Structure Underwater Imaging LiDAR (LIght Detection and Range) system (FSUIL) [4] was evaluated for investigating spatial and temporal heterogeneity of water optical components. Each impulse function measured by FSUIL is influenced by the source-detector configuration and by the signal attenuation due to changes on inherent optical properties (IOPs). Thus, for a fixed LiDAR system geometry and a complete characterization of the PMT response, IOPs may be retrieved from FSUIL waveforms. For this task, each raw waveform was simulated based on four shape parameters (i.e., slope, scale, amplitude, and noise baseline, hereafter P1, P2, P3 and P4, respectively) derived from a modified Weibull probability distribution function (MW). Unlike other empirical models, MW estimates may be influenced by multiple scattering and symmetry of main backscattering peak. By assuming a constant LiDAR system configuration, it is expected that P1 increases (i.e., more symmetrical or less skewed Gaussian shape) as multiple scattering is added to the total signal arriving to the detector. The parameter P2 is linked to the width of the Gaussian function and is expected to increase as P2 increases. The parameter P3 relates to the kurtosis of the probability distribution function and is expected to mainly vary with the contribution of retro-reflected and forward-reflected photons after the first and second collision, respectively. A final parameter extracted from MW simulations is P4, a variable associated to the signal `offset'.