Spatiotemporally Reconstructing the Flowing Water Surface to Infer Underwater Information
Video: What you see on the left is the 2D view of the flow of water in a water tunnel (think a river or canal), which has submerged coral canopies in it. This video depicts the slope map of the water surface as seen by a special type of instrumentation that I am implementing in my lab, called the Polarimetric Height Sensing or PHS imager. The video on the right is the 3D reconstruction of the same water surface I have generated from the data on the left at each time instance, using principles of optical metrology.
A Novel Remote Sensing Methodology
Accurate knowledge of underwater ecosystems – with adequate spatial resolution to differentiate between regions of differing depth, substrate or ecological habitat – is critical to accurate numerical modelling of coastal processes and shoreline evolution. But in-situ measurements of underwater roughness and topographies are often limited by the risk of damage, cost, spatiotemporal range, and accessibility. Existing techniques such as airborne LiDAR (Light Detection And Ranging, a non-intrusive altimetry technique that uses laser-light) that try to resolve submerged topographies directly but remotely are limited in resolution and are expensive. More importantly, they require transparent water flows thus typically failing in fluvial and littoral systems where sediment-laden waters render the flow optically opaque. The alternative to measuring bottom features directly is to infer the likely structure of the bottom by characterizing the water surface alone. Previous research has suggested that there is a strong link between flows generated by submerged bed features with their surface manifestations in shallow flows (Chickadel et al. 2009, Geophysical Research Letters). By leveraging a combination of deep learning methods and knowledge of geophysical flows and environmental engineering, we have demonstrated recently (Gakhar et al. 2020, Journal of Fluid Mechanics) that underwater bedforms in a flow cast water-surface signatures that carry important information about these bedforms thus allowing for a novel non-intrusive aerial remote sensing methodology for bedform inference through measurement of these signatures. Gakhar et al. also showed that although external physical processes such as winds can modulate the water surface, they do not necessarily eliminate the water-surface signature of the submerged bottom features & therefore the inference could still work. These results are a key stepping stone toward being able to reliably solve the inverse problem of inferring properties of the bottom boundary purely from their signatures on the free surface. Here too, however, existing measurement tools for quantifying these signatures are lacking; and, additionally, the surface expression of various kinds of bottom features is either not known or is poorly characterized.
Then can we somehow Capture or Quantify these Water Surfaces?
As a way to quantify these water-surface signatures, I am implementing a method (Polarimetric Height Sensing or PHS) for spatiotemporally reconstructing the two-dimensional height field of a flowing water surface. The instrumentation for PHS comprises a small, easily portable polarimetric camera that leverages changes in the polarization of light reflecting from the water surface. PHS has several advantages over techniques such as satellite imagery and surface PIV (Particle Image Velocimetry – an optical method to obtain velocity field by seeding the flow with tracers). First, addition of tracers to the fluid such as those needed for surface PIV (a method for measuring flow velocity by tracking motion of particles with the flow) is not necessary. Second, PHS does not require the ability to see through the water column making it deployable for both turbid and transparent flows. Third, completely non-intrusive, PHS does not require the placement of any instrumentation in the water column. Finally, PHS has sufficient dynamic range and is capable of measuring the free-surface deformation field at a fast temporal rate with high precision for relatively small (~meter scale or less) sized features that satellite imagery simply cannot resolve. Once this surface height map is obtained, the goal is to identify characteristic fingerprints of the submerged bed features using advanced spectral data analysis methods which are under development.
PHS, along with the analytical technique that I am working on would together open doors for a new remote sensing methodology that would enable mapping of underwater coastal ecosystems, and submerged coastal debris. The video here shows the 2D surface of water to the left as seen through one of the channels of the Polarimetric camera and the 3D reconstruction of the water surface on the right, is obtained using the data of the camera and principles in optical metrology.