A quantitative evaluation of Optical vs. SAR-based Wave Kinematic Bathymetry (WKB) for deriving ocean depth in the surf-zone

Project overview

The goal of this project is to derive bathymetry by WKB for four nearshore areas from both optical and SAR imagery, use existing hydrographic shallow water survey data as a source of ground-truth to compute the Root Mean Square Error (RMSE), and determine the strengths and weaknesses of each method.

• Step 1: Determine four Areas of Interest (AOI)

  • Select a variety of locations that feature a diverse set of conditions

  • For WKB, they must satisfy the conditions:

    • Publicly accessible hydrographic shallow water survey data
    • Swell-wave regime
      • Negligible effects from currents
    • An extended nearshore region of depths below 100 m

  • And they should vary by...

    • Latitude (turbidity)
    • Exposure to marine processes (depositional/erosional)
    • Seafloor features (reefs, sandbars, canyons, heavy slope)

• Step 2: Load and normalize datasets

  • Initialize each AOI with central latitude and longitude, filename header, link to CRM, and bounding box extents

  • Load CRM, extract important metadata and save in AOI object

  • For a range of days around CRM creation date, use CMEMS Wave Analysis and Forecast to identify times for each AOI when Mean significant wave height (SWH) greater than 1 m

    • Average of the highest one-third (33%) of waves (measured from trough to crest) that occur in a given period
    • Store swell period and direction data from CMEMS in AOI object for image selection and evaluation

  • Look for Sentinel-2 imagery from days when SWH > 1 m, and get image with best combination of factors for optical WKB

    • Higher SWH, low cloud coverage, wave direction toward solar azimuth, preferable solar elevation
      • Store this information for image selection and evaluation

  • Look for Sentinel-1 VV imagery from days when SWH > 1 m, and get image with best combination of factors for SAR WKB

    • Bragg waves on sea surface
      • Given by λB = 0.5λr sin θi, where λr is the radar wavelength, λB is the sea surface wavelength, and θi is the incidence angle
    • Velocity brunching due to motion of waves relative to SAR
      • Swell wavelengths need to be greater than cutoff wavelength given by Lmin = R√H/V, where R is the slant range of the wave, V is the SAR platform velocity, and H is the significant wave height
        • Lmin should be as low as possible
        • Only pixels that satisfy Lmin threshold should be used for analysis
        • The satellite-geometry based non-georeferenced VV measurement data should be used, and each pixel georeferenced with the associated XML file

  • Reproject imagery and CRM for each AOI into respective UTM zone.

• Step 4: Derive bathymetry

  • Optical

    • Randon transform > discrete Fourier transform > wave celerity > linear dispersion [1]

  • SAR

    • 2D Fast Fourier transform > wavelength estimation > linear dispersion [2]

  • Generate maps of derived depths using Mean Sea Level as the vertical datum

• Step 5: Evaluation

  • Chart characteristics from wave model

  • Compute root mean squared difference for each bathymetric derivation against CRM

  • Generate a difference map using RMS error for each pixel

  • Calculate & chart “visibility index” based on the unique requirements for identifying surface waves from optical and SAR imagery

    • Sentinel 1 - Backscatter and L-min
    • Sentinel 2 - "Glint score" & cloud coverage

Setup

Prereqs

• Python 3.11 (recommended)
• Git

Create and activate a virtual environment

python -m venv venv
source venv/bin/activate  # Windows: venv\\Scripts\\activate

Install dependencies

pip install -r requirements.txt

Optional: install dev tooling

pip install pytest ruff black mypy

Works cited

• [1] E. W. J. Bergsma, R. Almar, and P. Maisongrande, “Radon-Augmented Sentinel-2 Satellite Imagery to Derive Wave-Patterns and Regional Bathymetry,” Remote Sensing, vol. 11, no. 16, p. 1918, Jan. 2019, doi: 10.3390/rs11161918.

• [2] S. D. Mudiyanselage, B. Wilkinson, and A. Abd-Elrahman, “Automated High-Resolution Bathymetry from Sentinel-1 SAR Images in Deeper Nearshore Coastal Waters in Eastern Florida,” Remote Sensing, vol. 16, no. 1, p. 1, Jan. 2024, doi: 10.3390/rs16010001.

License

This project is licensed under the MIT License — see the LICENSE file for details. © 2025 Marcel Rodriguez-Riccelli