How the images are made
The bathymetry is real; the sonar images come from ApertureLab's simulator and beamformer.
The bathymetry is a 1.12 m grid of a submarine channel about 1,000 m
deep off Morro Bay, California, surveyed by an autonomous underwater
vehicle of the Monterey Bay Aquarium Research Institute and released by
the U.S. Geological Survey (Kennedy et al. 2021,
doi:10.5066/P97QM7NF,
public domain). Survey grids carry the marks of their own processing,
and at the low grazing angles of a sonar image a few centimeters of
that structure cast stripes of shadow. The grid is therefore smoothed
with a Gaussian of 3 m standard deviation before use, which removes most
relief shorter than about 10 m. The seabed's centimeter-scale roughness
comes from the simulator's own sand model; relief between that and
about 10 m, such as the walls of small gullies, is not represented.
Each line is 1,040 m long and flown as thirteen 80 m swaths that tile
it edge to edge. The vehicle follows the bottom at 20 m above the bed
beneath it, looking to starboard to 200 m of range, with the array the
rest of this site uses at 100 kHz: 32 receivers 40 mm long with 30 kHz
of bandwidth. The seabed is simulated with ApertureLab's
point-scatterer engine at scatterers
per square meter on the terrain the grid gives, with acoustic shadows
cast by the simulator, and every swath is formed by time-domain
back-projection and written as a GeoTIFF in the grid's own map
projection (UTM zone 10N), so the swaths land on the bathymetry in a
GIS.
The images are formed on one flat surface at the height of the bed
beneath the vehicle. Terrain far above or below that surface is drawn
at the wrong range: on line 1, the channel floor 60 m below the bench
is drawn about m too far out. Two
readouts describe what a line misses. "Hidden by terrain" is the share
of each image that no seabed reaches along a line of sight from the
sonar, computed over the same terrain. "Too dark to see" is the share
of 1 m patches more than 10 dB darker than the survey's lit seabed:
it includes everything hidden, and also slopes that tilt away from the
sonar and return too little at their low grazing angle. Every swath of
every line is beamformed with one fixed time-varying gain and drawn
with one brightness scale, so a swath that is half shadow is not drawn
brighter than its neighbors.
The map zooms from the whole line down to the sonar's own pixels:
2.5 cm on line 1, and 5 cm on the other three lines, which keeps the
page's download smaller. Every zoom level is cut from the same
full-resolution images with the same brightness scale.
For what the image surface does to objects standing above it, see
The surface of focus; for the imaging chain,
see Image formation.