Augment launch angles for monostatic fathometer scenarios

campreilly commented 9 years ago

The WaveQ3D model does not seem to be accurately computing eigenrays in 200 meters of water when the source is the same as the target.

Scenario:

bottom flat 200 meter, sand
sound speed isovelocity 1500 m/s
time_step = 0.001
time_maximum = 1.0
number_de = 181
number_az = 18
source depth 10 meters
only target is the source

Turn on DEBUG_OUTPUT_EIGENRAYS in wave_queue.cc to see eigenrays being created

wave_queue::build_eigenray() 
    target(0,0):
    t=0.0126707 inten=[1](28.4249) de=80.939 az=160
    surface=1 bottom=0 caustic=0
wave_queue::build_eigenray() 
    target(0,0):
    t=0.253438 inten=[1](63.3714) de=80.939 az=160
    surface=1 bottom=1 caustic=0
wave_queue::build_eigenray() 
    target(0,0):
    t=0.494203 inten=[1](79.226) de=-77.9156 az=160
    surface=1 bottom=2 caustic=0
wave_queue::build_eigenray() 
    target(0,0):
    t=0.506875 inten=[1](78.2781) de=80.939 az=160
    surface=2 bottom=2 caustic=0

Problems:

Eigenrays are missing for surface=0 bottom=1 path
There should be two eigenrays for surface=1 bottom=1 path, missing the one in which de is negative.
DE angles for these paths should be much closer to +/-90 than they are.

campreilly commented 9 years ago

Could this be related to the work that we did on Issue #36? That issue says,

"If the target is directly above or directly below the source, it is unclear if the eigenray calculation can accurately predict the propagation loss. In this case, all of the azimuths will contribute to the Gaussian beam."

Tibonium commented 9 years ago

Created a stand-alone test and got these results. Using the same setup that you have, but I have two targets separated from the source by +/- 5 meters.

=== shadow_eigenrays_test ===
creating ocean
computing max_time
propagating wavefront for 1.215
___eigenray for target(0,0) found___
    count = 1 time = 0.00316782
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 0 bot = 0
___eigenray for target(1,0) found___
    count = 1 time = 0.00316782
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (-80.1831, 0)
    srf = 0 bot = 0
___eigenray for target(0,0) found___
    count = 2 time = 0.00950694
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (-80.1831, 0)
    srf = 1 bot = 0
___eigenray for target(1,0) found___
    count = 2 time = 0.015838
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (-80.1831, 0)
    srf = 1 bot = 0
___eigenray for target(1,0) found___
    count = 3 time = 0.237597
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 0 bot = 1
___eigenray for target(0,0) found___
    count = 3 time = 0.243934
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 0 bot = 1
___eigenray for target(1,0) found___
    count = 4 time = 0.250271
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 1 bot = 1
___eigenray for target(0,0) found___
    count = 4 time = 0.256604
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 1 bot = 1
___eigenray for target(1,0) found___
    count = 5 time = 0.491033
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (80.183, 0)
    srf = 1 bot = 2
___eigenray for target(0,0) found___
    count = 5 time = 0.497369
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (80.183, 0)
    srf = 1 bot = 2
___eigenray for target(1,0) found___
    count = 6 time = 0.503703
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.183, 0)
    srf = 2 bot = 2
___eigenray for target(0,0) found___
    count = 6 time = 0.51004
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.183, 0)
    srf = 2 bot = 2
Finished propagation

Tibonium commented 9 years ago

Ran again with all the same parameters that you used and placed both targets on top of the source, which appear to give the correct results. (Note DE +/-90 are not value launch angles for eigenrays). Only the src->bottom->surface->target appears to have been shadowed.

=== shadow_eigenrays_test ===
creating ocean
computing max_time
propagating wavefront for 1.22
___eigenray for target(0,0) found___
    count = 1 time = 0.0126707
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (-80.1831, 0)
    srf = 1 bot = 0
___eigenray for target(1,0) found___
    count = 1 time = 0.0126707
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (-80.1831, 0)
    srf = 1 bot = 0
___eigenray for target(0,0) found___
    count = 2 time = 0.240768
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 0 bot = 1
___eigenray for target(1,0) found___
    count = 2 time = 0.240768
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 0 bot = 1
___eigenray for target(0,0) found___
    count = 3 time = 0.253438
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 1 bot = 1
___eigenray for target(1,0) found___
    count = 3 time = 0.253438
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.1831, 0)
    srf = 1 bot = 1
___eigenray for target(0,0) found___
    count = 4 time = 0.494203
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (80.183, 0)
    srf = 1 bot = 2
___eigenray for target(1,0) found___
    count = 4 time = 0.494203
    launch angles (de,az): (-77.9156, 0)
    arrival angles (de,az): (80.183, 0)
    srf = 1 bot = 2
___eigenray for target(0,0) found___
    count = 5 time = 0.506875
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.183, 0)
    srf = 2 bot = 2
___eigenray for target(1,0) found___
    count = 5 time = 0.506875
    launch angles (de,az): (80.939, 0)
    arrival angles (de,az): (80.183, 0)
    srf = 2 bot = 2
Finished propagation

Tibonium commented 9 years ago

This appears to be an error with the setup of the wavefront. In the monostatic scenario, the De angles are heavily weighted in the horizontal direction, thus leaving very few rays in the vertical directions, +/-90 with the neighbor ray roughly +/-71. As such after only a few time steps the spacing between these rays have grown so much that eigenrays are not being found, do to the lack of neighbor support.

When we change the DE spacing to linear, specifically -90 to +90 by 1, the wavefront produces all of the eigenrays that we would expect. There is a caveat to this as well, the time step of the wavefront cannot be too large or once again we experience shadow, or missing, eigenrays. Tests showed that for a time step of 0.1 many of the eigenrays were shadowed, where as with a time step of 0.01 and smaller, no eigenrays were shadowed.

Tibonium commented 9 years ago

We came up with an idea where in the monostatic scenarios, the wavefront DE fan can be augmented to include extra rays in the vertical directions. This would resolve this issue and not prevent any loss of data for any bistatic eigenrays that needed to be produced.

campreilly commented 9 years ago

Updated the title to reflect the work that still needs to be done. We may want to explore the option of using the GRAB ray fan scheme for this augmentation.

campreilly / UnderSeaModelingLibrary

Augment launch angles for monostatic fathometer scenarios #141