Hi,
I've been mulling this over a bit, looking for a solution to suggest
that is intermediate to the current list of parameters, such as those
listed, and saving more complete spectral information. We are all
pursuing research that we believe will eventually lead to models of
upper ocean turbulence, and integrally related chemical and biological
phenomena, where surface waves will play a significant role.
I was looking at the CDIP web page where they save and redistribute gobs
of spectral data, and noticing that in one part of their web site they
serve data in 9 standard frequency bins (whereas we are passing them
many more from OS Papa, and these are available by download of larger
compressed files). This may be more frequency bins than really necessary
to describe the Stokes drift profile well enough. On the other hand,
they only serve up (again just in this '9-band' presentation) just the
energy and propagation direction from reported angular Fourier
coefficients, which leaves out the angular spread in each frequency bin.
Said another way, this set is only reporting the magnitude of the energy
and the direction of the wave momentum, but omitting to state the
magnitude of the momentum. The more complete set could also be simply
stated as the trio of (1) the energy spectrum in each bin and the (2,3)
the two Stokes drift spectrum components in each frequency bin.
To do this for CDIP's 9 frequency bins from a model they we're talking
about reducing the spectrum at each saved grid point & time to 9x3=27
values. That's a lot less than typical 2D wave model spectra, but I
think we could get it smaller.
How many bins are enough? I'm not sure, but we can tell at this point
that to get the forcing right we will need to compute the profile of the
Stokes drift at the vertical resolution of the turbulence model. Such
models typically have O(1)m resolution, corresponding to the vertical
decay scale of waves with frequency f_{1m}=0.35 Hz , so the upper bin
should be for f>f_{1m}=0.35 .
Then the frequency bins chosen by CDIP (not really picking on them, it
just got me thinking) are static, which gives you problems when you're
applying the same decimation in varying forcing regimes. This is useful
in programming web sites or wave models, but it seems more efficient to
divide things up under some assumption of self-similarity. I think we
could therefore pick frequency bins delimited by the moments of the
energy spectrum reported as parameters, and the frequency bins therefore
adjust to best capture the wave properties as seas rise & fall. More
divisions of these frequency bins could be made as necessary, but I'll
stop here and try to summarize my 'list of wants':
A) 4 values: The Zeroth, First, Second and Third moments of the Scalar
spectrum, i.e. the integral of {f^m S(f)}df , where S(f) is the
frequency scalar spectrum, m=0,1,2,3 and where m=0 gives you energy.
Some of these are equivalent to H_s and various mean or dominant period
parameters. Let's call the moments, normalized on the zeroth (energy), '
f_m^m ', and define frequency bin boundaries using a subset of these and
the model or data upper and lower bounds f_{min}, f_{max}: [f_{min},
f_1, f_2, f_3, f_{1m}, f_{max}] , making 5 frequency bins.
B) 2 Values: The Vector Stokes drift [U_S, V_S] at the surface, i.e. the
integral of (16*pi^3/g)*{f^3 S(f,\theta) [sin(\theta), cos(\theta)]}df .
In other words, the (pseudo-)momentum of the waves.
C) 3x5=15 Values: The energy and Vector Stokes drift in each frequency
bin. Note the magnitude of the stokes drift is not simply proportional
to f_3^3.
This total is 21 values at each saved grid point in the wave model, but
since H_s and 2 different mean periods are (I think) equivalent to 3 of
these, it is an additional 18 values on top of other parameters, down
from 27 in the static CDIP frequency bins and probably more robust. It's
probably an improvement over a long-ago (early 90's) suggestion of M. Li
&/or C. Garrett to use just 2 exponentials.
The surface vector Stokes drift profile would then be combined with a
subsurface one reconstructed using 5 exponentials with coefficients
given by the bin averages and decay scales 8pi^2/f_3{i,j} where f_3{i,j}
is the 1/3 root of the 3rd moment with angular weighting, determined
from the magnitude of the Stokes drift and the energy in the bin [f_i,
f_j] .
Not sure how well all this would work. Could be the frequency bins I've
chosen are not the most effective and that other choices may be better,
so some testing could be done to improve this straw man. Some of this is
triggered by some work Baylor recently passed to me, so he may have some
modifications to suggest on the frequency bins and moments, where my
definitions here are off the top of my head.
Please chime in if you want to work together on this or if you think
this is TMI in any of several ways ...
-Ramsey Harcourt
[email protected]
206 221 4662
Baylor Fox-Kemper wrote:
Hi All,
Agree with Eric and Jim. We'd like to see either a spectrum or a
list of moments 0-3+, preferably directional, but just magnitude would
be nice.
We are trying to get at a Stokes drift climatology for the forcing
of near-surface wave-drivent turbulence in climate models.
Directional 3rd moment would be a nice direct measure of surface
Stokes drift. However, the whole spectrum is needed for direct
subsurface info. Jim is right that breaking is much harder. But, if
there commonly were spectra, even 1d for data lacking directional
sensors, this would be fantastic!
Furthermore, from a cf point of view, putting it out there that we
care about things other than significant wave height might stimulate
modelers and data to release these other variables.
We have found that the wave models do a great job on variables they
assimilate globally (e.g., significant wave height from altimeters).
They do an OK job with variables they assimilate sparsely (e.g., buoy
mean wave period). This is true even when the same models are run
*not assimilating data*. That is, if there's data to compare to then
the model algorithms and parameterizations improve to get it right.
Models do a significantly poorer job of getting the spectral shapes
right, since there is little ot no systematic data to constrain them
or train them against. So, if cf gives out a spectral format, then
some sites will keep them. Even if it's nondirectional, it can still
be used to constrain and train models.
Cheers,
-Baylor
On Wed, Oct 27, 2010 at 12:22 AM, Jim Thomson
<[email protected]> wrote:
Agreed with Eric, completely. The big difficulty is that many sites do not
have instruments capable of resolving directional spectra... some of these
sites even the scalar spectra are not very good. The high-frequency tail is
thought to be important for many of the quantities you list, so posting full
spectra is probably the best approach.
For example, bulk steepness (amplitude x wavenumber at the peak) is often less important
for wave breaking than "wave saturation" (spectral steepness given by the 5th
moment) at frequencies up to several multiples above the peak. Still, this is not
sufficient to prescribe breaking (else I would retire and go surfing). Estimating these
higher-order/dynamic quantities may be beyond the scope of the OS database, but at least
if full spectra were available, users could make their own estimates. This has been
working well for CDIP.
-Jim
On Oct 26, 2010, at 10:24 PM, Eric D'Asaro wrote:
Meghan et. al:
I think we would be missing a great opportunity if we used only the wave
variables proposed below. We have good dynamical reasons to think that waves
are important in upper ocean dynamics and not just for purposes of directing
shipping or filling up mandated data bases. I expect that in the future, ocean
models will be coupled to wave models (as is already done for tropical cyclone
models) and that historical wave data will be important in developing and
verifying these models. In another project we are struggling with trying to
reconstruct dynamically appropriate wave spectra from historical wave data that
is archived in such a way as to make proper reconstruction very difficult. It
would be a great waste if OceanSITES makes the same mistake.
Variables should be added to the list that are sufficient to
1. Resconstruct the important features of the directional wave spectrum,
including energy and momentum flux and their depth profiles
2. Provide indications of the wave steepness and breaking probabilities.
Specific suggestions are perhaps better made by others (Baylor?, Jim? Ramsey?).
-------- Original Message --------
Subject: Fwd: Seeking new CF standard names (waves and biogeochem)
Date: Thu, 21 Oct 2010 10:37:36 -0400
From: Nan Galbraith <[email protected]>
Reply-To: [email protected]
To: ots-dmt <[email protected]>
Hi all -
For anyone in OceanSITES who measures waves, there's a recent
discussion on the CF mailing list about new standard names. This
would be a good time to weigh in, if you haven't already, since we'll
want to use these terms in OceanSITES data files if and when we
include wave parameters.
You need to be a member of the CF list to post a reply. You can join
at
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
.
There's also been some discussion on biogeochem terms that might
be useful to some here.
Regards -
Nan
-------- Original Message --------
Subject: Re: [CF-metadata] Seeking new CF standard names (9) for sea
surface wave parameters
Date: Thu, 21 Oct 2010 11:09:29 +1100
From: andrew walsh
<[email protected]>
Hi Jonathon and CF metadata list,
Summarising our discussions thus far we propose:
2 new Cell methods:
root_mean_square
mean_of_upper_decile
and these new standard names:
sea_surface_wave_height (common concept)
sea_surface_wave_mean_crest_period
sea_surface_wave_significant_wave_period
sea_surface_wave_period_at_second_largest_peak_of_variance_spectral_density
sea_surface_wave_variance_spectral_density_zeroth_frequency_moment
sea_surface_wave_root_mean_square_amplitude_from_variance_spectral_density
The sea_surface_wave_height is a common concept (standard name) which may be
qualified by a cell_method attribute to realise the actual variable. The
sea_surface_wave_height
when combined with a cell method of:
time: mean
time: maximum
time: root_mean_square
time: mean_of_upper_decile
will describe the statistical wave height variables:
sea_surface_mean_wave_height
sea_surface_maximum_wave_height
sea_surface_root_mean_square_wave_height
sea_surface_wave_mean_of_highest_one_tenth_waves
respectively.
I have attached a spreadsheet which contains the names, descriptions and units
of the variables proposed.
Looking forward to getting final approval to add these to the CF name lists.
Andrew Walsh
Data Facilitator AODN
----- Original Message -----
From: "Jonathan Gregory"
<[email protected]>
To: "andrew walsh"
<[email protected]>
Cc:
<[email protected]>; "Mark Kulmar"<[email protected]>
Sent: Wednesday, October 13, 2010 20:21
Subject: [CF-metadata] Seeking new CF standard names (9) for sea surface wave
parameters
Dear Andrew
Yes, but depends on the community (list) accepting the idea of having a
Good. I think we are agreed then to propose these new cell methods:
root_mean_square
mean_of_upper_decile
and these new standard names:
sea_surface_wave_height
sea_surface_wave_mean_crest_period
sea_surface_wave_significant_wave_period
sea_surface_wave_period_at_second_largest_peak_of_variance_spectral_density
sea_surface_wave_variance_spectral_density_zeroth_frequency_moment
sea_surface_wave_root_mean_square_amplitude_from_variance_spectral_density
I agree that the last will avoid confusion with the RMS of wave height.
I agree also that this depends on users being comfortable with the concepts
being split into two attributes, in accordance with the usual CF practice,
but deciding on how to join them up as common_concepts will help. Comments
from others would help. It is good that Roy supports this compromise.
Best wishes
Jonathan
Cliquez sur l'url suivante
https://www.mailcontrol.com/sr/tjkP04wPWEHTndxI!oX7UpBlscM9Pl1fAM1p1zO6h070olZRfJ6KoHxitOmXf8!eLCPtl9EVgldLRkMuvIdUTQ==
si ce message est indésirable (pourriel).
<Proposed_wave_parm_names_for_CFlist.xls><Attached Message Part.txt>
_______________________________________________
CF-metadata mailing list
[email protected]
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
