[ecco-support] ECCO Mixed Layer Depths criteria
Martha Buckley
marthabuckley at gmail.com
Tue May 2 11:47:36 EDT 2017
Definitely not my idea! Lots of others have used it. In fact, the
definition used in the ECCO v4 "MXLDEPTH" is the density equivalent of
0.8C. But I agree that it is a good one.
Martha
On Tue, May 2, 2017 at 11:39 AM, Matthew Mazloff <mmazloff at ucsd.edu> wrote:
> Hi Martha
>
> Never tried a density equivalent of a temperature threshold — nice idea!
>
> Thanks!
> Matt
>
>
> On May 2, 2017, at 8:11 AM, Martha Buckley <marthabuckley at gmail.com>
> wrote:
>
> Thanks Matt! ECCO v4 does output DRHODR. I'd be curious about getting
> your code at some point, but for my present analysis, which involves
> comparing to other products (without DRHODR available) and observations, I
> can't really take that route. Unfortunately, I think my only choices are
> (1) use whatever arbitrary MLD definition is output by the various ocean
> data products I am using or (2) recompute MLD from some standard definition
> for all products using monthly mean T/S, neither of which is preferable.
> With that said, the focus of my work is not on mixed layer depths, it is
> on predictability of upper ocean heat content, so the accuracy of the MLD
> definition may not be key.
>
> Regarding the different criteria in different regions, I thought that
> using a density equivalent of a temperature threshold took care of this to
> some degree since the thermal expansion coefficient is smaller at low
> temperatures.
>
> Thanks!
> Martha
>
>
>
> On Mon, May 1, 2017 at 11:41 PM, Matthew Mazloff <mmazloff at ucsd.edu> wr
> ote:
>
>> Hello
>>
>> To get MLD from a model solution one should really look at first and
>> second vertical derivatives of density. Does ECCO v4 output DRHODR
>> diganostic? If so I can share code to get MLD.
>>
>> Observations are too noisy to look at vertical derivatives of density so
>> people typically use thresholds. But an appropriate threshold is regionally
>> variable so there are somewhat complicated algorithms for the global
>> problem (e.g. http://mixedlayer.ucsd.edu). So to compare to data it is
>> appropriate to use a threshold criteria.
>>
>> I have looked at MLD from a density threshold criteria, from the Holte
>> algorithm, and from DRHODR in the MITgcm. Really they aren’t qualitatively
>> different, but there are quantitative differences with the threshold
>> criteria tending to give deeper values by missing some restratification
>> events. I like the DRHODR method, but since it can’t be compared to obs
>> the threshold is nice
>>
>> I would suggest using a potential density cutoff is fine for most
>> applications, just tune it to your location. E.g. 0.3 may be fine in
>> subtropics but may want 0.2 in higher lats. MLD is very nonlinear, but
>> isn’t as bad as mixing *layer* depth, which is the instantaneous depth
>> of mixing and is very non-linear and should be calculated on the fly from
>> vertical diffusivity or KPP_HBL or some other method. MLD is sort of the
>> integrated mixing layer depth….so maybe even taking it from pot dense
>> calculated from monthly mean T and S is OK for an approximate average
>> value, though obviously not preferable!
>>
>> Hope this quick brain dump helps!
>>
>> Matt
>>
>>
>>
>>
>>
>>
>>
>> On May 1, 2017, at 7:44 PM, Fenty, Ian G (329C) <Ian.Fenty at jpl.nasa.gov>
>> wrote:
>>
>> Martha,
>>
>> I wouldn't say that there is a single MLD definition for ECCO v4 state
>> estimates. Yes, there is a MLD output field that uses the 0.8 deg C
>> criteria but I believe that MLD has no prognostic effect on the solution.
>> Model output can be used to calculate alternative MLDs. For example, Gael
>> provides three alternative MLD criteria (Kara, Suga, Boyer) to show the
>> wintertime and summertime MLD plots as part of the standard analysis
>> package.
>>
>> See pp 140-142 of
>> ftp://snowwhite.jpl.nasa.gov/data4/Release3//doc/standardplots.pdf
>>
>> In the Labrador Sea I used the following criteria to identify the base of
>> mixed layer: the shallowest depth level with a potential density exceeding
>> the profile minimum by 13 g m^−3.
>>
>> Note that ECCOv4 uses the GGL90 package not KPP so there is no KPPhbl
>> field.
>>
>> If you want exact monthly-mean MLDs from the state estimate using an
>> alternative criteria (e.g., potential density cutoff) then the forward
>> model would have to be rerun since the MLD diagnosed from monthly-mean T
>> and S are almost certainly not equal to the monthly mean of MLD because of
>> nonlinearities.
>>
>> -Ian
>>
>>
>>
>> On 5/1/17, 7:18 PM, "ecco-support-bounces at mit.edu on behalf of
>> Menemenlis, Dimitris (329C)" <ecco-support-bounces at mit.edu on behalf of
>> Dimitris.Menemenlis at jpl.nasa.gov> wrote:
>>
>> Hi Martha, I don't have answer for your question. I return discussion
>> to ECCO support, and also include Daria in cc, in case someone else has
>> advice for you.
>>
>> On May 1, 2017, at 6:39 PM, Martha Buckley <marthabuckley at gmail.com>
>> wrote:
>>
>>
>>
>> Thanks for the replies. It seems that all the ECCO products use the
>> 0.8C density equivalent.
>>
>>
>> The reason I am asking is that I am calculating predictability
>> timescales for temperature integrated over the maximum climatological
>> (e.g., wintertime) MLD for a variety of different ocean data products. I
>> am trying to be as comparable as possible between
>> products in the definition of MLD, but the MLD definitions for the
>> products varies wildly (e.g. 0.4 kg/m^3 for ECDA and 0.8 C for ECCO v4).
>> The question is whether the most comparable approach is to apply the same
>> criteria to all models/data (using for example
>> monthly output) or whether there is a better approach.
>> Thanks!
>> Martha
>>
>>
>>
>> ---------- Forwarded message ----------
>> From: Wang, Ou (329B) <Ou.Wang at jpl.nasa.gov>
>> Date: Mon, May 1, 2017 at 1:43 PM
>> Subject: Re: [ecco-support] ECCO Mixed Layer Depths criteria
>> To: "ECCO support list, wider membership" <ecco-support at mit.edu>
>>
>>
>> The same default parameters were used for ECCOv4:
>> hMixCriteria = -.8 _d 0
>> dRhoSmall = 1. _d -6
>> hMixSmooth = 0. _d 0
>>
>>
>> The criteria of the density equivalent of 0.8C is applied to the model
>> profiles at the model time step.
>>
>>
>> Ou Wang
>>
>>
>> From: <ecco-support-bounces at mit.edu> on behalf of "Menemenlis,
>> Dimitris (329C)" <Dimitris.Menemenlis at jpl.nasa.gov>
>> Reply-To: "ECCO support list, wider membership" <ecco-support at mit.edu>
>> Date: Monday, May 1, 2017 at 10:19 AM
>> To: "ECCO support list, wider membership" <ecco-support at mit.edu>
>> Subject: Re: [ecco-support] ECCO Mixed Layer Depths criteria
>>
>>
>>
>> Don’t know what criterion was used for ECCOv4
>> For cube92, the default parameters were used:
>>
>>
>> C hMixCriteria:: criteria for mixed-layer diagnostic
>> C dRhoSmall :: parameter for mixed-layer diagnostic
>> C hMixSmooth :: Smoothing parameter for mixed-layer diag
>> (default=0=no smoothing)
>>
>> hMixCriteria = -.8 _d 0
>> dRhoSmall = 1. _d -6
>> hMixSmooth = 0. _d 0
>>
>> C-- First method :
>> C where the potential density (ref.lev=surface) is larger than
>> C surface density plus Delta_Rho = hMixCriteria * Alpha(surf)
>> C = density of water which is -hMixCriteria colder than surface
>> water
>> C (see Kara, Rochford, and Hurlburt JGR 2000 for default criterion)
>>
>>
>>
>> See model/src/calc_oce_mxlayer for other options.
>>
>>
>> Another available definition (when using KPP scheme)
>> is KPPhbl, which is the “mixing” layer depth.
>> The daily maximum of KPPhbl is a very good proxy for
>> mixed layer depth most places, except seasonally-sea-ice-covered and
>> polar-winter regions.
>> (Kara’s criterion is also problematic for polar regions, which tend to
>> be salt-stratified.)
>>
>>
>>
>> On May 1, 2017, at 9:50 AM, Martha Buckley <marthabuckley at gmail.com>
>> wrote:
>>
>> One follow up question regarding mixed layer depths in ECCO.
>>
>>
>> What is the criteria for the variable "MXLDEPTH" in the ecco v4
>> product. I think I recall being told it is the density equivalent of
>> 0.8C. Two questions:
>> (1) Is this the criteria that was used? It the criteria applied to
>> model profiles at the model timestep?
>> (2) If so, isn't this criteria a bit large? 0.125 kg/m^3 or 0.5C seem
>> common MLD criteria when applying criteria to monthly-averaged data.
>> However, criteria that are smaller (~0.03 kg/m^3 or ~0.2 C) are typically
>> applied when using raw observational
>> profiles.
>>
>>
>> I know that the MLD criteria is somewhat arbitrary, but I want an idea
>> of the rational for the chosen criteria.
>> Thanks!
>> Martha
>>
>>
>> On Fri, Apr 28, 2017 at 6:04 PM, HONG ZHANG
>> <hong.zhang at ucla.edu> wrote:
>>
>>
>>
>>
>>
>> -------- Forwarded Message --------
>> Subject: Re: [ecco-support] ECCO2 Mixed Layer Depths ConcernsDate:
>> Fri, 28 Apr 2017 12:48:59 -0700From: HONG ZHANG
>> <hong.zhang at ucla.edu> <mailto:hong.zhang at ucla.edu>To: Dimitris
>> Menemenlis
>> <dmenemenlis at gmail.com> <mailto:dmenemenlis at gmail.com>, Moisan, John
>> R. (WFF-610W)
>> <john.r.moisan at nasa.gov> <mailto:john.r.moisan at nasa.gov>CC: Daria
>> Halkides
>> <halkides at esr.org> <mailto:halkides at esr.org>
>>
>> On 4/28/17 9:37 AM, Dimitris Menemenlis wrote:
>>
>>
>> Daria or Hong, would you have looked at MXLDEPTH depth in the “cube92”
>> solution?
>> ftp://ecco2.jpl.nasa.gov/data1/cube/cube92/lat_lon/quart_
>> 90S_90N/MXLDEPTH.nc/
>>
>>
>> John, why not use the ECCO-V4, which is data-constrained?
>> http://www.ecco-group.org/products.htm
>> ECCO v4 is on a 1-degree grid, which is closer to your desired 2x2
>> degree grid.
>>
>> Dimitris Menemenlis
>>
>>
>> On Apr 26, 2017, at 9:41 AM, Moisan, John R. (WFF-610W) <
>> john.r.moisan at nasa.gov> wrote:
>>
>> Hello ECCO2’ers,
>>
>> I would like to make use of the Mixed-Layer Depth model outputs from
>> the global model runs 1992-present (Nov. 2016) to drive some simple 1D
>> biogeochemical models for CDOM. I have binned to model outputs from the
>> jpg site into monthly 2x2 deg. grids. One thing
>> of concern that I have is the trends in the global MLDs (see attached
>> figure), which are more extreme in the SO. Is there an issue with
>> maintaining correct ocean boundary layers, esp. in the SO for this model?
>> Has this been observed yet?
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> Hi John,
>> You're right, MLD from cube92 run has bias in late stage
>> because it's a free run for the later period.
>> As Dimitris said, you can look at other ECCO product provided at the
>> website.
>> Please see attached fig showing MLD from ECCOv4 and ECCO llc270 for an
>> example,
>> where red line is for ECCOv4 (1992-2012) and blue line for ECCO llc270
>> (2001-2015) in upper plot
>> also MLD for cube92 is also shown in lower plot for reference.
>>
>> cheers
>> Hong
>>
>>
>>
>>
>>
>> _______________________________________________
>> ecco-support mailing list
>> ecco-support at mit.edu
>> http://mailman.mit.edu/mailman/listinfo/ecco-support
>>
>>
>>
>>
>>
>>
>>
>>
>> --
>> Martha W. Buckley
>> marthabuckley at gmail.com
>> http://sites.google.com/site/marthabuckley/home
>>
>>
>>
>>
>> _______________________________________________
>> ecco-support mailing list
>> ecco-support at mit.edu
>> http://mailman.mit.edu/mailman/listinfo/ecco-support
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> ecco-support mailing list
>> ecco-support at mit.edu
>> http://mailman.mit.edu/mailman/listinfo/ecco-support
>>
>>
>>
>>
>>
>>
>> --
>> Martha W. Buckley
>> marthabuckley at gmail.com
>> http://sites.google.com/site/marthabuckley/home
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> ecco-support mailing list
>> ecco-support at mit.edu
>> http://mailman.mit.edu/mailman/listinfo/ecco-support
>>
>>
>>
>
>
> --
> Martha W. Buckley
> marthabuckley at gmail.com
> http://sites.google.com/site/marthabuckley/home
>
>
>
--
Martha W. Buckley
marthabuckley at gmail.com
http://sites.google.com/site/marthabuckley/home
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.mit.edu/pipermail/ecco-support/attachments/20170502/def60352/attachment-0001.html
More information about the ecco-support
mailing list