[Met_help] [rt.rap.ucar.edu #97360] History for MTD Questions

Tue Nov 10 09:35:59 MST 2020

----------------------------------------------------------------
  Initial Request
----------------------------------------------------------------

I'm writing this letter because there are a couple questions about MODE-Time Domain (MTD) that I wish to have your expert opinion on. 

When I gave a presentation about my work, some people asked me about how the merging and splitting of objects are done. I'm running MTD on a single field (mtd -single xxx). So there is no matching between two fields. However, I believe there is still merging and splitting of objects in the same field. I did not do any configuration for merge_flag and merge_thresh, so I think the default setting about merging are used, which is the double thresholding merging with a merge threshold of 1.25. So the simple objects are merged when they are connected under the mask of 1.25 threshold, and split up when then are no longer connected under the mask of 1.25 threshold. Am I right?

If I use the merge_flag = ENGINE setup then the field is duplicated and compared to themselves using a fuzzy engine approach with the same setting from my configuration file. Will this be better than the double thresholding method?

When using mtd -single, the mtd output has only "simple objects", no "cluster objects" (The output files are mtd_20170501_000000V_2d.txt, mtd_20170501_000000V_3d_single_simple.txt, and mtd_20170501_000000V_obj.nc). But these "simple objects" are in fact clusters because they having gone through merging (even though no matching). Am I right? 

Thank you very much for your time and help!

Sincerely,
Yuxing

----------------------------------------------------------------
  Complete Ticket History
----------------------------------------------------------------

Subject: MTD Questions
From: John Halley Gotway
Time: Fri Nov 06 10:00:49 2020

Yuxing,

In case you haven't seen this, we've moved the MET User's Guide over
to a website. And here's the chapter about MTD:

https://dtcenter.github.io/MET/latest/Users_Guide/mode-td.html

I suspect your are getting confused by the configuration options which
are available for the 2D MODE tool versus those available for the
MODE-Time-Domain tool. Several of the MODE config options are NOT
currently available for MTD.

The double-thresholding logic does work the way you described for the
MODE tool. But the double-thresholding merging technique is not
supported for MTD. And while merge_flag is supported for MODE, it is
not supported for MTD.

I'd recommend looking closely at the default configuration files for
MODE and MTD here:

https://github.com/dtcenter/MET/blob/main_v9.1/met/data/config/MODEConfig_default

https://github.com/dtcenter/MET/blob/main_v9.1/met/data/config/MTDConfig_default

MTD only supports the options that are listed in that default
configuration file.

Regarding your last question on the definition of simple vs cluster
objects. I would say no, those are simple objects, not clusters. By
definition, a cluster is a group of one or more objects from one field
(e.g. fcst or obs) that has been matched to a group of one or more
objects from another field. Since you're not doing any matching, it is
not possible to have clusters.

It is perhaps a little confusing. For 2D MODE, a simple object is a
contiguous collection of grid points at which the object definition
criteria is met.

For MODE-Time-Domain, a simple object is a collection of all grid
points in space and time that are all part of the same space/time
object. At one time slice that 3D space time object may be split into
2 or more pieces, but it's part of the same space/time object because
at some other time, they were joined together.

Think of an MTD simple object like a ginger root. If you were to slice
the fingers off, the slices would appear to be separate. But they are
part of the same object.

Confused yet? Hope that helps.

Thanks,
John

------------------------------------------------
Subject: MTD Questions
From: John Halley Gotway
Time: Mon Nov 09 10:03:08 2020

Yuxing,

Yes, your understanding is correct. If at one time two 2D traditional
MODE
objects are separate objects... but at the next time, they are joined
together, they are part of the same 3D space/time object. Likewise, if
at
one time you have a large 2D traditional MODE object which at a later
time
splits into multiple pieces, they are part of the same 3D space/time
object.

To illustrate, I've attached an animated GIF I created by running the
NetCDF output from MTD through MET's plot_data_plane tool. I plotted
the
object field for each timestep.
There are 6 total 3D space time objects here. Notice, for example,
that the
gray one with object ID = 1, is split into many pieces at the first
time
step. But by the end of the time period, they have all joined
together.

Thanks,
John

On Mon, Nov 9, 2020 at 1:25 AM yunyx at cma.gov.cn via RT
<met_help at ucar.edu>
wrote:

>
> <URL: https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=97360 >
>
> Dear John,
> Thank you! I appreciate the detailed explanations, which clarifies
most of
> my
> confusions. I would still like a little clarification on the
following:
> "For MODE-Time-Domain, a simple object is a collection of all grid
points
> in
> space and time that are all part of the same space/time object. At
one time
> slice that 3D space time object may be split into 2 or more pieces,
but
> it's
> part of the same space/time object because at some other time, they
were
> joined
> together."
> Say my conv_thresh = 5 mm/h. By "joined together at some other
time", do
> you
> mean that at some time, the pieces are connected together after
applying
> the
> mask of 5 mm/h on the convolved field? Or in other words, they
fulfill the
> criteria of simple object for 2D MODE at this time step? So the
"ginger
> root"
> includes the life cycle of all the pieces that are ever connected
together,
> even if they were connected for just one time step?
> Thanks again and look forward to your reply!
> Sincerely,YuxingOn 11/7/2020 01:03， John Halley Gotway via
> RT<met_help at ucar.edu> (mailto:met_help at ucar.edu) wrote：
>
>   Yuxing,
>
>   In case you haven't seen this, we've moved the MET User's Guide
over to a
>   website. And here's the chapter about MTD:
>
>   https://dtcenter.github.io/MET/latest/Users_Guide/mode-td.html
>
>   I suspect your are getting confused by the configuration options
which
> are
>   available for the 2D MODE tool versus those available for the
>   MODE-Time-Domain tool. Several of the MODE config options are NOT
> currently
>   available for MTD.
>
>   The double-thresholding logic does work the way you described for
the
> MODE
>   tool. But the double-thresholding merging technique is not
supported for
>   MTD. And while merge_flag is supported for MODE, it is not
supported for
>   MTD.
>
>   I'd recommend looking closely at the default configuration files
for MODE
>   and MTD here:
>
>
>
https://github.com/dtcenter/MET/blob/main_v9.1/met/data/config/MODEConfig_default
>
>
>
https://github.com/dtcenter/MET/blob/main_v9.1/met/data/config/MTDConfig_default
>
>   MTD only supports the options that are listed in that default
> configuration
>   file.
>
>   Regarding your last question on the definition of simple vs
cluster
>   objects. I would say no, those are simple objects, not clusters.
By
>   definition, a cluster is a group of one or more objects from one
field
>   (e.g. fcst or obs) that has been matched to a group of one or more
> objects
>   from another field. Since you're not doing any matching, it is not
> possible
>   to have clusters.
>
>   It is perhaps a little confusing. For 2D MODE, a simple object is
a
>   contiguous collection of grid points at which the object
definition
>   criteria is met.
>
>   For MODE-Time-Domain, a simple object is a collection of all grid
points
> in
>   space and time that are all part of the same space/time object. At
one
> time
>   slice that 3D space time object may be split into 2 or more
pieces, but
>   it's part of the same space/time object because at some other
time, they
>   were joined together.
>
>   Think of an MTD simple object like a ginger root. If you were to
slice
> the
>   fingers off, the slices would appear to be separate. But they are
part of
>   the same object.
>
>   Confused yet? Hope that helps.
>
>   Thanks,
>   John
>
>
>

------------------------------------------------