A few general questions

[sophia-wright-blue]

Hello - thank you for releasing this repository - I have a few general questions that would help me understand the library better:

1. You mention the README that The particle tracking was made with oceanParcels, but I don't see Parcels being used anywhere in the code - could you please clarify how you've used Parcels for particle tracking.

2. You also mention that If you wish to read the data into python, contact me and I can show you how to access the Zarr files with the connectivity data - is this done with the script getEZfateFromOSN.py ?

3. In the python docs, you specify a maskfile:

`maskFile='EZfateData/EZfateFiles/ext-PSY4V3R1_mesh_zgr.nc'

and the comment says:

#get the file which contains the bathymetry and grid data for the mercator ocean model and for the connectivity files

could you point me the original url/data source where this netcdf file is available (not the OSN s3 bucket where you host it)

Thank you!

[JamiePringle]

1: The code for the parcels particle tracking is not on the GitHub, for it is long and complex and in the older parcels version version 2. However, if you want to view it, you are welcome to ask. I have attached the main particle tracking codes below, but there are another 5 scripts that are used to get to the final R matrices that I share. Ask if you want the whole pipeline. trackPointsScript.py is the core of the particle tracking, and 01_makeDispersalData_fromHabitat*py sets up the initial conditions and calls trackPointsScript.py. Again, if you wish to see the full processing tool chain, feel free to ask. It is not the most elegant code...

2: getEZfateFromOSN.py will get the data files, but there is also some python analysis code that is provided to fetch and start analyzing the data in python. See https://jamiepringle.github.io/EZfate/EZfate_python_guide.html for a description of how to get started. Vogt-Vincent has used this code to publish in https://www.science.org/doi/pdf/10.1126/sciadv.adr2545

3: EZfateData/EZfateFiles/ext-PSY4V3R1_mesh_zgr.nc is a slightly modified version of the file provided by Mercator ocean. We trimmed off the periodic padding points on the west side of the domain. This was a mistake, but it is now imbedded in our processing code. This is the version you want to use with my code. The original version, which is 1 or 2 grid-points wider in the east/west direction, can be found at https://tds.mercator-ocean.fr/thredds/psy4v3r1/psy4v3r1_dbb_00.html, specifically here: https://tds.mercator-ocean.fr/thredds/psy4v3r1/psy4v3r1_dbb_00.html?dataset=PSY4V3R1_mesh_zgr

01_makeDispersalData_fromHabitat_SURGE_splitYear.py
trackPointsScript.py

[sophia-wright-blue]

Thank you for the detailed response @JamiePringle .

Could you please share the remaining OceanParcels scripts please?

[JamiePringle]

I would be happy to, but can you first explain what you are thinking of doing with them? If you are just auditing them, that is fine, I will upload it as is. But if you want to replicate this work, perhaps we should talk about some changes I would make to the code that could save a great deal of time. This is particularly true if you using them for a smaller domain/higher resolution model.

Also, if you are using this code, could you send me your institution or company, so I can report it on my annual reports to the NSF? They like to know who is finding the work they fund useful. You can find my contact information on my web page (https://oxbow.sr.unh.edu), or you can post it below if you don't mind it being public. I don't require this, but it helps me a lot.

Jamie

[JamiePringle]

Just to re-iterate, I am not requiring you to tell me who you are or what you are doing with these scripts, but it would help me to know who is using them, and if I know what you are doing I might be able to save you a great deal of time.

[sophia-wright-blue]

sorry for the delay in my response - I was away for the weekend.

My primary interest is in Lagrangian Particle tracking (LPT) - I was working with the package Parcels, but their examples are all Europe centric

I came across your library when I was looking for applications of Parcels on the east coast

I would like to reproduce your work, so any guidance you can provide would help - if it's easier for you, it'd also help if you could point me to a different repo/example on using LPT with Parcels, or any other Python based LPT library that has been applied on the east coast.

I currently work as a Quantitative Researcher at a large investment bank in NY, and have begun to explore quantitative Oceanography applications because I'd like to work on something useful - I'm very familiar with advanced Math, Physics and coding.

Greatly appreciate your help!

[nvogtvincent]

My primary interest is in Lagrangian Particle tracking (LPT) - I was working with the package Parcels, but their examples are all Europe centric

For some reason I've been getting email notifications for this thread. EZfate is an excellent product and I'm sure Jamie's code will be useful for you to learn from. However, I'm not sure why the "Europe centric" examples from Parcels are a problem? There's no fundamental difference between particle tracking code applied to European waters vs elsewhere - code examples written for a European domain will work just fine for the US east coast (or indeed anywhere else) as long as the physics are appropriate for your question of interest (which applies to EZfate as well). I'm not in any way suggesting you shouldn't use EZfate, I would just suggest not discounting potentially useful code just because the domain they happen to have used isn't your domain of interest.

[sophia-wright-blue]

oh sure - I still intend to use Parcels, but my eventual objective is to learn and apply LPT to the east coast, where I live and can hopefully collaborate with researchers who are focused on this part of the ocean - LPT is just one part of my ocean modeling effort

[JamiePringle]

Sophia -- I will post the full tool chain to a separate issue thread over the course of the day. You are welcome to use it, but I would not just grab it blindly because 1) it use the old version of parcels, and 2) it is very heavily tuned to do large scale global runs that feed into EZfate -- it is not the way I would do it for a small regional example. In particular, for a small regional example I would keep the output in lat/lon space, not in grid coordinates, and I would write the output out much more frequently.

In general, would echo @nvogtvincent comments.

Finally, the hardest part of particle tracking is not the particle tracking code, it is the underlying source of the flow fields. The Mercator GLORYS model that I use for EZfate is probably the best for a global problem. However, if you are doing a regional problem, a regional ocean model will often be best. The recommendations I give here are perishable, but for now they are sensible. For the East Coast of the US, I would suggest the work of the Rutgers Ocean Modeling Group (https://gmd.copernicus.org/articles/13/3709/2020/ and https://tds.marine.rutgers.edu/thredds/catalog/roms/doppio/catalog.html , but I would talk to John Wilkins on there more recent projects, and when they will become operational). For the west coast I would start with the NOAA regional model https://coastaloceanmodels.noaa.gov/WCOFS/ , and I would talk to Alexander Kurapov about where that system is going.

I will mention your name the issue where I post the code to create the EZfate data, so you will get an update.

Jamie

[JamiePringle]

@sophia-wright-blue , the code is now posted as the second issue, you can follow this link

[sophia-wright-blue]

thank you so much for releasing the code and the detailed guideline @JamiePringle !

If I could ask just one more follow-up question to your comment about the underlying source of the flow fields - for a regional ocean model, how do you choose between the different options for the east coast:

Doppio (Rutgers model based on ROMS) vs FVCOM vs something that's based on NEMO (such as CIOPS-E)

if you're aware of a paper or documentation that evaluates the different regional models, please point me to those

Thank you!

[JamiePringle]

@sophia-wright-blue , in general, the exact model code (ROMS, NEMO, etc) does not matter as much as the configuration of the model, its forcing, care with boundary conditions, etc. All of the modern codes can be made to work.

One good paper is Wilkin J, Levin J, Moore A, Arango H, López A, Hunter E (2022) A data-assimilative model reanalysis of the U.S. Mid Atlantic Bight and Gulf of Maine: Configuration and comparison to observations and global ocean models. Progress in Oceanography 209:102919.

But of course, these things become rapidly out of date. I have no experience with CIOPS-E. I would talk to John Wilkin at Rutgers or someone else in his group (Levin and Lopez are good). Or Ruoying He at NC State. John is working on what I believe will become the NOAA East Coast modeling system, and his group does good and careful model inter-comparisons.

Again, one thing anyone who talks to you will ask you is "what are you trying to do with the model"? The answer of what is best depends sensitively on what your goal is. Especially if you are trying to use it to train some AI or ML scheme, it is worth spending some time figuring out what you will need to get from the data before you seek the data.

Jamie

[sophia-wright-blue]

thank you for all of all your thoughtful responses @JamiePringle - I learned a lot from this interaction and look forward to learning more from your work.