[julia-neme]

This one was comparing geostrophic surface velocities from sea level to those of the first cell instead of usurf vsurf, but otherwise OK. I've also updated intake loading.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

anton-seaice commented on 2026-07-16T02:00:18Z
----------------------------------------------------------------

Have you tested this with MOM6 ?

I guess you also need to get the right combination of

      dxCu: # dxCu is delta x at u points [L ~> m]

      dyCu: # dyCu is delta y at u points [L ~> m]

      dxCv: # dxCv is delta x at v points [L ~> m]

      dyCv: # dyCv is delta y at v points [L ~> m]

instead of dxu and dyu

it looks like uo and vo are 3d fields as well ?

julia-neme commented on 2026-07-20T02:49:38Z
----------------------------------------------------------------

True! I can replace with SSU and SSV,but I don't think they're available for any run. I'll add the caveat.

And regarding the right combination, since discovering that I can use .derivative() you don't need to think about combining interp and diff, so I might change the script to use that.

(see comment below)

anton-seaice commented on 2026-07-20T03:27:41Z
----------------------------------------------------------------

I think using .derivative would calculate quanitites at the cell centre, so the model velocities need to be interpolated to cell centres then?

SSU and SSV are just the top layer in uo and vo right? So it's ok to use them, just note it in the table ?

julia-neme commented on 2026-07-27T00:22:14Z
----------------------------------------------------------------

Ok, so now I've included more information on how to adapt to mom6 ! Also changed to using .derivative()

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

anton-seaice commented on 2026-07-16T02:00:20Z
----------------------------------------------------------------

Why id boundary = 'extend' ?

I think boundary = None might work

julia-neme commented on 2026-07-27T00:21:43Z
----------------------------------------------------------------

I am not aware of this condition, I only know extend (that just copies the nearest value) or fill (fills with something of your choice). I've actually changed it to fill with nans.

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

anton-seaice commented on 2026-07-19T23:31:51Z
----------------------------------------------------------------

can these two lines be combined ?

variable=['dxu','dyu']

these are just merged below anyway ?

julia-neme commented on 2026-07-20T02:47:29Z
----------------------------------------------------------------

I wanted that yes! But intake only opened one of them. So I had to open 1by1 and merge.

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

anton-seaice commented on 2026-07-19T23:31:52Z
----------------------------------------------------------------

Line #2.    detadx = grid.interp(grid.diff(ds['sea_level'], 'X', boundary='extend'), 'Y', boundary='extend') / ds['dxu']

It's a bit hard to follow this, but it looks like this:

1. takes the diff of the 'sea_level' field. This should interpolate to the middle of the top face.
2. interpolates those to the middle of the cell
3. then divides by the dxu - but surely dxu is the width of the top face?

shouldn't step 3 be either:

a. divide by dxt - the width across the middle of the cell, or

b. moved to between steps 1 and 2

(Assuming B-grid)

This calculation presumably would need changing for C-grid, to get rid of the .interp ?

anton-seaice commented on 2026-07-19T23:59:12Z
----------------------------------------------------------------

Hangon - i dont have that correct, ill revisit

anton-seaice commented on 2026-07-20T01:00:17Z
----------------------------------------------------------------

sea_level is a tracer/center point quantity, so this:

1. takes a diff across two center points to get the quantity in the middle of a face,
2. interpolates across two middle of face quantities to get the cell corner
3. divides by the width/length of the corner cell

so that is correct for B-grid,

julia-neme commented on 2026-07-20T02:48:10Z
----------------------------------------------------------------

Maybe I should change this to use the .derivative() function instead!

[anton-seaice]

Hangon - i dont have that correct, ill revisit

---

View entire conversation on ReviewNB

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

anton-seaice commented on 2026-07-20T00:51:13Z
----------------------------------------------------------------

@charles-turner-1 - this only works on older intake-catalogs, the new ones gives this error:

ESMDataSourceError: Failed to load dataset with key='ocean.fx.xu_ocean:3600.yu_ocean:2700.point'
                 You can use cat['ocean.fx.xu_ocean:3600.yu_ocean:2700.point'].df to inspect the assets/files for this key.

anton-seaice commented on 2026-07-20T01:05:57Z
----------------------------------------------------------------

I think there is a workaround for this already ?

charles-turner-1 commented on 2026-07-20T02:49:58Z
----------------------------------------------------------------

Yeah, there should be - in fact, I think it should already be in the newest environments.

I’ll double check when I’m off the plane. The code needs a more sophisticated fix anyway which is already on the todo list

julia-neme commented on 2026-07-27T00:22:48Z
----------------------------------------------------------------

This works in the current stable analysis for me?

[anton-seaice]

sea_level is a tracer/center point quantity, so this:

1. takes a diff across two center points to get the quantity in the middle of a face,
2. interpolates across two middle of face quantities to get the cell corner
3. divides by the width/length of the corner cell

so that is correct for B-grid,

---

View entire conversation on ReviewNB

[anton-seaice]

I am not sure what to do for C-grid / MOM6

This recipe calculates the geostrophic velocities at the cell corner (b-grid) point and compares to model speed at the cell corner.

So for C-grid, to calculate speed, should we also interpolate to the model velocities to the corner point to calculate speed?

[anton-seaice]

I think there is a workaround for this already ?

---

View entire conversation on ReviewNB

[julia-neme]

I wanted that yes! But intake only opened one of them. So I had to open 1by1 and merge.

---

View entire conversation on ReviewNB

[julia-neme]

So no change needed?

---

View entire conversation on ReviewNB

[julia-neme]

True! I can replace with SSU and SSV,but I don't think they're available for any run. I'll add the caveat.

And regarding the right combination, since discovering that I can use .derivative() you don't need to think about combining interp and diff, so I might change the script to use that.

(see comment below)

---

View entire conversation on ReviewNB

[charles-turner-1]

Yeah, there should be - in fact, I think it should already be in the newest environments.

I’ll double check when I’m off the plane. The code needs a more sophisticated fix anyway which is already on the todo list

---

View entire conversation on ReviewNB

[anton-seaice]

I think using .derivative would calculate quanitites at the cell centre, so the model velocities need to be interpolated to cell centres then?

SSU and SSV are just the top layer in uo and vo right? So it's ok to use them, just note it in the table ?

---

View entire conversation on ReviewNB

[julia-neme]

I am not aware of this condition, I only know extend (that just copies the nearest value) or fill (fills with something of your choice). I've actually changed it to fill with nans.

---

View entire conversation on ReviewNB

[julia-neme]

Ok, so now I've included more information on how to adapt to mom6 ! Also changed to using .derivative()

---

View entire conversation on ReviewNB

[julia-neme]

This works in the current stable analysis for me?

---

View entire conversation on ReviewNB

[julia-neme]

Thanks for the review @anton-seaice ! I've more info on mom6 adaptation and switched to using derivative.

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

anton-seaice commented on 2026-07-27T23:36:43Z
----------------------------------------------------------------

If the experiments you want to use do not have surface velocities as available diagnostics, you can use the 3D velocity fields and select the top cell. Technically, this represents an average of velocities within that first cell, and it is not exactly velocity at the surface, but it might be good enough for your purposes.

I'm moderately confident there's no difference in model world between average velocities in the top layer and surface velocities.

Also, does this need a dot point to say you need to interpolation the velocities to the corner point with MOM6 output ?

julia-neme commented on 2026-07-28T04:21:46Z
----------------------------------------------------------------

Ahh I thought surface velocity was evaluated at the surface, didn't know it was the cell average. I'll remove that bit. Re the interpolation with mom6, I'll correct that too, but its up to the user where they want to do the interpolation to.

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

anton-seaice commented on 2026-07-27T23:36:44Z
----------------------------------------------------------------

did you try boundary = 'none' ?

julia-neme commented on 2026-07-28T04:22:30Z
----------------------------------------------------------------

Yeah, but it just fills the boundary with values and I'm not quite sure how its doing that since I couldn't find it in the documentation. Filling with nans seems the more cautious approach.

[julia-neme]

Ahh I thought surface velocity was evaluated at the surface, didn't know it was the cell average. I'll remove that bit. Re the interpolation with mom6, I'll correct that too, but its up to the user where they want to do the interpolation to.

---

View entire conversation on ReviewNB

[julia-neme]

Yeah, but it just fills the boundary with values and I'm not quite sure how its doing that since I couldn't find it in the documentation. Filling with nans seems the more cautious approach.

---

View entire conversation on ReviewNB

[review-notebook-app]

View / edit / reply to this conversation on ReviewNB

anton-seaice commented on 2026-08-14T00:11:53Z
----------------------------------------------------------------

This step is kind of confusing. Whether or not its needed just depends on the limits above for u points and t points. I don't really have a good suggestion for it though

[anton-seaice]

Thanks @julia-neme

This looks good.

This notebook is complicated and has lots of changes between B-grid and C-grid models. If we are serious about it being usable on C-grid, I think it needs a B-grid and a C-grid version of the notebook