NCAR Climate and Global Dynamics Laboratory 2023 CESM Winter Working Group Meetings, 10 Feb 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2023 Ocean Model WWG- Day 1 PM Session

Description

No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).

A

B

Okay, we're gonna get started in just a minute um Alan. Do you want to channel sharing screenings.

B

C

D

On Sunday right, yes, okay,.

B

Today, not tomorrow, just another minute here this way.

E

B

B

Let's go ahead and get started, we're going to continue with some more talks on Mark six development, sorts of issues and first up is Ellen wallcraft we're.

A

Going to talk about remapping and regretting go ahead.

A

Oh wait: a minute we're not hearing Alan, you're, muted,.

B

B

See you in the meeting twice.

A

Alan once in presentation mode with no mic under.

B

F

Okay, can you hear me now yeah.

A

F

Okay, sorry, my mistake.

A

F

So the basic ale approach is for each time step. You solve the layered continuity equation, which moves all the layers, and then you use arbitrary, lagrangian and alera and Al method to move the layers back to where you want them, and this is in two steps: a re-grid step and then a remap step.

F

Regret steps, relax the layers, you, the layer structure, you want and there's several options there, but the ones I'm going to be talking about today are icon one from Mom, six and ebgen originally from Icom, and they both favor isotechnicals, that outcrop interfix that layers near the surface and then the remap step is.

F

Going from the original interfaces to the new interfaces on your new layers and the interfaces move, but the fluid doesn't move, so you choose interpolation or conservative with no new extrema. Normally, this doesn't change the solution, but the remapping is diffusive.

F

So in mom 6 we have high con one um and the regrading walks a monotonic vertical profile.

F

The source is Layer, Two, Sigma, 2 potential density, with a compressibility factor added and from these load and suzandalay I think this is you construct a vertical profile using piecewise polynomials?

F

Now the profile Canon prints will be from a remapping scheme like piecewise parabolic method H4, but the one that works is simpler than that it's p1m H2, which is linear between H2 interfaces, and this is the unique practical scheme at present.

F

The target is the interface Sigma, 2, potential density, plus a compressibility factor and the new interface adapts are simply the location of the targets on the on the profile and if the profile is monotonic, then there's a unique mapping from the targets to to the new interfaces and so the vertical remapping. We typically use piecewise parabolic method, H4, where H4 indicates a cubic polynomial for the interface values and, in addition, PPM Limits The Edge values to lie between the layer averages.

F

The H4 calculation doesn't have to lie between the layer averages because it involves four layers which can shift the universe, value outside the original range and and this approach produces smooth interfaces and you don't need an additional interface motor.

F

So why not use the remap paranormal for the re-grid I mean you know on the Facebook. That would seem to be the best approach. The problem is that ppmh4 eventually produces unstable layers.

F

Now it's possible that the now the original piecewise parallel method by Corella and Woodward I actually um had a more restrictive calculation for the edges and in addition, and in particular it um gives you a monotonic profile guaranteed and it- and it may well be that this one is okay, um even though the H4 is not now I've already said the new interface, that's the location of the targets on the profile, however, the actual new interface densities to get the actual new individual sensors, you would remap not densely but TNS to the new interface to the new interfaces.

F

So that's a remap step avail and then you would reapply H2 or H4 to the new layer, densities. I think this is, and that will give you the new interface steps.

F

um Now that's really complicated, but you can approximate it by holding the density profile fixed and just moving one interface, and this allows us to visualize the effect of what happens when you move the interfaces and it's the orange curve under the up on the in the upcoming plots.

F

So this is an actual profile from Mom 6 41 layers in the Gulf of Mexico and I'm, showing deeper than a thousand meters, and this this is p1mh2 for regrad and PPM H4 for remap, and the black curve is piecewise constant method, and this is to for you to see where the layers are. The horizontal lines are the interface depths and the vocal lines show you the layer, average densities and also the layer thicknesses.

F

um The orange squares are the H2 interface densities and the blue line is P1, M H2, that's that's the profile we're using for regrading now, if the target density on a given interface isn't the same as as the orange Square we'd move it up and down the interface along the blue line.

F

Until we have the right density, however, if we do that, what we actually get approximately at least is the orange curve, and you can see that, although the orange curve is monotonically increasing, it's often different to the blue curve and in particular on on the lowest layer here, they're actually quite a bit different.

F

So this might be an indication that um you know we're not doing something quite right now. Let's talk about the case which, which we know in the end, doesn't work which is PPM H4 for both regretting and remapping, and in this case the black squares are from H2 and you can see the H4 interface. Depths are actually quite different from H2, often and in addition, they can be discontinuous at the interfaces.

F

This is a general property of remapping piecewise polynomials, which they they must be discontinuous, at least sometimes on the interfaces and um that isn't an intrinsic problem. But you can see that these these, um these orange curves are not monody increasing um in two layers. They have kinks and then the lowest active layer here, it's actually discontinuous and the discontinentary is caused by the limiter in the PPM scheme.

F

So, on one side of that of that discontinuity is being invoked on, the other side is not being invoked and you can see in particular in this case, if you move far enough down into that lowest layer, you actually end up with an interface density which is lighter than when you start, which is not a good thing and and I don't know for sure, but this is. This is almost certainly why ppmh4 doesn't work in practice.

F

F

Woodward PPM: well, you can see here that it's that it fits most of all, it's monotonically, increasing in all layers and in in all, but one one interface case here. um The blue curve is very close to the orange curve.

F

So this is only one example, but it looks like this might be a very good candidate for um for the icon. One um regarding remapping um I won't say any more about that in in this talk, but it looks promising.

F

So the alternative regritter is, is hibgen from high com originally, but now also in mom six, and in this case the target is layer, average, Sigma, 2 potential density rather than interface density and regretting used.

F

The entrainment is trying to maintain isobacterial layers, and it does that if a lower is too heavy in trying from the layer above which is lighter, if it's too light in trying from the layer below which is denser now, of course, sometimes you want to move it in place, both ways which you can't do obviously, and in general, this doesn't give you a unique solution and- and um we often get thick than thick thin layer structure from this approach.

F

um A unique feature of hip drum, probably, is that it uses piecewise, constant method. The near I spent no layers. Pcm was the black curve on the previous previous slides.

F

This greatly simplifies in trademark and detrainment regretting and in particular, when you send a layer because because the layer density is constant throughout the layer, it doesn't change the low density, and this is a big advantage to the scheme um and I said it's a unique feature of hip gym, because PCM is almost never used in practice, because it's first order, accurate and and very very diffusive.

F

But in this case regretting does not affect most layers and it. If you don't regret, then there is no and there is no accuracy loss and there is no diffusion. Now for the.

G

F

Hebgen is using PPM with a wean like a window-based cubic Edge calculation, and this is the third cubic that we've seen associated with PPM.

F

This approach also produces noisy interfaces and it requires an interface smoother.

F

So so how do they compare hip, Gym machine, the source, and- and this also is close to the result- you want icon, one doesn't care, it just builds a profile.

F

That means that icon, one can regret from Z coordinates, and you can also use it as a vertical interpolator and that's taking advantage of in mom sex and hijin often produces this thick thin, thick thing.

F

Layer structure which icon never does, on the other hand, maintains ice pick on layers exactly and icon does not like on one is not if a layer changes adiabatically the lower average potential density is unchanged and so, and so there's no regretting, but the interface Financial density is changed because it depends on nearby the layer thicknesses as well as densities um and in general, hip gym moves in places less than icon. One.

F

So here's an example of that this. This is the regret, thickness change in meters per day, from a snapshot so on one timestamp and this layer and and and the color bar is plus or minus 50 meters per day. But 50 meters a day is only 17 centimeters per time Stone. So this layer is about 100 meters thick and it's moving on the order of a few 10 centimeters per diester.

F

Now in the case of hedge, and you can see that it actually is not moving at all, it's not being regretted at all, except where the layer intersects the Shelf slope in high con one.

F

It's being regretted nearly everywhere and there's lots of small scale structure in in this plot, which is surprising because in fact, the interface that we get in the end are quite smooth and then here's, the vertical section, which shows it's happening in in many of the layers.

F

So the reason I started looking at this. This issue is um several years ago: I ran a twin icon versus Mom 12 degree Global with 41 layer case I chose this case because it's similar to what both with navy and NOAA users for Global Ocean Prediction and the 10-year run, gave very similar results, except in the Atlantic and one of the differences. I think Atlantic was a milk.

F

The plot on the left here is from from high comma, and this is very similar to what you'd expect to see from an isopectinal or an isopectal like ocean model.

F

The mom result is showing a very shallow, a mark at 26 North. It's a 770, 770 meters, shallower than icon and and both are shallower than the rapid array. Now.

H

F

Layers isn't very many, but the fact is that a 75 layer om4 gives a similar string function. So it's not just the latest, not just the layers.

F

um The global case is expensive to run so I've been using an Atlantic test case where I use a sponge layer to monthly commodology at the southern and Northern batteries, and these string functions are similar to the ones I showed from the global model they're a little bit deeper, but but they're of the same character.

F

And often when you have a shallow string function is due to light overflow into the from the Nordic series and high comma is denser in in the deep ocean here in in this area, and it's slightly lighter and observed, but pretty close and um the moment is, is very light.

F

Foreign, a lot of cases um with this Atlantic test case and but the only mom input parameters which which are significant for this issue, are BBL efficiency, which is conversion of bottom drag to near bottom, vertical diffusivity, which we turn off and then and then thickness diffusion, which is laplacian interface diffusion, and this is using KH Ayla, which is from high com, and it is ISO layer diffusion.

F

You might be familiar with khthth, which is isopycnel uh diffusion, but otherwise, similar and probably k, h, d h would work here too and as an aside really, if you have um time varying layers and you're using thickness diffusion.

F

It's a good idea to turn on detangle interfaces which detangles interfaces um the other term, which is those the only two terms which are important for the Denmark's right overflow for the string function. In addition, adding a regret, filter to hike on one helps and in this case, we're using a regular time scale of 6000 seconds, which is five times notes.

F

Now in the in the hebgen case, we replace thickness diffusion, the blastian interface, smoothing with a fourth order. Interface filler and this filter currently only works with DT therm equals DT, um and when we and we look at those twin, runs.

F

The Atlantic mom sex hit hip gym case is virtually identical to the Atlantic icon. Result: it's in fact slightly done so, but the icon 1 result is drastically different from the original icon. One result, and now the um the density west of the red James Ridge is in fact almost exactly correct.

F

It is too dense, though um east of the Rick aims rich.

F

So this is a great result, except that when we run the same set of model parameters in the global case, the icon one case is still too light, and so and- and that seems to be because the density structure in the is not correct in that case, and of course, in this case, we're actually imposing that through the sponge layer, foreign function.

F

um The hike on One cases now is 700 meters deeper than before. Half of that is from the Denmark's right, overflow and half of it is from the regret, thrower I'm not entirely happy with the with the um the high com, the the hip General Zone.

F

It seems to be interacting poorly with a sponge layer at 28 South, um and so it looks a little weird um south of the Equator, but but generally it looks in the same character as the icon resolve So, based on what I've presented here and other evidence high com, one has more diet or diffusion than it should hip. Gen is better in that regard, but it has other issues, particularly for mom sex.

F

um We can improve on him. Gene eye is like Icom one and gets rid of the thick thin issue.

F

um It may not be the best for my mistakes. So far we haven't come up with it with a with a a hike on one approach with it with you know, with interface density targets, which works well.

F

On the other hand, I which targets layer density is an improvement, and so- and so you know it suggests there might be, there might be an approach that would work. Thank you.

B

James Brown.

B

A

You sort of mentioned this uh tangentially about the role of the vertical resolution itself um by comparing to own four completely different model. Have you explored vertical resolution in your own setups at all.

F

I think it's the case that icon one is more diffusive for thick layers and when you have more layers of their thinner and and and and so in that sense, it's um it's an adventure just to understand. You know to have more lawyers and I I haven't looked at that at a high resolution.

F

I've looked at it a little bit at lower resolution, um but I do plan on looking at say, for example, with the 75 layer set up at 12th degree Atlantic, for example, would it would it be better and and I think it might well be good.

B

Other questions.

B

See Me online, so okay, great thanks, Alan helpful next time.

D

D

So for presentation, toolkit for CSM, simple models framework and, in particular, I'll focus on how we can utilize this to set for idealized 96 applications within CSM, so just to give a brief overview of this project. Our goal is to streamline coupled and simple modeling within CSM and to enable hierarchical modeling, where we provide users with tools for modeling framework that allow them to adjust the model complexity in an incremental manner, and so we have these three tools.

D

uh Visual case channel is a graphical user interface, which is the centerpiece of this toolkits and then so. It's a GUI for creating a custom, cesm experiments in a consistent manner, so it guides the users through setting some high level CSM settings and then the second one is the python Library map sixpathy for generating idealized nonsticks grid and Optometry input for cdsm in particular, and um the third one is Lan tools for service data set and mass modification.

D

But for this talk, I'll focus on the first two tools, so I'll describe or demonstrate how these tools work. On a simple example, where I'll configure an idevice16 experiment, foreign works on jupyter notebooks and um here's the main screen main window of this GUI and what it does is it guides the users through uh the process of setting these uh uh options or settings for uh instantiating a cdsm experiment.

D

So one of the first things that needs to be determined is the initialization time and then also the users need to determine their models for each component, category, which are the atmosphere and Ice ocean plan of Glacier or uh landites and then the wave model.

D

So uh this is going to be a probably last month, 6 case, so I'll go and take mom as the ocean component and then perhaps I can force mom with the data, atmosphere and you'll notice that, as I make selections, some options become unavailable because of dependencies and constraints and I can click on these unavailable options to get a description of why they are and unavailable.

D

So they so that I can get back and adjust my settings or continue with um available options like in this case, I think stop blank and for actually remaining uh categories, I'm, picking the stop models which are basically inactive, placeholders and then, um after determining the components. The next is to determine physics and options for each model um from R6 we don't have any special physics or modifiers, but for data atmosphere.

D

I can select from these this list of uh modes, the Ada cam modes, so I'll go pick the normal air forcing for the ATM and so here's my final concept or component step, um which is one of the first things that we need to determine when we're creating an uh CSM experiment.

D

The next is to determine a grid, and here I have a list of predefined pre-existing grids which are guaranteed to be compatible with the concept that I've just defined, but I can also create custom grids using this toolkit and so Mark 6 is the only active component, so I have to create a custom map. 6 grid and I can start from scratch from a brand new idealized grid, but I can modify an existing one. So, let's see how I can create a brand new one and here you'll notice that I can have a global grid.

D

I'll click on it to get the reason of. Why that's the case and that's because land and Ice components are not active, they're, inactive, stuff components. So there's no way for me to hide a global amongst extreme people, and so I can I can't go over that Global uh case here. I can change the land or ice component, but otherwise I can go with a regional, Mark 6 grid and let's uh make this domain a non-reactive, visual right, original red and the next is to set the grip dimensions.

D

So our set is to be a 100 by 50, idealized Regional domain, and the resolution is going to be a 0.1 degree created.

D

So at this point uh for the mom sixth grade, I'm done with the visual case Jam, so I need to switch to mom6 battery library to further customize the grid and I can do that by clicking this launch Mark 6 battery uh button.

D

And what that button will do is it will uh Auto generate a jupyter notebook and uh here I'll pick the appropriate Journal, which is the visual case giant kernel, and so this is then also generated, um took the notebook that will use mom6 battery Library I'll import, the library and then the second step is to create the horizontal grid instance and all the parameters here for the grid is basically received from the GUI, which are the values I've provided so I'll uh execute this cell to generate this great object.

D

The horizontal grid object and the next is to configure the documentary and so again I'll instantiate a documentary object and the simplest case is to have a flat bottom photography. Here. I will demonstrate that I'll just have a 2000 meter deep flat bottom and then uh there are more options: more predefined. uh The Symmetry options, so I'll demonstrate another one in the next case.

D

For instance, I can have a bold shape with imagery, which is also present in the mom6 code, but here I'll show it in the context of man6 battery, so I'll have a maximum depth of 100 meters and the edges will be basically near Z level and then I'll uh pull out the set.

D

I, don't know how my height might uh mistake there. So um here's the depth field, basically a bowl, shaped um documentary and then I can also do manual changes manual modifications by basically using this point and click Interactive uh interface I can zoom in on certain locations, click on individual cells and change, their depth, values.

D

So if I set it to -5, it's basically going to be a dry cell above sea level and then yeah changing that to 50 meters and I can also so. This is the depth, but but I can switch between depth on mask and here's the mask corresponding to that field.

D

So I'll save my changes and then the final mom6 battery step is to save all these uh input files that mobstix and CSI is going to need, which are the horizontal grid, files, topography files and then the dsmf mesh file once I uh I execute that cell I'll get this load message confirming that the all the input files are generated and then my mind free to return to the GUI to finalize the remaining steps for the case.

D

So this is a simple idealized example. Just to uh demonstrate a few more examples: I'll just quickly go over Stone, so here's a more realistic domain I begin with an existing one, which is actually our Workhorse with and I can zoom in on certain locations and again make changes in the depth and mass by basically clicking and changing the values.

D

Thank you, but um going back to the the context of Idi, simple modeling and here is an example of defining a custom definitry where I uh use, numpy and x-ray to define a custom depth, which is a sinusoidal array and then I can also apply, for instance, ridges on this, the symmetry using the apply Ridge method that comes with the library which I think is a pretty common application and idealized ocean model in the world, and so here I'll demonstrate or show the output of that latest the last domain I've shown.

D

So this is the SSD field and the reason I'm showing this is to make two points. First of all, it's now pretty easy to to generate a idealized domain within a matter of hour or so and get the result. And the second point is that probably the first output you'll get will not look pretty, and so that's a seguating through the limitations of this tool.

D

Key so um again, the scope is limited to some higher level uh settings like the component set and physics options and horizontal grids for map 6 is the grid anthropography for ctsm is it? Is the surface data set and mask, and the goal of these tools are to ensure consistency of these high level intercomponent settings and to make sure that you can easily and quickly create a case instance and be able to compile it without getting any errors and perhaps run it for at least a few days.

D

But of course, these tools will not guarantee the medical stability or consistency or physical painfulness for each individual component. So then yeah it's going to be up to the user to refine and tune and calibrate things like the initial conditions, boundary conditions, finding their internal time, steps and mixing permanentizations Etc.

D

So how to obtain this tool? It's on a public repository on GitHub and I. Have a quick start documentation describing as you can, obtain it and run it and then, as for the next steps. So we are in the final year of this project, and so we have identified some final tasks like improving the robots and news about news. One feature that we're missing is to be able to save size, real input files, especially for a couple of ice in Ocean grounds.

D

So that's one of the very next tasks that I'll be working on and perhaps you can add the future for idealized photography inputs or can and we'll generate a Rich gallery of various examples for people to follow so I. We have some quick start documentation for a vision based gen and we have API documentation for um athlete library, but we want to have more comprehensive, more complete documentation by the end of this project.

D

But at the same time we are at the point where we would like people to try these tools and perhaps provide us with some feedback on how we can improve. What features would be better. So if you have any comments- and if you, if you'd like to try out these tools, just send us an email, we'll be happy to help, you get started and we'd be happy to hear any feedback or suggestions with that. All.

C

That's amazing: um that's really impressive! Oh well, quick question when you're doing the hand edits, does it store what you've edited for so that you have a record of that.

D

That we don't have, but it's now that you mentioned it, it might be a nice uh feature to add.

D

C

Really nice I'll.

D

Let you see this coming along um I think, most of the time we do these idealized Aqua, please there's no River runoff, but um are you? Is there a way to incorporate and then do the that's, the one, only piece that we don't do automatically within the opposite: the mapping generation between portion and and land reverse? Oh, so he's something that could be added to this.

D

Do you mean the mapping feature or yeah.

B

D

Like, let's say, I want to put a a river, you know in a location and then like I will have to then know how to do the Nike offline map and fire generation. Myself.

D

um Honestly, we have never thought about River runoff in context of idealized modeling, but uh nopsy, as we are aware, is planning to add some online mapping feature for brown right, or at least they are considering things so because we do things like the sobriety yeah, um so it you know we brought up to the esmf group, but it seems like he's, not something that will be solved anytime soon. That's my interesting.

D

So maybe that's something we can at least document how to do, but yeah it would be nice to use the same tool to generate the global configuration anyway. So in that sense would be useful. Yeah. That's.

A

Something happens to that thanks. Thank you, Dan and am right.

I

Yeah thanks a lot. This is super cool out there, um I guess, echoing Keith's comment, uh it would be neat to have a file saved or something that could preserve everything um from a particular experiment for reproducibility and I was wondering if you had any plans for um sort of idealized design of forcings. If you wanted to impose like a classic like sinusoidal forcing or something like that for wind, for instance,.

D

Yeah, that's something we think about and consider. In fact, uh Paul ball was uh asking about something like that. What we didn't have in cdaps, that's the uh data components for CSM is idealized mode, and one thing I did a month or so ago was to add and idealize that amount where you can basically provide some constant values for the ATM fields and that that will force the ocean or other components.

D

So it's very basic, it's just constant forcing, but we can certainly improve that and provide some more idealized cases, perhaps like science solution forcing or more idealized options, and so that's something on our radar.

A

Well, as it turns out, Paul has got his hand up next.

J

Hey, hey helper, I think.

H

J

um As as a as an early user here earlier, adopter of the um uh uh uh simple model tools that you've been building, the mom6bathy tool, in particular I want to uh just uh give it the thumbs up. It's it's been very helpful. The the early version of the script, I I, noticed that um one difference between the version I've been working with, and the one you demoed today is that you have both zonally reentrant and meridianally. Re-Entrant options is the really.

H

J

Grid now working and is it possible to do doubly reentrant at this point.

D

I just did a test case test experiment. It worked. It ran, but I can't comment on. uh It's like.

J

Physical physicality, okay,.

J

And then one other question too, uh following up I noticed that this defaults to a spherical grid, I think in the version I have of the tools you had Cartesian as an option. Is that is the Cartesian option still there or is it going away.

D

It's it had to go away, not because the tools I've been working on or not because Mom, six mom six actually supports Cartesian grants, but esmf does not so.

J

D

So we are basically limited to the spherical grades within CSM got it yes, my folks know about this and know that we are interested in having Cartesian grids but I'm, not sure. If there's a priority for them or not.

J

Sounds good. Thank you.

H

Good all right thanks.

B

And next up is we start so moving from the idealized to the purportedly.

A

Realistic questions.

A

Discovery is going to give us an update on where we are on our Workhorse model and uh moving forward to towards the csmetry release foreign.

K

D

Thank you and so yeah. This is um as we're moving towards fully cup simulations with customer interaction with the other working groups and WG for G chord, also, the the core and the landmarks as well. So the list of doctors is gigantic here. So we couldn't add everyone, but I want to thank everyone for it for helping with this, um so how we started doing that and walking through the the staff secretly the failures and just give you an idea of how this process started.

D

So this was part of uh we had this like Task Force Five. His goal was to provide preliminary evaluations on these fully couple simulations by the end of December 2032, and we had these compset in mind. So we had a Target that we want to choose, so it was an updated kind of basically expected content dive with the Specter element by core, with 58 protocol levels and Mom 6 size 6, and at the an E30 horizontal resolution for the atmosphere and co61 throughout the ocean.

D

So in 2022 the beginning of 2022, we ran these control keys, which is basically csm2 and we replace pathway. Well, so csm2 is claim six with 32 levels, final volume, size, five and then one six instead of Paul and and run for 100 years, and the parts on the right are what the the amwg folks produce. So this is what they typically look at.

D

So there's the global restaurant, which is the the heat balance in the top of the atmosphere, global and temperature surface and then like we're, particularly interested in the lab, see how the lab see was behaving on that um and then right away. You're like let's just go for our Target configuration, it's the one I mentioned before and try to run a case and the lab C starts freezing right away and you can see that with this plot that shows the labci extraction, then you get an increase in one seat action.

D

So that's indication and it's it's freezing and that's not what we want. So at this point we had done already 12 experiments shown here um and then we were like all right. We we're just changing too many things at once, like it's hard to find the culprints like what is what is the issue here? Why is it freezing, but it was not freezing with the control and then suddenly now it's freezing, so we took a step back. He said, instead of changing anything at once.

D

You're gonna have to go to the tedious process of changing one thing at a time and that's what we decided to do, starting with what we thought would be the the least disturbing change in the in that process, all right. So what what we have learned during that process and I'm gonna? No remember it was like 12 simulations at that point. uh Next time, I'll show some places will be 26, so between that first off there was a solidification of command.

D

So we in fully couple simulations you can have place, especially of course the highlights too, is where the CIS ciscillary devices for PSU and if the solidity of the cell is less than for PSU, the CIS will take more than what's available and we'll make that success.

D

So we've quickly came up with a hack on the ice model to avoid they have a limiting based on nests of food surface validity. But this is no longer a problem because Dave has been daily, has implemented, prognostics only size, both sizes, five and six, but we did run it in an issue. Also, there was some large shortly rescaling errors in size where size was the ATM physics were updated. There was some interesting pieces that arrived and then black led to this shortly.

D

A rescaling problem, I I, really actually what diabetes I'm not going to get into that, but it turned out to be an immunization problem and also got fixed and now, like the more interesting part, is by default. If, if you download this CSM and no run 1850 simulation, the CIS initial condition has this. It assumes that two and a half meters thick ice forward of 60 degree I, think it might be actually three degrees on the modern, Spirit two and a half of the South, um and this makes you very strong refreshingly write it.

D

You know the early stage of the simulation and we learned that a hard way by doing many times materializing, we need to use the spin out CIS condition in order to minimize the thresholding, the initial pressure and similarly with the default initial name. It's not bad. You have which is shown on the left here. So this is the coming from the land model.

D

You have very thick snow to begin with, like 20 plus meters, that when you start the monitoring cause a huge search of the liquid runoff right away. In the beginning and as an example, we figure out the model for 58 years. This is how the same field looks like, so that thickness has been produced by quite a lot. So also, the knee strategy is to use a somehow spun up initial condition for this no deck we are working on, I. Think Adam Harrison. Is the person leading this?

D

How should properly initialize this note that, like just I, think it takes a much longer time, the future years to to equilibrate that? So that's something we're working on that hopefully will in the future will have a well spun up initial condition for that all right. So now, after going through this, we were at Ground 26, the family 26. So as I recap, we have the control on the top. We have the Target that we were.

D

We tried in the beginning at the bottom, and now the 26 family was basically the last steps like we were very close, so I think that's in red. Here is what we want, and the blue is what we began with. So it will come with the ATM, the ice and the purpose of the run. For how many years we have so the 26th family, we were approaching the target configuration, so those are the runs I'm gonna Focus today, um in particular like when we we did 26c.

D

There was like the only thing we were missing was the CIS model being size, five, instead of size, six, um they were able to run for 100 years and that we did a few problems of those with a small motivation with the temperature of the atmosphere, thanks to the minus order of 10 to the minus 14., um because the street there's a two problems of of 360 and then we have a 26e where we switch to size, 6 and 26 H, where we switched side six and also is a advanced there's, no physics, that's a run that has gone for for further, which is 244 units.

D

All right, so just so among time, series of annual and March lab labs that other sea ice area, so the top one is the annual annual mean. Now they both say the same thing. So you see these brands that where it goes above, like 0.5.6, is when the live Series. So all the all the rounds that showed up above that value they're frozen, not good.

D

um The only one that hasn't Frozen is 26 h, which is the latest one. It has we stopped at 244 um and yeah, so it's either like the per limbs of 36C. They all froze um yeah. So it is because of that that the winners of the person that we had a joke, like different people, were running through the simulations and they built it was the one that 26 8.

D

So he gets the the trophy of time, a model fast food cup simulations and trophy generated by concept: diffusion, algorithms, AI that we created this for him.

I

D

For me, we'll still make sure that he obeys all the conservation laws and that did I run before we handling the trophy surgery.

D

So how do you know it's not going to freeze later? We don't yeah. It was always a question how long we have to register anyway. Moving to how the number two way to fix the adapt. It is criteria so I'm, just showing a few of them 2016 at the bottom left. So these are the averages of the year that I used to do.

D

The averages are shiny, for instance, for each case, so this is 2016, and you can clearly see that this blue law- that's not signal that got frozen, see a little bit of convection with another big Seas for comparison. This is the Argo based display, adapts and I added this for Nikki, because yesterday as how we were doing in terms of nuclear depths, just to have the Ops by so we can evaluate so.

D

The 2016 on the bottom left again has been frozen, but it was seriously convection uh where the East Greenland current is and then finally 26 age, like anything, has too much compaction in the laboratory scene, foreign.

D

Looking like I'm just showing the time series here for um um the 26th family of drums, and you can already see the ones that go really low, those are the ones that froze and the aim off basically shuts down. But what's interesting here is 26c, so the solid black it went down when the labs started, freezing but then start coming back up, and we want today about that.

D

We rent for longer um it didn't the the last thing was too Frozen, but like I don't know, maybe we should have continued that maybe we would get out of the increasing State back to your normal State. That's the main HPC. Another thing is the 26h and the 26g. They were very similar around right here, uh year, 178 or so, but then 26 G, which the one I ran just froze after close to you 200.

D

So in order to get to win- um and this is just at 26- so what first row of 40.5 North now is it 26 North is a similar story now moving to Canada drift, starting with the live C. Only and again you can get a sense of the the runs that freeze, because you get this very cold bias near the surface and precise the beginning of the freezing. So all the runs are freeze have that including 26g end of the run here now, 26 age, you know it's in the office.

D

It is a very warm bias um right. I, don't think we can treat that as a program, because no, we change the the field. There's no, there's no physics as well. So we don't understand like we need to understand why? What is the? What is causing that? What was different about that? That um it could be just random like, but anyway, I'll tell a little bit more about that towards.

H

D

Oops now this is the same thing for civility and there's not know where to add, but you you know instead of getting them, you get this freshly signal very strong near the front surface when it freezes over and again the one that doesn't freeze, and it is a saltier virus and including out the global temperature drift and just to make the point that we have these cold bias in the surface, warm biased at the depth and a similar pattern was obtained when they added this.

D

The Spectrum element came SE to csf1, so the response was very similar. I um I was when we were developing this, like they brought it up the plots from at that time, at least remarkably similar the the way, the global bias. The temperature looks like right. So to summarize, we perform multi-sentry physical programs with the target configuration for csn3.

D

This runs 26, the 26 Channel. You have reasonable balance at the top of the atmosphere, click balance and the library there's C faces towards the end of the simulation through output, one which is 26h. Why? We don't really know? We have a hypothesis that this system supports multiple equilibrium stakes and we are close to a state that leads to freezing.

D

So every time you perturb we go to today that state that just it's freezing, but in any case we really need to understand what's causing our perceived free history, a lot of which is freezing these grants. We have some ideas on how to do that, which is the first one is to continue the food policy budget for the lab, C and Frank and Ian are actually working on this. We have also another things we want to try out going on.

D

The next steps is to low-handed fruits to couple to wave model with watch3 and then enable the mixing via the back, the C lagging board option. We also will redo the mixing coefficients and Implement an option to apply that dependent puzzle, picking.

C

D

Future now moving to New and revised configurations in the end of last year in the future, thanks to Frank Brian Frank was Lucio for helping with this. We got together and within like a month or so. We revised the two thirds degree of creation. So there were a few things that we wanted to do like shift. Designer velocity point should be right at the equator. We we went through the topography in 19. It made a lot of changes that hopefully improve the solutions and also change the domain size to optimize the beautiful position.

D

To give you the number of processors on our nodes. um Also, we have two linear configuration, one added a mini quarter resolution and the idea of putting this together is to leverage the work in the NBC. So these rounds are, we always almost completed the full cycle of the GRE.

D

We have a z star in the icon, vertical we're going, and so it should be done in the next few days and then we also have added resolving 112 resolution that we are still working on on that and all of these configurations will now be and then documented on GitHub. So the ideas we have a single repository that goes from the great generation to the math generation to even regretting the data sets that will be used as the observational data set. That will be used to validate that right.

D

So yeah stay tuned for that, and then just a few words on the GA. The data simulation I got this like from then Ms, so mob6 start development is in progress for the global tutorials configuration. We anticipate to have a functional capabilities. We depart algorithms for assemble adaptive inflation by the end of 2033 and focusing on the 20th degree. For now, and then you know having uh da work course, release of csm3 plus Dot enabled for CSA and he and CIS and land and there's also a NSF proposal that was submitted to do high resolution.

D

Regional simulation is remote, seek start plus BTC marble with this idea of Ratty rapid prototyping of model grids when you similar lines to what Albert was showing force and files and observations, plus, GA and python on the back end and Community improved is on priorities for csnga capabilities is highly welcome at least email damage. If you have any any suggestions or comments, and with that I'll take your questions. Thank you.

H

Bill well, I think it's great that you're running these runs now so I'll know more power to you trophies and everything aside. But what you didn't show is the Northwood heat transport, the Atlantic still too low, or is that changing at all and if there's not enough Heat going north?

H

Are you trying to find here a compensating error to keep the lap C or do you think your physics is getting better yeah.

D

To be honest, I didn't look at the property transport very carefully on these events. um I I, don't think they are like terribly.

H

In your labsee budget should do this too right you're going to do a budget. Yes,.

D

Right but yeah I, don't think they're like too too low like too much too um it's getting better. Then okay,.

A

That's good in the fourth run, the heat transport, the long six is actually a little higher than it wasn't hot, but the same force in data set. So it's a little closer to the observations. I also, don't I think we've looked at it in these 26 series, but I, don't remember. October yeah.

D

But we should look at this: the one 26 age, the one that ran for a really long questions. Sorry.

A

So questions and discussion is open, but I mean this particular discussion period is really about. You know what creating feedback to you guys about what should we be doing? Moving towards the release like? Are there big gaps right now or if you have suggestions like what should we be doing to try to get past some of the bringing hurdles and which of the missing pieces that we're still lacking are the highest priority?

A

Put in I'll just follow one more thing to kill what we've looked at so far is, it seems you know in the lab C. Basically, you put buoyancy released from this model, we're putting buoyancy in mostly in the summertime through fresh water, and then you have to remove that buoyancy through cooling to come back through it, and so it's not necessarily that we're lacking heat transport into the lab sea, we're putting I think we're putting too much fresh water in during the Melt season, and it's been difficult.

A

Problem is sort of diagnose where that pressure water is coming from, because it's coming from all directions. You've got it coming in the ocean in the East Greenland current you've got it coming through the archipelago. You've got it coming in the ice itself transporting in there and melting. So it's a it's fairly complicated little system.

D

So it was a size six, but we had done one one of the 26 cameras in size, six and that froze and then um new, snow physics and I I. Don't know what that means. There's no physics! That's a question.

H

Well, that was just snow on ice, so what not snow on land or are.

D

They even saying this: yes,.

H

On the CX, okay,.

L

The strength might have covered it, but uh I was thinking. I was trying to understand the labrador sea freezing and having more fresh water, because that's kind of countering what I think of.

D

G

Ice form and you get more sailing Waters.

A

Well, those those drain out: okay, okay, but then, when the ice melts, it leaves the fresh water at the top. So if you're you know, over the period of a year, you tend to stratify the water column and salinity by forming ice.

B

Other comments or questions.

M

I got it yes, so the 1200 degree and a quarter degree are those of me to the the 12th degree is really just like a refinement on a quarter degree. So it's really a three by three inset. That's nice, but that'll.

A

Be really helpful: the center scalar points lineup. That would be really nice and print started. They said 136 as well, so you started to finance resolution and designated it to get back to a quarter degree.

A

We can't really do that at two-thirds, because we have to go to equatorial stretching right and more than about a quarter degree.

D

So have you done any couple drawings of the two-thirds.

K

K

Yeah and I guess uh there could be loads here said one like the solution for the core bias with the.

K

But um we never had any issue with the high-rise, but it was always true that anyone anybody was always like very important at.

M

The surface and we don't know why you don't want it and.

G

M

Are the Target parameterizations for the quarter degree in 12 degree.

D

I mean hopefully um a lot of Works being done in the Simplicity of a degree. That's why I read about that um I think for the 112 by the week of rain at 110th using the parts grade, we migrated at your mom and ran that we were missing a retraction Munich as much as possible with pop. We don't have everything in place, I think some of the big scene options- and you know right now- yeah um yeah, we'll have to think through those.

H

Yeah, just are you: do you know what the atmosphere people have done to this problem, because they did quite a few things early on and they've taken them all out. They were pretty crazy things they were doing or do you know what they've done or keeping track of what the atmosphere is doing as well. Yeah.

D

Basically, in the end of the year, we were meeting sometimes twice a week with them um so yeah, but now, like we've, decided to take a break. Okay, it's working group will work on its own, but there's things we want to try it and then we'll merge again in a month or so. It's good some.

A

Of those early experiments is sort of 1 to 16, mostly because I was skipped or things like playing with the Greenland topography and.

B

A

And that kind of stuff, so that's what I meant yeah is the the SE. They were using a pretty aggressive, smoother to generate their topography. So we had, you know elevations and 500 meters of the cert. The atmosphere's surface was 500 meters above the ocean of the over the lavender sea. So by just by the adiabatic lapsulate, it was colder yeah. But that's right, they've gone through and done as much as they can to fix those kind of things um and I think they're now kind of deep into microphysics trying to to not.

C

Radiation coconut had a question in the chat.

A

In the chat, any.

B

A

Related work again from empty line: Square HP.

A

I think that's an aspiration. uh It's not in at this point, yeah.

D

So for the the hpe will quick, we had a very limited time to play with the training, the data and we we try to train for the two thirds and did not get good results, even though at the initial stage. So we move forward with a high resolution configuration and we're using gfgs configuration for that. However, that doesn't mean like we cannot do it in the two thirds I think we just need to invest more time and look at different features and yeah.

D

We need in-house expertise to be able to do that and run that on on our machine, so I'm I'm hopeful about you know once we we go through this approval Concepts and do it in on four half of degree. We might be able to go to the beginning again and train the data to work on the two thirds using perhaps different features that we were using for. For those other cases.

A

I'll, just maybe wrap up unless someone, but um that you know basically we're human resource limited here with um you know it sounds like end. Car is a huge group, but it's it's a couple of us doing this work and the more eyes we have looking at the solutions, the more things we find out, the more ideas we get about it move forward.

A

The model is available and this basically this configuration now you can run these cases out of the box yourself and people are, you know, Ian has students who have been doing work with the model and you know getting papers out with this model right now, so um you know we're working towards csm3 release, but it is available right now and the more people we have working on it. The faster we'll get towards get to the Cs I really encourage anyone who has an interest to pick it up and pitch in.

D

Beer is also running yeah.

M

Bill is it that run. Is it that the model I'm running yeah.

H

D

I just got it for.

H

H

B

There any other uh questions or comments online participants.

B

So let's take a.

A

Well, we'll take a 15-minute break so we'll start about um to uh wait. Yeah! Oh, no! We're ahead of schedule! No problem! Okay, so we'll be back at 2, 40., uh Mountain, Time and we're gonna have a session of three talks on uh Regional ocean modeling and uh there's one change in the schedule. I'll give you a heads up now that Melissa Moulton is under the weather, so we brought in a pinch. Hitter uh Walter Torres from UW Applied Physics lab will be speaking in her place at uh three o'clock.

H

C

The time of that one.

A

I didn't look at Walter's slide text three.

C

A

C

Walter's talking basically on.

A

Modeling uh wave-driven circulation around the Palmyra, atoll I think.

N

It'll be an idealized case, but it's based on where I think.

A

So so anyway, so we'll be back, ask um 240.

E

D

L

E

B

um I just want to say.

A

That um we're moving into some talks that um are at least most of them are not using our code base they're, not running mom, six or pot, but I think it's still totally appropriate for this working group to be talking to people especially doing near shore. Coastal stuff that are interested in climate impacts, be talking to them about what those problems are, how we can support and stimulate that work and how our models can interface um in a downscaling way or whatever.

A

So to start off, we've got uh Baylor talking about I.

M

Was thinking about Coastal inundation, Coastal inundation? Yes, so um that's exactly what um I I wanted to talk about today, and this is a very um early stage kind of application.

M

We thought we would kick the tires on doing a different kind of simulation with the tool we've been building for near-term forecasting not for climate scale, applications and see what it would take to switch around and and see what happens if we try and use it for a climate projection so um and there are lots of people to credit most especially Johnny Benoit, who is an undergrad who set ups the simulation um as his senior thesis project, based with lots of help from the group for people who had developed this near-term system.

M

Okay, so here is the simulation domain on the left: um that's Rhode Island and a little bit of Massachusetts. You can see Cape caught in a little bit of Connecticut and Long Island New York to the left.

M

um The Open Water boundaries are shown here in the red line going around the outside and then they're a dozen or so rivers that are shown in little purple dots in different places, which are caged um and to come back and point in there a couple of water, a wastewater treatment plants that are also in there that have fresh water sources.

M

um Normally, we would drive the system with something like the uh North American mesoscale Meteorological model, or we would use um a regional climate model that had been downscaled from something else um like the one that's done by Bob, Beardsley and um Dartmouth and um on the right hand side you can see. So you you have a question of how you invalid data model at this scale.

M

So there are actually a couple dozen all right well, depending on which year you're looking they're about a dozen um in C2 instruments that are throughout that domain, when they're on we can validate against them and then the other thing we've been doing is you've. So the grid resolution here is 50 meters within narragans at Bay and about 100 to 200 out in the on the shelves.

M

um The land satellite has the thermal band that you can use to reconstruct sea surface temperature if you calibrate it for your local conditions and that's what's shown here on the right. The blue line is the model. The red dots are the landsat scenes and then the black dots are the landsat scenes. If you correct them to the moment at which uh the satellite was overhead, so you're pulling it together, and so we do a pretty decent job of getting at least a seasonal cycle in this one year right.

M

So that's our validation protocol and there are a couple papers on this. Why do we even want a forecast system on this small scale? It turns out that um in Akash sanu's paper he showed that there's actually a 10 to 40 day predictability for this system, so we could predict the conditions like beach closures or something a week or two or maybe even three in advance, which is clear, economic value.

M

So we're converting this into a forecast system is one part, but I want to talk about the climate projection side, okay and thanks to Justin small here are the references from Justin's slide this morning that show um what the eye has forcing our idea was well, we would just grab the flexes. Our eye has been drawing this thing with it, because the models were the weather models were using drive. It aren't that different in resolution. We use the Navy model normally for the officer organizations, so.

L

B

Where the is forcing.

M

um Looks like so I'm sorry that light blue washes out, but there's a Time series of temperature going like this, and this model is expensive. So we chose three years: um 2018, 1917 and 2100 to simulate, and so we pull the forcing out of I Has for those there's happened to be years where there was high frequency, surface conditions and high frequency RC flexes. We ended up using the bulk formula and using a high frequency surface conditions in the offshore condition. You also have to reconstruct something like a barotropic tide.

M

Add that on top and you have to get a bear Clinic profile- that's from the monthly means Landing in temperature, depth and kind of piece it together to give you a sidewall condition, but it seems to work. Okay, then. Similarly, there's uh the surface change in salinity and so there's not as big of a signature in the salinity and I ask. But there is one, um so it's getting basically a little saltier and I'll come back to that in a minute.

M

So, basically, here's the bulk over the region inputs on the left that outputs on the right and what are the kinds of things you'd like to say? Well, here's the air temperature changing and here's the surface temperature of the bay responding, here's the surface temperature of the ocean. It's also in that here's, the surface salinity going up. Here's the surface salinity going up notice.

M

These numbers are not the same because they're fresh water sources in the bay, it's an estuary, and so we have other sources between the ocean water and what's going on deeper in the bay that you need to think about, and then there are other responses that are kind of less. The stratification goes up, which is sort of interesting and probably impactful for things like hypoxia in the future. So we'd like to do that.

M

A little better and sea surface height goes up by about a meter not because of I Has, but because we're adding on top a global sea surface um increase um based on you know a whole where it's estimated until you get here the three different years in temperature time. Series The seasonal cycle is still the dominant thing, but you can see that there is a dis. You know we're moving a few degrees from one to the other.

M

This is an RCP 8.5 scenario, so it's a pretty substantial warming um uh and then here's where you start to get an interest is where is it warming? Well, it's not actually warming. It's warming most on the Shelf a little bit in the lower Bay and then it's cooler up here at the top.

M

We think that may have to do with the way we're forcing the rivers the way that we have the temperature of the inflowing river water, which is very approximate to begin with, and you can play the same game with salinity and you get these interesting sort of hot spots of where their salinity increases, um which would be an environmental concern, but we're still not sure that we're doing the rivers right so I'll come back to that and here's the density, stratification change.

M

So it's getting it's a non-trivial pattern and it is denser at the bottom than at the top, which would tend to track more fluid in between being refreshed with oxygen, leading to hypoxia.

M

um Other kinds of sediment uh driven biological processes on the on the Shelf might be a problem, and you can play funny games like the present day high or the high tide in 1970 is almost lower than the low tide and 2100..

M

L

M

We're not doing wedding and drying, so the magnitude of the time doesn't change much because I, basically we're adding a meter, but the beach doesn't recede. But we would like to do wedding and drying and see if there's like a different, resonant response of the way he's running around the bay and if the tides change dynamically.

M

So we in present day we run and we drive the thing with observed: River gauges. um The USGS will give us a forecast a couple weeks ahead, but they don't give us a forecast 100 years ahead. So we need hydrology in order to figure out what's going on with the rivers in relationship to the changing precipitation right now, we're just using the same river, forcing so essentially a climatological river in all three cases. But we know that the precipitation goes up substantially in our region over that uh over the century.

M

And so we have been experimenting with the idea of going to all of the river catchments, like this figuring out how much rainfall happens on them, taking something like a two-week average and then seeing if that's correlated with the river flow, and it does really well shockingly well in Winter, where you know this is sort of observed and predicted with that kind of simple scaling in summer. It doesn't do well, I think it's because we're not including evaporation.

M

So if we had some simple parametrization of evaporation, we might have a really simple way to predict the rivers and then be able to do rivers in 2100 with no observed gauges. So there's.

C

M

Of room to improve in something like this, we learned a lot of things about what would go unstable, um particularly using the flux. Forcing didn't right really was not helpful. We spent a lot of time trying to get back to work. Switching over to the bulk formula worked much better.

M

um We had a lot of challenges in missing data that had not been thought to be recorded, particularly the subsurface data. um I. Think if I was going to do this again, I would have. You know talked to folks who are setting up a simulation like this at the beginning and said: could we have 3D Volume for just this tiny region, which is it's like a dozen grid points, but we want them as profiles of depth as well.

M

um One of the cool things that Johnny did was he actually turned on and off the breaking point power plant, which was the largest coal burning plant in North America, and we had observed its signature. It shut off about 10 years ago, and you can see that in the satellite record that one part of the bay gets colder when that happens, and you can turn it on in this model and suppose it lasted longer and play games with like management decisions and what the impacts would be versus climate change, which is an interesting comparison point.

M

um We got a lot of proposals out to try and do the better job of this. We have a lot of other applications like dropping a large editing simulation inside of this and going down to the sub meter scale resolution for places in the bay that are hard to resolve. So that's that's. The fun we've been having so far.

D

Thank you, Baylor.

D

That's just for me, yeah.

E

um I was curious about doing the open value conditions that um sidewall I think yeah.

E

M

So we're using CSM, which I think we had as I recall, we had daily surface conditions, including velocity, and then we had like monthly below, and so we essentially took the daily information and projected it into like a barotropic mode and then had a bear Clinic mode that we were expecting to be evolving at a slower scale.

M

On top of that, we also had to add in the tides and sea level rise, so we were messing around a lot with the conditions on the walls and coming up with something that seemed like most comparable to what we would normally force it with in present day.

E

We're um using the open, dining conditions were you able to reproduce like the time periods of water level, height and other things that were for observations, because I think yeah.

M

Yes, so normally we can, we actually get the we get the tides right. There are five different tide pages in that region, and so, if we're doing modern day where we have high frequencies sidewalls, we do a good job of reproducing the tides and we're essentially getting consistency with that same uh approach with the future forcing, but the temperature and salinity conditions along the wall are changing as high ASP is changing. Essentially.

D

Who have a question from Justine online yeah.

O

Yeah I had to question about how it's, basically, how good do you think the eye has data is in the future and I know it's it's kind of impossible to answer that, but it it's very different to the the low resolution simulations for the uh change in SST and and one other thing I just wanted to add. Is there were three Ensemble members done so it you might want to take that into account as well. You're, probably looking at one one member, um so I wondered there's one other paper. That's looked at high resolution.

O

um Climate change in that region from uh gfdl group I think they did more than one paper a one was by Saba. uh So.

M

O

Wondered if you had any thoughts on that.

M

um I didn't I didn't know about the Ensemble. We definitely would have we're probably going to run out of this case, but we had some inner annual variability that made some of the some of the interpretation difficult and it would have been nice to have been able to average across an ensemble and smooth that out a bit um yeah. The bias versus the low resolution is part of the par for the course. Normally.

M

The way that you would do this is you'd use a regional downscaling model to get to the kind of kilometer scale from a one degree model, and then, from there use that to force this 100 meter model, we were trying to skip the first step on the principle that it would have more legs eventually, but all of the biases of ihasper coming along into this interpretation, but it's still better. This is we I still think this is one of the better potential ways to get climate change.

M

uh Inferences for this region um as well and I, didn't know about gfdl stuff about to look into that.

O

Yeah I wasn't trying to say the is, has a lot of bias. It's just more. The uncertainty compared to other models.

M

Yeah we don't have I mean it would be lovely to know what other scenarios looks like weaker, uh forcing scenarios as well and so yeah. We don't have a good way to measure uncertainties the we know how to measure the uncertainties that are due to this step of the downscaling, but we don't have a good sense of how to do the uncertainties on that forcing side um that we need something like ensembles or something smarter um that works robustly.

A

K

So we have as distance C and get free lcp-85 ensemble. uh We are going to uh seven more soon member, and we also have also a few seats. Four five and six zero excellent c0 is actually avoiding well.

G

The O2 I think um all available okay.

D

So these will have high frequency output for forces. They will.

K

Have the same style, uh so you will still have like monthly uh three dimensional options: uh data, but your life frequency for the full service, yeah, maybe more a little bit more high frequency data, uh the Bruce lct85 that we.

M

Generally yeah, that would be great to.

K

Look at I think there is a huge model on top of what was available from that one. But if anything, I.

M

Think, there's a little bit more good, good and I think the method that we used I would be comfortable saying that that was a reasonable way to get between the two, but we could test that more extensively in the modern work. When we have the high frequency like if we just sub-sampled it the same way, how bad? How much error does that introduce.

K

Yeah I was going to ask like because I was surprised. We brought to be sure where you show that the low tide is trees. So you correct products because it's not like in the highest data.

M

Set yeah! No! No! So we we add that on on top along with the tides so um but- um and that would be important if you were going to do inundation so we're we're headed that direction. But we're data set all.

D

Right one more question like from then yeah.

I

Thanks if I missed, this I was just wondering if you could say something about the forecast da mode.

M

Oh yeah, so um the forecast mode, so we basically have been experimenting with Kind of Perfect models um and trying to simulate the uh observation array that we have and see how how much, how many constraints it uses.

M

um We um we think there's going to well. We know from like Diagnostics that there is predictability in principle 10 to 40 days depending on season and like how rainy it's been things like that. So we think there's a valuable product there to exist, but we do not have one that's running with the real data yet for a variety of technical reasons like the buoy data doesn't go online in real time, which we are not negotiating with Partners to try and get access to real-time data.

D

Sure last question: for a bit before we move to the next talk, so this is from lorisa I'm going to read because she's under the weather, thanks for the interesting talk really, but you see advantages the advantage of downscaling within CSM in the future, for example, from a global model to a nested amongst the grid versus applying boundary conditions offline to a different model like realms.

M

So our two-way Nest thing probably wouldn't do a whole lot in this case. There's not a lot of exchange from the mouth of the day out or shelf even also um we've done. Other control runs where you do just in the Inner Bay versus the larger domain, and there are interesting kind of local features there, but it doesn't really change the conditions in the sound very much so I. Don't think there would be a whole lot of benefit to a two-way coupling.

M

If that was the intent of the question yeah, then we'll keep messing with CSM, we'll probably drive it with other things, too.

M

um Are we interesting to contrast just from an expense perspective versus like an e3sm with a regional refinement, because then we have to run the whole thing for the whole 100 years, which is a different just it's like a different planting game to figure out what it would look like so, um but we haven't tried to. We haven't been thinking about that very hard. Yet thank.

O

D

Let's move to our next speaker, Valkyrie Flores, which we'll talk about the interactional we've driven with trash Jacks and requirements, yeah.

N

Hey guys, I'm, not Melissa who's on the schedule, but I'm gonna start. You know with me and Sharon um just check it out back at Applied, Physics, it's not back at uh um and what we're going right now is Currents, but most of this talk is going to be on. So my PhD work, which was on way driven close on weeks and particularly um the interaction of multiple repost Jets so broadly in my science objectives, are to understand fundamental hydrodynamics behind wave-driven circulation features. Send it like examples.

N

I just mentioned on the right um equation: three pass Jets um and grip currents which arise due to patterns of wave, breaking and topographic interaction, and um often I try to develop like simplified analytical models for circulation patterns.

N

So we can understand it very deeply um and I want to understand both what happens to the Jet and plume once it exits the surf zone um and um how it's formed to begin with, um once it exits the surf's Stone, um the Jetter plume undergoes spreading and deflection, depending on the physics at hand and potentially a multi-jet or multi-plume interaction, which will be the focus of the talk here and some of my other work focuses on the consequences of the circulation features for ecological conductivity.

N

So this is a uh schematic overlaid on a picture of Maria in French Polynesia that shows the General Physics of wave driven close on reefs waves break at the Reef press, which Drive currents into the Lagoon and uh the currents have gone, polluted and come out through these uh Reef passes and that those are gone to form. Jets um here are some other examples to read past jets in other parts of the world um transportation and Palau and Micronesia, and in Western Australia.

N

You can see this uh Gap features, kind of periodically spaced um up and down the coast or around an island we'll skip that. So.

K

Here's another schematic.

N

View of the circulation field kind of abstracted, um so we think about the waves breaking driving flow into the Lagoon out as a jet and then what happens to the jet um is uh This research. We assume it will deflect due to coriolis kind of like Rivers, but it's it's sort of an intermediate scale um where these features are more. You know 100 meters kilometer size. So it's a bit of an intermediate region uh that intermediate rostery numbers so um we're actually not sure what it does.

N

So that will be the focus of This research and we're also interested if there's consecutive um repast tests that if they deflect um they actually might run into each other um and have some interaction, uh Mutual interaction down coaster around an island.

N

So yeah, that's the uncertainties if whether they can drive meaningful along the shore transport and if there's a interaction with neighboring Jets.

N

uh Those questions are partly motivated by some long-term observations on Maria now in the French Polynesia, where there's a long-term ecological research site and um they've been doing measurements there. Almost 20 years, uh certain Moorings and a few low pass the data it turns out. There's this residual counterclockwise float that they measure different Moorings around the island of the order of a couple centimeters per second um that's coherent in time.

N

So um the kind of our hypothesis was that maybe there's a mechanism by which these uh read past Jets come into uh bounce with their assortation and maybe drive a coherent counterclockwise flow. This float is also shown to uh be related or there's also a counterclockwise Trend in connectivity and uh parentage for clownfish on the island. So there's interesting ramifications for it um ecologist here so to investigate this.

N

The instead of going um straight into realistic modeling, we decided to um do an idealized experiment and kind of make an abstraction of a similar uh electronic problem. That's a similar scale and similar with silly peaches, says Maria. So we set up like a 20 kilometer uh radius island with 12, equally spaced um Reef process around it.

N

There's there's 12 Jets around Mariah too, um and with like a you, know, analytically prescribed the Symmetry and um we could take the toy model and bring it at several different latitudes um and you know planets like uh coreless, forcing but and also change. The bottom drag um we're kind of into Coral, restore particularly hijack environments, so we're interested in on what the effects of the flow on the flow might be.

N

If we uh for different um degrees of bottom up, as some correlates are very rough, some uh Growers are not uh rough at all and um if Coral Ridge degrade, it has an effect on the roughness too, oh and so I should say the simulations we're doing in ROMs, but we're using poets where couples uh wrongs and uh Swan. Oh, we need the way of course.

N

So um this is the problem setup, um but because of the radial symmetry of the problem, we really don't need to simulate the whole island, uh we're just taking a slice of the pie and uh getting a periodic boundary condition so um giving giving one slice uh periodic conditions the same as doing the whole island physically and um yeah. That's the setup, so some of the possible show will be just that inset, but it's meant to represent the circulation on the whole island.

N

So here's a the simple set of the kinematic results from uh a run with coriolis on and like intermediate drag, you see the waves uh coming in from offshore um they're high at the they break and then it dissipates to the Lagoon.

N

um This uh results in free service setup that drives the flow again like into Lagoon and out as a read fast jet, and we can see in this particular simulation at 30 degrees, south um we're in southern hemisphere lens. So things will bend left. um uh So we have to see a pretty strong deflection into the jet uh quite near shore, which was a surprisingly interesting.

N

So the next thing we can do. We can look at the results across the parameter space that I describe the numerical experiment, so uh the y-axis here is latitudes, are increasing uh going down and the x-axis is uh there's a the velocity field um related to the bottom roughness Z naught. So um you could go one at a time we can see if we stay on the um High drag, which is 0.25 centimeters column.

N

um Oh, okay, sorry, the the left panel is um the radial velocity component and the right panel is the azimuthal velocity component.

N

Polar coordinates are natural for an annulus kind of setup, um uh so we can see as we increase the latitude, we get a increasingly stronger deflection, which is to be expected and as we uh go right in parameter space with higher drag, we get weaker and weaker flows um and but interesting stuff starts to happen at um at those lower drag and the kind of bottom left of the parameter space that the uh lower drag values and uh higher rotation, where the jet not only forms the Coastal Current, but it does run back into itself.

N

um If we kind of look at the minus 22 and a half South and across the range of drive, we can see that in this in the hijack case, like 16 centimeter bottom roughness case it deflects, but then it dissipates before encountering the next jet. So there's actually not like a strong interaction. But if we uh look at the quarter centimeter on bottom ruptus, then you can see the um there's a more intense recirculation region, uh intensified, Jet and um clearly strong interaction um with with itself. But you know this.

N

This is again modeling, like neighboring Jets um put a certain proximity to another.

N

So that's an uh it's interesting results and we can abstract it um by Computing like an Excursion distance of the jet, um and also uh quantify the interaction um as a Coastal Current to the a jet outflow of momentum ratio, and you can see the genetics version uh distance that changes both with rotation and with with drag um it's uh the more drag you get. The less far makes it out and also um with increasing rotation. It doesn't make it as far out.

N

um We also see that there's a much stronger, uh Coastal Current to Jet momentum ratios in the bottom left of the parameter space, where there's both um uh strong rotation and low of drag so so it'd be nice to like uh it's fairly clean results. So it would be nice to reproduce it with an analytical model. So um we just want to see if we can come up with something back on the back of the envelope to describe it.

N

So there's already been work done on like kind of the inshore part of part of this problem, where uh you know estimating the initial jet output speed based on um the reap affiliatory and the incident wave field. um So we can get like a a v-not for the jet and then also if we, uh we assume, like a conservation of potential vorticity, there's been work to say that we can kind of know the magnitude of the shape of the Longshore current constrained by the flux. The initial flux too.

N

um So the part I'll show you guys now is um um where I picked up and then is once this Coastal Current is formed. How far will it make it um until it interacts to the next jet so um so you're, just using a simple momentum balance um in the natural coordinate system? We assume the balance between three terms.

N

We went to the non-linear introduction, a bottom drag and a pressure gradient, but uh in an annular domain you can't have like a net integrated net pressure, gradient, an asymmetical Direction. So we kind of ignore that term. You might be locally important in a greater sense.

N

So then, then, all you get is a uh exponential decay, um um we're controlled by the depth that the current is at and the drag coefficient um so that, let's gives us kind of an e-folding scale of how far a jet will make it down down the coast, and um we can say that for some values of the uh of that e-folding scale, there will be strong interaction and then uh for some values there will be weaker interaction and um we can also calculate the speed at that point uh down at a certain point um uh in the asking of the direction, and that gives us sort of the our our models or um analytical version of the across flow interaction.

N

Velocity ratio uh bco on DJ and then we can um I won't show you guys the nitty-gritty, the analytical model, but that's um maybe later, if you guys, but um the the point is that we can just we can reconstruct uh probably the trends and, um if not quite the magnitudes, um but we think Bradley's capturing the general visits the problem and it Compares favorable to the numerical Results issue about it.

N

um So kind of revisiting the uh necessary pattern, clockwise flow, we think we'd uh come up with the.

N

uh With this figure, um they thought it might be uh some sort of um heating uh coriolis balance, where we kind of demonstrate that it's um that could very well be a barotropic effect, and here's just some uh streamlines that show very uh well that uh in ideal conditions you can get this sort of uh flow on the um on the four Reef. Oh that's, coherent,.

D

So how much time.

C

I guess a little bit right, yeah.

N

Yeah I guess: I'll talk a little bit. I'll talk a little bit about um the work I'm doing now um with Melissa and Chris. um It's a project on the spreading of um excuse me of uh plumes with different densities in the surrounding environment.

N

So um our approach has been so far to develop a Evolution equation for the spreading or diffluence of the force plume, and now that we have that we plan to uh carry out a series of idealized modern simulations using uh Coast again um to look at how yeah cooler warm plumes evolve differently in the mechanisms responsible for how they spread, dispersed um it'll, be again like a systematic, systematic variation of stratification and uh maybe geometry and wave conditions. So I'll just show some quick plots.

N

um There are like philosophical uh similarities, between uh weight, driven, plus and briefs um rip currents um because it's not financially controlled, but here's kind of the difference of the scale that we're looking at. You know a reef, the the where the jet exits might be a whole kilometer away from the coast, but in a rip current it's going to be more like 100 meters, so there's much larger.

N

So there's some interesting differences and we're doing work to kind of understand that in the broader framework, but for the rip current, you know you get um from something like this, where yeah again like way, German plus, um but uh but in this case and on reeks a lot of drag, is going to be a lot more important than an unrip current. So you might have radiation.

N

Stress, experience a planet, stronger role, um so yep we're kind of finding that the the we're interested in the Longshore extent of the current cell and um yeah it's like is there. You know, there's some limiting distance at which the uh rip current can entrain flow. So we'll work on that right now and before we get to the spreading stuff just to get the initial bear Tropic Dynamics right first before we do anything with spreading, but uh yeah um I think that's where I'll stop that, but yeah thank you guys have for having me.

N

G

Good yeah, um thanks enjoyed your talk. uh You should I think that the coriolence parameter does play a role right that was early, but then, in your sort of idealized model, I didn't see. Coriolis showing up did I miss something um okay, silence.

N

Oh, oh sorry that was the stream wise, um okay, yeah, yeah yeah, um so I was just looking because, because in in the uh uh normal direction, it'll be it's very very strongly geosphere um yeah, but yeah, um it's nice, because when you rotate it to stream, live stream normal. You can like ignore the.

H

Yeah, so can I ask a question in the wave Dragon you you have a wave model in here. Yes,.

O

Yes, but you have.

H

The outer boundary conditions of a wave model- and you have sort of like swell coming in- is that the.

N

Idea, yeah and.

H

It comes at right angles to the coast everywhere,.

N

H

Therefore, that's why you can't set up pressure grades in the ants missile Direction everything's symmetrical. But if you had just the waves coming from One Direction, then you pile the water up, maybe and kick it so then you could get pressured, it would be quite different and then obiquity of the Waves would be another parameter. Correct.

N

Yes, yeah and in practice, um uh you're going to have stronger weaker swell from the north Northern uh southern end.

C

Of the island yeah.

N

H

Publicity of that and the other question is these: these Reef exits these reefer I mean it would seem to me that there's something that the current stop. The coral from growing The Reef there yet.

N

It's really interesting because the.

H

Coral people must they take the spot away or something it just because they have to get away somewhere, but it doesn't keep growing there. So is.

N

That part of the puzzle so Coral or nutrient uptake of a coral is understood to kind of limit its growth and in tropical regions. It's a a very interesting poor, so they actually it's it's a positive relationship with flow speed and coral growth up to a point up into a point where there's like significant stress, but that would induce breakage um and usually that will be from like wave like large wave events like hurricanes um versus like strong fluid I.

N

Don't this is a little yeah, my wheelhouse, but as far as I know, it's actually like favorable.

M

Yeah, so your age and it's in your scaling, yes, is that a tune parameter or that something that you could know like under sea level rise or something like that. Could you infer what the.

N

Future ages would be for like the Stevie on age yeah, so because we know the bathymetry we like, we can estimate where the where the Jet's going to be based on the conservation of PV. We we know that we know that Excursion is not in depth yeah, but that whether that would like change with sea level right, like kind of like where, where it spins off it's more in the tens of meters or 100, meter range I'm, not sure if one meter, but what would sea level rise is certainly going to have a first order.

N

Impact is on like the openness, the wave breaking and like the the physics. So it's it's a very interesting connection to me between these. You know breaking is happening in this much water and then these kind of large-scale features that develop. But if you all of a sudden increase the depth that could you know, have uh big effects on but kind of indirectly,.

D

One more question from Julia on the back: um I'm sorry.

E

um Thanks for your talk, I enjoyed it. um I was curious about the drag event, and so I was curious. You know in this week to have more creepy areas and more Sandy areas, and if you considered, like facially, variable, drag or yeah yeah, um we.

N

I didn't talk today, but we definitely for realistic applications that matters. We found that for um doing a wave model like on the North sword. Maria it was it improved. The scale to have some sort of spatially variable drag, um but I I would think that I would yeah I would use that for a realistic model, but maybe for an idealized model. It tends to be that, like in the deeper areas where drag would not, there would not be as much Coral in those kind of deep Sandy channels.

N

You would have less drag anyway in a better place, so, but but it is likely very important.

D

All right thank.

C

You very much again, please.

D

Show this please go ahead and show your screen our next next speaker is Giovanni cejo from UC Boulder he's going to talk about Regional mode, csm16 configurations for the Caribbean Sea.

L

um Hi everyone uh thanks for having me um my name is Giovanni I'm, a graduate student at CU I work with Samantha, video and I've been working. Today we could have one Frank on setting up a regional bum, six computation clinical DNC um previous studies, uh I've been working on multi-observational work. You know Caribbean Sea, looking at the Amazon clothing tracks and how it drives as an antibiability in the conveyancy and and for this step of our uh my project, we decided to uh head into more of a modeling component.

L

So that's how this configuration configuration is born. um So um we've with the better part of last year, we've been developing. This quarter of a degree configuration that you can see here. um It has 65 sea, star, vertical levels, um initial and boundary conditions uh from glories at a portable degree, also uh daily.

L

um He implemented some Indian restoring and function layers uh using monkeys, um the fourth scene coming from Bia bronoff. We moved away from our Dia into Globus because of some issues with getting the North Atlantic Basin, Network, skewing or selling the signature in the model, um and then we have chlorophyll climatology uh we're stimulating 20 years 2000 to the end of 2019., um and uh so this run that I'm, showing you actually uh was run overnight by Phoenix sunscrew blood this morning, so they're fresh out of the open uh and yeah.

L

So uh first off, we have like a full 20-year mean so some of the surface fields from the simulation, the black box kind of like highlighting or remaining audio focus. uh Once we start looking, uh uh we have a properly working model that we feel comfortable, analyzing and studying processes, um and the first thing that you guys can probably notice. So we have a model output on the left.

L

Side of the middle panels are the actual glorious data um that we're also using uh 4D model and then the difference between both um and on the differences. You can see that we're getting some uh features along the Asian boundary in temperature um salinity uh doesn't look too bad. There's. Definitely some differences um in the tropical Atlantic and in the area of the Amazon River uh runoff facing uh which uh may be relevant as we move forward uh to look at in a little bit uh more detail um seeing surface side.

L

We have something um suspicious, also going on on the eastern boundary a mixed layer. Depth in the model it seems like Mom, tends to have a shallower, uh makes Lego depth overall than uh or glorious free analysis um going back to the eastern boundary.

L

If we look at the uh meridiano component of the law cities, we see that there's definitely some uh flows along that boundary uh in the model, um and so hopefully that's not to be washed out um for the people online, um and so some tweaking uh that we may have to do uh still at the open, but at the decimal boundary um and then also relevant, is that we see that the Amazon River plume may be moving a little bit slower than a glory.

L

So we'll need to find a way to run these the actual observations and determine how that might play a role in the spreading of the bloom Waters that say approach the Caribbean Sea um um just to highlight uh decent boundary. Again. If we look at a season of the conversation uh that feature Landings and Boundary, it seems to be persistent. So it's not something that's happening uh any point in the simulation and and kind of like uh filling out that area like that.

L

So it seems to be fairly consistent um other than that I'm talking about some uh well-known things that we can quickly identify the equatorial counter. Current uh appears along the right uh latitudes, maybe pushing a little bit to the North in some cases.

L

But there seems to be some larger discrepancies on the sound of velocities along the coastal shelf uh of South America, which again potentially relevant to us in terms of how that affects uh the attraction of the plume into the Caribbean um in terms of runoff, which of course, again really important for us uh at the seasonal scale.

L

It looks everything seems to be happening uh when it's shipped uh uh lower uh Plum Waters uh during the winter increased interest, training reaching the maximum sometime in the summer and then slowly decreasing their extent uh during fall in the winter, uh and you can kind of like see here also the effect of paper towel in the exporting. Some of that fresh wire um across the Atlantic first in East, um so, um interestingly, uh temperatures and this happens across several fields at the surface, the largest differences don't always happen along the same season.

L

For example, uh We've uh I saw that I noticed that the largest uh differences in temperature tend to occur during uh the string, while the largest differences in salinity appear to happen during fall, which is during the maximum runoff from the Amazon asset um in the qvm.

L

So interesting that this larger discrepancies are happening at different points of the year um and even make clear depth also happens at a different point of the year than temperature, um so to get an idea of how the models actually performing within a region of the interest more or less, um we took some uh graded Argo data from the Ramekin Wilson product. We extracted some of that data for decentricular event that more or less a little idea.

L

That's highlighted in magenta just to see uh you know both how our model, our monsaics and glorious, are doing against those observations. Temperature looks perfect basically, but what we get to salinity there's some uh obvious differences there, um we're not sure at the moment how many outward floats are actually in that little box, so that may influence um some of that. uh The.

H

L

Greener product from order, so something to look forward to actually uh what's happening. But um yeah, uh we're on on a good side are sees the seasonal patterns of seas within the group, yet weren't uh properly aligned with either Pilates or Argo. um When we had uh DNA for runoff. So at.

O

Least, we know.

L

That load size is helping with that um make sure your dad, um as I mentioned, it seems like Mom six tends to uh drive a relatively shallower uh mixer that uh across the 20 years, um and then you will be at least within the Caribbean Sea appear to perform rather well.

L

So um it's kind of like the current state at the surface of our configuration as I mentioned, this is fresh out of the oven, so I haven't finished, I haven't even looked at any vertical sections or layers to see what's happening there, um and so hopefully, looking at some of the transports. Some of the vertical layers um kinetic energy Maps will help us uh understand what the model is doing. Why is it doing uh behaving in this way?

L

And you know hopefully, within the next few weeks, uh improve 30 different boundary and some of the selling it fluxes in terms of the Amazon River runoff and how to express sort of romantic and interpre BMC uh yeah uh kind of like a quick overview of our state of our simulations of thinking.

D

B

B

A

The Argo uh time series plots those young labeled as sea surface right.

L

B

L

Four arguments would be near yeah.

A

Yeah, so that's my question: are you really are you comparing apples and.

L

Apples here it's 2.5 uh decibels.

O

A

That's the shallows, but just does Argo really measure salinity to 2.5 meters, I thought they turned off deeper than level. That's a good question and I think. Well. As you said, you want to move into the subsurface comparison and subsurface, but yeah yeah. That's a good question, especially well. Maybe in this one we expect pretty strong salinity stratification and something could change a lot into a 2.5 and 10 years. Yeah.

L

Especially during the previous episode.

D

So yeah any other questions for everybody.

D

It's not now we have some time to discuss about Regional virtual modeling in.

M

Csi so I have a kind of just closely related to thank this product, which.

F

Is I saw somebody was.

M

Using the Dubois Montague old statement set come on just released a new 2022 version. That's got 15 years more Argo for mixed layer. Death um I can share the DOI with if anyone who wants it, but it's much cleaner than the previous ones. They're now there's.

C

Some data, that's actually data, all the.

M

Good points this time and then um the but the 10 meter versus surface grid cell issue and the definition of mixed layer. So if you're doing a Delta really matters for exactly the point that Frank just brought up, um Anne-Marie trege has been looking into that. There's can be huge density differences between two and a half meters and 10 meters depending on season, and you know how high.

H

And they have to turn Argo off against biofile. That's why they do it yeah.

H

They keep pumping to the surface that doesn't work outside.

D

So we should Philadelphia want to see it's the option to complete relatives to 10 years. Yes,.

M

Where that's one of the things we're going to put into the whole map for our cmap 7 requirements is to calculate them consistently because they are inconsistent, which nobody saw coming basically.

D

Okay, Regional ocean modeling in CSM suggestions, comments. What would you like to see happen.

C

D

C

A

Like I said just come back to the comment I made at the beginning, you know, but clearly we can't recover all scales in a single code base or with this with the resources we have in this effort, and we have to somehow um no. That means we have to decide like down to what scales. Are we really going to work to make the model appropriate?

A

And you know Giovanni, you know- is sort of going down to the tennis he's intending to go down to the tennis kilometer range I.

C

A

um And I think that's a a resolution, we're pretty comfortable going towards you know, but should we be thinking about pushing Mount six down even further um and supporting it uh with a appropriate parameterization suite at that resolution, or should we, you know, be working in the mode we've just heard about in a couple of talks where we sort of provide, you know, conditions at the on the Shelf outside the surf zone, certainly, um and how? How can we um stimulate that kind of interaction and support of the community and those kinds of calculations is.

H

There support for non-hypnosed credit.

M

But don't understand so this is one of the things that we wanted to look at is how small you should go before it actually really starts to matter. We found that there was no difference than Narragansett Bay anyway, at five meter, horizontal resolution with non-hydrostatic it only it happened. Oh no, sorry, a 10 meter resolution. It comes between 10 meters and 2 meters, so it's a long way off, but that's a shallow bay.

M

So you know whatever um it depends on what what structure you're looking at, but I think that the try non-hydrostatic is coming more from conceptual side than the pragmatic side, at least in the ocean codes in the atmosphere. It made me make sense on much of a larger scale.

D

We have a comment.

O

From Justin uh they uh internal waves uh need the non-hydrostatic and uh trying to remember scale, hundreds of meters um scale, internals nonlinear waves, so.

C

O

They're of interest to people, then you need the non-hydrostatic.

M

Yeah, that's a good point Justin and actually the the way that we did the offshore boundary conditions specifically filters out the internal waves. So we would have had no idea if there were tidal boars banging around. You know it to 100 meters back, but they were all gone from our sidewall conditions.

A

So you were using monthly data monthly.

M

Data that was good.

A

Enough, well, that was going to be my question. Yeah we routinely you know, especially, of course, the resolution, but even the final resolutions, just because of storage constraints, we typically, we won't save very high frequency. Is that limiting the kinds of problems that people can do um or applicable science in the coastal zone.

M

Yes, um it is, this was a workaround, but I think it would have been nice to have had an option to say we're really interested in this one year. This one region can we can we, you know, write CPU our proposal together to get just that little box rerun. You know we're keeping the high frequency in one small region. That kind of capability would would really have helped that we would.

M

We could have taken a lot of uncertainties off the table by doing that, but I wouldn't think that you know these kind of calculations aren't done very often, and one reason is because everybody who's running the coastal model needs their little region at high resolution and they are all doing different experimental designs and whatever so it would have to be in a collaborative. You know, kind of request phase, but the capability to do that and like that sort of thing, would be really helpful. Well,.

K

Yeah yeah actually I think it's more. We need to make sure that we say that we start fighting Etc, because the the story I know is really. The limitations like rewarding the 110 to a degree for one year uh in Saving, advocacy, output for your or your region or even globally is is not that could be the biggest. So it's actually much easier to do that um than like trying to find everybody and like, as you say like there is people like doing normal application in everywhere.

K

So we cannot say that yeah I mean the amount of data that we're not saving at just like private, daily three-dimensional being on the ocean. It's just like subtractable, like even what we have know. It's like. We will be issue to try to cut the data like uh accessibility, for so it's much easier to rerun like a segment and save what you need in like trying to save what proportionally.

K

Maybe somebody sometimes will need that support, but but I asked there was some issue that, with the first point of integration that was running in China, where we have like clearly, uh they didn't like archive properly. Like all, do we stopped by so there is like. So that's more like something that need to be make sure that, like the people that, while those simulation just like working on them properly and save uh every like the result, file like, for example, case directory also so so that people can socialize make like a utk. And we.

M

Were in the segment to say that I can see that we need to do that.

M

I guess another thing to consider is so we added in the tide person, but there have been discussions at this meeting or similar ones about having tithes and cdsm at the 136 degree. That probably is a very good idea at the 112th it's kind of like marginal, but you know is: is there any thought about that kind of stuff.

A

I, don't think anyone I, don't think anyone hears it's not a top of our list, because I think we're sort of assuming that there's a not project at gfdl was doing that. So we can hopefully leverage that if we need to do that so ads around it's working well,.

C

There's people with their hands up online.

A

P

Right Rob go ahead, so I might have missed that last little bit. uh My question was going to be regarding um what Baylor mentioned: Regional modelers running their little region at very high resolutions, um I'd like to do that um um so I'm I've been around user for for uh since its Inception and uh it'd be great to switch to another model.

P

um Is there like and and that, and if you are interested in going to higher resolutions, looking at the work that's been done using romance might be a good model for what's what you can do in Regional model conviction configurations, but is there any? Is there a lot of interest in that in model 6 development, or is the focus definitely on large-scale regions or Global modeling like how? How much do you want? How much would you like to replace ROMs.

D

Plus one so there's enough for going on right now uh to adopt Prime six as a regional model throughout Norway. So all the the Fisheries.

D

um People are moving towards Norm six and the big momentum going on in a week bye week, every other week meeting, that's led by Charlie stock from from gfgl I. Think it's like close to 40 people attending those abilities and just like this uh to 2006 or so different domains and put together um and a lot of ROMs expertise being transferred to to Mom six. You know Kate had some is a good example. She might be online but others so yeah.

D

The idea is to be able to do the same kind of Regional simulations that you do with ROMs with mob.

P

Okay, okay, thumbs up! So that's that's good foreign! Thank you.

B

C

Q

So a couple of things first I can confirm that there is a significant effort underway um to put uh Tides into mom six, um and this is including doing the the self-attraction and loading online using the full spherical harmonics um in a way that we have uh graciously been offered by uh the the folks down at Los.

Q

Alamos, who kind of have have had a really nice little breakthrough within about the last year and so I think we are looking very much at the prospect of having the ability to run tides and looking at how they will evolve, for instance, in a changing climate.

Q

um The other thing I can confirm is that as a part of that effort, um uh hewang and and in particular and Brian arbic have been working on um looking at what we can do to represent the tides accurately without having to push to very high resolution and and again we're pretty optimistic that, at least for the barotropic tides that we can do pretty well um using some things that he Wong I think is going to be describing in a paper coming forth without having to push to exceptionally High resolutions um at a 12th of a degree.

Q

We think that we can do a reasonably good job with the barotropic tides compared to say a 25th or a 50 50th of a degree, and so again the the prospect of having self-consistent ties in climate models is something that is coming and we expect will probably be of great interest to the CES Community um once it it gets there, um although it at this point, it's not clear that it would really need further investment from from cesm for that to happen. So that's the first one on Baylor's questions about the tides.

Q

uh The second one I can confirm. First, that we are exactly as Baylor described, um putting together Regional mom, 6 configurations. This is um been funded by NOAA um and they're they're. We expect there to be hiring going on of permanent positions along those lines. uh As for the question of of displacing ROMs, um it's really more of a question of providing capabilities that are analogous to what's in.

Q

What's in ROMs, we're not trying to take away anybody's anybody's job, there are things that ROMs is being used for right now that are really nice and and I. Think highly of the ROMs people.

D

Thanks Bob today, there's been some chatting some conversation in the chat as well. uh This session is not over by the way. There's one more talk tomorrow by my router Rutland.

H

C

D

um I guess: okay, we still have anything else about Regional modeling in CSM. If not, it started for a year.

A

B

Tomorrow morning,.

A

And we'll be back in this room and online.

E

A

B

uh Yeah see everybody tomorrow, a.

A

Little bit early on schedule for the beer, but that doesn't mean we can't go.

A

It's already been.