Comments (11)
I've done a simple GMM analysis on surface speed, \sqrt{u^2+v^2}, averaged over the last 10 years of the RYF simulation. Interestingly, for 2 clusters the model successfully highlights the extent of the ASC hugging the coast, though we need to do more analysis on whether this truly is the ASC or some other features that are being lumped together.
from asc_and_heat_transport.
Interesting! It looks like a mixture of ASC and coastal current that is picked up by the 2 cluster version. And you need 5 or 6 clusters to include the Weddell Sea where surface velocities are low.
A little off topic: but interesting that the Weddell Gyre is a consistent feature for all higher clusters, while the Ross Gyre isn't.
from asc_and_heat_transport.
The calculation of the along-slope velocity does only depend on the local (grid cell) bathymetric slope and can be calculated for every grid point. So no need to extract any isobaths beforehand - in fact I calculate the along-slope component first and then select an isobath, if I remember correctly.
The along-slope velocity will have neg and pos values which might help in isolating what I call the "reverse ASC" (= ACC) in West Antarctica.
In general, I'd prefer an approach where we don't constrain the input for the GMM too much. Creating a mask based on bathymetry values makes sense, I think. E.g., Fig 2b in Huneke et al. 2022 shows the velocity for a mask shallower than 2500 m. Maybe 3000 m would be a good starting point. But if we start to select too narrow values/individual isobaths, then we make already lots of assumptions. My hope was that the GMM approach would (ideally) do exactly that for us.
Finally, Fig 5b shows the barotropic velocity, I think that's a better approach as the bottom velocity. The bottom will be large in the bottom-intensified regions (where there is DSW export), but close to zero elsewhere.
from asc_and_heat_transport.
@wghuneke That would be amazing! I used the time-mean velocity in the final 10 years of the RYF run, so perhaps the monthly along-slope velocity in those final 10 years would be good?
I can calculate the time-mean myself, or otherwise explore seasonal/annual variability.
from asc_and_heat_transport.
That's a good point @wghuneke! So far I have just used the speed, so it doesn't take into account any directionality (which would clump things into different clusters). Perhaps n=4 is better because it removes the coastal current, and also highlights a weak current in the Ross and Weddell Seas... I'm guessing the Ross Gyre has more of a sub-surface signal, which is why it isn'y highlighted at the higher cluster numbers...?
One option for further exploration could be to use the full-depth speed profiles, or to add the surface temperature and salinity in our clustering algorithm. Or perhaps we just use the cluster outputs above as our definition and do diagnostics based on that?
from asc_and_heat_transport.
Maybe another test could be to run the GMM analysis with along-slope speed instead of the actual speed, just for purposes of identification.
from asc_and_heat_transport.
I'm not sure what you mean @willaguiar. The along-slope speed would be at a fixed contour (e.g., 1000m), so isn't that the current definition of the ASC anyway? I was thinking the purpose of using GMM was to objectively define the ASC without needing ad-hoc solutions like that? Or perhaps I don't understand the purpose of calculating the along-slope velocity!
Also, I did a similar clustering analysis by feeding in surface speed, temperature and salinity. Due to computational complexity I was only able to get 10 clusters. But basically the algorithm is more inclined to highlight surface water masses than speed, probably because 2 out of the 3 variables are T and S so they are outweighing the influence of speed in defining clusters.
from asc_and_heat_transport.
You have a point @taimoorsohail . I thought of using the same velocity transposing process in the whole domain ( just extending northwards and southwards of the 1000m isobath) to make all U,V follow along/across coordinates instead of I,j in the whole domain. But I now realize it it would be tricky in the Antarctic peninsula.
Do you think "cleaning up" the domain before running the GMM could refine the results further? ( for example, only using the surface where the maximum ocean depth is smaller than 2500m , so the algorithm is forced to focus on the region of interest)
from asc_and_heat_transport.
@willaguiar - yes - cleaning up would definitely help by masking the deeper regions. I had another idea - what if we performed the clustering of speed in isobath vs distance along contour space? That is, we calculate the speed along each isobath from 0 to 3000 or 4000m, and then cluster that speed to identify the isobaths along which the ASC signal is evident? This would have the advantage of identifying the isobaths over which averaging/integration should occur, so it would enable further analysis that isn't constrained to the 1000m isobath.
In Fig 5b of Huneke et al., 2022, they plot the bottom velocity (I think?), so perhaps that's the best thing to cluster? The only issue is I'm not sure how best to calculate the along-slope velocities for a range of isobaths, but I think you or @wghuneke would have the code to do that?
from asc_and_heat_transport.
@taimoorsohail - do you want me to prep a file of the along-slope velocity? What time period + frequency did you use for your GMM analysis?
from asc_and_heat_transport.
Just a note that we're not following up on the ASC definition anymore. We decided to use the isobath definition and potentially repeat the calculation for different isobaths.
from asc_and_heat_transport.
Related Issues (20)
- Fix zonal convergence of heat transport around peninsula [ Task 5] HOT 3
- Cross-slope heat transport calculation method HOT 6
- Cross-slope heat transport using different methods to extract an isobath HOT 6
- Poster for IUGG
- Identifying ASC Regimes and variability HOT 45
- Zonal convergence bias in vertical profiles of CSHT HOT 6
- Analysis of individual sub-regimes HOT 6
- Alternative plots visualizations HOT 6
- Figure discussion for paper HOT 4
- Recreate figures with time-mean regime masks HOT 3
- Statistics using longitude bins before averaging over distinct ASC regimes HOT 29
- Depth normalisation HOT 21
- Reduce bin size when longitude-binning CSHT and U_along HOT 4
- Presenting these results at conferences HOT 11
- Missing data at longitude edges HOT 10
- Choice of vertical coordinate / depth range of importance HOT 6
- New visualisation of cross-slope properties in Antarctic margins? HOT 7
- Bining of ASC speed into sigma increase maximums HOT 4
- Defining sigma levels according to water mass properties HOT 18
- Explaining spatial variation of correlations in density space HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from asc_and_heat_transport.