2024-03-15 - March 15, 2024 #52

dprincz · 2024-03-15T22:07:12Z

dprincz
Mar 15, 2024
Maintainer

Work management and coordinated planning

Al brought up work management processes, in the context of tools and research branching back out and losing the connection to bridge innovations back to the community
Could consider work planning but expressly write down necessary developments and expected deliverables to coordinate this, to the extent that priorities are identified
The community is currently lacking proper documentation and papers to reference updates (e.g., MOUNTAINMESH, etc..)

Workflow discussion

@AlaBahrami brought up the divergence from the initial vector-based workflow relative to the Model Agnostic Framework (MAF) development
@bdavisongit mentioned Zelalem is looking at @kasra-keshavarz's workflow
@AlaBahrami would like to see the initial vector-based workflow and MAF converge
Al's perception is that this has already happened; this should be obvious from the MAF training to take place next week (March 21), and @kasra-keshavarz notes the core has evolved from that work (ref)
@AlaBahrami notes that it is not clear how to contribute to a main version of "meshflow"

Contributions and licensing

For contributions of scientific code to MESH, we cannot accept models using GPL as it's incompatible with the Open Government License - Canada (there's no problem if they instead use Lesser GPL, LGPL)
@kasra-keshavarz notes "meshflow" is still considered 'pre-release', not even yet at 1.0 version; this might be why the management/contribution process isn't as clear
@mee067 notes a group met earlier this week about the workflow, and that this process might be starting

MOUNTAINMESH

To get MOUNTAINMESH to work for the Yukon, @mee067 had to use an additional script from Zelalem
Some of the parameters take a long time to calculate; requires the DEM, so might require additional data than the MERITHydro/blue-line datasets
At some point, this process should be integrated in MAF
Al notes the GIStools part of MAF should be able to calculate at least some of the required parameters
@AlaBahrami brought up that the codes go back and forth between RScript and Python; it would be useful if the tools in both languages performed identically

Model Agnostic Framework (MAF)

Al asks ECCC to help converge to MAF for the MESH workflow
ECCC/NHS confirmed MESH is still being maintained for internal and external clients; ongoing workplanning includes updating the ECCC models (following pending Innovation Cycle 4, due this spring)

Geospatial fabrics and processing

There are slivers that exist in the MERITHydro geospatial fabric that creates artefacts that cause issues when running the model; these have to be corrected using workflows (i.e., systematically)
Sanity check for basins that are too small, which either invalidate the model physics or needlessly increase computational time
There is a need to identify the smallest permissible basin size/shape in MAF; @kasra-keshavarz notes Zelalem shared his updates; in time, these will be integrated into MAF
Geofabrics should also be delineated to WSC gauge locations; Bruce brough up the UWaterloo product - @sujata91 is doing a comparison of that and other geofabrics, including MERITHydro
An issue is identified with the way MESH connects a subbasin to a gauge location, as the centroid is used by MAF to provide subbasin coordinates; MAF could use subbasin outlets to write the subbasin coordinates to file, as opposed to the centroid

Grid versus vector-versions

@mee067 notes the vector version does not perform as well as the grid version, in terms of simulated outputs
Under future scenarios, the vector and grid versions show different progression of peakflow under future climate
Vector-based setups introduce variable-resolution, which may invalidate the GRU assumption for parameter transferability
As part of MAF, EASYMORE might not be considering variable scale when mapping variables, which could be distorting the forcing or other variables; @mee067 suggests a need for harmonization than simple spatial interpolation
@AlaBahrami did work a few years ago to compare the grid and vector-based versions, and validated variables were consistent between both versions
Issues with invalidating parameterizations, or assumptions of the parameters, might require reassessing geofabrics and considering the creation of subbasins of similar size

Actions

@kasra-keshavarz and Zelalem will continue meeting to converge different workflows; @fuadyassin will also get involved
@AlaBahrami, @mee067 and @fuadyassin will follow-up on comparing the grid and vector-based versions
Al to initiate work reassessing geofabrics, like MERITHydro, to ensure the derived products are model-compatible; @sujata91 will follow-up with Rodolfo
@dprincz to continue and complete the migration of r1860-ME changes into MESH-Dev with input from @mee067 as required
Al will explore a strategy to create a new Slack space and retire the GWF one (contains too much stale information)