Add a year long global simulation to run 1x per week

[kmdeck]

Purpose

Add a year long global simulation to run 1x per week

Note that to minimize NaNs, I have decreased the resolution at the surface from to 50cm (in reality, we want something like 2.5 cm), and increased ac_canopy by 10x (in reality, ~10^3). Once we resolve the NaNs in this simulation through future PRs, we can decrease both of these.

Example output after 7 days:
soil water content

gpp

Content

Add a year long global simulation to run 1x per week [pipeline.yml, experiments/long_runs/land.jl, diagnostics code]

Use an updated set of soil parameters with the correct lat/lon and higher resolution [changes to richards_runoff.jl, Artifacts.toml, and soil params file name in land.jl]

Fix bug in bucket model which did not update drivers in callback [global_bucket_function.jl]

Update NEWS.md

To-do

This PR:

• restore buildkite Manifest
• review

Future PRs - summarizing Alexis' comment below and adding some -

1. Figure out NaNs - (a) if we make ac_canopy larger, fewer NaNs. (b) if we make the thermal timescale for phase change longer, we get fewer NaNs (c) if we make the timestep smaller, we get fewer NaNs -> issue with timestepping? better implicit solver? One other question: why NaNs, and not domain errors?
2. Improve/debug diagnostics - more variables, perhaps we want monthly averages
3. Make better plots, maybe animations
4. Use mask in diagnostics so we only get continents
5. Use units that are reasonable for the community

Even later: in ClimaAnalysis, have a function to generate a local dashboard where we can use a slider to change time, a menu to change variables, maybe a play button for animations

---

• I have read and checked the items on the review checklist.

[Sbozzolo]

I scheduled this run to on Mondays starting at 23.35

(It would be good to name Buildkite pipelines without spaces)

[AlexisRenchon]

I can fix the diagnostics on Friday the latest!
We can coordinate and I'd directly push my commit to this PR

[AlexisRenchon]

Ok I made some changes and pushed my commits, GPP diagnostics is written, and plots are produced, it runs well on the cluster.

However, GPP becomes NaN progressively? (not sure if this is a simulation bug or a diagnostic writing bug?)

I think next step for this PR (or a next PR) is:

• add other diagnostic variables (need to fix things a bit in src/diagnostics, e.g., I forgot a lot of ! at the end of compute function name)
• make better plots, maybe animations?
• fix the bugs (NaNs for examples) and make the simulation 1 year long

And for later, maybe in ClimaAnalysis, have a function to generate a local dashboard where we can use a slider to change time, a menu to change variables, maybe a play button for animations?

It would also help to express GPP in units we are familiar with

[AlexisRenchon]

Just a thought that eventually the weekly CI run of this could push a web dashboard to a server (the existing Microsoft Azure VM or another).

I think that would be very cool. We will always be able to interact with global outputs of latest run from the web, convenient for discussions etc. I might work on this with the viz students this summer...