Experiment scripts and plots for Paper I - Intro to land model

[kmdeck]

Purpose

This outlines and tracks PRs which add experiments and plotting of results for use in our paper introducing the land model (read only link).

Components

Simulation scripts @kmdeck
Diagnostics @AlexisRenchon
Plotting @kmdeck (single site) @AlexisRenchon (global)
Performance @Sbozzolo
Implicit solver @juliasloan25

Results and deliverables

High quality plots demonstrating the model output for:

• standalone soil: freeze/thaw, layered soil, and evaporation
• standalone soil+ snow: bare soil site (Barrows Alaska)
• integrated land model (soil/snow/canopy): ozark site (LWP during drought, GPP, soil water content, ET, etc)
• integrated land model (soil/snow/canopy): global run with ERA5 forcing - energy fluxes (latent and sensible heat, SWout), water fluxes (evapotranspiration, runoff), carbon fluxes (GPP, autotrophic and heterotrophic respiration)

Task breakdown

A preliminary list of PRs and a preliminary timeline of PRs, milestones, and key results.

• Upgrade all existing paper plots for soil to use CairoMakie: Paper plots #565 @kmdeck
• Add global run of soil/canopy model: Global run of complex land model with ERA5 forcing #591 @kmdeck @Sbozzolo
• Add bare soil site even if without snow: bare soil site #444 @kmdeck
• Diagnostics @AlexisRenchon
• Plotting of diagnostics @AlexisRenchon
• Optimization of models @Sbozzolo
• Plots showing computional performance of model (metrics TBD) @Sbozzolo
• Plot demonstrating implicit solver - timestep vs conservation error @juliasloan25

Reviewers

@braghiere

Performance/implicit solver plots

Beta Give feedback

1. CPU and GPU scaling of soil+canopy model (plot CPU and GPU on same axes, show time taken by each component) - see example plots in comments
2. timestep vs RMSE and time to solution for both soil and soil+canopy (can use this to comment on ideal timestep to use for simulations)
3. (in appendix) Float64/Float32 water and energy conservation over time

[braghiere]

Thanks @kmdeck! I updated the list of plots in Results: integrated land model (soil/snow/canopy): global run with ERA5 forcing.

[AlexisRenchon]

Thanks @kmdeck! Here are some thoughts:

• I think eventually we should separate generating simulation outputs and plotting from those outputs (especially for global run, as we may run simulations on the cluster, and then make plots on local computer). I understand the next step is me working on diagnostics =) we should agree on standardized output file (netcdf probably, with long and short name of variables, units, etc. I think Gabriele has a framework for this)
• I am not sure where and how yet, but having a place where we gather plotting functions would be good for reproducibility. We currently have https://github.com/CliMA/ClimaLand.jl/blob/main/lib/ClimaLandSimulations/src/utilities/makie_plots.jl which contains plotting functions that take inputs and simulation outputs as arguments, and return a plot. This make it easier to remake the same plot with different inputs and outputs (e.g., different start and end date, or different site). We could have a discussion on the best place and practices to do this. It would be great to have a library of plots (and analysis) that improves over time and we can go back to it easily!

Note these are just thoughts and not a priority for the papers.

[juliasloan25]

I added a tasklist "Performance/implicit solver plots" listing the plots we discussed during the land meeting today. Here are the reference/example plots for the scaling information, from Renato and Gabriele: