FATES calibration update

Rosie Fisher, Jessie Needham, Charlie Koven, Jennifer Holm, Adrianna Foster, Daniel Kennedy, Katie Dagon, Ryan Knox, Keith Oleson, Kjetil Aas, Marcos Longo, Jackie Shuman, Greg Lemieux & the FATES team (and the CLM team! etc. etc. )

Background

FATES has a ‘default’ set of parameters, but these are largely derived from CLM defaults and are not widely trusted

Calibration Cascade

Use a set of reduced complexity modes of FATES to divide up parameter space into manageable chunks…

Calibration Cascade

​

Sat Phenology

NOCOMP

Fixed Biogeography

Full-fat-FATES

Albedo bias

Default FATES

One at a time ensemble

One at a time ensemble

This looks about right! (trade off of time vs bias correction

Albedo bias

Default

Thin layers

Albedo bias

Albedo bias

Thin layers + no clumping

So far so good, but what now?

1. More detailed tuning of albedo & raditive transfer parameters.
2. Calibration of next stage of cascade (gas exchange, hydrology (GPP, ET, etc.)
3. Parallel tuning of next stages of cascade

​

​

So far so good, but what now?

• More detailed tuning of albedo & raditive transfer parameters.
• Calibration of next stage of cascade (gas exchange, hydrology (GPP, ET, etc.)
• Parallel tuning of next stages of cascade

​

​

​

So far so good, but what now?

• More detailed tuning of albedo & raditive transfer parameters.
• Calibration of next stage of cascade (gas exchange, hydrology (GPP, ET, etc.)
• Parallel tuning of next stages of cascade

​

We need to compare output to (at least) all relevant ILAMB data products

We need to run in ensemble mode

We need to do this -a lot-

​

FATES calibration repo

https://github.com/adrifoster/fates-global-cal

Respiration / carbon use efficiency / productivity assessment

Ryan vs Atkin vs growth respiration

https://github.com/adrifoster/fates-global-cal/blob/main/jupyter_ppe_scripts/FATES_mainenance_respiration_ensemble.ipynb

‘mega-ensemble script’

• Check out code
• Builds code
• Makes ensemble of cases
• Makes parameter files
• Modifies parameter files
• Points cases at files
• Submits jobs
• Plots output
• Runs ensemble through ILAMB
• Accesses ILAMB output (bias maps etc.)
• Plots modifiable ensemble plots of model-data comparisions
• Handles all filenames, paths, setup for multiple parallel ensembles.

​

outlook

• New scripts should allow rapid iteration of parameter testing
• CLM(FATES) has comparable error magnitudes (for these data products) to CLM5 already in SP mode.
• We just pushed a new parameter set (by Jennifer Holm) that we will use as a baseline pending calibration updates
• New parameter update method to formalise iterations to the parameter file, including ILAMB metric reporting.

Leaf layer thickness vs. increase factor vs. albedo (CERES product)

Leaf layer thickness vs. increase factor vs. albedo (GEWEX.SRB product)

CERES albedo bias with one at a time changes in RTM parameters per Majasalmi and Bright ranges

Leaf layer thickness vs. increase factor^:: SurfaceNetSWRadiation (CERES product)

Vcmax vs Ball Berry slope: latent heat flux bias (GBAF product)

Vcmax vs Ball Berry slope: GPP bias (GBAF product)

outlook

• New scripts should allow rapid iteration of parameter testing
• CLM(FATES) has comparable error magnitudes (for these data products) to CLM5 already in SP mode.
• We just pushed a new parameter set (by Jennifer Holm) that we will use as a baseline pending calibration updates
• New parameter update method to formalise iterations to the parameter file, including ILAMB metric reporting.
• ​

To Do

• Short Term
• Compare with CLM5.1 DONE
• Run entire ILAMB comparison for radiation
• Switch to using Jennifer's parameter set. DONE
• Push ILAMB results on to some sort of website (in script)

Write code to multiply FATES-SP LAI by 1/clumping.

Acquire new ILAMB scoring system

Debug scoring w/ Nate

Make priors from LHC from M&B and leaf angle papers. DONE

​

Make different rtm grid,

• Generate sparse grid capability
• Data assimilation? (Tristan Quaife, Kris Aalstad)
• Quick and coarse vs. cutting edge and slower (r.e. CLM6/CESM2.