djk2120 / clm5ppe Goto Github PK

starting with global mean annual mean LAI, build a framework for emulation

• test different emulation methods, see ESEm
• test emulator performance vs. training dataset size
• "leave one out": remove one member, train on the remaining ensemble, and predict the held-out run
• check reliability of uncertainty estimates

See McNeall et al. 2020

sparsegrid global accumulation/averaging does not include Antarctica, therefore if you are comparing to other global products, you need to mask out antarctica

for standard 2deg output this can be achieved via:
f = 'clusters.clm51_PPEn02ctsm51d021_2deg_GSWP3V1_leafbiomassesai_PPE3_hist.annual+sd.400.nc'
sg = xr.open_dataset(f)
la = sg.landfrac*sg.area

Can we leverage data catalogs (e.g., intake-esm) for managing and sharing the CLM PPE output? See also catalog creation with ecgtools and esm_catalog_utils. Lots of example notebooks through ESDS and Pythia.

create a notebook to calculate metrics of interest, which will write to a shared netcdf

Separate commonly used functions (e.g., amean, gmean) into a shared analysis library so we don't have to copy/paste them into individual notebooks.

Maybe we could perform sensitivity analysis on global avg LAI (2005-2014), using the VISCOUS methodology, and compare to Linnia's existing R scripts.

Python package:
https://www.essoar.org/doi/abs/10.1002/essoar.10512586.1

Reference paper:
https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2020WR028435

@linniahawkins @katiedagon

should be noleap to comport with CESM

@djk2120 the modifications to rank_plot and top_n in analysis.py, specifically the indexing by minmax, appear to not work with the OAAT data format being read in by get_exp. I think this has something to do with how the minmax coordinate is set up. Do you have another processed data location for OAAT output that works with the new rank_plot function?