Bias Correction

Aligns CMIP extrapolated CT/SA to an observational climatology on the ISMIP grid and writes bias‑corrected monthly CT/SA. This page covers the classic geographic‑space method; a timeslice/projection approach also exists for basin‑wise T–S corrections.

Key behavior (classic)

  • Single invocation writes both scenarios: run once with the future scenario argument; internally it reads historical and future extrapolated files, computes the bias using the configured historical window, and writes bias‑corrected monthly CT/SA for both scenarios.

  • Climatology: prefer the 06_nov (v2) set (e.g., zhou_annual_06_nov).

  • Inputs must be extrapolated CMIP and extrapolated climatology on the ISMIP grid.

CLI

Run after remapping and extrapolation (Steps 3–4):

ismip7-antarctic-bias-corr-classic \
  --model <MODEL> \
  --scenario <FUTURE_SCENARIO> \
  --clim <CLIM_NAME> \
  --workdir <WORKDIR> \
  --config <CONFIG>

Inputs under <WORKDIR>:

  • CMIP extrapolated monthly: extrap/<MODEL>/{historical,<future>}/Omon/ct_sa/*_{ct,sa}_extrap_*.nc

  • Climatology extrapolated: extrap/climatology/<CLIM_NAME>/*_{ct,sa}_extrap.nc

Outputs:

  • Bias‑corrected CT/SA: biascorr/<MODEL>/{historical,<future>}/<CLIM_NAME>/Omon/ct_sa/*_{ct,sa}_biascorr_*.nc

Algorithm (classic)

  1. Build a model climatology from CMIP extrapolated monthly ct and sa over the configured historical window ([biascorr] climatology_start_year/_end_year).

  2. Read the extrapolated reference climatology (ct, sa) on the ISMIP grid.

  3. Compute bias = model_climatology - reference for each variable.

  4. Subtract the bias from every CMIP extrapolated monthly file; preserve coordinates and ISMIP bounds; write outputs per scenario.

Configuration

[biascorr]
climatology_start_year = 1995
climatology_end_year   = 2024
time_chunk = 12

Set time_chunk to balance IO/memory. The climatology window should match the reference period intended for alignment and be fully covered by the CMIP historical input.

Python API

from i7aof.biascorr.classic import biascorr_cmip

biascorr_cmip(
    model='CESM2-WACCM',
    future_scenario='ssp585',
    clim_name='zhou_annual_06_nov',
    user_config_filename='my.cfg',
)

See also: