WrPMIP Regional Simulations (Phase 1)

Phase Experiment Input Historical
1 a. Regional Baseline (low temporal resolution) Climate, CO2 CRUJRA (1901-2000) Not simulated Not simulated
b Regional baseline (high temporal resolution) Climate, CO2 CRUJRA (2000-2021) Not simulated Not simulated
c. Regional OTC Warming Climate, CO2 CRUJRA (2000-2021) Not simulated Not simulated
d. Regional Snow Fence Warming Climate, CO2 CRUJRA (2000-2021) Not simulated Not simulated

Regional Simulation Details

At the regional scale we plan to look at model agreement for permafrost extent and ecosystem carbon fluxes and abiotic variables in response to warming. This will facilitate an assessment of uncertainty across models, space, and time using functional benchmarks and ILAMB assessments (still in development). As an important baseline, control and perturbation runs at the regional scale can then be compared to future site-level simulations for model ensemble uncertainty and coherence between perturbation responses and anthropogenic climate change across scale.

Several key expectations are set for core regional simulations of the WrPMIP:

  1. We will need to standardize initial regional simulations as much as possible:
    1. We will provide 0.5 x 0.5 degree climate forcing product for consistent climate forcing (at hourly, 3-hourly, or 6-hourly temporal resolution as needed):
      1. Products we have access to:
        1. CRUJRAv2.3 (1901-2021)
        2. CRUNCEPv8 (1901-2016)
        3. GSWP3v2 (1901-2014)
        4. GSWP3-W5E5 (1901-2019)
        5. MERRA2 (1980-2022)
        6. ERA5 (1950-2019)
      2. We will decide on a single consistent product with CRUJRA a likely candidate having the longest and most recently updated historical period of 1901-2021 (published 2022.09.30).
    2. Standardize grids to 0.5 degree:
      1. We welcome participation for models that don’t meet this requirement.
      2. We will regrid output to 0.5 degree if/when necessary for comparison.
    3. Hourly outputs for recent historical period (1980-2020):
      1. Allows high temporal assessment of ecosystem flux based on diurnal, daily, monthly, seasonal responses in comparison to site-level data.
      2. Most of the historical time period, and all future simulations, can be lower temporal resolution but hourly outputs across experimental warming trials are a key focus of the work.
    4. NetCDF outputs and how we will format them:
      1. We will likely use the NETCDF4_CLASSIC format where data is stored in an HDF5 file using only netCDF 3 compatible API features – allowing backward compatible reading for older code. 
      2. We will likely need each variable as individual netCDFs based on data size if we output hourly data in the 1980-2021 time period to cover multiple warming experiments.
  2. We will similarly implement model warming at the regional and site-level, options to discuss below:
    1. OTC (air and surface soil warming):
      1. Scale aerodynamic resistance (calibration?)
      2. Other suggestions?
    2. Snow Fences (deep soil warming):
      1. Scale the insulating properties of snow (similar to aerodynamic resistance)
      2. Increase snow depth (potentially remove snow in spring, requires timing and programming to remove)
      3. Other suggestions?
    3. A simple optimization step may be pertinent to achieve warming similar to field experiments:
      1. We could run short regional simulations across equally spaced parameter space (i.e. air resistance changed by 0.25, 0.5, 1, 1.5, 2 times current) for several simulation years and look for average increase of air temperature response within the 0.5 – 1.5 °C warming expected by ITEX chambers.
      2. Snow fences could similarly increase snow insulation or snow depth over a several year regional simulation and look for expected range of soil warming.
  3. Spin-up using the first 20 years of chosen reanalysis product
    1. From talking with other MIPs we know spinup may be an issue that will need greater discussion.
    2. Spinup will likely recycle or average the first 20 years of the reanalysis product from 1901-1920, though we can discuss other options like random climate.
    3. We aim to not overly prescribe what each model does here as long as the outcome of representative permafrost carbon pools and stable state conditions are met.
    4. We will rely on modeling groups to report specifics in terms of the number of repeated years of climate data, how many iterations, the final rate of pool change, and any model specific aspects of the spin-up process.

There are several issues with implementing warming in regional simulations considering contrasting years of when experiments were implemented both across and within grids. For simplicity, we will choose an initial year at which warming started for the oldest experiments of each warming type (OTCs and snow fences). Outputs will then simply be aligned at the beginning of warming implementation to compare relative warming responses. We would have to assume that regional climate forcing is biased and that experimental warming at sites are more influenced by site-level changes induced by warming manipulations compared to decadal changes in air temperature over the last 40 years of experimentation. Large assumptions, but by taking this approach we would vastly simplify regional warming implementation in models to provide a regional warming baseline. We can discuss better alternatives at the meeting, like setting up many single-point simulations that start at the implementation of warming for each experiment.