Skip to content

Latest commit

 

History

History
128 lines (88 loc) · 7.85 KB

README.md

File metadata and controls

128 lines (88 loc) · 7.85 KB
Raw

MimiFAIRv2.jl

This is a work-in-progress repository for a Julia-Mimi implementation of the FAIRv2.0 simple climate model. The model description paper can be found at FaIRv2.0.0: A Generalized Impulse Response Model for Climate Uncertainty and Future Scenario Exploration.

Preparing the Software Environment

To add the package to your current environment, run the following command at the julia package REPL:

pkg> add https://github.com/FrankErrickson/MimiFAIRv2.jl.git

You probably also want to install the Mimi package into your julia environment, so that you can use some of the tools in there:

pkg> add Mimi

Running the Model

The model uses the Mimi framework and it is highly recommended to read the Mimi documentation first to understand the code structure. The basic way to access a copy of the default MimiFAIRv2 model and explore the resuts is the following:

using Mimi 
using MimiFAIRv2

# Create an instance of MimiFAIRv2.
m = MimiFAIRv2.get_model() 

# Run the model.
run(m)

# Access the temperature output.
fair_temps = m[:temperature, :T]

# Explore interactive plots of all the model output.
explore(m)

The get_model() function currently has the following keyword arguments:

  • emissions_forcing_scenario: One of the RCMIP scenarios from the original FAIRv2.0 paper. Current options include "ssp119", "ssp126", "ssp245", "ssp370", and"ssp585". The default is "ssp585".
  • start_year: The model has an option to be initialized at different time periods, however this is only currently set up to start in 1750.
  • end_year: The model can be run out to 2500 (the default final year).
  • TCR: The transient climate response (default = 1.79).
  • RWF: The realized warming fraction, defined as the TCR/ECS ratio (default = 0.552).
  • F2x: The forcing from a doubling of CO2 (default = 3.759).



Python vs. Julia temperature comparison

Running a Monte Carlo Simulation

Overview

See monte_carlo.jl for the script and details on running a Monte Carlo Simulation, and don't hesitate to contact the developers with any questions, or post an Issue on Github.

This file contains functions to run a Monte Carlo with MimiFAIRv2 using the constrained parameters from Leach et al. (2021). The first function, create_fair_monte_carlo, loads the constrained parameter samples from Leach et al. (2021) and cleans them up so they can be easily passed into Mimi-FAIRv2.0. It then creates a second function, in the script called fair_monte_carlo that performs the actual analysis. This function returns a dictionary with temperature, atmospheric co2, ch4, and n2o, and radiative forcing. Adding other outputs is doable please add an Issue on Github if you would like this to be done. This returned function can optionally take a vector of vectors to force co2, ch4, and n2o emissions trajectories for the n_samples runs, providing these in the native MimiFAIRv2 emissions units, using optional arguments co2_em_vals, ch4_em_vals, and n2o_em_vals respectively.

The required constrained parameter files are uploaded to Zenodo with the citation Frank Errickson, & Lisa Rennels. (2021). MimiFAIRV2 Large Data File Storage (0.1.0-DEV) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.5513221

Function Details

The create_fair_monte_carlo function is the primary user-facing function provided for the monte carlo simulation and has the signature and function arguments as follows:

function create_fair_monte_carlo(
    n_samples::Int; 
    emissions_scenario::String="ssp585", 
    start_year::Int=1750, 
    end_year::Int=2300, 
    data_dir::String = joinpath(@__DIR__, "..", "data", "large_constrained_parameter_files"),
    sample_id_subset::Union{Vector, Nothing} = nothing,
    delete_downloaded_data::Bool = true
)

Function Arguments:

  • n_samples: Number of samples to randomly draw from original constrained parameter values (max = 93,995).
  • emissions_scenario: Current options are: "ssp119", "ssp126", "ssp245", "ssp370", "ssp585"
  • start_year: First year to run the model (note, Mimi-FAIR requires user-supplied initial conditions if not starting in 1750 so this is not yet supported for a year other than 1750).
  • end_year: Final year to run the model.
  • data_dir: Location of the data files
  • sample_id_subset: IDs of the subset of samples to use from original constrained parameter values (each ID between 1 and 93,995 inclusive). If this argument is set, it will be used instead of the step to create n random indices. The length should match n_samples, and if it is longer we will choose the first 1:n.
  • delete_downloaded_data: Boolean (defaulting to false) to recursively delete all downloaded large data files of constrained parameters at the end of the script. Should be set to false if one will be running this several times and want to avoid re-downloading each time.

Calling create_fair_monte_carlo as follows:

my_fair_monte_carlo = create_fair_monte_carlo(1_000)

will return another function, in this case named my_fair_monte_carlo, that performs the actual analysis. This function returns a dictionary with temperature, atmospheric co2, ch4, and n2o, and radiative forcing. Adding other outputs is doable please add an Issue on Github if you would like this to be done. This returned function can optionally take a vector of vectors to force co2, ch4, and n2o emissions trajectories for the n_samples runs, providing these in the native MimiFAIRv2 emissions units, using optional arguments co2_em_vals, ch4_em_vals, and n2o_em_vals respectively.

function fair_monte_carlo( ; 
    co2_em_vals::Union{Nothing, Vector{Vector{T1}}}  = nothing,
    n2o_em_vals::Union{Nothing, Vector{Vector{T2}}} = nothing,
    ch4_em_vals::Union{Nothing, Vector{Vector{T3}}} = nothing
) where {T1, T2, T3}

Function Arguments:

  • co2_em_vals: A vector with n_samples elements, each of which is a vector spanning the full time of the model being run and holding an exogenous stream of co2 emissions in the native FAIRv2 units, GtC (convert to GtCO2 with multiplier 44/12)
  • n2o_em_vals: A vector with n_samples elements, each of which is a vector spanning the full time of the model being run and holding an exogenous stream of n2o emissions in the native FAIRv2 units, MtN (convert to MtN2O with multiplier 44/28)
  • ch4_em_vals: A vector with n_samples elements, each of which is a vector spanning the full time of the model being run and holding an exogenous stream of ch4 emissions in the native FAIRv2 units, MtCH4.

An important note on emissions sources

Also see data/scripts/mimifairv2_python_data_explainer.ipynb

In this repository we include two sets of rcmip emissions data. The first set are entitled

  • data/python_replication/raw_data_from_python/rcmip_sspxx_emissions_1750_to_2500.csv (raw files from Python scripts)
  • data/rcmip_sspxx_emissions_1750_to_2500.csv (inputs to MimiFAIRv2)

which use FAIRv2.0 code to directly pull RCMIP outputs.

The second set rebase emission-driven forcings & species with natural emissions to 1750, including ["so2","nox","co","nmvoc","bc","nh3","oc","nox_avi","methyl_bromide","methyl_chloride","chcl3","ch2cl2"], as indicated in notebooks/RCMIP/RCMIP-experiments.ipynb from the FAIRv2.0.0-alpha from the replication code here: paper replication code here: https://github.com/njleach/leach-et-al-2021/tree/a4f8c73f3b45c8e82e70c5906f73e6ebad327954 and are labeled

  • data/python_replication/raw_data_from_python/rcmip_sspxx_emissions_1750_to_2500_rebased.csv (raw files from Python scripts)
  • data/rcmip_sspxx_emissions_1750_to_2500_rebased.csv (inputs to MimiFAIRv2)

To perfectly replicate the FAIRv2.0 paper code, rebased versions should be used. Thus the default of this repository will be to use the rebased files.