The feedback model is an idealized Bayesian model that uses Turing.jl (version 0.43) for MCMC and largely follows Turing's Bayesian differential equations tutorial: https://turinglang.org/docs/tutorials/bayesian-differential-equations/. An in-depth description of all aspects of the model (including parameters and the formulation of the random walks) can be found in the methods section and the appendix of the corresponding paper.
Model Files:
The file named "Base Case" represents the standard formulation of the model
The file named "Increased Uncertainties" represents the model formulation where CenCO2PIP uncertainties are doubled
The file named "Constant Weathering Sensitivity" represents the model formulation where wT is constant
The file named "Other Forcing Random Walk" represents the model formulation where O is allowed to vary as a random walk
The file named "Increased Degassing Bounds" represents the model formulation where degassing bounds are increased by 10%
Plots File:
The model exports posterior parameters (for each time step). To create each plot, the random walks must be recreated from the posteriors, which is done in the file named "Final Plots".
Data Files:
The Data required for the model and plotting is as follows:
CenCO2PIP data converted to doublings: "cenco2pip_500kyr_testrev_ppm_12_copy"
Base Case: fullchain_final_TU1_1500_copy
Increased Uncertainty: fullchain_final_loose_1500_copy
Constant Weathering Sensitivity: fullchain_final_constant_w_1500_copy
Other Forcing Random Walk: fullchain_final_O_rw_1500_copy
Increased Degassing Bounds: fullchain_final_inc_G_1500_copy
Other Notes: The model uses CO2System.jl to create the DIC scaling factor. See: https://github.com/mvdh7/CO2System.jl for instructions on installing