Replication code for Moore et al., "Mimi-Page".

This repository holds all code required to replicate the results of "Mimi-PAGE - An Open-Source Implementation of the PAGE09 Integrated Assessment Model".

Software requirements

To run the replication code, you need to install:

• version v0.6.4 of Julia
• and R

Preparing the software environment

On the Julia side of things you need to install a few packages. These consist of the packages required by Mimi, along with RCall, which is used to access R for graphing. You can use the following julia code to do so:

Pkg.add("Mimi")
Pkg.add("Distributions")
Pkg.add("DataFrames")
Pkg.add("CSVFiles")
Pkg.add("RCall")

Pkg.pin("Mimi", v"0.4.1")

On the R side of things you also need to install a number packages, which are all included in the tidyverse environment. From R code, run the following:

install.packages("tidyverse")

You also need to download the file Rallfun-v35.txt from here and place it in the src folder.

Cloning the repository

This git repository uses a git submodule for Mimi-PAGE. To ensure the submodule gets properly downloaded, make sure to use the git --recurse-submodules option when cloning the repository. That is, you can clone the repository from the command-line with:

git clone --recurse-submodules https://github.com/anthofflab/paper-2018-mimi-page-replication.jl.git

If you cloned the repository without that option, you can issue the following two git commands to make sure the submodule is present on your system: git submodule init, followed by git submodule update.

Running the replication script

To recreate all outputs for this paper, run the main.jl file in the folder src with this command:

julia src/main.jl

Input files

data/excel-performance.csv

The data in this file was generated by manually running the Excel version of PAGE with the @RISK plugin. We started by opening the original PAGE Excel file PAGE09 v1.7 default.xlsx in the 64 bit version of Microsoft Excel 2016 on Windows 10. We used @Risk version 7.5.1 Build 146. The different rows in the file correspond to runs where we modified two settings in the Simulation->Settings dialog: the Number of Iterations and the Multiple CPU Support setting. Each run was initiated by clicking the Start Simulation button. The elapsed time is the number @RISK reported in the @RISK Simulating window.

The runs were performed on a system with an Intel Core i7-4770 @ 3.4GHz CPU that had 16 GB of main memory.

data/PAGE09_mc_policya_100k.csv

This contains the raw Monte Carlo output from the Excel version of PAGE09 for each of four main outputs from the Excel PAGE '09 model: td (total damages), tac (total adaptation costs), tpc (total preventative costs), and te (total effect, combining these three).

src/mimi-page/output/mimipagemontecarlooutput.csv

To generate the Monte Carlo comparison graph, you will need to run the montecarlo.jl script in Mimi-PAGE. This will generate a file src/mimi-page/output/mimipagemontecarlooutput.csv, which includes the values for a number of output metrics for each Monte Carlo run.

Result tables

All result tables will be stored in CSV files in the folder results. The following files will be created by running the main script:

• deterministiccomparison.csv: Reports the difference between Mimi-PAGE results for a set metrics and the corresponding metrics in PAGE '09, computed by deterministiccomp.jl, which is included in main.jl.

• performance.csv: This combines the Excel performance metrics with Mimi-PAGE performance metrics generated when the mccomp.jl script is run, which is included in main.jl.

Result plots

All results plots are saved in the folder results in various formats.