DICE-like models in Python

This is a little application put together by Kyle Duffy to run simulations for integrated assessment models (IAMs) derived from DICE (Nordhaus et al. 2016). The current version suffers some optimization issues due to high dimensionality and weaknesses of scipy.optimize, but simulation is currently functional and the code supports SICE models (Ikefuji et al. 2019) and a SICE variant where temperature dynamics from Golosav et al. are substituted in place of those from Ikefuji et al. Really, any number of additional variants could easily be added by the determined user.

Setup and usage instructions

At the moment this project depends on numpy, pandas, and scipy, but I've put together a little script so you don't have to configure everything yourself. To use the code, first make sure you have Python 3 installed. Any version of Python 3 should work, but I recommend using the most recent version, which is 3.9.4 at the time of writing. (Or better yet, use pyenv to manage your versions of Python).

Once Python 3 is installed, you can run

python configure_python.py

in your command line environment (e.g. Terminal on mac). Note that if you're on Mac, Python 2 comes preinstalled out of the box. To make sure you're using Python 3, you can run

python3 configure_python.py

instead.

If you're on Windows, open PowerShell as an administrator (it's important you're in adminstrator mode) and run

python.exe configure_python.py

If your configuration was successful, you'll get a confirmation and it will tell you how to activate the newly created virtual environment. Once you've done that (you should see the name of the virtual environment appear in parentheses before your prompt, e.g. (DICE)), you can interact with the code using the solver.py module. To get a help message, run

python solver.py -h

on Mac and Linux or

python.exe solver.py -h

on Windows.

Follow the usage instructions and run as many simulations as you want! To change the parameters, you can edit the default values in specs.py. To get the original version, just re-download the original file form this repository. (This is pretty clunky, but I don't currently have the time to wrap this up with a better UX)

Next steps

This is still a work in progress, and there are a number of things that can be done to improve the codebase. Some places to start are:

• replace scipy.optimize with a symbolic solver like CasADi or some implementation of IPOPT that finds a better optimum
• update the interface for changing parameters to be more user-friendly
• add a receding horizon to simulate an infinite horizon