Data models and scripting languages for field-agent based modelling

This repository holds a Jupyter notebook demonstrating the Daisyworld model implementation in Campo, a YAML file to create the Python environment required to run the model, and necessary scripts for pre- and postprocessing.

You can run the Jupyter notebook locally on your own computer.

More information will be given in the GIScience 2023 workshop.

Running on your machine

How to install

A few steps are required to run the Jupyter notebook. General information on Jupyter notebooks and manuals can be found here. The user guide and short reference on Conda can be found here.

1. You will need a working Python environment, we recommend to install Miniconda. Follow their instructions given at:

   https://docs.conda.io/en/latest/miniconda.html

2. Open a terminal (Linux/macOS) or Miniconda command prompt (Windows) and browse to a location where you want to store the course contents.

3. Clone this repository, or download and uncompress the zip file. Afterwards change to the giscience2023 folder.

4. Create the required Python environment:

   Linux/macOS:

   conda env create -f environment/environment.yaml

   Windows:

   conda env create -f environment\environment.yaml

The environment file will create a environment named giscience2023 using Python 3.11. In case you prefer a different name or Python version you need to edit the environment file.

How to run

Activate the environment in the command prompt:

conda activate giscience2023

Then change to the notebook folder. You can now start the Jupyter notebook from the command prompt. The notebook will open in your browser:

jupyter lab daisyworld.ipynb

Running the LUE model

You can download the model script and input data here. The model script can be executed e.g. with

python ./model.py --hpx:threads=4 --write "50,50" input_path output_path

For output_path create an output folder first, e.g. output In case of crashes on Linux systems you can run like

LD_PRELOAD=$(find $CONDA_PREFIX -name libtcmalloc_minimal.so.4) python ./model.py --hpx:threads=4 --write "50,50" input_path output_path

Output visualisation

You can display the output as raster timeseries with aguila. First, change the lines 224-225 in model.py into

            self.write_raster(discharge, f"discharge_{timestep}.tif")
            self.write_raster(self.snow, f"snow_{timestep}.tif")

re-run the model and display the output with

aguila --timesteps=[1,365] output/discharge

Further reading

Background on DaisyWorld:

https://en.wikipedia.org/wiki/Daisyworld

Scientific literature about Campo and LUE:

M.P. de Bakker, K. de Jong, O. Schmitz, D. Karssenberg (2017). Design and demonstration of a data model to integrate agent-based and field-based modelling. Environmental Modelling & Software, 89, 172-189, DOI: 10.1016/j.envsoft.2016.11.016.

K. de Jong, D. Karssenberg (2019). A physical data model for spatio-temporal objects. Environmental Modelling & Software, 122, 104553, DOI: 10.1016/j.envsoft.2019.104553.

K. de Jong, D. Panja, M. van Kreveld, D. Karssenberg (2021). An environmental modelling framework based on asynchronous many-tasks: Scalability and usability. Environmental Modelling & Software, 139, 104998, DOI: 10.1016/j.envsoft.2021.104998.

https://campo.computationalgeography.org/