Quick start

Installation

Clone the repository :

git clone [email protected]:os-climate/hazard.git
cd hazard

Then use either pdm (recommended):

pip install pdm
pdm config venv.with_pip True
pdm install

Or virtualenv:

python -m venv .venv
source .venv/bin/activate
pip install -e .

Usage

In a virtual environment

A command line interface is exposed with the package. For example, this code snippet will run a cut-down version of a "days above temperature" indicator and write the output to $HOME/hazard_example :

source .venv/bin/activate
mkdir -p $HOME/hazard_example
os_climate_hazard days_tas_above_indicator --store $HOME/hazard_example

In a docker container

First, build the image.

docker build -t os-hazard-indicator -f dockerfiles/Dockerfile .

Then, you can run an example the following way. In the example, we save the data locally to /data/hazard-test-container in the container. To have access to the output once the container finished running, we are mounting /data from the container to $HOME/data locally.

docker run -it -v $HOME/data:/data os-hazard-indicator os_climate_hazard days_tas_above_indicator --store /data/hazard-test-container

In a CWL (Common Workflow Language) workflow

At the root of the repository you'll find hazard_workflow.cwl - This contains a definition to invoke the os_climate_hazard CLI within a CWL environment.

Structurally, it goes:

# Define the class (CommandLineTool)
# Hints (We define the docker image to use)
# Requirements (Networking access, resource access, and environment variables)
# Inputs (Map to CLI args and environment variables)
# Outputs (Where to mount and save results)
# Command (What to run)
# Arguments (What to pass directly to the CLI as args)

There is also hazard_workflow_input_example.yml which shows the structure of inputs expected.

Copy hazard_workflow_input_example.yml, rename to hazard_workflow_input.yml, and provide the relevant input values.

You can then invoke the CWL with:

$ cwltool hazard_workflow.cwl#produce-hazard-indicator hazard_workflow_input.yml

If successful, you'll find indicators generated in indicators/ in the repo root.

If you don't have cwltool installed you can find it here: cwltool installation

Arguments parsing in CLI

The CLI for this package is built on top of google's fire package. Arguments passed to the command line are parsed not based on their declared type in python but their string value at runtime. For complex types such as lists, this can lead to confusion if internal list elements have special characters like hyphens.

This is an example of command that works for the argument gcm_list (note the single and double quotes in that argument value)

os_climate_hazard degree_days_indicator --store $HOME/data/hazard-test --scenario_list [ssp126,ssp585] --central_year_list [2070,2080] --window_years 1 --gcm_list "['ACCESS-CM2','NorESM2-MM']"

And this is an example that does not :

os_climate_hazard degree_days_indicator --store $HOME/data/hazard-test --scenario_list [ssp126,ssp585] --central_year_list [2070,2080] --window_years 1 --gcm_list [ACCESS-CM2,NorESM2-MM]

Contributing

Patches may be contributed via pull requests from forks to https://github.com/os-climate/hazard.

All changes must pass the automated test suite, along with various static checks.

The easiest way to run these is via:

pdm run all

Hazard modelling

For more modelling-specific information, see HAZARD.md.