Quick Start

The package provides five command-line entrypoints. From a cloned repository with the project dependencies installed (see the installation instructions in the project README), run each entrypoint with Poetry:

poetry run python -m ice_shelf_state_and_fate.timeseries_module_entrypoint --help

Replace --help with the arguments shown below to run a command. If the package is installed in your active Python environment, you can omit poetry run. Each entrypoint also accepts --help to list its arguments and defaults. Input Zarr stores may be local paths or http://, https://, or s3:// URLs. Output paths are local directories.

The commands can be combined into a processing workflow:

Extract a timeseries from one or more source datasets.
Optionally preprocess multiple timeseries together.
Analyse covariates in the preprocessed store.

Dataset summaries and timeseries-to-JSON conversion are separate operations and can be run on their own.

Timeseries extraction

Run timeseries_module_entrypoint to extract a measurement from each input dataset and save a timeseries Zarr store in a standardised format. Repeat --dataset_paths for each input dataset:

poetry run python -m ice_shelf_state_and_fate.timeseries_module_entrypoint \
   --label "Ice velocity" \
   --dataset_paths data/velocity_2020.zarr \
   --dataset_paths data/velocity_2021.zarr \
   --output_path output/velocity.zarr \
   --measurement_name velocity

--measurement_name names the measurement variable in the input datasets. Optionally provide --measurement_uncertainty_name if the input also contains an uncertainty variable, and --mask_json_path to supply a GeoJSON mask. The default timeseries type is cumulative (use derivative for a derivative); aggregation defaults to mean by month. The available aggregation strategies are mean, median, max, min, std, IQR, and None; aggregation periods are year, quarter, month, and day.

Preprocessing timeseries

Run timeseries_preprocessor_module_entrypoint to combine one or more timeseries Zarr stores into the format expected by covariate analysis. Repeat --input_paths for each store:

poetry run python -m ice_shelf_state_and_fate.timeseries_preprocessor_module_entrypoint \
   --input_paths output/velocity.zarr \
   --input_paths output/temperature.zarr \
   --output_path output/preprocessed_timeseries.zarr \
   --time_period year \
   --aggregation mean

If --time_period is omitted, the lowest frequency in the input timeseries is used. The aggregation defaults to mean and accepts mean, median, max, min, and std.

Covariate analysis

Run covariate_analyser_module_entrypoint on the preprocessed store to write analysis results as JSON:

poetry run python -m ice_shelf_state_and_fate.covariate_analyser_module_entrypoint \
   --input_path output/preprocessed_timeseries.zarr \
   --output_path output/covariate_analysis.json \
   --analysis_type corr

The analysis type defaults to corr. Other supported values are cross_corr and causal.

Dataset summary

Run dataset_summary_module_entrypoint to create a JSON summary of a GeoZarr dataset. --max_res controls the maximum summary resolution and defaults to 20:

poetry run python -m ice_shelf_state_and_fate.dataset_summary_module_entrypoint \
   --input_path data/dataset.zarr \
   --output_path output/dataset_summary.json \
   --max_res 20

Convert a timeseries Zarr store to JSON

Run timeseries_zarr_to_json_module_entrypoint on a store whose variable names match the timeseries_summary metadata in src/ice_shelf_state_and_fate/timeseries_zarr_to_json_module/variable_mappings_local.yml (for example, mass_balance or discharge). The generic extraction store produced above is not compatible with this converter.

poetry run python -m ice_shelf_state_and_fate.timeseries_zarr_to_json_module_entrypoint \
   --input_path data/ice_shelf_timeseries.zarr \
   --output_path output/ice_shelf_timeseries.json