gridstats

gridstats is an Oceanum library for computing gridded statistics over large oceanographic and climate datasets. Pipelines are defined in YAML and run as a single CLI command. Computation is lazy and out-of-core via xarray and dask, so datasets of arbitrary size are handled without loading them into memory.

Features

• YAML-driven pipelines — source, operations, and output are all declared in one config file
• Out-of-core — processes arbitrarily large grids lazily; spatial tiles: keeps peak memory bounded
• Rich stat library — aggregations, quantiles, exceedance, return period values, directional stats, distributions, and more
• Multiple output formats — NetCDF or Zarr
• CF-compliant — output variables are automatically annotated with standard names, units, and long names
• Extensible — register custom stat functions and loaders via decorator or package entry point

Installation

Requires Python ≥ 3.10.

pip install gridstats

For intake catalog support and cloud storage (GCS):

pip install "gridstats[extra]"

Quick start

1. Write a config file

# stats.yml
source:
  type: xarray
  urlpath: /data/hindcast/waves.zarr
  engine: zarr
  sel:
    time: {start: "2000-01-01", stop: "2020-12-31"}
    latitude: {start: -50, stop: -30}
    longitude: {start: 160, stop: 180}

output:
  outfile: ./wave_stats.zarr

calls:
  - func: mean
    dim: time
    data_vars: [hs, tp]

  - func: quantile
    dim: time
    data_vars: [hs]
    q: [0.5, 0.90, 0.95, 0.99]

  - func: rpv
    dim: time
    data_vars: [hs]
    return_periods: [10, 50, 100]
    distribution: gumbel_r

2. Run it

gridstats run stats.yml

The output dataset will contain variables like hs_mean, tp_mean, hs_quantile, and hs_rpv, each with CF-standard attributes.

3. Use the result

import xarray as xr

ds = xr.open_zarr("wave_stats.zarr")
print(ds)

Available stat functions

All calls accept a group: key (month, season, year) to compute statistics per calendar period.

Grouping and spatial tiling

calls:
  # Monthly mean
  - func: mean
    dim: time
    data_vars: [hs]
    group: month

  # Quantile with spatial tiling to control memory
  - func: quantile
    dim: time
    data_vars: [hs]
    q: [0.95, 0.99]
    chunks: {time: -1, latitude: 50, longitude: 50}
    tiles: {latitude: 10, longitude: 10}

Plugin system

Register a custom stat function in your own package:

from gridstats.registry import register_stat
import xarray as xr

@register_stat("my_stat")
def my_stat(data: xr.Dataset, *, dim: str = "time", **kwargs) -> xr.Dataset:
    ...

Or declare it as a package entry point so it is discovered automatically:

[project.entry-points."gridstats.stats"]
my_stat = "my_package.stats:my_stat"

CLI

Usage: gridstats [OPTIONS] COMMAND [ARGS]...

Commands:
  run         Run a stats pipeline from a YAML configuration file.
  list-stats  List all registered stat functions.

License

MIT — see LICENSE.