Vignettes

Run a complete analysis: Context → Question → Source → Pipe → Figure → Interpretation → Evidence. Each vignette treats the pipe as an executable scientific statement and makes its assumptions and boundaries visible. For a short introduction to the grammar, start with Learn.

Executable real-data notebooks

These eight notebooks use the same reviewed PRISM extract. Each tells a scientific story and includes working code, figures, Data used, Reproduce and See also sections. Read the shared structure.

01 · You have an array

Build a scientific cube

Add coordinates, units, and provenance; compare a map with a pixel history; then rotate the cube.

Begin with NumPy → 02 · You have observations

Make locations comparable

Reshape rows into a cube and use one clean verb to standardize every location through time.

Begin with pandas → 03 · You have several variables

Ask two questions of one Dataset

Select aligned variables, preserve their meanings, and compose a separate pipe for each question.

Begin with xarray → 04 · You want a readable method

Write analysis as a sentence

Compare direct and piped calls, combine built-in and ordinary functions, and see the minimal grammar.

Learn the core pipe → 05 · You want possibilities

Explore the verb gallery

Compare means, variance, anomalies, standardized values, project functions, and model-ready shapes.

Browse working verbs → 06 · You care about episodes

Follow cold from value to event

Turn measurements into states, states into events, and events into a spatial relationship.

Follow the event story → 07 · Your project has a method

Give the project its own verb

Encode a scientific rule as a small callable factory and keep project assumptions visible.

Build a custom verb → 08 · Your cube is larger

Scale the same analysis lazily

Keep the grammar unchanged while Dask delays computation until the final result is needed.

Follow the lazy workflow →

Keep the analytical sentence short

result = (
    pipe(cube)
    | v.anomaly(dim="time")
    | v.mean(dim=("y", "x"), keep_dim=False)
).unwrap()

Put source acquisition and preparation before the pipe. Explain the baseline and interpretation after it. Remember that the semantic trace covers the pipe, not that preparation or later work. See anomaly and mean for their canonical argument reference.

Explore a noun

Three different objects, the same grammar. Each lesson uses frozen real source data, explains the baseline and limitations, and produces three inline figures.

Terrain · USGS 3DEP

Read a landscape at its native scale

Inspect a real hillside, center on a local baseline, and reduce the map to a west–east profile.

Explore elevation → Networks · Overture and OSM

Compare mapped roads carefully

Keep native segments and classes, clip an explicit area, and measure length with two small project verbs.

Explore roads → Water · USGS stations

Keep the observations and their evidence

Inspect real discharge, subtract a one-day mean, and reuse the pipe at three identified stations.

Explore streamflow →

References: elevation · roads · streamflow.

Other analyses and educational material

Collection	What to expect
Working Lands	Executed notebook: observed hot-and-dry weather in South Dakota
South Dakota Decision Lab	One executable analysis; other questions are labeled dependency designs
Research workflows	Domain workflow directories and analysis reports
Fire VASE	Observed FIRED/gridMET examples and explicit renderer limitations
Synchrony	Methods and project vocabulary; not a second core grammar
Recipes	Task-oriented live-data code; provider access may be required
Examples and how-tos	Additional source and workflow guidance

Reproduce

From the repository root:

python -m pip install -e ".[vignettes]"
python scripts/run_vignettes.py

The runner executes clean copies of all twelve supported notebooks, verifies static plots, and leaves source notebooks unchanged. MkDocs renders the same code and figures on the site. Validation records the real-data checks; live-data recipes are not covered by the offline claim.