Documents · Verb reference

landscape_change_signature

Compare complete neighboring center landscapes along separate axes.

Callable type: Grammar verb / pipe stage · Browse: Synchrony and comparison

Usage

from cubedynamics import verbs as v
v.landscape_change_signature(*, metric='delta_s', deadband=0.02, near_tie_epsilon=0.005, min_overlap=25)

Arguments

Argument	Meaning	Default
metric	Pair field used to build each center landscape.	'delta_s'
deadband	Absolute interval around zero treated as neutral for sign comparisons.	0.02
near_tie_epsilon	Tolerance used to report near-tied values in rank diagnostics.	0.005
min_overlap	Minimum shared comparison locations required for a landscape pair.	25

Accepts

A sparse pair Dataset that retains enough neighboring center landscapes for the requested overlap threshold.

Returns

Separate east-west and north-south magnitude, rank, deadbanded sign, gradient, range, and valid-overlap diagnostics; it does not create one composite index.

Order / grammar behavior

This is a branch from pair relationships, not a reduction of stack heterogeneity. Keep both products distinct.

Minimal example

Run from the repository root after python -m pip install -e '.[vignettes]'. Uses the checked observational PRISM fixture; no network is required.

from pathlib import Path
import xarray as xr
import matplotlib.pyplot as plt
from cubedynamics import pipe, verbs as v

# Frozen, reviewed PRISM observations; run from the repository root.
path = Path("tests/fixtures/real_data/prism_boulder_january_2024.nc")
with xr.open_dataset(path, engine="scipy") as observed:
    cube = observed["tmax"].load()
assert cube.attrs["units"] == "degC"

with xr.open_dataset(path, engine="scipy") as observed:
    temperature = observed[["tmin", "tmax"]].isel(y=slice(0, 12), x=slice(0, 12)).load()
pairs = (pipe(temperature) | v.local_synchrony_pairs(lower_var="tmin", upper_var="tmax", max_radius_km=30, window_days=30, min_t=3)).unwrap()
result = (pipe(pairs) | v.landscape_change_signature(metric="delta_s", min_overlap=3)).unwrap()
result["normalized_rmse"].plot(col="orientation")
plt.show()

Works with