WFM4Cities

WFM4Cities is a research-oriented Python package for constructing DEM-inspired urban scalar surfaces and analyzing them with Progressive Threshold Scanning (PTS). It supports a transparent point-to-surface workflow, a headless API and command-line interface, and an optional bilingual desktop interface.

简体中文说明

What the package does

• checks and visualizes point patterns;
• interpolates scattered observations with scipy.interpolate.griddata(method="cubic");
• fills cells outside the cubic interpolation hull with nearest-neighbor values;
• min--max normalizes an analytical surface to ([0,1]);
• scans an ordered threshold sequence;
• calculates projected area (A(h)), normalized excess volume (V(h)), connected-component count (N(h)), and largest-component ratio (L(h));
• exports regular grids, full response trajectories, fixed summary features, and overview figures;
• provides optional GeoPandas/PySAL spatial diagnostics.

WFM4Cities treats the vertical coordinate as the value of a specified urban attribute (for example, density, activity intensity, or building volume). It does not infer the meaning of that variable.

Scientific scope

The water-flooding metaphor is an explanatory analogy for moving a horizontal threshold through a scalar surface. WFM4Cities is not a hydrodynamic flood model, a general DEM-production system, or an operational planning platform. Threshold order is not time. Hydrological applications require additional terrain, rainfall, drainage, flow-path, water-depth, and exposure information.

Installation

Python 3.10--3.13 is supported. Python 3.11 is the recommended reference environment.

Install from the repository

git clone https://github.com/zhangzuocughustccnu/WFM4Cities.git
cd WFM4Cities
python -m venv .venv

Activate the environment:

# Windows PowerShell
.venv\Scripts\Activate.ps1

# macOS/Linux
source .venv/bin/activate

Then install:

python -m pip install --upgrade pip
python -m pip install -e .

Install the released package

python -m pip install PTS-WFM4Cities

The PyPI distribution is named PTS-WFM4Cities. The historical 2.6 release is still published under the shorter name WFM4Cities, which belongs to a different PyPI account, so pip install WFM4Cities installs that older release. Always install PTS-WFM4Cities to get version 3.0.

Optional spatial diagnostics

python -m pip install "PTS-WFM4Cities[spatial]"

Conda environment

conda env create -f environment.yml
conda activate wfm4cities-3.0

Tkinter is included with standard Python on Windows and macOS. On some Linux distributions it is a separate operating-system package (for example, sudo apt install python3-tk on Debian/Ubuntu).

Quick start

Desktop interface

wfm4cities gui

Running wfm4cities with no subcommand also opens the interface. The historical WFM4Cities command remains available after installation; the python WFM4Cities.py launcher also remains available in a source checkout.

Deterministic headless demonstration

wfm4cities demo --output-dir demo_results --seed 42

This writes:

• source_points.csv;
• surface_normalized.csv;
• pts_response.csv;
• pts_features.json;
• wfm4cities_overview.png.

Analyze an existing grid

The input must be a headerless numeric CSV matrix:

wfm4cities scan my_surface.csv --output-dir scan_results

Use strict thresholding to exclude zero background cells at (h=0):

wfm4cities scan my_surface.csv --rule strict --output-dir scan_results

Construct a surface from points

The CSV must contain x and y columns. An optional column named value, density, or z supplies the vertical value. If it is absent, WFM4Cities calculates a local neighbor count.

wfm4cities points observations.csv --grid-size 100 --method cubic

Python API

from wfm4cities import (
    SimulationConfig,
    build_surface,
    extract_features,
    progressive_threshold_scan,
)

config = SimulationConfig(
    num_points=150,
    noise_ratio=0.30,
    num_clusters=4,
    grid_size=100,
    seed=2026,
)
model = build_surface(config)
result = progressive_threshold_scan(
    model.grid_z,
    threshold_count=51,
    threshold_min=0.0,
    threshold_max=0.95,
    rule="inclusive",
    connectivity=8,
)
features = extract_features(result)
print(features)

The main response arrays are available as result.A, result.V, result.N, and result.L.

Canonical structures (M1--M4)

The four canonical structures of the accompanying study are available as one-click presets. One cluster means monocentric and several clusters mean polycentric; a small cluster standard deviation means concentrated and a large one means dispersed.

from wfm4cities import MODEL_LABELS, build_model_surface

surface = build_model_surface("M3")      # or "M1", "M2", "M4"
print(MODEL_LABELS["M3"])                # M3 monocentric-dispersed

The presets are the stochastic counterparts of the deterministic benchmark surfaces: they generate point patterns with the characteristic signature of each structure, not the archived benchmark grids themselves. In the desktop interface they are the four buttons of the Canonical structures panel.

Importing your own observations

Any table of two-dimensional coordinates plus one attribute can be imported. The external margin is removed, the extent is affinely mapped onto the standard square domain (0--100 by default) and the attribute is min--max normalised to 0--1, mirroring the night-time-light preparation of the accompanying study. The aspect ratio is not preserved unless preserve_aspect=True, and the applied scales are always reported.

import numpy as np
from wfm4cities import standardize_points

points = np.array([[0, 0], [4, 0], [4, 3], [0, 3]])
values = np.array([1.0, 2.0, 3.0, 4.0])
mapped, normalised, report = standardize_points(points, values)
print(report.scale_x, report.scale_y)

In the desktop interface choose Import points and pick the x, y and attribute columns; the status bar reports the mapping that was applied.

Interface language

The desktop interface starts in English and the Language button switches it to Chinese; every label, panel title and status message follows. The strings live in wfm4cities.i18n, so a further language is a table away.

Fonts and colour scheme

wfm4cities.fonts selects Times New Roman for Latin text and SimSun/STSong for Chinese, using the copies already installed on the system, and registers extra font files from the folder named by PTS_FONT_DIR. No proprietary font file is redistributed with the package.

Every colour lives in wfm4cities.palette, so the interface, the command line and the exported figures always agree.

from wfm4cities.palette import SURFACE_RAMP, surface_cmap

Threshold conventions

Two explicit rules are available:

• inclusive: retain cells satisfying (z \ge h). This is the default and matches the controlled archetype analysis.
• strict: retain cells satisfying (z > h). This excludes zero-valued background cells when (h=0).

The denominator of (A(h)) is the fixed analysis domain, not the number of currently positive cells. Components use eight-neighbor connectivity by default; four-neighbor connectivity is also available. (V(h)) is normalized by total intensity at (h=0).

See Method conventions for the exact definitions.

Project structure

WFM4Cities/
├── src/wfm4cities/       # maintained implementation
├── src/WFM4Cities/       # compatibility namespace for 2.x users
├── tests/                # automated regression tests
├── examples/             # executable examples and sample data
├── docs/                 # method, API, architecture, and release notes
├── legacy/               # unmodified historical single-file script
├── pyproject.toml        # package metadata and dependencies
├── environment.yml       # recommended Conda environment
├── CITATION.cff          # machine-readable software citation
└── LICENSE               # MIT license

The maintained package does not create Tk windows or force a Matplotlib backend when imported. GUI code is loaded only when requested.

Reproducibility and testing

Create a development environment and run:

python -m pip install -e ".[dev]"
ruff check .
pytest
python -m build
python -m twine check dist/*

requirements-tested.txt records the exact numerical library versions used for the 3.0.0 release-candidate verification. pyproject.toml remains the authoritative dependency specification.

Method demonstration

The physical immersion demonstration used to explain the threshold analogy is available at YouTube. It is explanatory only and does not define the numerical algorithm.

Authors

WFM4Cities was developed by Zuo Zhang, Yukun Jiang, and Zhe Wang. See AUTHORS.md for details.

Citation

Use GitHub's “Cite this repository” control or the metadata in CITATION.cff. Add the article DOI to the citation metadata after the associated paper is published; no DOI is invented in this repository.

Contributing and support

Please read CONTRIBUTING.md before submitting a pull request. Reproducible bug reports can be filed through the repository's issue templates. Security-sensitive reports should follow SECURITY.md.

License

WFM4Cities is distributed under the MIT License. The historical script is retained for provenance and is covered by the same repository license.