Environment methodology 環境方法論
The environment check reports four measurable dimensions of a site. This page documents exactly where each number comes from, how accurate it is, and where it stops being meaningful. One design rule governs all of it: facts are facts, interpretations are labelled interpretations — and there is no single opaque "environment score".
The four dimensions, at a glance
| Dimension | Source | Resolution | Known limits |
|---|---|---|---|
| Wind — monthly speed & prevailing direction | NASA POWER (MERRA-2 reanalysis), monthly climatology WS2M/WD2M | ~0.5° grid (~50 km) | Vector-averaged monthly means, not a 16-sector wind rose; grid, not an anemometer at your door |
| Rain — monthly daily mean | NASA POWER PRECTOTCORR (corrected gauge) | ~0.5° grid | Wet-day frequency needs daily statistics (not in the free monthly layer) |
| Sun — noon elevation by month, day length, azimuth | Our own kernel (VSOP87D + nutation + ΔT), computed in your browser | Exact for time & place | Device clock matters: the sun moves 0.25°/minute (see the solar compass for the uncertainty budget) |
| Terrain & water — skyline, white-tiger check, nearest mapped water | SRTM 30 m (DTM) via opentopodata; OpenStreetMap waterways | 30 m terrain; OSM varies | DTM has no buildings and flattens small urban relief (demonstrated on the Forbidden City case); OSM completeness varies, culverted streams missing |
Verification, not vibes
- Sun: the same kernel is frozen-tested against the NOAA GML Solar Calculator on 58 vectors (max difference 0.015° azimuth / 0.004° elevation) plus sunrise/sunset agreement over 20 cities × 12 months — re-run in CI on every change.
- Wind & rain: every parse is pinned to a frozen 6-city × 12-month golden set from the POWER API itself (144 checks), plus climate sanity anchors (Hong Kong's January NE monsoon and annual easterlies; Beijing drier than Hong Kong drier than Reykjavík).
- Terrain & water: sampling math unit-tested with synthetic hills; every landmark case page cross-checks its printed verdict against a live kernel recomputation.
The academic lineage (and its honest thinness)
Quantified, GIS-based feng-shui analysis is a real but thin research tradition — isolated papers across ~25 years, which largely do not cite each other. Representative entries we build on:
- Um (2009, IJGIS) — terrain-derived feng-shui parameters with spatial prioritisation; the strongest precedents for grave-site quantification. doi:10.1080/13658810802055954
- Park (2008) — a quantitative model of feng-shui landscape structure (slope, aspect, fractal dimension, "water-acquisition factor") from 1:5000 maps.
- Whang & Lee (2006, Landscape & Ecological Engineering) — Korean bi-bo woodlands and ponds reframed as landscape-ecology planning tools. doi:10.1007/s11355-006-0014-8
- He et al. (2026, Humanities & Social Sciences Communications) — a geocomputational model that treats feng-shui site criteria as computable terrain variables for Ming–Qing imperial tombs. doi:10.1057/s41599-026-07917-2
- Fang, Pan & Wu (2022, arXiv) — a sensor-based "feng-shui compass" measuring ambient room environments; early-stage, and a useful reminder of where measurement must not slide into health claims. arXiv:2210.13672
What this literature does not contain: a validated single-number "environment score", any peer-reviewed automatic dragon-vein detection, or evidence that any of these measures predict health, wealth or outcomes. We will not pretend otherwise.
What we refuse (and why)
- No single environment score. Wind, rain, sun and terrain are different physical quantities; averaging them into one number would be numerology. Each dimension is reported with its own units, source and threshold.
- No school judgments smuggled into data. 藏风聚气 (gathering wind, holding qi) or 白虎抬头 (white tiger raising its head) appear only as labelled interpretations with rule ids and calibration grades — never as if they were measurements. See the rule provenance register.
- No health or fortune claims. Environment data ≠ "living here is good for you". Cultural and educational use, printed on every page.
- No direction from photos. Facing direction comes from a compass or the solar compass — never from an image.
Changelog & known gaps
- Wind frequency roses (16 sectors, hourly statistics) — planned; needs a heavier data layer than the free monthly climatology.
- River curvature judgments (jade-belt / reverse-bow) — deliberately not computed until curvature analysis meets a measured sitting direction; today we report the nearest mapped water as a fact.
- DSM (buildings included) for urban canyon effects — out of scope for the free tier; the DTM limitation is printed wherever terrain appears.