Verified sandbox evidence · Singapore · Tokyo 23 wards · Vienna

Built-form indicator verification

Inspect the approved achieved GPR calculation and the recalculated building-footprint compactness indicator. This surface visualises existing evidence and keeps the production-readiness gaps visible.

Validated evidence, not a production promotion
GPR is available at H3 R10. The approved direct district GPR output is not yet materialised. Compactness has passed the sandbox recalculation checks, but its runtime binding and acquisition graph remain pending.
GPR building-contribution cells
Median achieved GPR, H3 R10
Median district Moran's I, R
Low-sample districts, R of

H3 evidence and spatial quality

Both maps display native R10 calculations. GPR cells are present only where the source recorded a building contribution. Footprint coverage uses continuous non-water support and retains valid zero-coverage cells. Permanent water and invalid support are omitted from the coverage map, not converted to zero.

Achieved density

GPR at H3 R10

GFA divided by H3 cell area. Empty map space is not evidence of zero GPR.

Morphology input

Building-footprint coverage at H3 R10

Footprint area divided by non-water H3 support. Valid zero values remain visible.

Structure-filter audit still needed. The approved contract retains habitable buildings and excludes non-habitable structures. The current GPR summary does not expose included and excluded counts by source class, so this page does not imply that those counts have been visually verified.

GPR comparison across cities

The curves compare achieved GPR among H3 cells with building contribution. The pooled top 1% is clipped to preserve detail in the main distribution. City-wide GPR is calculated from total attributable GFA over the city boundary area and is not the mean of H3 ratios.

Primary floor-height assumption: m. Sensitivity assumption: m. Singapore height-censored GPR is a lower bound. Tokyo anomaly treatment remains an open decision. The source site_cov field is a summed-footprint diagnostic and can exceed 1 where source bodies overlap.

District morphology through Moran's I

Moran's I measures whether districts contain spatially similar clusters of footprint coverage. A high positive value means similar coverage values are spatially clustered. It does not, by itself, distinguish compact development from clustered low-density sprawl, so footprint intensity and the map pattern must be interpreted alongside it.

Interpretation Negative: neighbouring coverage differs Near zero: weak spatial pattern Positive: similar coverage clusters White boundaries: fewer than 30 Moran cells

District geometry is simplified by m for browser rendering only. Moran's I is read from the validated outputs calculated with the full upstream geometry.

Resolution sensitivity

R10 remains the primary grid. R11 is recalculated natively as a sensitivity check, then compared at the district level. Points on the diagonal are stable between resolutions. Tokyo departs most strongly, with many R10 districts below the 30-cell diagnostic threshold.

resolution check. Spearman rho ; median absolute Moran change ; R10 low-sample districts ; R11 low-sample districts .

Robustness and contract status

Singapore's 0.41 versus 0.34 support threshold is highly rank-stable at both resolutions. Full-cell versus city-clipped support is also rank-stable overall, although edge districts can have larger local changes. These checks support retaining full-cell R10 as the primary specification while preserving the sensitivity evidence.

Dynamic World compactness remains archived as a valid but different land-cover construct. This page does not change the capacity registry, outcome runtime definitions, city configurations, source readiness, or evaluation eligibility.