Skip to content
Canopulse

Research

How the index is built and where it is uncertain

Canopulse documents its sources, processing steps, and limitations openly. Each part can be inspected and updated as new data is released.

Input data layers

Official and established public datasets.

Landsat land-surface temperature is native 100 m resolution and was resampled to 30 m for spatial alignment. Resampling does not increase the native information resolution.

Surface heat is the daytime temperature of roofs, pavement and ground measured by Landsat satellites on summer mid-mornings (about 10 to 10:30 a.m. local time). It runs much hotter than air temperature and does not measure air temperature or nighttime heat.

Tree canopy comes from a 30 m satellite model (NLCD Tree Canopy Cover 2025). It misses many street trees and small canopy in dense urban areas, so values are usually lower than a city's own 1 m canopy assessment. Use it to compare tracts, not as a local canopy inventory.

Components, layers and records are different counts. Five weighted components come from five component datasets. Adding the census tract boundary layer, which every component is carried on and none is weighted by, gives six published data layers. Those layers draw on six official sources, and the source registry lists exactly those six.

  • Tree canopy

    NLCD 2025 Tree Canopy Cover (TCC v2025-6)

    202530 m raster
  • Surface heat

    Landsat 8/9 C2L2 ST_B10 summer composite

    2026100 m native, delivered at 30 m
  • Impervious

    Annual NLCD 2025 Impervious Surface

    202530 m raster
  • Social vulnerability

    CDC/ATSDR SVI 2022 (RPL_THEMES)

    2022Census tract
  • Pollution burden

    CalEnviroScreen 5.0 Pollution Burden

    2026Census tract

Confidence framework

Honest about uncertainty.

Each tract receives a confidence label, a data-quality label computed as the weakest-link minimum across data vintage, native spatial resolution, in-tract coverage, and QA status. The rubric defines four labels. This release uses two of them: every fully scored tract is Moderate, so the label does not separate one scored tract from another.

High0 tracts
Not attainable in this release
Moderate9,041 tracts
Every fully scored tract
Low0 tracts
No tract meets this rule in this release
Insufficient88 tracts
Provisional and unscored records

Why no tract reaches High. High requires every contributing layer at a native resolution of 30 m or finer with a vintage within two years. Landsat land-surface temperature is natively 100 m and was resampled to 30 m for alignment, which does not increase its information resolution, and both CDC/ATSDR SVI and CalEnviroScreen Pollution Burden are tract-level composite indices. Each of those three triggers the Moderate rule, so no tract in this release reaches High.

A thermal source at 30 m native resolution, and component inputs that are measured rather than composite, would make High reachable. Both are data availability questions rather than changes to the index.

Robustness

Weight sensitivity checks

  • Formula recomputation: reproduce.py recomputes every released score, class and rank from the published components
  • Equal weights (0.2 each) compared with the published weights
  • Each weight raised or lowered by 20%, others rescaled (10 scenarios)
  • Leave one component out, remaining weights rescaled (5 scenarios)
  • 1,000 random weightings around the published weights (weights only; input data not varied)
  • Spearman correlation matrix of the five components (9,041 fully scored tracts)

Run by Canopulse. Not an independent review. These checks test sensitivity to the weights only. Alternative normalizations, regional percentiles and input uncertainty were not tested.

Read the robustness report | Reproducibility package

Research Foundation

A research-driven approach to urban climate screening.

Canopulse is a documented geospatial screening index for finding California tracts where low canopy, surface heat, paving, social vulnerability and pollution burden overlap.

Urban canopy risk

Synthesizing tree-canopy gap, heat, and equity into one prioritization signal.

Surface heat and vulnerability

Where summer surface heat and social vulnerability overlap. Surface heat is satellite land-surface temperature, not air temperature.

Pollution burden

CalEnviroScreen 5.0 Pollution Burden percentile, used as published, is the EBS component.

Tract-level analysis

Designed for the 11-digit US census tract to support subsequent local investigation.

Open methodology

The formula, source registry, model registry, scoring and robustness scripts, and release checksums are published for review. The raster processing pipeline is not published.

Transparent scoring

Every tract carries a data-quality label, the weighting checks are published, and every input is traced to its source.

Summary

Each UCRI score can be traced to its five public inputs, fixed weights and frozen release.

Development Timeline

From concept to a statewide screening release.

The current release and earlier releases are kept separate. Each release retains its own source records, assumptions, and reproducibility materials.

  1. 01

    Research concept

    Formalize the screening question across canopy deficit, surface heat, impervious surface, social vulnerability, and pollution burden.

  2. 02

    LA historical baseline

    Freeze the Los Angeles County release (May 2026) for historical reproducibility. Its definition is recorded in the historical model registry linked from the Technical Archive, and it is not the current model.

  3. 03

    California data foundation

    Extend the registered tract foundation to all 58 counties and 9,129 California census tracts while preserving explicit missing-data states.

  4. 04

    July 2026 release

    Release California UCRI with CalEnviroScreen 5.0 Pollution Burden as EBS. Its record counts, sources and signed verification bundle are preserved in the Technical Archive.

  5. 05

    Current October 2026 release

    Refresh tree canopy (NLCD 2025), impervious surface (Annual NLCD 2025) and surface heat (multi-pass summer 2026 Landsat): 9,041 fully scored, 66 provisional, and 22 water-only tracts not scored.

    Current stage
  6. 06

    Next: invite external methodology review

    Invite outside methodology review and support outcome-validation research. No independent review has been commissioned or completed, and Canopulse's own sensitivity checks are not presented as validation.