Method and Validation Center
Release integrity, separate from outcome validation.
California UCRI is generated through deterministic source, join, formula, ranking, and output checks. Those controls support reproducibility. They do not establish field validity, causal impact, eligibility, or project outcomes.
9,129
California tract records
9,041
Fully scored
66
Provisional
22
Not scored
Current release
California UCRI release integrity
Current public release: October 2026. Pollution burden source: CalEnviroScreen 5.0 Pollution Burden percentile, released July 1, 2026.
CDS
Canopy deficit
25%
HES
Surface heat
25%
ISS
Impervious surface
15%
SVS
Social vulnerability
20%
EBS
Pollution burden
15%
Source reconciliation
Missing records remain visible
The official CES 5.0 table contains 9,106 unique tract records. California geometry contains 9,129 records. Twenty-three special or offshore-style tract GEOIDs have no CES 5.0 row. Canopulse leaves those EBS values missing and applies the documented provisional or unscored rules.
Control register
Exact checks and current status
Complete means the automated release check passed. It does not mean an outside reviewer validated the method or its outcomes.
Official source checksum
Complete
The CalEnviroScreen 5.0 CSV is pinned to SHA-256 3abb90971997cc151afe8c0ba46a01f544d1f8aaf4bc8b5d1f72c6d46007a296.
Source-row uniqueness
Complete
9,106 official tract rows were parsed with unique tract identifiers and non-missing PollutionP values.
Tract reconciliation
Complete
9,106 California tract GEOIDs matched. The 23 records absent from CES 5.0 remain missing; no values were invented.
Formula reproduction
Complete
The fixed 25%, 25%, 15%, 20%, and 15% component weights reproduce released scores within rounding tolerance.
Score-kind accounting
Complete
9,041 fully scored, 66 provisional, and 22 unscored records reconcile to 9,129.
Ranks and priority bands
Complete
State and county ranks are assigned only to fully scored records. Provisional and unscored records remain outside released ranks.
Geometry and tiles
Complete
Official 2020 tract geometry is delivered through scale-dependent PMTiles and does not depend on score recomputation.
Independent methodology review
Not done
Outside reviewers are invited to check the method. None has been commissioned or completed, and none is claimed.
Outcome and field validation
Not complete
Canopulse does not represent screening scores as validated project outcomes, causal effects, or field conditions.
Weighting sensitivity
Internal robustness checks
Canopulse recomputed the ranking of the 9,041 fully scored tracts under 1,016 alternative weightings to see whether the same tracts stay on top. These checks test sensitivity to the chosen weights. They do not test whether UCRI predicts heat illness or any other outcome.
Run by Canopulse. Not an independent review.
91.3%
Top-class tracts kept with equal weights
Of the 1,814 tracts in the top class (Very high screening priority), this share stays in the statewide top 20% when all five components get the same 20% weight.
97.2%
Kept after any single weight changes by 20%
Raising or lowering any one weight by a fifth, and rescaling the others, keeps at least this share of top-class tracts in the top 20%.
95.3%
Kept across 1,000 random weightings
In the typical random draw around the published weights this share stays in the top 20%; in the 5th percentile draw it is 90.9%.
1,432 of 1,814
Top-class tracts that hold in 95% of draws
These tracts stay in the top 20% in at least 95% of random weightings. The other 382 top-class tracts are less stable. Statewide, 791 tracts, in the top class or not, are in the top 20% in between 5% and 95% of draws.
0.997
Median rank correlation
Across random weightings the statewide order of tracts barely changes; a value of 1 would mean an identical order.
0.56
Strongest link between two components
No two components are redundant. The closest pair is canopy deficit and surface heat, well below the 0.7 level that would suggest overlap.
Agreement with Landsat temperature, CDC SVI or CalEnviroScreen is not independent evidence, because those datasets are inputs to UCRI. The report lists what these checks show and what they do not.
Using these data in your own work? Cite the dataset.
Automated QA gates
27 internal release and package QA gates
These automated checks detect inconsistencies in the release, its packages and its documents. They are internal quality controls, not independent methodology review, field validation, or proof of project outcomes.
Delivery gates (18)
Run against every generated package. Any failure stops the package from being delivered.
README present
Every package explains its own contents.
All required files present
No deliverable is silently missing.
Inventory contains only approved files
A package contains only the files it is meant to contain.
No QA, test or debug artifact names
Internal scratch files never reach a package.
No foreign GEOID in inventory
A package contains only the geography it was generated for.
Filename rule
Package names follow the documented, parseable convention.
PDF carries embedded maps
The report is never delivered without its cartography.
Locator shows the full county fabric
The locator map is geographically meaningful, not a blank frame.
Context map has a meaningful neighborhood
The neighborhood map contains real surrounding tracts.
Component panels show real local variation
Choropleths are not flat fills that convey nothing.
No scored visuals for insufficient data
An unscored tract never receives score-like graphics.
CSV re-parse matches the canonical row
Exported data is re-read and compared, not assumed.
GeoJSON re-parse valid with correct GEOID
Geometry exports reopen cleanly and identify the right tract.
Workbook re-read: sheets, freeze, filter, values
The spreadsheet is reopened and its sheets, frozen headers, filters and values are checked.
Score reproduction
The weighted formula reproduces the publicly released UCRI.
Manifest inventory matches files
Every listed file exists and every file is listed.
Status wording consistent
Published, provisional and unscored language never drifts.
ZIP integrity re-read
The finished archive is reopened and re-hashed against its manifest.
Release gates (9)
Run against the frozen data release itself.
Frozen data checksums
Every frozen file matches its recorded SHA-256.
UCRI reproduction gate
Published scores recompute from their components within tolerance.
Canonical consistency
The build is internally consistent across every derived index.
Public claims check
Public counts, labels and claims match the release metadata.
Route and link scan
Every internal link resolves to a real route or asset.
Security regression
Headers, auth guards and access rules hold.
Geometry quality
Tract geometry is valid and correctly attributed.
CanoProof trace
Certificate pages include every data file, font and verifier they need.
Release manifest check
The published manifest matches the tree it describes.
Current release files: source manifest (SOURCE_MANIFEST.json) and package checksums (SHA256SUMS.txt). Superseded releases are listed in the Technical Archive.
What these checks do not prove
They do not prove medical risk, causation, regulatory status, funding eligibility, official designation, parcel suitability, intervention effectiveness, or real-time conditions. UCRI remains a tract-level screening tool that requires current local evidence and professional review before project decisions.
UCRI classes are not official designations. They do not show SB 535 disadvantaged community status, AB 1550 low-income community status or eligibility for any program. Check the official CalEPA and California Climate Investments maps for those.
Surface heat. 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.
Canopy. 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.
Scale and region. Every value is a tract average, so it does not describe individual residents, blocks or parcels. Statewide percentiles rank hot, low-canopy desert and Central Valley tracts high by construction. The method page explains these limits.