Trust Center
A serious climate platform should show how trust is earned.
What is released, how CalEnviroScreen 5.0 enters the model, how incomplete records behave, what is not validated, and how the method should be used.
Dataset scope
9,129 California 2020 census tracts indexed; 9,033 fully scored; 74 provisional; 22 unscorable.
Current source
California UCRI uses statewide percentiles and CalEnviroScreen 5.0 Pollution Burden percentile for EBS.
Source policy
Metric values trace to named public source products, recorded vintages, and the frozen release artifacts.
Report policy
Reports provide screening context and do not establish site feasibility, eligibility, compliance, or project outcomes.
9,129 tracts
Statewide screening release
58 counties and 724 cities/places represented in the current California UCRI build.
Deterministic
Build QA gate
Source checksum, row uniqueness, tract join, score-kind accounting, ranking, priority-band, and output-count assertions run during release generation.
Versioned
Model registry
The public registry names the current formula, source releases, component definitions, exact technical ID, and archive policy.
Release controls
Controls that keep the statewide release inspectable
Stable IDs, scale-based geometry, release preservation, report traceability, and visible missing-data rules keep the operating model reviewable.
Stable IDs
Every tract resolves to an 11-digit GEOID, a stable tract_id, and a canopulse JSON ID.
Statewide build gate
The current release is generated only after source checksum, tract counts, formula reproduction, ranking sanity, geometry coverage, and component plausibility assertions pass.
Release preservation
Previous model artifacts remain unchanged in the Technical Archive for reproducibility and are not used by current public workflows.
Scale-based geometry
The statewide view uses county boundaries. Tiled 2020 tract geometry appears automatically at close zoom without a county-opening step.
Report auditability
California reports carry model version, frozen date, generated date, source provenance, limitations, and a stable audit ID.
Released geography only
County, city/place, and census tract outputs use released records. ZIP/ZCTA and parcel claims are not exposed.
Missing data
Incomplete records are labeled provisional, insufficient, or not scored rather than forced into the official ranking.
Human review
Field checks, parcel ownership, right-of-way constraints, and community input remain required before siting work.
What we do not claim
Clear boundaries are part of the product.
Canopulse provides inspectable screening context. Decisions still require current source review, local evidence, field conditions, applicable program rules, and community context.
Risk register
Known risks are disclosed, not hidden
This turns a weakness into trust. Serious users know every geospatial product has limits; Canopulse names them clearly.
Source vintage differences
Every report lists the observation year for each source. Values are not presented as same-day conditions.
Tract aggregation
Tract averages support comparison at census-tract scale and do not describe every block, property, or person inside the tract.
Historical release substitution
Normal workflows use only the current July 2026 release. Historical artifacts remain segregated in the Technical Archive.
Incomplete source records
Provisional and unscored records remain visible and are excluded from the frozen fully scored ranking.
Unsupported geography
The product exposes county, city/place, and tract outputs only. ZIP/ZCTA and parcel support are not represented.
Inspect the source and model registries next.
The source registry lists official products, vintages, access paths, costs, update cadence, and risk notes. The model registry binds the current release to CalEnviroScreen 5.0 and its source checksum.