HeatAQ Nexus
Chronological and spatial forecasting benchmarks for temperature, PM2.5 and compound events.
Open benchmark →HeatSafe integrates deterministic analysis, geospatial workflows, compound-hazard research, forecasting benchmarks, uncertainty quantification and optional AI explanation layers.
Open a simple guide to free household tools, plain-language learning pages and important safety limits—no account, purchase or paid AI required.
Every track is designed around explicit data identity, baselines, uncertainty, validation and reproducibility.
Chronological and spatial forecasting benchmarks for temperature, PM2.5 and compound events.
Open benchmark →Transparent multi-hazard formulations with exposed assumptions and sensitivity analysis.
Research questions →Trends, anomalies, heatwaves, hot nights and cooling-degree metrics.
Climate analysis →Concentration-first summaries, explicit AQI standards and cautious wildfire context.
PM2.5 workflows →Land-surface temperature, regional workflows and geospatial validation.
Urban-heat workflow →Experiment manifests, checksums, model cards, data cards and versioned artifacts.
Methods →Run one immutable specification to produce complete metrics, uncertainty figures, paper-ready reports, checksums and a self-contained candidate release.
Open experiment orchestrator →Deterministic core, local files, synthetic data, CPU baselines and browser tools. No paid AI API.
Optional local language models explain existing outputs without changing decisions or measurements.
Optional provider-neutral cloud models and credentialed data connectors using user-managed secrets.
Public tools translate the scientific core into inspectable decisions and local-first workflows.
Balance heat removal with PM2.5, smoke, humidity, wind and airflow.
Open tool →Identify likely heat pathways without producing a certified score.
Open tool →Expose energy, price and duty-cycle assumptions.
Open tool →Inspect thermal lag and observational response without causal overclaiming.
Open tool →The repository includes typed APIs, CLI workflows, Docker, CI, CodeQL, tests, provenance contracts, citation metadata and a research roadmap.
Explore the San Diego County station-level forecasting dataset, its verified research archive, citation, temporal coverage and scientific limitations.