What Canada-specific problems — wildfires, Arctic and permafrost thaw, bilingual public services, housing supply shortages, extreme weather, and Indigenous community infrastructure — could be addressed by combining Google's 2025–2026 open-source AI releases (Gemma 4, EmbeddingGemma 2, TimesFM-3, WeatherNext 2 and WeatherNext Cyclones, AlphaEarth satellite embeddings, Groundsource, LangExtract, MedGemma 1.5, A2A, ADK 2.0) with Canadian open data? Name the AI capability and the Canadian datasets for each.
The evidence is a catalogue of 11 distinct Google releases from 2025–2026 (15 rows once spelling and version aliases are merged) and 149 candidate AI-to-dataset pairings over 117 distinct Canadian open datasets across five federal data hosts; no single dataset backs more than about 3.4% of candidates. These are opportunity portfolios, not deployed solutions: AI alone does not solve structural problems. One caution up front: the pairing rows contain no source-linked AlphaEarth entry, so no Canadian dataset pairing is claimed for it here.
Detect fires, map fuels and burn scars, and forecast local risk and resource demand.
AI capability TimesFM-3 for quantile fire-weather and smoke forecasts; WeatherNext 2 for atmospheric drivers; Gemma 4 as an offline field assistant; A2A + ADK 2.0 to coordinate specialist agents.
Canadian open data ECCC daily climate observations (dry, hot, windy conditions), MSC RAQDPS wildfire-smoke plumes, NRCan SCANFI v2 forest attributes, HRDPS high-resolution forecasts.
Workflow Feed station observations and smoke-plume grids into TimesFM-3 for station-level risk quantiles; an ADK pipeline routes exceedances to dispatch agents over A2A; crews query a local Gemma 4 offline.
Caveats CWFIS hotspot feeds, the National Burned Area Composite and FBP fuel layers were requested but are absent from this evidence set; the nearest source-linked datasets are shown. Forecasts need Canadian calibration before operational use.
Monitor ground-temperature trends, snow cover and infrastructure exposure; recover historical observations.
AI capability TimesFM-3 for ground-temperature and snow trend forecasting; Groundsource's document-to-record methodology and LangExtract to structure historical reports.
Canadian open data ECCC homogenized daily temperature and precipitation (AHCCD), CanGridT gridded air temperature, Canadian Historical Daily Snow Depth Database, 1 km snow-cover fraction maps, CMIP5 scenarios including sea-ice and snow-depth change.
Workflow Trend-forecast northern station series with TimesFM-3; LangExtract mines decades of reports into structured thaw observations, cross-checked against gridded products.
Caveats Satellite or embedding products do not measure subsurface ice: any thaw signal must be calibrated against ground observations. Dedicated permafrost ground-temperature layers are absent from this evidence set.
Estimate English/French service demand, retrieve forms and media locally, extract bilingual notices, orchestrate referrals.
AI capability Gemma 4 multilingual on-device answering; EmbeddingGemma 2 private multimodal retrieval; LangExtract for source-grounded extraction of bilingual bulletins; A2A + ADK for referral workflows.
Canadian open data MSC city-page weather feed with EN/FR region names, Meteocode bilingual forecast XML, Statistics Canada historical census geography files.
Workflow Index bilingual government content with EmbeddingGemma 2 for offline retrieval; Gemma 4 answers in either official language; ADK routes referrals between EN and FR service agents over A2A.
Caveats The 2021 Census official-language fields were requested but are absent from this evidence set. 140-language support is not equal quality per language, and Indigenous-language service needs are not an English/French problem.
Forecast starts, permits and rents; find bottlenecks in planning documents; map addresses and monitor build-out.
AI capability TimesFM-3 for multivariate housing-series forecasts; LangExtract over planning and permit documents; EmbeddingGemma 2 for local search over address and planning records; A2A + ADK pipelines.
Canadian open data Canadian Housing Statistics Program tables (Statistics Canada), National Address Register API.
Workflow Forecast CHSP series with covariates in TimesFM-3; LangExtract surfaces approval bottlenecks from municipal documents; the National Address Register anchors everything to addresses.
Caveats CMHC starts/completions and Rental Market Survey rows are absent from this evidence set. Forecasts support decisions; they do not remove regulatory, financing, labour, land or infrastructure constraints.
Localized probabilistic forecasts, cyclone risk, flood evidence, alert summarization and coordinated response.
AI capability WeatherNext 2 for medium-range hazards; WeatherNext Cyclones for Atlantic track/intensity; TimesFM-3 for hydrometric quantiles; Groundsource for flash-flood evidence; LangExtract on alerts; A2A + ADK for response.
Canadian open data MSC GeoMet OGC API, current weather alerts, 1 km radar composite, GDPS/RDPS/HRDPS model output, real-time hydrometric data, HYDAT annual peak levels, NRCan flood-susceptibility mapping (current and future climate), AAFC extreme-weather heat and precipitation indices.
Workflow WeatherNext 2 and Cyclones run against GDPS/RDPS baselines; TimesFM-3 turns hydrometric gauges into quantile flood forecasts; Groundsource-style mining of public reports builds flash-flood training evidence; ADK executes alert-triggered playbooks over A2A.
Caveats WeatherNext Cyclones covers tropical and post-tropical systems, chiefly relevant to Atlantic Canada — not derechos, ice storms or prairie hail.
Community-governed monitoring and prioritization of water, housing, connectivity, roads and health access in remote settings.
AI capability Groundsource + LangExtract for auditable extraction from infrastructure and inspection reports; Gemma 4 + EmbeddingGemma 2 for offline, locally held retrieval; MedGemma 1.5 for validated medical-document and imaging support; A2A + ADK for cross-agency coordination.
Canadian open data CanEcumene 3.0 GIS database of community-oriented geography (NRCan/FGP).
Workflow A community-owned stack: local embeddings and a local Gemma 4 keep data on-site; extraction stays source-grounded and auditable; MedGemma 1.5 is used only inside clinically validated, human-reviewed pathways.
Caveats Indigenous data sovereignty comes first: OCAP® and CARE-aligned governance, consent, local ownership and human review. ISC drinking-water-advisory, Community Well-Being and broadband rows are absent from this evidence set; open public data does not remove governance obligations.
A curated shortlist of 26 strong source-linked pairings out of 149 candidates; the semantic problem labels in the candidate set are noisy and off-topic rows were omitted.
| Portfolio | Dataset | Publisher | Paired release | Source |
|---|---|---|---|---|
| Wildfires | Regional Air Quality Deterministic Prediction System (RAQDPS) | Meteorological Service of Canada (MSC) | TimesFM-3 | eccc-msc.github.io |
| Wildfires | Climate - Daily Observations (climate-daily) | MSC GeoMet | TimesFM-3 | api.weather.gc.ca |
| Wildfires | SCANFI v2: the Spatialized CAnadian National Forest Inventory data product | Natural Resources Canada | TimesFM-3 | open.canada.ca |
| Wildfires | High Resolution Deterministic Prediction System (HRDPS) | Meteorological Service of Canada (MSC) | TimesFM-3 | eccc-msc.github.io |
| Arctic & permafrost | Adjusted and Homogenized Canadian Climate Data – Daily Temperature and Precipitation (AHCCD – daily T&P) | Environment and Climate Change Canada | TimesFM-3 | open.canada.ca |
| Arctic & permafrost | Canadian Gridded Homogenized Surface Air Temperature (CanGridT V4) | Environment and Climate Change Canada | TimesFM-3 | open.canada.ca |
| Arctic & permafrost | Canadian Historical Daily Snow Depth Database | Environment and Climate Change Canada | TimesFM-3 | open.canada.ca |
| Arctic & permafrost | Daily snow cover fraction maps over Canada of the period of 2006-2010 from 1km resolution NOAA AVHRR imagery | Natural Resources Canada | TimesFM-3 | open.canada.ca |
| Arctic & permafrost | Global climate model scenarios dataset based on an ensemble of global climate model projections from the Coupled Model Intercomparison Project Phase 5 (CMIP5) | Environment and Climate Change Canada (ECCC) / Meteorological Service of Canada (MSC) | TimesFM-3 | eccc-msc.github.io |
| Bilingual services | City Page Weather [experimental] (MSC GeoMet - GeoMet-OGC-API, citypageweather-realtime collection) | GeoMet du SMC (Meteorological Service of Canada / ECCC-MSC) | EmbeddingGemma 2 | api.weather.gc.ca |
| Bilingual services | Restored 1971 census enumeration area boundaries for Canada's largest cities (Geographic files and documentation: 16-510-X2017001) | Statistics Canada | TimesFM-3 | statcan.gc.ca |
| Bilingual services | Meteocode forecast data in CSV format | Meteorological Service of Canada (MSC), Environment and Climate Change Canada (ECCC) | TimesFM-3 | eccc-msc.github.io |
| Housing supply | Canadian Housing Statistics Program (CHSP) | Statistics Canada | TimesFM-3 | statcan.gc.ca |
| Housing supply | National Address Register (NAR) | Statistics Canada (Data Integration Infrastructure Division) | EmbeddingGemma 2 | statcan.gc.ca |
| Extreme weather | Weather alerts data available on MSC GeoMet (Current-Alerts layer) | MSC (Meteorological Service of Canada) / ECCC | EmbeddingGemma 2 | eccc-msc.github.io |
| Extreme weather | North American Radar Composite (1 km) | Environment and Climate Change Canada | Groundsource | open.canada.ca |
| Extreme weather | Global Deterministic Prediction System (GDPS) | Environment and Climate Change Canada (Meteorological Service of Canada) | TimesFM-3 | eccc-msc.github.io |
| Extreme weather | Regional Deterministic Prediction System (RDPS) | Environment and Climate Change Canada | TimesFM-3 | open.canada.ca |
| Extreme weather | Real-time Hydrometric Data | MSC Open Data (ECCC) | TimesFM-3 | eccc-msc.github.io |
| Extreme weather | Hydrometric Historical Data (HYDAT) - Annual Maximum and Minimum Instantaneous Water Level or Flow | MSC GeoMet | TimesFM-3 | api.weather.gc.ca |
| Extreme weather | Flood Susceptibility Mapping | Natural Resources Canada | TimesFM-3 | open.canada.ca |
| Extreme weather | Future - Flood Susceptibility Mapping | Natural Resources Canada | TimesFM-3 | open.canada.ca |
| Extreme weather | Extreme Weather Indices: Heat | Agriculture and Agri-Food Canada | TimesFM-3 | open.canada.ca |
| Extreme weather | Extreme Weather Indices: Precipitation | Agriculture and Agri-Food Canada | Groundsource | open.canada.ca |
| Extreme weather | MSC GeoMet - GeoMet-OGC-API | Meteorological Service of Canada (MSC) and Environment and Climate Change Canada (ECCC) | ADK 2.0 | eccc-msc.github.io |
| Indigenous infrastructure | The Canadian Ecumene (CanEcumene) 3.0 GIS Database | Federal Geospatial Platform (FGP), Natural Resources Canada | TimesFM-3 | open.canada.ca |
Method: built from a Google release catalogue (15 rows, 11 distinct 2025–2026 releases after merging aliases) and 149 candidate AI–Canadian-dataset pairings over 117 distinct dataset URLs on five federal hosts, compiled 2026. Matrix cells are editorial placements of exact row capabilities into the six asked portfolios with a maturity judgment; the table shows a 26-row curated shortlist and the remaining candidates were cut for space. No deployment outcomes or Canadian accuracy figures are claimed. CWFIS, CMHC, census-language, drinking-water-advisory and AlphaEarth pairing rows are absent from the evidence set and are flagged where relevant.