Canola source watch

Build the read from admitted facts. Watch everything else.

This is the source board for Sprint 1. It separates official facts, interpretation candidates, and derived proxies before anything reaches a farmer-facing market call.

V1 guardrail

Climate, satellite, and map layers are not Canola V1 inputs yet.

Canola V1 can show stale, empty, lagged, and proxy sources. It cannot quietly turn those sources into advice.

The first market read stays anchored to the live Canola packet RPC until each new lane has a source contract and freshness proof.

Current read

Canola V1

Deterministic output from get_canada_thesis_packet('Canola', ...). No LLM thesis prose.

Source truth

Live packet RPC

Facts, warnings, stale rows, empty rows, and proxy rows stay visible before interpretation.

Run ledger

Waiting

source_runs exists, but collectors still need to write run summaries before health can roll up.

Weekly spine

CGC first

CGC movement data drives the Canada market read before slower compiled or map-derived sources.

Core Weekly Spine

These sources come before slower compiled releases and map-derived stress signals.

Canada movement first

cgc_weekly_stats

Canadian Grain Commission weekly statistics

Admitted

Cadence: Weekly, after the CGC Thursday publication.
Controls: Deliveries, shipments, terminal receipts, terminal exports, stocks, producer cars, process rows, grades, regions, crop year, grain week.
Why first: This is the Canada movement anchor. It tells us what is actually flowing before we interpret acres, drought, or farmer behavior.
Guardrail: Use the canonical packet/RPC rows, not stale market_analysis prose. Show week, crop year, units, and missing worksheets.

grain_monitor_weekly

Grain Monitor weekly logistics report

Admitted

Cadence: Weekly, with natural lag versus current CGC week.
Controls: Port unloads, vessel timing, grain-in-store, rail performance, corridor capacity, logistics pressure.
Why first: It explains why CGC movement is happening, especially when export pace looks weak but the real issue is unloads, vessels, or rail.
Guardrail: Preserve report date and lag. Do not smooth it into the same week as CGC if the source is behind.

cgc_producer_cars

Producer cars and railcar staging

Admitted

Cadence: Weekly, Thursday schedule after CGC movement data.
Controls: Producer-car allocations, shipments, destinations, and farmer-direct rail pressure.
Why first: This is one of the cleanest ways to see farmer-direct logistics pressure outside the elevator channel.
Guardrail: Producer-car rows are cumulative and forward-looking in places. Treat worksheet and metric names exactly.

cftc_cot

CFTC Commitments of Traders

Admitted

Cadence: Weekly Friday release for Tuesday positioning.
Controls: Managed money, commercial hedging, net position, week-over-week positioning pressure.
Why first: It gives futures positioning context, which matters when cash movement and futures reaction diverge.
Guardrail: Canola is direct only where mapping is direct. Proxy mappings must stay labelled and cannot carry the same confidence.

aafc_statscan_supply

AAFC / StatsCan supply-disposition and crop size

Watch next

Cadence: Periodic through intentions, seeded area, model estimates, and final production.
Controls: Seeded acres, harvested acres, yield, production, imports, exports, domestic use, carryout, stocks-to-use.
Why first: This sets the season-size baseline, but it is not the weekly flow source. A lot of the grain movement truth still rolls back to CGC.
Guardrail: Label survey/model/outlook/final status. Do not treat compiled annual values as fresher than CGC weekly movement.

Candidate Weather And Satellite Lanes

Admission means the lane can affect the read. Watching means it can appear as context only.

Verified research snapshot: 2026-05-04

aafc_canadian_drought_monitor

Canadian Drought Monitor

Watch next

Cadence: Monthly. AAFC metadata says usually by the 10th for the previous month.
Best use: Official drought-class benchmark for prairie yield risk and public-source traceability.
V1 decision: Good benchmark, not live enough for V1 stress detection. Show lag explicitly if displayed.
Failure mode: Consensus product, not a raw sensor feed. It can confirm the big picture but will miss fast changes.

AAFC CDM ArcGIS service

aafc_agroclimate_maps

AAFC Agroclimate maps

Watch next

Cadence: Near-real-time weather observations are assembled within about 12 hours, then mapped as rolling or fixed time frames.
Best use: Precipitation, temperature, GDD, SPI, SPEI, soil moisture, and drought-index context.
V1 decision: Useful next source family, but it is a weather lane. Do not let it into Canola V1 until one layer is admitted cleanly.
Failure mode: Large family of layers. Each map needs its own units, time window, geography, and failure mode.

AAFC Agroclimate maps How AAFC builds agroclimate maps

aafc_vegdri

Vegetation Drought Response Index

Research

Cadence: AAFC service describes weekly creation at 1 km resolution from MODIS NDVI, SPI, and static landscape inputs.
Best use: A stronger crop-stress layer than raw NDVI because it tries to isolate drought-driven vegetation stress.
V1 decision: Promising, but not admitted. The AAFC image service metadata observed on 2026-05-04 ended at 2025-08-30, so freshness must be solved first.
Failure mode: If stale, it is a historical benchmark only. If rebuilt in GEE, it becomes a derived proxy and must be labelled that way.

AAFC VegDRI image service AAFC VegDRI Canada research summary

aafc_smos_soil_moisture

Satellite soil moisture

Watch next

Cadence: Created daily from SMOS passive microwave data and averaged weekly, biweekly, or monthly.
Best use: Surface-layer moisture anomaly for drought, waterlogging, seeding conditions, and early yield-risk context.
V1 decision: Best AAFC satellite candidate after acres/production. Start as a map/read-only signal, not a thesis driver.
Failure mode: Coarse 0.25 degree grid and top-of-soil signal. Quality flags remove rain, snow, and radio-frequency interference.

AAFC soil moisture image service Open Canada dataset

nasa_servir_esi

Evaporative Stress Index

Research

Cadence: NASA/SERVIR products are commonly distributed as 4-week and 12-week ESI rasters.
Best use: Fast drought stress warning from evapotranspiration anomalies and land-surface-temperature behavior.
V1 decision: Useful early-warning proxy, but not AAFC. Keep outside Canola V1 until Canadian coverage, cadence, and access are proven.
Failure mode: Strong for flash drought, but source family, coverage, and raster access path must be pinned before farmer-facing use.

Drought.gov ESI data page SERVIR ESI downloads

gee_derived_drought_stack

GEE derived drought stack

Deferred

Cadence: Depends on the chosen ingredients: MODIS NDVI is 16-day, MODIS LST is daily, SMAP/SMOS-style soil moisture is coarse but frequent.
Best use: A closer-to-live proxy layer that can sit beside AAFC official products and flag stress before monthly maps catch up.
V1 decision: Build later as a derived source with formulas and validation, not as a substitute for official AAFC/StatsCan records.
Failure mode: Derived models can look authoritative while being wrong. They need back-testing against AAFC drought classes and yield outcomes.

GEE MODIS NDVI catalog GEE MODIS LST catalog NASA-USDA soil moisture in GEE

GEE answer

Yes, but it should be treated as a derived early-warning lane.

GEE can help get closer to live drought stress by combining MODIS NDVI, MODIS land-surface temperature, SMAP/SMOS-style soil moisture, and gridded precipitation. That is useful, but it is a model output, not an official source.

The clean architecture is official AAFC and StatsCan as the benchmark, GEE as the fast proxy, then Canola V1 as the consumer only after validation.

Admission test

A source enters the read only after these are pinned.

Official URL or explicit proxy label.
Cadence, lag, date field, and crop-year mapping.
Units and geography down to the display level.
Collector or API path with freshness proof.
Known failure modes shown in the read, not hidden.