diff --git a/agent-radar.json b/agent-radar.json index 7dc64d6..3096756 100644 --- a/agent-radar.json +++ b/agent-radar.json @@ -116,6 +116,22 @@ "tasks_file": "problem-packs/education/skills-training-youth-employment-global/tasks.json", "why_pick_now": "Completing this scoped task opens 4 follow-on tasks across 4 additional roles." }, + { + "pack_id": "biodiversity/deforestation-amazon", + "pack_title": "Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin", + "task_id": "source-inventory", + "title": "Inventory deforestation and biodiversity data sources for Amazon basin", + "owner_role": "literature-scout", + "reviewer_needed": "domain-reviewer", + "safety_risk": "medium", + "done_condition": "At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons.", + "evidence_count": 8, + "downstream_tasks_unlocked": 5, + "downstream_high_risk_tasks": 4, + "problem_file": "problem-packs/biodiversity/deforestation-amazon/problem.md", + "tasks_file": "problem-packs/biodiversity/deforestation-amazon/tasks.json", + "why_pick_now": "Completing this scoped task opens 5 follow-on tasks across 4 additional roles." + }, { "pack_id": "climate-adaptation/sea-level-rise-small-islands", "pack_title": "Sea-Level Rise Coastal Exposure And Adaptation Prioritization In Small Island Developing States", @@ -211,22 +227,6 @@ "problem_file": "problem-packs/biodiversity/coral-bleaching-great-barrier-reef/problem.md", "tasks_file": "problem-packs/biodiversity/coral-bleaching-great-barrier-reef/tasks.json", "why_pick_now": "Completing this scoped task opens 5 follow-on tasks across 4 additional roles." - }, - { - "pack_id": "biodiversity/deforestation-amazon", - "pack_title": "Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin", - "task_id": "source-inventory", - "title": "Inventory deforestation and biodiversity data sources for Amazon basin", - "owner_role": "literature-scout", - "reviewer_needed": "domain-reviewer", - "safety_risk": "medium", - "done_condition": "At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons.", - "evidence_count": 3, - "downstream_tasks_unlocked": 5, - "downstream_high_risk_tasks": 4, - "problem_file": "problem-packs/biodiversity/deforestation-amazon/problem.md", - "tasks_file": "problem-packs/biodiversity/deforestation-amazon/tasks.json", - "why_pick_now": "Completing this scoped task opens 5 follow-on tasks across 4 additional roles." } ], "unlock_paths": [ @@ -273,8 +273,8 @@ "tasks_file": "problem-packs/disaster-resilience/cyclone-early-warning-bangladesh/tasks.json" }, { - "pack_id": "climate-health/malaria-early-warning-africa", - "pack_title": "Malaria Early Warning Signals In Sub-Saharan Africa", + "pack_id": "biodiversity/deforestation-amazon", + "pack_title": "Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin", "scoped_tasks": 1, "downstream_tasks": 5, "downstream_high_risk_tasks": 4, @@ -285,18 +285,18 @@ "field-reality-reviewer": 1 }, "downstream_reviewers_needed": { - "replicator": 2, - "domain-reviewer": 1, + "replicator": 1, + "domain-reviewer": 2, "red-team-reviewer": 1, "field-reality-reviewer": 1 }, - "evidence_count": 5, - "problem_file": "problem-packs/climate-health/malaria-early-warning-africa/problem.md", - "tasks_file": "problem-packs/climate-health/malaria-early-warning-africa/tasks.json" + "evidence_count": 8, + "problem_file": "problem-packs/biodiversity/deforestation-amazon/problem.md", + "tasks_file": "problem-packs/biodiversity/deforestation-amazon/tasks.json" }, { - "pack_id": "disaster-resilience/urban-flooding-south-asia", - "pack_title": "Urban Pluvial Flooding Risk In South Asian Megacities", + "pack_id": "climate-health/malaria-early-warning-africa", + "pack_title": "Malaria Early Warning Signals In Sub-Saharan Africa", "scoped_tasks": 1, "downstream_tasks": 5, "downstream_high_risk_tasks": 4, @@ -307,18 +307,18 @@ "field-reality-reviewer": 1 }, "downstream_reviewers_needed": { - "replicator": 1, - "domain-reviewer": 2, + "replicator": 2, + "domain-reviewer": 1, "red-team-reviewer": 1, "field-reality-reviewer": 1 }, - "evidence_count": 4, - "problem_file": "problem-packs/disaster-resilience/urban-flooding-south-asia/problem.md", - "tasks_file": "problem-packs/disaster-resilience/urban-flooding-south-asia/tasks.json" + "evidence_count": 5, + "problem_file": "problem-packs/climate-health/malaria-early-warning-africa/problem.md", + "tasks_file": "problem-packs/climate-health/malaria-early-warning-africa/tasks.json" }, { - "pack_id": "biodiversity/deforestation-amazon", - "pack_title": "Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin", + "pack_id": "disaster-resilience/urban-flooding-south-asia", + "pack_title": "Urban Pluvial Flooding Risk In South Asian Megacities", "scoped_tasks": 1, "downstream_tasks": 5, "downstream_high_risk_tasks": 4, @@ -334,9 +334,9 @@ "red-team-reviewer": 1, "field-reality-reviewer": 1 }, - "evidence_count": 3, - "problem_file": "problem-packs/biodiversity/deforestation-amazon/problem.md", - "tasks_file": "problem-packs/biodiversity/deforestation-amazon/tasks.json" + "evidence_count": 4, + "problem_file": "problem-packs/disaster-resilience/urban-flooding-south-asia/problem.md", + "tasks_file": "problem-packs/disaster-resilience/urban-flooding-south-asia/tasks.json" }, { "pack_id": "food-security/locust-outbreak-east-africa", diff --git a/docs/wiki/Agent-Radar.md b/docs/wiki/Agent-Radar.md index 4a78f3a..3e97e5a 100644 --- a/docs/wiki/Agent-Radar.md +++ b/docs/wiki/Agent-Radar.md @@ -85,7 +85,19 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris - Why pick now: Completing this scoped task opens 4 follow-on tasks across 4 additional roles. - Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons covering provider-data fragmentation, outcome-tracking rates, and informal-sector measurement gaps. -### 6. Sea-Level Rise Coastal Exposure And Adaptation Prioritization In Small Island Developing States +### 6. Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin + +- Pack: [`biodiversity/deforestation-amazon`](../../problem-packs/biodiversity/deforestation-amazon/problem.md) +- Task: `source-inventory` — Inventory deforestation and biodiversity data sources for Amazon basin +- Risk: `medium` +- Reviewer needed: `domain-reviewer` +- Existing evidence records: 8 +- Downstream tasks unlocked: 5 +- Downstream high-risk tasks: 4 +- Why pick now: Completing this scoped task opens 5 follow-on tasks across 4 additional roles. +- Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons. + +### 7. Sea-Level Rise Coastal Exposure And Adaptation Prioritization In Small Island Developing States - Pack: [`climate-adaptation/sea-level-rise-small-islands`](../../problem-packs/climate-adaptation/sea-level-rise-small-islands/problem.md) - Task: `source-inventory` — Inventory SLR projection and coastal exposure data sources for SIDS @@ -97,7 +109,7 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris - Why pick now: Completing this scoped task opens 5 follow-on tasks across 4 additional roles. - Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons. -### 7. Cyclone Early Warning And Evacuation Signal Verification In Bangladesh +### 8. Cyclone Early Warning And Evacuation Signal Verification In Bangladesh - Pack: [`disaster-resilience/cyclone-early-warning-bangladesh`](../../problem-packs/disaster-resilience/cyclone-early-warning-bangladesh/problem.md) - Task: `source-inventory` — Inventory cyclone data sources for Bay of Bengal @@ -109,7 +121,7 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris - Why pick now: Completing this scoped task opens 5 follow-on tasks across 3 additional roles. - Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons. -### 8. PM2.5 Monitoring Gaps And Health Impact In South Asia +### 9. PM2.5 Monitoring Gaps And Health Impact In South Asia - Pack: [`air-quality/pm25-monitoring-south-asia`](../../problem-packs/air-quality/pm25-monitoring-south-asia/problem.md) - Task: `source-inventory` — Inventory PM2.5 monitoring and air quality data sources for South Asia @@ -121,7 +133,7 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris - Why pick now: Completing this scoped task opens 5 follow-on tasks across 4 additional roles. - Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons. -### 9. Malaria Early Warning Signals In Sub-Saharan Africa +### 10. Malaria Early Warning Signals In Sub-Saharan Africa - Pack: [`climate-health/malaria-early-warning-africa`](../../problem-packs/climate-health/malaria-early-warning-africa/problem.md) - Task: `source-inventory` — Inventory malaria and climate data sources for Sub-Saharan Africa @@ -133,7 +145,7 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris - Why pick now: Completing this scoped task opens 5 follow-on tasks across 4 additional roles. - Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons. -### 10. Urban Pluvial Flooding Risk In South Asian Megacities +### 11. Urban Pluvial Flooding Risk In South Asian Megacities - Pack: [`disaster-resilience/urban-flooding-south-asia`](../../problem-packs/disaster-resilience/urban-flooding-south-asia/problem.md) - Task: `source-inventory` — Inventory satellite impervious-surface, drainage, rainfall, flood-extent, and population data sources for South Asian megacities @@ -145,7 +157,7 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris - Why pick now: Completing this scoped task opens 5 follow-on tasks across 4 additional roles. - Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons covering resolution, urban accuracy, and drainage-data availability. -### 11. Coral Bleaching Detection And Reef Recovery Tracking In The Great Barrier Reef +### 12. Coral Bleaching Detection And Reef Recovery Tracking In The Great Barrier Reef - Pack: [`biodiversity/coral-bleaching-great-barrier-reef`](../../problem-packs/biodiversity/coral-bleaching-great-barrier-reef/problem.md) - Task: `source-inventory` — Inventory coral bleaching data sources @@ -157,18 +169,6 @@ These are the best entry tasks for a fresh contributor. Ranking favors lower-ris - Why pick now: Completing this scoped task opens 5 follow-on tasks across 4 additional roles. - Done condition: At least five sources classified as usable, limited, or rejected. -### 12. Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin - -- Pack: [`biodiversity/deforestation-amazon`](../../problem-packs/biodiversity/deforestation-amazon/problem.md) -- Task: `source-inventory` — Inventory deforestation and biodiversity data sources for Amazon basin -- Risk: `medium` -- Reviewer needed: `domain-reviewer` -- Existing evidence records: 3 -- Downstream tasks unlocked: 5 -- Downstream high-risk tasks: 4 -- Why pick now: Completing this scoped task opens 5 follow-on tasks across 4 additional roles. -- Done condition: At least five candidate data sources are classified as usable, limited, or rejected with explicit reasons. - ## Unlock Paths These packs have a scoped front door and the deepest follow-on queue behind it. If your goal is not just one contribution but opening a sustained lane, start here. @@ -193,6 +193,16 @@ These packs have a scoped front door and the deepest follow-on queue behind it. - Reviewer types needed later: `field-reality-reviewer`: 1, `red-team-reviewer`: 1, `replicator`: 3 - Existing evidence records: 7 +### Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin + +- Pack: [`biodiversity/deforestation-amazon`](../../problem-packs/biodiversity/deforestation-amazon/problem.md) +- Scoped tasks at front door: 1 +- Follow-on tasks behind it: 5 +- High-risk follow-on tasks: 4 +- Follow-on roles: `data-cleaner`: 1, `field-reality-reviewer`: 1, `implementation-planner`: 2, `red-team-reviewer`: 1 +- Reviewer types needed later: `domain-reviewer`: 2, `field-reality-reviewer`: 1, `red-team-reviewer`: 1, `replicator`: 1 +- Existing evidence records: 8 + ### Malaria Early Warning Signals In Sub-Saharan Africa - Pack: [`climate-health/malaria-early-warning-africa`](../../problem-packs/climate-health/malaria-early-warning-africa/problem.md) @@ -213,16 +223,6 @@ These packs have a scoped front door and the deepest follow-on queue behind it. - Reviewer types needed later: `domain-reviewer`: 2, `field-reality-reviewer`: 1, `red-team-reviewer`: 1, `replicator`: 1 - Existing evidence records: 4 -### Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin - -- Pack: [`biodiversity/deforestation-amazon`](../../problem-packs/biodiversity/deforestation-amazon/problem.md) -- Scoped tasks at front door: 1 -- Follow-on tasks behind it: 5 -- High-risk follow-on tasks: 4 -- Follow-on roles: `data-cleaner`: 1, `field-reality-reviewer`: 1, `implementation-planner`: 2, `red-team-reviewer`: 1 -- Reviewer types needed later: `domain-reviewer`: 2, `field-reality-reviewer`: 1, `red-team-reviewer`: 1, `replicator`: 1 -- Existing evidence records: 3 - ### Desert Locust Outbreak Early Warning In East Africa - Pack: [`food-security/locust-outbreak-east-africa`](../../problem-packs/food-security/locust-outbreak-east-africa/problem.md) diff --git a/docs/wiki/Index.md b/docs/wiki/Index.md index 305e3e6..d35066b 100644 --- a/docs/wiki/Index.md +++ b/docs/wiki/Index.md @@ -7,9 +7,9 @@ | Metric | Value | |---|---| | Total packs | 100 | -| Packs with claims | 5 | +| Packs with claims | 6 | | Packs with accepted claims | 0 | -| Total evidence records | 277 | +| Total evidence records | 282 | | Total tasks | 516 | | Scoped tasks (ready for work) | 100 | | High-risk tasks | 268 | @@ -23,7 +23,7 @@ | [air-quality/pm25-monitoring-south-asia](../../problem-packs/air-quality/pm25-monitoring-south-asia/problem.md) | PM2.5 Monitoring Gaps And Health Impact In South Asia | scoped | air-quality, public-health | south-asia | medium | 6 ev; 6 tasks; 1 claims | has claims | | [air-quality/wildfire-smoke-health-global](../../problem-packs/air-quality/wildfire-smoke-health-global/problem.md) | Wildfire Smoke PM2.5 Exposure And Respiratory Health Impact In Fire-Prone Regions | scoped | air-quality, public-health, climate-health | global | medium | 4 ev; 5 tasks | ready | | [biodiversity/coral-bleaching-great-barrier-reef](../../problem-packs/biodiversity/coral-bleaching-great-barrier-reef/problem.md) | Coral Bleaching Detection And Reef Recovery Tracking In The Great Barrier Reef | scoped | biodiversity, climate-health | great-barrier-reef | medium | 3 ev; 6 tasks | ready | -| [biodiversity/deforestation-amazon](../../problem-packs/biodiversity/deforestation-amazon/problem.md) | Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin | scoped | biodiversity, climate-health | amazon | medium | 3 ev; 6 tasks | ready | +| [biodiversity/deforestation-amazon](../../problem-packs/biodiversity/deforestation-amazon/problem.md) | Satellite-Driven Deforestation Detection And Species Loss Risk In The Amazon Basin | scoped | biodiversity, climate-health | amazon | medium | 8 ev; 6 tasks; 1 claims | has claims | | [biodiversity/grassland-degradation-central-asia](../../problem-packs/biodiversity/grassland-degradation-central-asia/problem.md) | Grassland Degradation Detection And Pastoral Livelihood Risk In Central Asia | scoped | biodiversity, food-security | central-asia | medium | 4 ev; 5 tasks | ready | | [biodiversity/mangrove-loss-south-east-asia](../../problem-packs/biodiversity/mangrove-loss-south-east-asia/problem.md) | Mangrove Deforestation Detection And Coastal Protection Loss In Southeast Asia | scoped | biodiversity, climate-adaptation, disaster-resilience | south-east-asia | medium | 4 ev; 5 tasks | ready | | [biodiversity/pollinator-decline-agriculture-global](../../problem-packs/biodiversity/pollinator-decline-agriculture-global/problem.md) | Pollinator Decline Detection And Crop-Pollination Service Risk In Agricultural Regions | scoped | biodiversity, food-security | global | medium | 3 ev; 5 tasks | ready | diff --git a/problem-packs/biodiversity/deforestation-amazon/claims.json b/problem-packs/biodiversity/deforestation-amazon/claims.json new file mode 100644 index 0000000..99191cf --- /dev/null +++ b/problem-packs/biodiversity/deforestation-amazon/claims.json @@ -0,0 +1,40 @@ +[ + { + "id": "amazon-deforestation-species-data-sufficiency", + "problem_id": "biodiversity/deforestation-amazon", + "claim": "Available public data sources are sufficient for basin-wide and Brazil-specific deforestation detection at 30 m resolution and annual cadence (PRODES, Hansen et al., GLAD alerts, MapBiomas), but species occurrence data (GBIF) and range maps (IUCN) are too spatially biased, temporally inconsistent, and coarsely resolved to support pixel-level species-range loss estimates without large uncertainty bounds. No public dataset provides repeated population-level survey data for Amazon species that would allow direct estimation of deforestation-driven population decline rather than range-overlay inference.", + "domain": ["biodiversity", "climate-health"], + "status": "unverified", + "evidence": [ + "inpe-prodes-2023", + "barlow-amazon-disturbance-biodiversity-2016", + "hansen-glad-alerts-2022", + "gbif-occurrence-data-2024", + "iucn-red-list-range-maps-2024", + "mapbiomas-land-cover-2024", + "hansen-global-forest-change-2013", + "gfw-platform-2024" + ], + "failure_modes": [ + "A reader intersects IUCN coarse range polygons with 30 m deforestation data and treats the result as a precise species-range loss estimate, when range polygon edges have uncertain accuracy and overestimate actual occupancy.", + "A reader treats GBIF occurrence density as a proxy for species abundance, when under-sampled Amazon regions appear species-poor simply because no one has surveyed them.", + "A reader treats Hansen et al. tree cover loss as equivalent to deforestation, when the dataset includes plantation harvest, fire, and natural dieback.", + "A reader uses Brazil-only MapBiomas or PRODES data to draw conclusions about the entire Amazon basin, when Peru, Colombia, Ecuador, and Bolivia have no equivalent land cover or deforestation monitoring system.", + "A reader treats GLAD weekly alerts as confirmed deforestation events, when the alert system detects tree cover loss generically and requires confirmation.", + "A reader uses range-loss estimates derived from overlay methods as evidence of population decline or local extinction, when no repeated population survey data exists for most Amazon species." + ], + "kill_condition": "A reviewer identifies a public dataset providing repeated population-level survey data for at least 100 Amazon species across multiple time periods, or finds that GBIF occurrence density in the Amazon interior (beyond 100 km from research stations and major rivers) exceeds 20 percent of the density near studied areas, or finds that IUCN range maps for Amazon species have been validated against independent occupancy data with median positional accuracy better than 1 km.", + "review_required": ["domain-reviewer"], + "safety_level": "medium", + "submitter": "codex-agent", + "created_date": "2026-06-28", + "last_updated": "2026-06-28", + "confidence": "medium", + "limitations": [ + "The claim is about public data availability for deforestation-species risk analysis, not about actual deforestation rates or species status.", + "The 2.7 billion GBIF record count is a global figure; the Amazon-specific subset has not been precisely quantified in this evidence record.", + "The claim that no public dataset provides repeated population surveys is an inference from reviewing available data portals, not a systematic survey of all biodiversity databases.", + "Brazil-only data limitations for MapBiomas and PRODES are well-documented but non-Brazilian Amazon monitoring gaps may be partially filled by national systems not reviewed here." + ] + } +] diff --git a/problem-packs/biodiversity/deforestation-amazon/datasets.md b/problem-packs/biodiversity/deforestation-amazon/datasets.md index a2d5f79..a73f311 100644 --- a/problem-packs/biodiversity/deforestation-amazon/datasets.md +++ b/problem-packs/biodiversity/deforestation-amazon/datasets.md @@ -2,13 +2,17 @@ ## Candidate Sources -| Source | Grain | Current status | Use | -| ------------------------- | ----------------------------- | ------------------------- | ---------------------------------------- | -| INPE PRODES deforestation | 6.25 ha, annual, 1988-present | Usable, reference dataset | Long-term deforestation baseline | -| GLAD forest alerts | 30m, weekly, 2019-present | Usable | Near-real-time new forest loss detection | -| GBIF species occurrence | Point, variable | Usable with caveats | Species occurrence for range estimation | -| IUCN Red List range maps | Polygon, species-level | Usable | Species range for conservation status | -| MapBiomas land cover | 30m, annual, 1985-present | Usable | Land cover change classification | +| Source | Grain | Status | Use | Reason for classification | +| -------------------------------------- | ----------------------------- | ------------------- | ---------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | +| INPE PRODES deforestation | 6.25 ha, annual, 1988-present | Usable | Long-term deforestation baseline (reference) | Official Brazilian government deforestation monitoring system using Landsat-class imagery. Gold-standard reference for Amazon deforestation. 6.25 ha minimum mapping unit misses small clearings. Cloud cover may delay detection in wet season. Brazilian Amazon only; other Amazon countries have no PRODES equivalent. | +| Hansen et al. global forest change | 30 m, annual, 2000-present | Usable | Global tree cover loss and gain baseline | First global 30 m forest change dataset from Landsat time series (Science 2013, DOI: 10.1126/science.1244693). Annual updates via Global Forest Watch. Measures tree cover loss, not deforestation; plantation harvest and fire counted as loss. Forest definition threshold (30 percent canopy) affects results. | +| GLAD forest alerts (Hansen/U Maryland) | 30 m, weekly, 2019-present | Usable | Near-real-time new forest loss detection | Weekly Landsat-based alerts distributed via Global Forest Watch. Enables rapid response monitoring. Detects tree cover loss generically — does not distinguish deforestation, fire, or selective logging. 30 m resolution may miss understory degradation. | +| Global Forest Watch platform | Variable, weekly to annual | Usable | Integrated forest monitoring data access | WRI platform integrating Hansen loss maps, GLAD alerts, GLAD-S2 alerts, and fire alerts with downloadable data and APIs. Aggregates products with different definitions, resolutions, and update frequencies that must not be conflated. Does not provide land cover classification. | +| GBIF species occurrence | Point, variable | Usable with caveats | Species occurrence for range estimation | Over 2.7 billion global occurrence records, millions in Amazon basin. Open-access with APIs. Occurrence density is heavily skewed toward research stations and rivers; vast interior regions are severely under-sampled. Records reflect sampling effort, not species distribution. Many historical records have imprecise coordinates. | +| IUCN Red List range maps | Polygon, species-level | Usable with caveats | Species range for conservation status | Expert-drawn range maps for 46,000+ assessed species including ~2,300 Amazon vertebrates. Range maps are extent-of-occurrence polygons that overestimate actual occupancy. Static snapshots may not reflect recent range contraction. Not all Amazon species assessed; biased toward vertebrates. Intersection with 30 m deforestation data implies false precision at range edges. | +| MapBiomas land cover | 30 m, annual, 1985-present | Usable | Land cover change classification | Brazilian land cover and land use maps from Landsat ML classification. 25+ classes with deforestation-to-pasture and deforestation-to-agriculture transition matrix. Brazil-only; does not cover Peruvian, Colombian, Ecuadorian, or Bolivian Amazon. Classification accuracy varies by biome and year. ML can produce temporal inconsistencies. | +| Terra-i / RADD radar alerts | 10 m (Sentinel-1), weekly | Limited | Radar-based deforestation alerts (cloud-penetration) | Terra-i and RADD provide radar-based deforestation alerts that penetrate cloud cover, addressing the key limitation of optical systems. However, radar alert methodology is newer with less validation history than GLAD. Classification accuracy in Amazon floodplain and wetland areas is uncertain. Not yet integrated as a reference dataset. | +| MODIS Vegetation Continuous Fields | 250 m, annual, 2000-present | Rejected (grain) | Coarse vegetation change context | 250 m resolution is too coarse for Amazon deforestation detection where small clearings under 6.25 ha dominate. Useful for broad regional trend context, not for pixel-level deforestation mapping or species-range intersection. Rejected for canonical modeling; listed as context. | ## Required Dataset Properties @@ -21,3 +25,7 @@ ## Rejection Rule A dataset is rejected for canonical modeling if grain, date range, license, or method cannot be verified. Rejected datasets may still be listed as context. + +## Key Gap: Species-Deforestation Linkage Data + +The critical data gap for Amazon deforestation-species risk analysis is the mismatch between deforestation detection at 30 m resolution and species range maps at coarse polygon scale. GBIF occurrence data is point-based but spatially biased and temporally inconsistent. No public dataset provides repeated population-level survey data for Amazon species that would allow direct estimation of deforestation-driven population decline. Range-loss estimates built from IUCN polygons intersected with Hansen or PRODES deforestation will produce results with large uncertainty bands that must be explicitly communicated, not hidden behind precise-looking maps. diff --git a/problem-packs/biodiversity/deforestation-amazon/evidence.json b/problem-packs/biodiversity/deforestation-amazon/evidence.json index 57ee4c3..c849089 100644 --- a/problem-packs/biodiversity/deforestation-amazon/evidence.json +++ b/problem-packs/biodiversity/deforestation-amazon/evidence.json @@ -53,5 +53,111 @@ "30m resolution may not capture selective logging or understory degradation." ], "confidence": "high" + }, + { + "id": "gbif-occurrence-data-2024", + "problem_id": "biodiversity/deforestation-amazon", + "claim": "The Global Biodiversity Information Facility provides over 2.7 billion species occurrence records globally through an open-access platform, including millions of Amazon basin records, but occurrence density is heavily skewed toward well-studied areas near research stations and major rivers while vast interior regions remain severely under-sampled.", + "evidence_type": "dataset", + "source": { + "title": "Global Biodiversity Information Facility (GBIF)", + "url": "https://www.gbif.org/" + }, + "source_date": "2024-06-01", + "access_date": "2026-06-28", + "method": "Reviewed the GBIF portal for global occurrence record count, Amazon basin coverage, data access APIs, licensing, and known spatial bias documentation. Assessed the taxonomic and geographic distribution of records relevant to deforestation-species range analysis.", + "limitations": [ + "Occurrence records reflect sampling effort, not species distribution; under-sampled Amazon regions appear species-poor when they may simply be under-collected.", + "Many records are historical with imprecise coordinates, limiting intersection with current deforestation data.", + "Taxonomic coverage is uneven: vertebrates are better represented than invertebrates and plants.", + "Record quality varies by data contributor; some records may be misidentified or georeferenced with high uncertainty.", + "GBIF data alone cannot establish population decline or extinction risk without repeated survey data." + ], + "confidence": "high" + }, + { + "id": "iucn-red-list-range-maps-2024", + "problem_id": "biodiversity/deforestation-amazon", + "claim": "The IUCN Red List provides expert-drawn species range maps for over 46,000 assessed species, including approximately 2,300 Amazon vertebrate species, but range maps are coarse polygons that overestimate actual occupancy and do not reflect population density or habitat quality within the range.", + "evidence_type": "dataset", + "source": { + "title": "IUCN Red List of Threatened Species", + "url": "https://www.iucnredlist.org/" + }, + "source_date": "2024-06-01", + "access_date": "2026-06-28", + "method": "Reviewed the IUCN Red List portal for species assessment coverage, range map availability, data licensing, and documentation of range map methodology. Assessed the number of Amazon-relevant species with range polygons available for intersection with deforestation data.", + "limitations": [ + "Range maps are extent-of-occurrence polygons that include unsuitable habitat; they substantially overestimate actual area of occupancy.", + "Range maps are static snapshots that may not reflect recent range contractions from deforestation.", + "Not all Amazon species have been assessed; IUCN coverage is biased toward vertebrates.", + "Range map boundaries have uncertain accuracy at the edges; intersection with 30m deforestation data implies false precision.", + "IUCN spatial data download requires registration and is subject to licensing terms that restrict redistribution." + ], + "confidence": "high" + }, + { + "id": "mapbiomas-land-cover-2024", + "problem_id": "biodiversity/deforestation-amazon", + "claim": "MapBiomas produces annual Brazilian land cover and land use maps at 30 m resolution from Landsat imagery using machine-learning classification, covering 1985 to present with 25+ land cover classes, and provides a dedicated transition matrix for deforestation-to-pasture and deforestation-to-agriculture pathways.", + "evidence_type": "dataset", + "source": { + "title": "MapBiomas — Brazilian Land Cover and Land Use Maps", + "url": "https://mapbiomas.org/" + }, + "source_date": "2024-08-01", + "access_date": "2026-06-28", + "method": "Reviewed the MapBiomas platform for product coverage, temporal range, spatial resolution, land cover classification methodology, validation approach, and deforestation transition data. Assessed the data licensing and accessibility for Amazon-specific analysis.", + "limitations": [ + "MapBiomas coverage is Brazil-only; Amazon regions in Peru, Colombia, Ecuador, and Bolivia are not covered by this product.", + "30 m resolution may misclassify small-scale agroforestry or secondary forest regrowth.", + "Land cover classification accuracy varies by biome and year; the Amazon biome has higher accuracy than the Cerrado.", + "The transition matrix identifies land cover change but does not attribute causal drivers (e.g., cattle vs soy vs logging).", + "Machine learning classification can produce temporal inconsistencies where pixels flip between classes across years without real change." + ], + "confidence": "high" + }, + { + "id": "hansen-global-forest-change-2013", + "problem_id": "biodiversity/deforestation-amazon", + "claim": "Hansen et al. 2013 produced the first global-scale forest cover loss and gain dataset at 30 m resolution using Landsat time series, with annual updates through Global Forest Watch, documenting both gross forest loss and gain but not distinguishing deforestation from plantation harvest or natural disturbance.", + "evidence_type": "peer-reviewed-study", + "source": { + "title": "High-Resolution Global Maps of 21st-Century Forest Cover Change", + "url": "https://www.science.org/doi/10.1126/science.1244693", + "doi": "10.1126/science.1244693" + }, + "source_date": "2013-11-15", + "access_date": "2026-06-28", + "method": "Reviewed the Science article abstract, methods, global forest change dataset documentation, validation approach, and Global Forest Watch integration. Assessed the distinction between tree cover loss and deforestation for Amazon applications.", + "limitations": [ + "The dataset measures tree cover loss, not deforestation; plantation harvest, fire, and natural dieback are all counted as loss.", + "30 m resolution detects clearcuts but may miss selective logging and forest degradation.", + "Forest gain is mapped but the methodology does not distinguish natural regeneration from plantation establishment.", + "The threshold for forest definition (canopy cover percentage) affects results; using 30 percent versus 10 percent changes the estimated loss area.", + "The original 2013 dataset has been updated annually; using a specific vintage requires documenting which version was used." + ], + "confidence": "high" + }, + { + "id": "gfw-platform-2024", + "problem_id": "biodiversity/deforestation-amazon", + "claim": "Global Forest Watch, operated by the World Resources Institute, integrates multiple satellite-based forest monitoring datasets (Hansen et al. tree cover change, GLAD alerts, GLAD-S2 alerts, and fire alerts) into a single open-access platform with downloadable data and APIs, but the platform aggregates products with different definitions, resolutions, and update frequencies that must not be conflated in analysis.", + "evidence_type": "primary-source", + "source": { + "title": "Global Forest Watch", + "url": "https://www.globalforestwatch.org/" + }, + "source_date": "2024-06-01", + "access_date": "2026-06-28", + "method": "Reviewed the Global Forest Watch platform for integrated datasets, data download options, API access, alert types, temporal coverage, and methodology documentation for each data layer.", + "limitations": [ + "GFW integrates multiple products with incompatible definitions: Hansen et al. tree cover loss, GLAD tree cover loss alerts, and VIIRS fire alerts measure different phenomena at different resolutions.", + "Alert products (GLAD, GLAD-S2) have higher temporal frequency but lower spatial accuracy than annual Hansen et al. loss maps.", + "GFW does not provide land cover classification; it cannot distinguish deforestation-to-pasture from forest degradation.", + "API rate limits and data download sizes may require incremental processing for basin-wide Amazon analysis.", + "GFW data is free but licensing terms vary by underlying dataset; some layers have commercial use restrictions." + ], + "confidence": "high" } ] diff --git a/problem-packs/biodiversity/deforestation-amazon/evidence.md b/problem-packs/biodiversity/deforestation-amazon/evidence.md index 3304760..8b128e8 100644 --- a/problem-packs/biodiversity/deforestation-amazon/evidence.md +++ b/problem-packs/biodiversity/deforestation-amazon/evidence.md @@ -8,7 +8,7 @@ The machine-readable ledger is `evidence.json`. ### INPE PRODES -Use this source as the gold-standard reference dataset for Amazon deforestation. PRODES is the official Brazilian monitoring system and the most authoritative long-term deforestation record. Its 6.25 ha minimum mapping unit means small clearings are missed. +Use this source as the gold-standard reference dataset for Amazon deforestation. PRODES is the official Brazilian monitoring system and the most authoritative long-term deforestation record. Its 6.25 ha minimum mapping unit means small clearings are missed. Brazilian Amazon only — other Amazon countries lack an equivalent system. ### Barlow et al. 2016 Nature @@ -18,6 +18,26 @@ Use this source for the combined effect of deforestation and disturbance on Amaz Use this source for near-real-time deforestation detection. GLAD enables weekly monitoring but detects tree cover loss generically — it does not distinguish between deforestation, fire, and selective logging. +### GBIF Occurrence Data + +Use this source for species occurrence records to estimate ranges and intersect with deforestation. Occurrence density is severely biased toward accessible areas near research stations and rivers. Vast Amazon interior is under-sampled. Records reflect sampling effort, not species distribution. Cannot establish population decline without repeated survey data. + +### IUCN Red List Range Maps + +Use this source for expert-drawn species range polygons to intersect with deforestation data. Range maps are extent-of-occurrence polygons that overestimate actual occupancy. Static snapshots may not reflect recent range contractions. Intersection with 30 m deforestation data implies false precision at range edges. Biased toward vertebrates; many Amazon species are not assessed. + +### MapBiomas Land Cover + +Use this source for land cover classification and deforestation-to-land-use transition analysis. Brazil-only coverage. 30 m resolution with 25+ classes and a transition matrix. ML classification can produce temporal inconsistencies. Accuracy varies by biome and year. Does not attribute causal drivers. + +### Hansen et al. 2013 Global Forest Change + +Use this source as the global 30 m forest change baseline. Measures tree cover loss, not deforestation. Plantation harvest, fire, and natural dieback all counted as loss. Forest definition threshold (canopy cover percentage) affects results. Annual updates available via Global Forest Watch. + +### Global Forest Watch Platform + +Use this source for integrated access to multiple forest monitoring datasets. Aggregates Hansen loss maps, GLAD alerts, GLAD-S2 alerts, and fire alerts with different definitions, resolutions, and update frequencies. These products must not be conflated in analysis. Does not provide land cover classification. + ## Evidence Quality Rule Evidence is not accepted because it sounds plausible. It is accepted when the source, method, limitations, and confidence are explicit enough for a reviewer to attack.