Project Sankofa — Aframso Enclave Intelligence Programme

Proceed to G1 needs assessment within the three CEO-set boundaries: wider-enclave scope with farm/block drill-down; cassava primacy treated as a hypothesis to scientifically validate; purely desktop method (no soil lab or fieldwork spend). Partner/EOSDA/naming questions logged as elicitations, to be resolved during needs assessment.

Proceed with Option C (in-house desktop programme building permanent capability by delivering the Study). Before G2 promotion: resolve the EOSDA elicitation (paid data augmentation yes/no), the partner elicitation (KNUST/CSIR/Ghana Met vs in-house first), and the programme name; and clear the Phase 0 blockers (NDVI provenance trace, enclave boundary polygon definition). Fieldwork remains a WONT (BR-008) reversible only by explicit CEO gate decision.

Promote to Projects for delivery per the charter: phases P0-P4, desktop-only constraint (fieldwork WONT reversible only by CEO gate decision), dual-track NDVI fix (EOSDA free tier + in-house Sentinel-2, retire GEE), Sekyere Central (GH0239) boundary with farm/block drill-down, every finding confidence-rated, all layers Data Trust-registered, module contracts updated with every behavior change.

Recorded retroactively same-day: the P0-P4 phase milestone structure (created at G2 promotion) served as the de facto delivery management plan and baseline. No separate G3 planning pack was produced before execution began; this gate closes that governance gap for the audit trail rather than pretending it happened in sequence.

Accept Project Sankofa's delivered scope: Sekyere Central enclave boundary registered; 45-year observed climate baseline + 2050 CMIP6 projections (~+2C total warming, HIGH confidence); desk-researched human layer with an independent cassava/1982-83-drought cross-validation; EcoCrop-style crop suitability verdict (cassava+yam tie at 1.00, robust through 2050); 9 recommendations (R1-R9) and a community engagement plan. Accepted as a desktop-only Study with explicitly stated confidence levels and honest open items (within-season timing, block-level resolution, economics, field validation, forest-loss cross-check) -- not as a field-validated or economically modeled result.

Close Project Sankofa. Benefits review against the 6 benefits committed at G2: B1 (evidence-based block decisions) - partially achieved, decision-support now exists but block-level differentiation was found to have no real signal at this data resolution (honest limitation, not a failure). B2 (cassava hypothesis tested) - ACHIEVED, the core deliverable: cassava+yam verdict at 1.00 suitability, robust through 2050. B3 (honest data foundation) - ACHIEVED, NDVI pipeline fixed and live, all layers provenance-tracked. B4 (intelligence-to-action loop) - design ACHIEVED, implementation NOT built (BR-005 code remains open work, tracked separately, does not block closure). B5 (capability spillover to consultancy/EIRAS) - PARTIAL, positioned as R8 but not yet monetized. B6 (community positioning) - PARTIAL, engagement plan + 4 output specs delivered, production of materials is a downstream comms/HITL task not yet executed. Post-closure open items handed to standing operations, not re-opened as project scope: BR-005 implementation, community engagement material production, within-season timing refinement, block-level future assessment if ever needed, economic modeling, forest-loss satellite cross-check.

AUDIT FINDINGS (verified 2026-07-12, evidence inline): 1. SPATIAL BASE PACK IS REAL AND SUBSTANTIAL: ~11.5 GB on disk at data-trust/07-spatial-ghana-base across 18 categories (terrain 8.1G, landcover 1.0G, forest 785M, soils 704M, population 365M, infrastructure 301M, climate 238M, plus boundaries/hydro/wetlands/protected/hazard/governance/constraints/manifests). Built for Ghana-wide EIRAS screening resolution. Housekeeping drift noted: empty duplicate hazards/ dir beside hazard/. 2. GIS AGENT IS ALIVE AND PRODUCING DECISION-GRADE OUTPUT: gis-analysis agent runs every 3 days; verified readings 2026-06-28 through 2026-07-10 contain per-run weather (Open-Meteo), soil moisture at 3 depths, ET

OPTIONS CONSIDERED: A. Fix defects only (clipping/UI) - cheap but answers none of the climate/crop/human questions. B. Commission an external consultant study - one-off report, no standing capability, external spend, contradicts GAB's AIOPS identity (we sell exactly this kind of capability). C. In-house desktop programme that builds permanent capability by delivering the Study (RECOMMENDED) - honours the CEO's desktop-only constraint, compounds into consultancy/EIRAS capability. D. Option C plus EOSDA paid data augmentation - decide via the open EOSDA elicitation before G2; EOSDA is the main way to buy data depth without fieldwork. COSTS (Option C): agent/CEO time; compute/storage (base pac

BENEFITS PLAN B1 EVIDENCE-BASED FARM DECISIONS: any block decision (plant/fallow/switch/invest) answerable with referenced evidence in minutes from the OS. Measure: decision-support queries answerable per block; baseline today ~0 (NDVI estimated, no climate futures). Owner: CEO/Farm. Timing: from P1 onward, full at P3. B2 CASSAVA HYPOTHESIS SCIENTIFICALLY TESTED: sourced, confidence-rated statement of whether cassava primacy survives the 2050 climate trajectory, and what alternatives rank where. Measure: published suitability model + rankings. Timing: P3. B3 HONEST DATA FOUNDATION: zero conflation of estimated vs measured data in briefings; ndvi_source visible everywhere; every layer Data

PROJECT SANKOFA CHARTER (basis for G2) PURPOSE: Scientifically understand the Sekyere Central enclave (climate past/present/future, land, people) to evidence-base GAB's crop and investment decisions and prepare the community for climate change; build permanent GIS intelligence capability in GAB OS by delivering that Study. Principle: build the factory by making the first product. SCOPE IN: Sekyere Central District (GH0239) boundary with drill-down to farm/blocks; 30+yr climate baseline + downscaled projections to 2050 (2+ scenarios); desktop soil/terrain/water layers; crop suitability modelling testing the cassava-primacy hypothesis vs 4+ alternatives; desk-researched human layer (census/G