Transparent Cross-Sector Emissions Scenarios Under Parameter Uncertainty

Small organisations and researchers need emissions tools that are more inspectable than large integrated assessment systems and more rigorous than opaque spreadsheet totals. This preprint presents an open Python prototype that represents greenhouse-gas estimates as traceable activity-by-emission-factor calculations, exposes scenario assumptions, propagates uncertainty, tests the stability of mitigation rankings, and reports explicit interpretation boundaries. The software includes deterministic scenario comparison, one-at-a-time sensitivity analysis, reproducible Monte Carlo uncertainty, a least-cost abatement optimiser, a synthetic physics-informed neural benchmark, and a held-out comparison against 320 US EPA eGRID natural-gas plants. The results support a limited methodological claim about transparent, provenance-first evaluation; they do not constitute contemporary European validation, forecasting accuracy, or realised emissions reduction. Educational research only; not a regulatory inventory or operational decision system.