Replication package for "When decoupling is a series break: auditing modeled dairy methane in Minas Gerais, Brazil"

Reproducible package for the article "When decoupling is a series break: auditing modeled dairy methane in Minas Gerais, Brazil". It audits a two-point municipal inventory of modeled dairy enteric methane (2010 vs 2021; balanced sample of 848 of the state's 853 municipalities) and tests it against the annual 2010–2024 series of the Municipal Livestock Survey (PPM) and the 2006 and 2017 Agricultural Censuses. Contents: the assembled 2010 and 2021 municipal snapshots from IBGE SIDRA (retrieved 25 August 2026); the annual 2010–2024 series of cows milked (SIDRA table 94) and milk output (table 74) for all 853 municipalities (retrieved 24 September 2026); the analysis pipeline, configuration, unit tests and the complete generated build, including a SHA-256 manifest of inputs and outputs. Emissions are modeled with Equation 3 of Congio et al. (2022), with Equations 8 and 13 as confined and grazing sensitivity scenarios, GWP100 = 27.0 (IPCC AR6) and 365, 335 and 305 days of basal exposure. The build reports the reduced form of the aggregate, an exact Fisher decomposition with a herd-size/composition split, the complete Tapio taxonomy with class-stability tables, and the series-comparability test (src/annual_series.py): annual state totals, sub-period decompositions around the 2016→2017 step, the municipal timing of the largest annual decline, and Figure 4. The graphical abstract is drawn by a deterministic script from the build outputs. Version 2 adds the herd-size and composition split, the reduced-form terms, the exposure sensitivities, the class-stability table, the annual 2010–2024 series and the series-comparability test. Version 1 corresponds to an earlier manuscript version based on the two-point comparison only. Code: MIT licence. Data derived from IBGE public statistics: CC BY 4.0 (see LICENSE.md).