From Methodology Narrative to Code: A Worked Example

Composting (CH₄ and N₂O), U.S. Greenhouse Gas Inventory

Published

June 26, 2026

What this is

This page shows a single source category — composting — compiled two ways at once: as a methodology narrative written in plain language, and as the R pipeline generated from it.

The narrative is the source of truth. A domain expert authors it, and it describes the methodology completely and precisely— data sources, factors, calculation steps, decision rules, and uncertainty — without any code. The R pipeline beneath it is a derived artifact: generated from the narrative and following a fixed set of project conventions (here, a set of instructions for coding in R). When the methodology changes, the narrative changes, and the code is regenerated. The methodology narrative persists and outlives the code.

This narrative was distilled from the composting source category’s published methodology. The pipeline was then generated from that narrative.

I selected composting because it’s a simple category: one activity driver (the mass of waste composted), two gases, and an IPCC Tier 1 method. That makes it a clean illustration of the workflow from end to end.

The code shown here demonstrates the narrative-to-code step. A production compilation adds a validation layer — which is part of the workflow but not shown in this example. Further, production code will also include various outputs: writing to CRT tables, Quarto reports, ggplot figures, etc., which are not included here.

The methodology narrative

This is the authored artifact — the source of truth. Everything in the code section that follows was generated from it.

Conforms to Joe’s Claude SKILL.md Methodology Narrative Standard v0.4. Emission-factor values are the 2006 IPCC defaults.

Identification

  • Category: Composting — CH₄ and N₂O
  • Sector: Waste (IPCC 4.B, Biological Treatment of Solid Waste)
  • Inventory edition: GHGIA 2026
  • Author: [Waste source lead]
  • Last revised: [date]

What this category covers

This category covers methane and nitrous oxide emitted at commercial composting facilities, where organic waste is composted to divert it from disposal in municipal solid waste landfills.

It does not cover: carbon dioxide from composting (biogenic, excluded from national totals by convention); emissions from the landfilling of waste that is not composted ([landfill category]); or home or non-commercial composting, which is not estimated.

Data sources

  • Quantity of waste composted, from EPA’s Advancing Sustainable Materials Management: Facts and Figures report (formerly MSW: Facts & Figures). Short label: epa_asmm_composted. Primary activity data, covering 1990–2018, reported on a wet (as-received) weight basis.
  • U.S. resident population, from the U.S. Census Bureau annual population estimates (U.S., regions, states, and Puerto Rico). Short label: census_population. Used to extrapolate waste composted beyond the last measured year.

Emission factors and other fixed values

  • IPCC default CH₄ emission factor for composting: 4 g CH₄ per kg waste (wet weight). Source: 2006 IPCC Guidelines, Vol. 5, Ch. 4, Table 4.1.
  • IPCC default N₂O emission factor for composting: 0.3 g N₂O per kg waste (wet weight). Source: 2006 IPCC Guidelines, Vol. 5, Ch. 4, Table 4.1.
  • Global warming potentials (AR5) for CH₄ and N₂O (dimensionless).

Both emission factors are on a wet-weight basis, matching the moisture basis of epa_asmm_composted; the two must agree, since a wet-weight factor applied to dry-weight activity data (or vice versa) is wrong by roughly a factor of two.

Units

Datasets:

  • epa_asmm_composted — native: thousand short tons; canonical: Gg.
  • census_population — native: persons; canonical: persons (no conversion).

Dimensioned constants:

  • CH₄ emission factor — native basis: g CH₄ / kg waste; canonical basis: Gg CH₄ / Gg waste.
  • N₂O emission factor — native basis: g N₂O / kg waste; canonical basis: Gg N₂O / Gg waste.

Both factors are normalized to their canonical bases at ingestion, alongside the activity data, so the calculation steps are unitless. Emissions are reported in Gg of CH₄ and N₂O, and as CO₂-equivalent for summary tables.

Connections to other categories

  • Upstream: none. Composting takes no inputs computed by other categories.
  • Downstream: category totals feed the Waste-sector and national CH₄ and N₂O summaries.

(Composting diverts waste that would otherwise be landfilled, but there is no computational handoff between the two categories — each is estimated from its own activity data.)

How the calculation works, step by step

  1. Establish the mass of waste composted for each year:
    • Measured years (through the last year ASMM provides, currently 2018): take the reported quantity directly from epa_asmm_composted.
    • Extrapolated years (after the last measured year): hold the waste-per-person ratio fixed at its value in the last measured year (last-measured-year mass ÷ that year’s census_population), and multiply it by each later year’s census_population to project that year’s mass composted. (The ratio is frozen at the last measured year, not recomputed from each prior extrapolated year — those are equivalent, but freezing is the clearer statement of intent.)
  2. Apply the CH₄ emission factor to the mass composted to estimate methane emissions.
  3. Apply the N₂O emission factor to the mass composted to estimate nitrous oxide emissions.
  4. Apply the AR5 global warming potentials to express the totals as CO₂-equivalent.

Rules and judgment calls

  • Activity-data source by year: waste composted is taken from epa_asmm_composted for measured years and population-extrapolated (step 1) for years after the last measured year. The boundary is “the last year ASMM reports” (currently 2018), not a fixed calendar year — it moves forward if a replacement activity-data source is adopted. This is the only conditional logic in the category.
  • Tier selection: the IPCC default Tier 1 method is used because facility-specific and nationwide data on composting methods and waste types/amounts are not publicly available. Tier 1 applies across the entire time series.

What this category produces (outputs)

  • CH₄ and N₂O emissions from composting, by year, totaled and arranged for the inventory database, with CO₂-equivalent applied via the AR5 GWPs.

Uncertainty

Uncertainty is estimated using IPCC Approach 1 (error propagation), as recommended for a Tier 1 method with default emission factors; a Monte Carlo approach may be adopted in a future edition. The two parameter uncertainties are combined via the IPCC error-propagation equation to a combined uncertainty of approximately ±58% on national emissions.

Parameter Distribution Range Basis
IPCC default emission factors (CH₄, N₂O) — ±50% IPCC 2006 default (Vol. 5, Ch. 4)
waste composted (activity data) — ±30% IPCC 2006 guidance

Known changes and things to watch

  • The ASMM report was discontinued; its final edition covers 2018 data. After the last measured year the mass composted is extrapolated from population growth (step 1) rather than measured, until a replacement activity-data source is identified.
  • The uncertainty method is currently the simple Approach 1 error propagation; a move to Monte Carlo is flagged as a possible future improvement.

References

  • GHGIA 2026, Waste chapter, Composting section.
  • 2006 IPCC Guidelines, Vol. 5, Ch. 4 (Biological Treatment of Solid Waste), Table 4.1; Vol. 1, Ch. 3 (uncertainty, error-propagation equation).
  • [ASMM and Census population data documentation].

The generated R code

Everything below was generated from the narrative above, following the project’s R conventions: a {targets} pipeline, {pins} for inter-stage handoff, {syrinx} for canonical-unit normalization at ingestion, tidyverse style, and unit logic confined to the ingestion and output boundaries so the calculation core is unitless.

list(

  ## Data Retrieval ----------------------------------------------------

  tar_target(composting_waste,            pin_read(board, "composting_waste")),
  tar_target(us_population,               pin_read(board, "us_population")),
  tar_target(composting_emission_factors, pin_read(board, "composting_emission_factors")),
  tar_target(gwp_ar5,                     pin_read(board, "gwp_ar5")),

  ## Transform Data ----------------------------------------------------

  tar_target(
    composting_activity,
    get_composting_activity(composting_waste, us_population)
  ),

  tar_target(
    composting_emissions,
    get_composting_emissions(composting_activity,
                             composting_emission_factors,
                             gwp_ar5)
  ),

  tar_target(
    composting_invdb,
    format_composting_invdb(composting_emissions)
  )
)
#' Mass of waste composted by year, in canonical units (Gg).
#'
#' Measured years come directly from ASMM. Years after the last measured year
#' are extrapolated by holding the waste-per-person ratio fixed at its
#' last-measured-year value and scaling by population (narrative step 1).
get_composting_activity <- function(composting_waste, us_population,
                                    last_measured_year = 2018) {

  pop <- dplyr::select(us_population, year, population)

  measured <- composting_waste |>
    dplyr::filter(year <= last_measured_year) |>
    dplyr::select(year, waste_composted)

  # Frozen waste-per-person ratio at the last measured year
  ratio <- measured |>
    dplyr::filter(year == last_measured_year) |>
    dplyr::left_join(pop, by = "year") |>
    dplyr::summarise(r = waste_composted / population) |>
    dplyr::pull(r)

  extrapolated <- pop |>
    dplyr::filter(year > last_measured_year) |>
    dplyr::mutate(waste_composted = ratio * population) |>
    dplyr::select(year, waste_composted)

  dplyr::bind_rows(measured, extrapolated) |>
    dplyr::arrange(year)
}
#' CH4 and N2O emissions from composting (Gg), plus CO2-equivalent.
#'
#' Activity (Gg waste) x emission factor (Gg gas / Gg waste) -> Gg gas.
#' Every input is canonical, so this calculation carries no units.
get_composting_emissions <- function(composting_activity,
                                     composting_emission_factors,
                                     gwp_ar5) {

  composting_activity |>
    tidyr::expand_grid(composting_emission_factors) |>   # year x {ch4, n2o}
    dplyr::mutate(emissions = waste_composted * ef) |>
    dplyr::left_join(gwp_ar5, by = "gas") |>
    dplyr::mutate(emissions_co2e = emissions * gwp) |>
    dplyr::select(year, gas, emissions, emissions_co2e)
}
#' Arrange composting emissions into the InvDB output structure.
#' Emissions are Gg internally; converted to Tg at the output boundary.
format_composting_invdb <- function(composting_emissions) {
  composting_emissions |>
    dplyr::mutate(
      year           = forcats::as_factor(year),
      category       = "composting",
      sector         = "waste",
      emissions_tg   = emissions / 1e3,        # Gg -> Tg, output boundary
      emissions_co2e = emissions_co2e / 1e3    # Gg CO2e -> Tg CO2e (MMT)
    ) |>
    dplyr::select(category, sector, year, gas, emissions_tg, emissions_co2e) |>
    dplyr::arrange(gas, year)
  ## InvDB
}

Each data source has a paired metadata file and ingestion script. Unit normalization happens here, at the ingestion boundary — including the emission factors, whose g / kg basis is scale-sensitive and so is converted to the canonical Gg / Gg basis rather than carried as a dimensionless constant.

Activity data

# composting_waste_metadata.yml
dataset_id:    composting_waste
source:        "EPA Advancing Sustainable Materials Management: Facts and Figures"
native_unit:   thousand_short_tons
canonical_unit: Gg
moisture_basis: wet
time_coverage: "1990-2018"
# composting_waste_ingest.R
# thousand short tons (wet) -> Gg (canonical), long by year.
composting_waste_ingest <- function(path, metadata) {
  readr::read_csv(path, show_col_types = FALSE) |>
    pre_clean() |>
    apply_metadata_labels(metadata) |>
    to_canonical_all() |>                 # waste_composted: thousand_short_tons -> Gg
    dplyr::transmute(year = as.integer(year), waste_composted)
}

Emission factors

# composting_emission_factors_metadata.yml
dataset_id:    composting_emission_factors
source:        "2006 IPCC Guidelines Vol. 5 Ch. 4 Table 4.1"
moisture_basis: wet
factors:
  ch4: { value: 4.0, native_basis: g_per_kg, canonical_basis: Gg_per_Gg }
  n2o: { value: 0.3, native_basis: g_per_kg, canonical_basis: Gg_per_Gg }
# composting_emission_factors_ingest.R
# g emission / kg waste  ->  Gg emission / Gg waste.
# g/kg is scale-SENSITIVE (different unit top and bottom), so it is normalized
# here at ingestion, NOT carried as a dimensionless constant (standard v0.4 §5.1).
composting_emission_factors_ingest <- function(metadata) {
  tibble::tibble(
    gas       = c("ch4", "n2o"),
    ef_native = c(4.0, 0.3)
  ) |>
    dplyr::mutate(
      # (1e-9 Gg/g) / (1e-6 Gg/kg) = 1e-3
      ef = ef_native * 1e-3
    ) |>
    dplyr::select(gas, ef)
}