Civitas Systems

Specimen

What a deliverable looks like

Client work is confidential, so instead of a redacted excerpt, here is a specimen: a short methods briefing built entirely from public, published sources — my own ERL paper — in the exact format every client deliverable follows. The analysis is real; only the client is missing.

Civitas Systems · Methods briefing · Specimen (public sources only)

Can targeting a small set of power plants meaningfully cut electricity-sector CO2?

QuestionWhether plant-level targeting — rather than economy-wide instruments alone — can materially reduce global electricity-sector CO2 emissions. Short answerYes, and the size of the lever is quantified: emissions are so concentrated that instruments aimed at the worst few percent of plants address most of the sector. BasisGrant, Zelinka & Mitova (2021), Environmental Research Letters 16 094022 — peer-reviewed, open access; 2018 plant-level data. I was the data engineer on this work. ConfidenceHigh on the distributional finding; moderate on any individual plant's rank (see soft spots).

Findings

  1. Emissions are extremely concentrated. SourceGrant, Zelinka & Mitova (2021), resultsdoi.org.
  2. Concentration is national, not just global — and growing. SourceGrant, Zelinka & Mitova (2021), abstract & resultsdoi.org.
  3. The policy lever is large. SourceGrant, Zelinka & Mitova (2021), abstractdoi.org.

Method (what produced these numbers)

An updated edition of the CARMA plant-level inventory: official emissions registries (U.S. eGRID, the European Pollutant Emission Register, counterparts in Australia, Canada, and India) integrated with global plant- and company-level data. Where plants don't report, values were estimated with gradient-boosting models fit by ten-fold cross-validation on the reporting plants (3,019 observations with observed capacity factor; 2,581 with observed emission factor). My role: the data engineering — integrating 16 databases, geocoding 30,000+ facilities, and the gap-fill modeling.

Soft spots (stated, not buried)

  • Model-filled values are estimates, not measurements; the paper flags that where predicted emissions run close together, individual plants' relative rankings are less certain.
  • The headline findings are distributional — they hold across tens of thousands of plants and do not depend on any single plant's exact rank.
  • Data year is 2018; the 17–49% range spans three different policy instruments, not one estimate.

Provenance ledger

Claim in this briefingSourceKind
Top 5% of plants → 73% of emissions; 14.6× even distributionERL 16 094022, resultsPeer-reviewed, quoted on One number
75–89.6% in the most skewed national distributions; growth over timeERL 16 094022, abstract & resultsPeer-reviewed
17–49% eliminable depending on instrumentERL 16 094022, abstractPeer-reviewed
Data sources and GBM gap-fill detailsPaper's data & methods section; case studyPeer-reviewed + first-person

Every client deliverable — inventory, model, network analysis, or briefing like this one — ships in this format: a summary box a decision-maker can read in a minute, findings with their receipts attached, the method stated, the soft spots named, and a ledger that answers "where did this number come from?" before anyone has to ask.

Want one of these about your question? Start here, or read how an engagement runs.