Outline: The final-quarter hurdle for a new temperature record

The prose will render this outline. If the prose changes the argument, revise and review the outline first. Stage: outline only; independent approval pending. No charts or prose have been made.

Intended location: Factory stories/2026-temperature-record-outline.md, stories/2026-temperature-record-src/, later stories/2026-temperature-record-make-charts.mjs, SVGs and stories/2026-temperature-record.md. Intended DataHub blog slug: 2026-temperature-record-hurdle; draft/preview only. Source snapshots belong to this story, not a new dataset repository. Inventory and publications checks found climate/CO₂ stories but no competing annual-temperature-hurdle story.

The argument, in one sentence

On either equal-month or calendar-day weighting of C3S's monthly anomalies, 2026 would need an October–December mean about 0.17–0.18°C above the warmest previous final quarter to overtake 2024 in this ERA5 monthly record; repeating that record quarter would still leave the annual anomaly slightly lower.

What this story is about

A concrete answer to “What would the last quarter have to deliver for a new annual temperature record?” The timely hook is the C3S September bulletin released in October, which says 2026 is second for January–September and a repeat of late 2023 would bring it close to 2024. This story quantifies the hurdle and “close”, with a historical comparator the reader can inspect. Audience: readers interpreting temperature-record headlines, climate journalists and educators. The contribution is conditional arithmetic and scale, not a prediction of the year-end winner.

The observations are global monthly surface air temperature anomalies from ERA5, with each month compared with its own 1991–2020 climatology. All chart calculations stay in that baseline. A quarter is October–December, not the September–November meteorological season. The series is not local weather, sea surface temperature, or a count of days crossing a threshold. An annual ranking is not the Paris Agreement's long-term warming measure. The analysis tests no cause of the warmth and derives no probability of exceeding the hurdle.

Reader questions, recorded before the outline

What would the remaining months have to average? Answer: the calculated tie hurdle is about 1.02–1.03°C above 1991–2020; to overtake, exceed the unrounded hurdle.
Has a final quarter ever been that warm? Answer: no in this 1940–2025 final-quarter record; the highest was 2023 at 0.85°C.
How is 2026 doing so far? Answer: January–September is second under both averaging choices, behind 2024; means are about 0.62°C versus 0.70–0.71°C.
Would simply repeating the record quarter suffice? Answer: no; the resulting annual anomaly is about 0.67–0.68°C, versus 0.72°C for 2024, a gap of less than 0.05°C.
Does a hurdle above past quarters make a new record impossible? Answer: no. History supplies a comparison, not a physical upper bound or a forecast.
Why could late-year warmth matter? Answer: three months supply roughly a quarter of the annual average; a year-to-date lead or deficit can change. C3S also reports developing El Niño conditions; that is attributed outside context, not an explanation inferred from these rows.
How does this relate to the widely quoted pre-industrial figures? Answer: they use another baseline; no chart mixes it with 1991–2020. Explain the distinction briefly, without adding an unnecessary second numerical scale.
How dependable is the exact hurdle? Answer: conclusion survives both weighting choices and source rounding. No measurement uncertainty is supplied here; the range denotes two calculation conventions, not a confidence interval.

Argument, in order

Finding / headline chart. With nine months observed, the calculated October–December tie hurdle is 1.0303667°C under equal-month weighting or 1.0237232°C under calendar-day weighting. Display 1.02–1.03°C and state that this is the range across the two conventions. The hottest previous October–December is 2023: 0.8497333°C / 0.8496859°C, both displayed as 0.85°C. The hurdle exceeds the historical maximum by 0.1806333°C / 0.1740374°C; display 0.17–0.18°C. This answers the opening question; it does not estimate the likelihood of reaching it.
What it is. C3S/ECMWF's globally averaged ERA5 monthly surface air anomaly series, January 1940–September 2026. Use the supplied anomaly, not absolute 2t, for the calculation. A +1°C anomaly means a global mean one degree above the same months' 1991–2020 average; it is not a statement that the year's coldest months must have the highest absolute temperatures.
Why the hurdle is high / second chart. January–September 2026 is second under both conventions: 0.6160667°C / 0.6167945°C; 2024 is first at 0.7055000°C / 0.7049015°C. The remaining quarter must offset that accumulated deficit as well as match 2024's annual target. Show the cumulative paths and September endpoints so the reader can see that recent warmth is only part of the annual calculation.
Wrinkle and counterevidence. A literal repeat of October–December 2023 gives 2026 an annual anomaly of 0.6744833°C / 0.6754959°C, compared with 2024's 0.7196417°C / 0.7193628°C: less than 0.05°C behind. C3S's phrase “close to” is consistent with this result. 2026 is already second through September, not a cool year. A past maximum is not a ceiling: the bulletin reports strengthening El Niño, so the story must not recast historical comparison as impossibility or a forecast. The month-weighting uncertainty is explicit next to the finding, not hidden at the end.
Method last. Link the frozen input, source bulletin, provenance and executable calculation. Explain both weightings in a sentence; dates through September only. Treat the figures as conditional calculations on this vintage of the monthly anomaly series. No claim that the two conventions exhaust every detail of official annual classification. The human voice pass remains outstanding.

Calculation and checks

Reproducible statistics, calculation script, provenance.

Let A24 be the 2024 full-year anomaly under the chosen convention and S26 the January–September 2026 anomaly under the same convention. The tie hurdle is:

Equal months: Q = (12*A24 - 9*S26)/3.
Calendar days: Q = (365*A24 - 273*S26)/92. The 2024 target itself uses its actual 366-day calendar; 2026 uses 365. Compare the resulting Q with day-weighted historical quarters. Both are anomaly means under their stated conventions, not averages of absolute temperatures.
Overtake means Q4_2026 > Q; equality ties this calculation's target. Never say the rounded label is an exact cutoff.
Repeating 2023 Q4 substitutes its three anomalies into October–December 2026, retaining the observed 2026 first nine months. Their October–December day counts coincide.

Executed checks: 1,041 unique, consecutive monthly rows, January 1940–September 2026; 86 complete historical final quarters, 1940–2025; all five numerical fields finite. The 1991–2020 climatology and pre-industrial offsets are fixed for each calendar month. ano_pi = ano_91-20 + offset_pi holds exactly; 2t - clim_91-20 = ano_91-20 differs by at most 0.0001°C, consistent with the individually rounded source columns. Do not recompute anomalies from rounded absolute fields.

The source provides four decimal places, not an observational uncertainty estimate. Assuming nearest-0.0001 rounding, propagated input rounding of the tie threshold is at most 0.00035°C (equal month) or 0.00034674°C (day weighted); the historical-quarter mean's input rounding is at most 0.00005°C. Even the lower hurdle minus its rounding allowance exceeds the higher record-quarter estimate plus its rounding allowance by more than 0.17°C. These are bounds on rounding only. No statistical/physical uncertainty claim is made. Recent reanalysis data can be revised; the frozen vintage is identified.

The annual pre-industrial conversion uses a fixed 0.88°C offset in C3S's method; monthly offsets vary. The story deliberately avoids numerical pre-industrial conversion, eliminating a second scale that could confuse the comparison. The statistics.json audit includes explicitly named annual conversions for reviewers only, not planned prose claims. Do not average the monthly ano_pi values and call that the annual pre-industrial result.

Outside context

C3S September bulletin, “Year to date and outlook”: the January–September ranking is only a partial indication; C3S reports strengthening El Niño and says a repeat of late 2023 would put 2026 close to 2024. Attribute to C3S and keep any causal/forecast language within that attribution. The chart arithmetic does not test or reproduce an El Niño forecast. No long detour into ENSO is needed to answer this question.
C3S methods: global ERA5 surface air temperature is 2m air temperature over land and ocean; the full global ranking record begins in 1940. Actual temperatures carry more uncertainty than anomalies. The page specifies baseline conversion but does not settle temporal weighting for this calculation.
Hook dating: news page displays 12 October 2026 with a 9 October dateline. Describe this as the September bulletin released in October. Do not silently select one of the two dates or imply October temperatures have been observed.

Chart plan

Both charts use Observable Plot rendered to static SVG. Every annotation is generated from checked rows or reproducible derived statistics. Direct labels replace lookup legends. Show source and baseline on the charts. Ranges below denote the two temporal averaging conventions; never draw them as a physical uncertainty band.

#	Chart	Data (resource + fields)	Transform, gaps, dates	Purpose
1	“2026's final-quarter hurdle is above every previous final quarter”. Historical October–December anomaly against year; horizontal tie-hurdle strip above the history.	Frozen `global-monthly.csv`: `month`, `ano_91-20`; `statistics.json`: both `modes.*.q4_history`, `q4_tie_threshold`, `q4_record`, `threshold_minus_q4_record`.	Use all 86 complete Q4s, 1940–2025; validate exactly October, November, December per year. Show paired equal-month/day-weighted values as a small vertical segment at each year (most overlap); label the 2023 maximum 0.85°C. Add two horizontal rules at exact tie hurdles, joined only by a pale strip labelled “Tie hurdle: 1.02–1.03°C; averaging choices, not uncertainty”. Annotate “0.17–0.18°C above the previous maximum” and “Hotter to overtake”. Date by calendar year and say October–December. No 2026 actual Q4 point, no interpolation beyond observed history, no missing as zero. Y-domain must include all negative historical anomalies, zero, both hurdle rules and labels (roughly −1.1 to 1.25°C; derive minimum). The strip is a condition, visually distinct from observations; no forecast line.	Answer the opening question and place its scale in the complete historical final-quarter record. Show the same conclusion under both conventions.
2	“Repeating late 2023 would bring 2026 close to 2024”. Cumulative January-to-month anomaly for 2024 and 2026, followed by the explicitly hypothetical repeat-2023 finish.	Same source fields; `statistics.json`: both `modes.*.cumulative_recent` filtered to 2024/2026; `scenario_cumulative_2026_repeat_2023_q4`; year-to-date ranks and annual targets.	X=calendar month1–12. Each actual cumulative point uses January through that month; 2024 has12 actual points, 2026 has9. Draw equal-month paths as solid2024 neutral/2026 blue, with direct labels; add a very light envelope spanning the day-weighted alternatives, labelled in caption as sensitivity, not statistical uncertainty. Connect the Sep2026 point to the three hypothetical Oct–Dec cumulative means with a dashed blue path labelled “If Oct–Dec repeats 2023”. Mark October–December as hypothetical with light background; actual line ends in September. Label September2026 0.62°C; second Jan–Sep, September2024 0.70–0.71°C; first Jan–Sep; December2024 0.72°C, hypothetical December2026 0.67–0.68°C, endpoint gap “less than 0.05°C”. Include zero; upper domain at least maximum cumulative value+annotation room. January1 monthly keys represent calendar months, not daily observations. No invented Oct–Dec2026 actuals. Do not display2023/2025 paths merely because the calculation JSON includes them.	Show why the recent monthly rise does not settle the annual ranking, and present the strongest weakening fact: repeating the historical record quarter would put the annual result close to2024.

Review request and gate

Independent reviewer must first record five to eight reader questions without reading this outline; then inspect interest, source rights and coverage, arithmetic under both conventions, chart feasibility/readability, counterevidence, and whether the proposed writing earns the reader's time. Reproduce the numbers from CSV with a separate scan rather than importing this script's result. Return APPROVED or numbered, line-referenced corrections.

The coordinator will record the exact committed revision and outline SHA-256 with the reviewer identity, reproduced numbers and verdict. Approval is required before charts or prose. Current staging work is uncommitted and is not approved. If the weight caveat prevents a clear accurate story, narrow the headline to the transparent anomaly calculation rather than imply an official forecast or ranking.

Voice

Follow the DataPressr voice guide: finding first, short factual sentences, no rhetorical question, no record-as-doom language. Human author's voice pass is a later, outstanding stage.

Friction notes

A calculation can be numerically correct while importing an unstated aggregation convention. Here both conventional choices are disclosed and tested; neither is attributed to C3S without evidence. The next stage must preserve this distinction visibly.
Source monthly pre-industrial offsets and the annual offset are not interchangeable. Staying in 1991–2020 avoids a needless conversion and keeps every historical comparison on one basis.
A threshold outside historical experience is not an impossibility claim. The strongest counterpoint is included: a repeat of the record quarter leaves a small annual gap.
Proposed two-chart design needs an actual mobile-width check; the second chart has several endpoint labels. If they collide, use a small two-row endpoint annotation below the plot rather than dropping the convention range or counterevidence.
No prose exemptions are decided yet. Counts and methodological constants above are audit material, not automatic prose content. Every data number used in prose must be on the charts or explicitly audited under the story skill's exemption rules.