Climate state

Realized warming. Total long-term warming includes anthropogenic influences and natural variability; it is distinct from an individual year’s observed temperature. Anthropogenic warming estimates the human-induced contribution.

Historical benchmark. Crossing years use the realized observational centered 20-year mean. Its rate uses centered 10-year means around the crossing. These blue benchmarks also appear on anthropogenic views as an observational comparison. A centered 20-year benchmark is available through 2015.

Seven late crossings. The preindustrial 1.0°C benchmark includes 93 observed crossings and seven estimates from quartic fits to the last 25 available centered 10-year means. These estimates are identified in the downloads; their model uncertainty is not included in the reported ranges.

Method comparisons. Red segments extend 15 years backward from the selected method’s mean first crossing. Their slopes are endpoint derivatives of quadratic fits to 15 past annual smoothed outputs. Crossing ranges condition on the members that cross.

Display period. Temperature series start in 1965. Earlier rows remain in the downloads. Smoothed anthropogenic estimates before 1965 require special caution: FaIR applies a prescribed early-period mean and contracts its current samples, and GWI is unavailable before 1950. The pinch is a feature of those method outputs, not evidence of unusually small physical uncertainty. This warning does not invalidate the annual observational ensemble or other realised methods.

2025 estimates. Levels use current method outputs. Rates use corrected retrospective outputs for 2011–2025; GWI uses its current outputs as an explicit temporary retrospective proxy. The red 15-year line is anchored at the current level and fitted rate. Level ranges combine the members’ calibrated Gaussian distributions; rate ranges include fit and member variability.

Probabilities. The three assessment windows use calibrated current probabilities already saved by the paper pipeline. Outside those windows, and for the fourth threshold, the display uses Gaussian approximations with the same calibrated means and uncertainties. Rate switches flatten their coloured segments without changing the reported rates or underlying data. Short red downward and blue upward ticks mark the respective crossing years. Hover inside a grey box, away from plotted lines, for its benchmark crossing-range explanation.

The bottom curves use the paper’s cumulative-maximum and one-year Gaussian smoothing convention, with annual probability increments located in the preceding year-long interval; they are density proxies, not a joint first-passage model.

Satellite calibration. Values are anomalies relative to 1981–2010. The reference thresholds are shifted by 0.7°C, following the source pipeline. Both the calibration and its threshold labels change together.

Background lines. The Temperature series menu switches only the grey/black curves between annual observations, selected-method smoothed outputs, and observational centred 20-year means. The latter are retrospective and currently end in 2015; on anthropogenic views they still represent realised observations. In selected-method smoothed mode, hover or tap a member (or choose its number) to show one pale-yellow approximate 95% calibrated uncertainty ribbon. It uses ±1.96 calibrated standard errors for that member and excludes the spread across ensemble members. Tap again or use Clear selection to dismiss it. Numerical estimates, benchmarks, and probabilities do not change.

Distribution scales. All three panels default to 0–0.3 probability/year. Bars exceeding the ceiling have // labels with their actual heights; hover values retain the full fraction. The magnify switch gives the rightmost threshold an explicitly labeled finer y-axis and does not renormalize its probabilities.

Rates over time. Open the optional panel from the warming-rate card. The blue line uses the same centered-mean OLS benchmark as the crossing segments and ends in 2015 because its centered windows require ten subsequent years of observations. The red line defaults to retrospective, trailing 15-year quadratic endpoint rates; these are revised using the full observation record, not forecasts made at the time. An optional faint dotted red line shows rates fitted to annual current outputs. The retrospective red line remains visible. Both series and both method conventions are included in the rate-history CSV. The panel is interactive only and loads on demand.

Browser interaction. Charts use the exported Python results, with zooming and a data cursor in the browser. Python calculations run locally; they are not rerun by the browser.