How do you know?
Keep a reported measurement, its context, and its interpretation in separate columns. A single interval can be worth examining without carrying a causal conclusion by itself.
Record an observation
Use the supplied excerpt as a claim to inspect, not as a verdict.
Evidence readout
A careful summary of what this record can support.
−0.50°C over 1940–1970
In the Northern Hemisphere, as described by the supplied excerpt.
Keep reported observations distinct from causal conclusions
Read the explanation
The saved example records a reported negative half degree Celsius change. A hypothetical negative one degree input doubles its magnitude from half a degree to one degree. Bars share four hundred pixels per degree of absolute magnitude; the displayed minus sign still records the cooling direction. The tool formats two decimal places but does not fetch temperature records, estimate uncertainty or validate that the supplied historical observation is accurate. Three region choices and three carbon dioxide description choices permit nine context combinations. Bars use fifty pixels per count, showing three region choices and nine pairs. These options are metadata, not nine measurements or causal controls. Changing the carbon dioxide description is preserved in export but does not change the displayed temperature finding. The interval is user text; the tool does not compute an annual trend or infer climate causation. Four input fields are retained locally: temperature change, period, region and carbon dioxide description. The export contains three top-level fields: observation, interpretation and note, with the four inputs nested in observation. Bars share one hundred twenty pixels per field. Reset restores the original supplied values and stores them locally. A record and its uncertainty boundary should travel together; neither the stored observation nor a JSON export proves a causal explanation.