What a temperature anomaly means

A graph shows a positive temperature anomaly, prompting a reasonable question: How can there be an anomaly if it was cold that day? The answer lies in what is being measured, where the average is taken, and what reference it is compared against. An anomaly describes a difference, not the temperature a person would feel when stepping outside.

A subtraction with qualifiers

The basic operation is observed temperature minus reference temperature for a comparable period. If we use a fictional example in which a place has an April average of 14 °C and the April being studied reaches 15 °C, its anomaly is +1 °C. We have not said that every day had that difference or that all nearby places behaved the same way.

The Copernicus Climate Change Service defines an anomaly this way. The qualifiers are essential: variable, region, data period, and reference period. Comparing a month with an annual reference would mix different questions. Comparing sea-surface temperatures with air temperatures also requires knowing how the dataset is defined.

Why the reference period changes

A series can be expressed relative to a historical reference or a more recent one. Changing the reference shifts zero. If applied consistently to the same series and periods, it does not erase the temporal evolution of the values. The Copernicus questions and answers explain the role of these periods.

Imagine another invented series: two years have average temperatures of 12 and 13 °C. Relative to a 10 °C reference, their anomalies would be +2 and +3 °C; relative to 11 °C, they would be +1 and +2 °C. In both cases, the difference between the years remains one degree. This helps explain a shift in the axis without attributing it to a change in the observations.

Converting between real reference periods requires more care than this example. The variable, spatial coverage, and averaging method must be preserved. Adjustments used by a particular product should not be applied automatically to any other graph.

A global average needs a map

Global temperature is not obtained by choosing a thermometer representative of the planet. Observations and estimation methods with geographic coverage are combined. A grid makes it easier to organize the information and weight it by area. If we averaged stations without considering their distribution, a region with many instruments could receive a disproportionate weight.

The NOAA global anomalies tool describes the use of land and ocean data. Other products, such as reanalyses, integrate observations using physical models. Sharing a word in the title does not mean all datasets measure exactly the same thing or treat areas without observations in the same way.

A year, a decade, and a street

Temporal averages answer different questions. A monthly anomaly can show variations that become smoothed when calculating an annual average. An average over several years makes it possible to examine a longer timescale. The Copernicus temperature indicators show these differences in scale.

A cold afternoon does not determine how we interpret a global trend, just as a warm afternoon is not enough to calculate one. It would also be incorrect to take a global figure and announce that all municipalities have warmed by exactly that amount. Averages summarize distributions with regional and temporal differences.

Reading the graph responsibly

Before sharing a screenshot, it is useful to retain the full title, legend, reference period, and link to the original product. Then we can check what the colors represent, whether the curve corresponds to land or land and ocean, and whether the latest datum is preliminary. A truncated scale can change the visual impression without changing the numbers.

Uncertainty does not mean that any interpretation is equally valid. It indicates specific limits associated with observations and methods. This guide explains how to read those limits; it does not provide a current temperature figure or an annual record. For that, we must consult the dated publication, identify its version, and retain the definitions that accompany the data.

A real example of a definition: GISTEMP

The NASA GISS frequently asked questions specify that GISTEMP uses 1951–1980 as its reference. That choice belongs to this product; it is not a universal convention that we can impose on every series. NASA also explains why it constructs regional averages from station anomalies rather than presenting a supposed absolute global temperature obtained through a simple subtraction.

This nuance helps us better read two graphs that appear to disagree: before discussing their curves, let us check whether they share the same variable, coverage, and reference. A difference in zero is not by itself a difference in trend. Nor should we assume that every discrepancy is due to zero: different observations and methods may be involved. Comparison requires preserving the product definition, not choosing the graph that best fits our impression.