Climatologists have long debated what caused the early 20th-century warming: natural variability or external forcing. Some models point to solar activity and reduced volcanic dust, others to greenhouse gases and ocean currents. This work stands apart: nearby in the archive are studies of weather on a lone planet, El Niño predictions, and Antarctic ice memory, but there are almost no direct topical neighbors. This is a measurement, not an opinion: the nearest neighbor is at distance 0.514, and there are no works in immediate proximity. Nevertheless, the thematic group of the map indicates a connection to gravitational waves and baryon acoustic oscillations, highlighting the interdisciplinary nature of climate data.
The authors took land temperature anomaly data from NOAA and NASA GISTEMP for 1880–2017. They divided the globe into six continents, seven latitude zones (excluding Antarctica), and eight seasons. For each dataset, they applied principal component analysis to extract the main modes of variability. Then, using the reverse arrangement test, they checked when the warming trend becomes statistically significant. Additionally, they constructed cumulative anomaly profiles and found the points of greatest change by minimizing the sum of variances on both sides of the point.
The first principal component for continents explains 75.2% of the variance, for zones 81.7%, and for seasons 87.6%. It shows the overall course: a rise until the 1940s, a decline until the late 1950s, a plateau until the mid-1970s, and a sharp rise afterward. The second component for continents is related to the North Atlantic Oscillation and describes short-period differences between North America and Eurasia. The third component, which includes the early 20th-century warming, has a quasi-period of 60–70 years and correlates with the Atlantic Multidecadal Oscillation (coefficient 0.411). For zones, the second component also correlates with the AMO (0.380), and the third with the Southern Oscillation Index (0.479). For seasons, the second component correlates with the AMO (0.557), and the third with the winter Niño 3.4 index (0.358). The reverse arrangement test showed that the early 20th-century warming is significant across all continents except Oceania, and in all Northern Hemisphere seasons, but in none of the Southern Hemisphere seasons. In South America, significant warming has continued since the 1930s without interruption, while in North America and Europe, modern warming became significant only around 1990.
This work differs from its archive neighbors in that it applies the reverse arrangement test to temperature series for the first time and combines it with principal component analysis and cumulative profiles. This reveals that the early 20th-century warming was not homogeneous but regionally and seasonally selective. Neighboring works, such as those on weather on a lone planet or El Niño prediction, address other aspects of climate but do not provide such a detailed breakdown by continents and seasons. Here, it is shown that different regions have different transition points to significant warming, which is important for climate change attribution.
The topic could develop toward a more detailed analysis of the causes of regional differences. For example, the same methods could be applied to climate model data to check whether they reproduce the observed patterns. It is also promising to study the connection of the identified components with other indices, such as the Pacific Decadal Oscillation or the North Atlantic Oscillation over longer intervals.
The results could influence climatology, meteorology, and oceanography, as they refine the spatiotemporal structure of 20th-century warming.
The next step could be to test how well climate models reproduce the identified differences between continents and seasons.
This work is directly related to the unresolved problem of attributing the early 20th-century warming: the role of natural variability versus external forcing remains unclear. Identifying regional and seasonal patterns provides new constraints for models. Furthermore, the connection to the Atlantic Multidecadal Oscillation raises the question of the predictability of climate fluctuations on decadal timescales.
The early 20th-century warming in the Northern Hemisphere was so pronounced that in the 1930s, the Arctic experienced a significant reduction in sea ice, comparable to modern levels.