Oceans warmer than we thought in the early 1900s, and it is likely due to buckets
Research supported by the EU-funded AI4PEX and XAIDA projects has brought to light errors in estimates of ocean temperatures from the first half of the 1900s. In their paper published in ‘Nature’, researchers suggest that correcting this bias would result in showing a more modest warming trend in the early 20th century. Global mean surface temperature (GMST) is an important indicator of climate change needed to guide climate policies such as the Paris Agreement. Calculated using a combination of sea surface temperatures with near-surface air temperatures over land, it provides valuable insight into climate variability and change. However, GMST records from the early 1900s might not be so reliable given how much measurement technologies and practices have changed since then.
Consistency throughout… almost
A research team led by first author Sebastian Sippel from AI4PEX and XAIDA project partner Leipzig University, Germany, reconstructed GMST from either ocean or land data. They found that while their reconstructions yielded a consistent picture of global temperature variability and long-term changes in most years, this was not the case for the estimates for the first half of the 20th century. From 1900 to 1930, ocean-based estimates were found to be about 0.26 °C colder than land-based values. “Some years in this period were notably cold owing to volcanic activity and internal variability. But such a prolonged cold period does not appear in land data, even with recent homogenization efforts leading to slightly cooler land baseline period temperatures, and with new exposure bias-correction methods,” the authors write, adding: “The pronounced ocean cold anomaly contradicts the physical theory of land–ocean warming patterns, and cannot be explained by internal variability or forcing.” The scientists conclude that the anomaly stems from uncorrected sea surface temperature biases, potentially caused by varying source data and unaccounted changes in measurement practices. In fact, before World War II, sea surface data was gathered by collecting water samples in ship buckets. “The transition within the early record from wooden to canvas buckets around the late nineteenth century, combined with the shifting patterns of shipping routes and shipping fleets, complicates systematic bias adjustments and adds to their uncertainty,” the authors explain. The study proposes that, although estimates of global warming since the mid-19th century are not affected, correcting the ocean cold bias would result in a more modest warming trend in the early 20th century and a lower estimate of variability over decades. It would also lead to greater agreement between simulated and observed warming than current datasets suggest. The XAIDA (EXTREME EVENTS: ARTIFICIAL INTELLIGENCE FOR DETECTION AND ATTRIBUTION) project ends in 2025. AI4PEX (Artificial Intelligence and Machine Learning for Enhanced Representation of Processes and Extremes in Earth System Models) ends in 2028. For more information, please see: AI4PEX project website XAIDA project website
Keywords
AI4PEX, XAIDA, global warming, ocean, land, temperature, global mean surface temperature