Identifying and correcting the World War 2 warm anomaly in sea surface temperature measurements

This is a Preprint and has not been peer reviewed. This is version 3 of this Preprint.

Add a Comment

You must log in to post a comment.


There are no comments or no comments have been made public for this article.


Download Preprint


Duo Chan, Peter Huybers


Most foregoing estimates of historical sea surface temperature (SST) feature warmer global-average SSTs during World War 2 well in excess of climate-model predictions. This warm anomaly, referred to as the WW2WA, was hypothesized to arise from incomplete corrections of biases associated with rapid changes in measurement instruments and protocols. Using linear mixed-effects methods we confirm highly significant offsets among specific groups of bucket and engine-room-intake SST measurements that, upon correction, reduce the WW2WA by 0.26°C (95% c.i. 0.15 to 0.38°C). Furthermore, SST measurements during WW2 coming from buckets are reportedly warmer at night than day, and controlling for this evident bias reduces the WW2WA by another 0.05°C (0.02 to 0.08°C). Adjusted SSTs give a more stable and smoothly evolving record of historical warming with a WW2WA of 0.09°C (-0.01 to 0.18°C) that is consistent with internal variability in climate models.



Climate, Oceanography, Oceanography and Atmospheric Sciences and Meteorology, Physical Sciences and Mathematics


Sea surface temperature, Bias, Correction, Model-data discrepancy, World War II


Published: 2020-08-20 14:15

Older Versions

GNU Lesser General Public License (LGPL) 2.1

Additional Metadata

Data Availability (Reason not available):
All datasets used in this study are publicly available as follows: ICOADS3.0 ( ds548.0/), HadSST2 (, HadSST3 and a 100-member ensemble ( data/download.html), HadSST4 and a 200-member ensemble ( uk/hadobs/hadsst4/data/download.html), ERSST4 ( data.noaa.ersst.v4.html), and ERSST5 ( data.noaa.ersst.v5.html). Groupwise adjusted SSTs and codes required to reproduce key results in this manuscript are available from the authors upon request and will be posted on Harvard Dataverse and Github upon publication.