Sea temperature data, some collected by a fleet of drifting and diving probes, shows that a decade-long slowdown in global surface warming masked a coincident rise in ocean temperature below 300 feet, according to NASA JPL researchers.
In the JPL study, climate scientists Veronica Nieves, Josh Willis and William Patzert pored over sea temperature data dating back two decades.
A large portion of this information was collected by the Argo array -- a network of more than 3,000 automated sea probes that can dive deeper than a mile, take temperature and salinity readings, then return to the surface where they transmit the information to orbiting satellites.
"They're the oceans' weather balloons," Willis said of the devices.
What researchers discovered was that during the period of the hiatus -- from roughly 2003 to 2013 -- sea surface temperatures in the Pacific and Indian oceans rose more slowly than they had in previous years.
However, heat was actually accumulating in a layer of water just below the surface, in an area between 300 and 1,000 feet deep.
This layer of warming showed that even though the rise in global average surface temperature had slowed, the ocean continued to absorb heat generated by greenhouse gasses, authors said.
"Basically what happened is the heat missing from the surface went to a subsurface layer in the Pacific and Indian oceans," said Nieves, the lead study author.
The precise mechanism by which the oceans trapped heat in this middle layer of water remains unclear, but researchers say it occurs on a decadal timescale.
Patzert said that this subsurface heating gradually "piled up" in the western Pacific Ocean, then "leaked" into the Indian Ocean.
The idea that the Pacific Ocean absorbed heat that would have otherwise led to a rise in global average surface temperature is not new.
However, the JPL study is the first to use only direct observations to describe the phenomenon and specify a precise area of warming.
Lead author Nieves said that previous studies have relied on model-based data, or a combination of models and observations, to frame their conclusions.
"This can lead to completely different results," Nieves said.