These temperatures were 17.5C on March 24, 2015 and 18.3C, on February 6, 2020, at the Esperanza station on the peninsula.
Clem said it had generally been assumed that ozone depletion, and the ozone hole, had been responsible for strengthening the westerly winds and enhancing warming on the eastern side of the peninsula, leading to the collapse of the Larsen A and B ice shelves further north.
"We have shown that may not necessarily be the case and that there is another mechanism which may help us predict what might happen with the Larsen C ice shelf."
Clem said the research suggested Larsen C's future, and the global effects of sea-level rise that could occur after ice-shelf collapse, hinge on the variability of convection in the tropics, itself determined by other factors affected by climate change.
"Surface melt can have a devastating effect on ice shelves. Intense surface melting and ponding can cause ice shelves to collapse through a process called hydrofracture, in which the surface melt water flows downward through small cracks and fractures, refreezes, and expands, mechanically breaking apart the ice shelf over a period of weeks."
The Larsen Ice Shelf, on the eastern peninsula, has experienced a dramatic series of collapses since the mid-1990s due to this process, including the most northern section Larsen A in 1995, and the larger Larsen B section to its south in 2002.
Farther south, the largest remaining section, the Larsen C ice shelf, was now thinning.
Larsen C was Antarctica's fourth largest remaining ice shelf, spanning about 44,200 km2.
The loss of these ice shelves in recent decades had triggered rapid thinning and acceleration of the glaciers that once fed into them and has resulted in an accelerated pace of sea-level rise contribution from the Antarctic Peninsula, Clem said.
"In our study, we show for the first time the pattern, which results from the convection near Fiji, leads to a large and deep area of low pressure off the coast of West Antarctica and a strong high-pressure system north of the peninsula over Drake Passage.
"Together, these features transport very warm and moist air from middle and sub-tropical latitudes of the eastern South Pacific to the Antarctic Peninsula in the form of intense atmospheric rivers.
"We find that variability in deep convective activity over the central tropical Pacific region explains 40 per cent of the year-to-year variability in total summertime Larsen C surface melt and 50 per cent of the variability in the total number of intense atmospheric rivers making landfall during summer."
In both cases, the atmospheric rivers arrived at the Antarctic Peninsula a few hours before the record temperatures occurred.
The research team included experts from the University of Valparaiso, the Centre for Climate and Resilience Research in Santiago, the Universidad de Concepcion and the British Antarctic Survey.