Contrasting continental temperature responses during MIS4 and MIS3 off northeastern Brazil inferred from branched GDGTs

Multiproxy records from marine sediment core GL-1248, off northeastern Brazil, indicate persistent humid conditions during Marine Isotope Stage (MIS) 4 and episodic moisture pulses associated with Heinrich Stadials (HS) during MIS3. These hydroclimatic changes have been linked to weakening of the Atlantic Meridional Overturning Circulation (AMOC), heat redistribution in the tropical Atlantic, and southward displacement of the Intertropical Convergence Zone (ITCZ). However, the continental temperature response remains poorly constrained. This study reconstructs mean annual air temperature (MAAT) using branched glycerol dialkyl glycerol tetraethers (brGDGTs) and a tropical Bayesian neural network calibration developed for tropical mineral soils. BrGDGT source indicators, including #ringstetra and the ΣIIIa/ΣIIa ratio, together with compositional similarity to modern tropical soils, support a predominantly soil-derived contribution of brGDGTs and the application of the tropical soil calibration. MAAT estimates were further assessed against lignin phenols, which indicate terrestrial vascular plant organic matter, and the BIT index to assess whether the reconstructed temperature signal systematically tracked terrestrial organic-matter input or changes in GDGT distributions. The results reveal divergent thermal patterns between MIS4 and MIS3. During MIS4, MAAT increases toward HS6 and remains elevated beyond the event toward the MIS4–MIS3 transition. Sea surface temperature (SST) derived from Mg/Ca, previously published for the same core, also increased toward HS6 but declined afterward, while MAAT remained elevated. This ocean-continent decoupling suggests different temporal expressions of oceanic and continental thermal responses. During MIS3, MAAT shows more event-specific responses, with warming in some intervals occurring near the end of or extending beyond individual HS. Significant differences across selected HS to post-HS transitions further support this variability. At selected intervals, elevated MAAT coincided with increased tropical rainforest pollen in previously published records from the same core. However, MAAT did not systematically covary with lignin phenols or BIT, indicating that variations in terrestrial input and GDGT distributions do not systematically control the reconstructed temperature signal. Collectively, these results indicate a more persistent continental thermal response during MIS4 and more episodic, event-specific responses during MIS3, showing that ocean, continent, and vegetation did not warm in step during AMOC-driven ITCZ reorganizations. This temporal decoupling implies that continental temperatures over near equatorial regions cannot be read directly from marine SST, and that the pace of the continental response to abrupt climate change was itself set by the background climate state, persistent under full-glacial MIS4, but pulse-like under interstadial MIS3.