A holistic view of tropical modes of variability as drivers of humid heat

Extreme humid heat threatens human health by limiting the body’s ability to cool through sweating. Its impacts are greatest in the tropics and subtropics, where high population density coincides with hot and humid conditions that are projected to intensify under climate change. While temperature extremes in the mid-latitudes have been widely studied, the drivers and predictability of tropical humid heat remain poorly understood.

We identify historical humid heat extremes in reanalysis across tropical and subtropical land. We use logistic regression to holistically examine the relationships between hot-humid days and the major modes of tropical variability.

We find that ENSO exerts a dominant influence on humid heat extremes across much of the tropics at interannual timescales, acting through combined effects on temperature and humidity. The Indian Ocean Dipole and the Atlantic modes further modulate the extremes.

On shorter timescales, the MJO is the dominant driver of humid heat variability in the regions surrounding the Indian Ocean. Humid heat peaks according to a fine, regionally varying, balance between increased humidity and longwave warming in the active MJO phases and increased shortwave warming in the suppressed MJO phases.

Over central Africa, north-west South America and parts of the Maritime Continent, Kelvin waves dominate over the MJO. The divergent and easterly phases of Kelvin waves increase humid heat predominantly through temperature increases, driven adiabatically through subsidence and diabatically through shortwave warming. Rapid transitions between the convergence and divergence Kelvin wave phases tend to constrain the duration of humid heat extremes, typically to no more than three consecutive days. Rossby and WMRG waves only dominate over smaller regions of the sub-tropics. 

This study has significantly advanced understanding of the drivers of tropical humid heat and highlights pathways for improved prediction. Our findings have important implications for model evaluation, seasonal outlooks, and the design of early warning systems.