In this study, we train machine learning models on tree rings, speleothems, and instrumental rainfall to estimate seasonal monsoon rainfall over India over the last 500 years. Our models highlight multidecadal droughts in the mid-17th and 19th centuries, and we link these to historical famines. Using techniques from explainable AI (artificial intelligence), we show that our models use known relationships between local hydroclimate and the monsoon circulation.

To improve the simulation of surface mass balance (SMB) that influences the advance–retreat of ice sheets, we run a snow model, the BErgen Snow SImulator (BESSI), with transient climate forcing obtained from an Earth system model, iLOVECLIM, over Greenland and Antarctica during the Last Interglacial (LIG; 130–116 ka). Compared to the simple existing SMB scheme of iLOVECLIM, BESSI gives more details about SMB processes with higher physics constraints while maintaining a low computational cost.