Abstract:

My research combines speleothem records (cave deposits) and climate models to investigate hydroclimate change during the Late Pleistocene and Holocene. This approach also provides a framework for evaluating climate model performance across past climate states, with the goal of refining projections of future (hydro)climate change. To illustrate this, I present a 45,000-year multiproxy record (stable isotopes and trace elements) from a stalagmite in central Vietnam, which captures Southeast Asian autumn monsoon variability. The record shows drier conditions during the Last Glacial Maximum, followed by a rapid shift to wetter conditions during the Deglaciation that persisted into the Holocene. By pairing this record with climate model simulations, I demonstrate that sea-level, driven by glacial ice volume fluctuations, is the dominant control on monsoon rainfall on glacial timescales. Additionally, I developed a novel semi-quantitative rainfall reconstruction using calcium isotopes (δ⁴⁴Ca), an emerging and largely underutilized proxy in speleothem science. Unlike most traditional speleothem proxies, which provide qualitative hydroclimate reconstructions, speleothem δ⁴⁴Ca reflects changes in rainfall amount. My δ⁴⁴Ca record indicates that central Vietnam experienced rainfall amounts ~50% lower than modern levels during the Last Glacial Maximum. These findings align closely with climate model simulations, providing confidence in both the reconstruction and the ability of climate models to accurately simulate precipitation change in Southeast Asia across the Deglaciation. This highlights the potential of using speleothem δ⁴⁴Ca for both quantifying past rainfall and validating climate model performance. At Princeton, I plan to expand the use of δ⁴⁴Ca by generating a global network of speleothem δ⁴⁴Ca records and comparing them with climate model simulations. My research will focus on identifying proxy and model biases, providing critical insights towards how these data products should be used, improved, and integrated (i.e. data assimilation) in the future.