Seasonal Lake Surface Water Temperature to Mean Annual Air Temperature Relationships and Applications: An Analysis of 1000 Lakes

Lacustrine sediments are important archives of past climates. Many lacustrine-based proxies (e.g., clumped isotope thermometry, oxygen isotope thermometry, and GDGTs) record seasonal lake surface water temperature (LSWT), depending on site characteristics (e.g., latitude and elevation). However, LSWT is a climatic variable not typically simulated by climate models or easily comparable to non-lacustrine proxies in contrast to mean annual air temperature (MAAT). Building on prior seasonal LSWT-to-MAAT (lake-to-air temperature) transfer functions (Hren and Sheldon, 2012) that utilize regression models and a LSWT dataset of 88 lakes, we work with new satellite-derived datasets that encompass roughly 1000 latitudinally distributed lakes that capture the global diversity of lake systems and climatological regimes. We apply machine learning techniques to impute missing temperature values in a LSWT dataset from GloboLakes and construct seasonally specific lake-to-air temperature transfer functions from regression models that encompass large ranges in latitude (55°S to 82°N) and elevation (-415 to 5751 masl) based on three intervals of proxy seasonality: (1) April-June, (2) June-August, and (3) April-October. We further improve our transfer functions by incorporating latitudinal and elevation-specific components, the two greatest influences on LSWT and MAAT. Using existing proxy records of LSWT, we demonstrate that seasonally specific lake-to-air temperature transfer functions enable the more accurate reconstruction of MAAT for various climatological and hydrological settings.