Westerly winds across the North American Mid-Continent during the last glacial maximum; a new data-model assessment

Last Glacial Maximum (LGM) proxy evidence of wind direction across the North American midcontinent comes primarily from loess deposits, and overwhelmingly suggests surface winds had a dominant westerly component. However, paleoclimate model simulations over the last several decades have raised the possibility of easterly winds, due to the presence of an anticyclone over the Laurentide ice sheet or, near the ice margin, katabatic winds off the ice sheet. Here we reassess model-data agreement using nine global circulation models participating in the LGM experiment of Paleoclimate Model Intercomparison Project 3 (PMIP3) and a compilation of previously published loess paleowind directions. An agreement metric that accounts for agreement by chance, Gwet's AC1, is calculated for each model-proxy wind direction pair to quantitatively assess model performance. We find high agreement in 7 of 9 models in the winter season (December-January), suggesting westerly winds are present in the midcontinent in most models. We also find high agreement in spring (March-May) and fall (September-November) in 5 of 9 models, supporting LGM loess deposition and westerly winds in these seasons. Models with the weakest agreement have lower spatial resolution and northeasterly flow off the ice sheet, whereas models with the best agreement have higher resolution and westerly or northwesterly flow off the ice sheet. We only find evidence for a southward shifted jet stream in the summer season (June-August), suggesting different pressure patterns over the ice sheet and adjacent land, rather than a jet stream south of the ice margin, is likely the mechanism forcing LGM surface wind direction in mid-continental North America.