Birds have adapted to climatic and ecological cycles to inform their Spring and Fall migration timings, but anthropogenic global warming has affected these long-establish cycles. Understanding these dynamics is critical for conservation during a changing climate. Here, we employ a modeling approach to explore how climate spatiotemporally affects bird occurrence on eBird surveys. Specifically, we train an ensemble of multivariate and multi-response random forest models on North and South American climate data, then predict eBird survey occurrence rates for 41 migrating passerine bird species in a Northeastern American ecoregion from 2008-2018. In October, when many passerines have begun their southward winter migration, we achieve more accurate predictions of bird occurrence using lagged climate features alone to predict occurrence. These results suggest that analyses of machine learning model metrics may be useful for identifying spatiotemporal climatic cues that affect migratory behavior. Lastly, we explore the application and limitations of random forests for prediction of future bird occurrence using 2021-2040 climate projections.