Predicting seed production is challenging because many plants produce highly variable crops among years (i.e. masting), but doing so can inform forest management, conservation, and our understanding of ecosystem trajectories in a changing climate. We evaluated the ability of an existing model to forecast masting in an ecologically and culturally important tree species in the southwestern United States, Pinus edulis. Annual seed cone production was predicted using cross-validation techniques on two unique out-of-sample datasets, representing different collection methods and spatial scales (cone scars and cone counts). We then hindcasted this model into the historical past to evaluate whether seed production has declined with the onset of extreme drought conditions in western North America. The evaluated model had fair skill, with root-mean-squared error of 6%. The model had better skill predicting the interannual variability within a site than among sites (i.e. within years). Hindcast analyses indicated recent (2000-2024) mean annual cone production was 30.6% lower than in the past century (1900-1999). Mast forecasts are within reach, but much room remains for improvement. Forecasts may be a powerful tool to anticipate the effects of climate change on forests and woodlands.