A statistical model to forecast tropical cyclogenesis (TCG) over the Atlantic Basin up to 48 hours in advance is described. The model was trained on four seasons (1998-2001) of tropical cloud cluster activity during the Atlantic hurricane season. Eight large-scale predictors of TCG were derived from the NCEP-NCAR reanalysis and applied to each case.

Although the probability of detection was low (30-50%), false alarm rates were only 2-3%. Heidke skill scores ranged from 0.4 - 0.5 for each forecast period. Composite and case study results suggest three major areas for improvement: 1) The identification of more developing cases, 2) The addition of a climatological predictor, and 3) A better representation of the moisture conditions within the cloud cluster and its environment. Each will be discussed.