The applications of two very different methods for deducing (downscaling) tropical cyclone activity from NCAR/NCEP reanalysis data explain, respectively, 60% and 65% of interannual variations in Atlantic tropical cyclone frequency during the period 1980-2006. Yet, when one of these methods is applied to the output of simulations using a global climate model forced by observed sea surface temperature over the same period, far less variance is accounted for, and the upward trend seen in both the observations and the downscaled NCAR/NCEP reanalysis is largely absent. Moreover, when this downscaling technique is applied to ERA40 re-analysis data, the amount of variance explained is comparable to that of the global climate model, and again the upward trend is largely absent. This proposal seeks support for an effort to understand the physical reasons for these discrepancies, and by so doing to advance our understanding of environmental control of tropical cyclone activity and its relationship with climate change. We propose to undertake a comprehensive analysis of the physical causes of the variability and trends seen in various downscaled tropical cyclone metrics, focusing on the disparity among the reanalysis-driven and global climate model-driven results. We here present a few hypotheses for the discrepancies and a plan to test these, with the goal of identifying model and/or reanalysis biases that may be affecting the results. To the extent we are successful, we can begin to assess the ability of climate models to predict future variations in tropical cyclone activity resulting from natural and anthropogenic climate variability, while at the same time increasing our understanding of the fundamental environmental controls on tropical cyclone activity.