A generalized linear model based on Poisson regression has been used to assess the impact of environmental variables modulating tropical cyclone frequency in six main cyclone development areas: the East Pacific, West Pacific, North Atlantic, North Indian, South Indian, and South Pacific. The analysis covers the period 1980-2009 and focuses on widely used meteorological parameters including wind shear, sea surface temperature, and relative humidity from different reanalyses as well as aerosol optical depth for different compounds simulated by the Goddard Chemistry Aerosol Radiation and Transport model. Circulation indices are also included. Cyclone frequency is obtained from the International Best Track Archive for Climate Stewardship. A strong link is found between cyclone frequency and the relative sea surface temperature, Atlantic Meridional Mode, and wind shear with significant explained log likelihoods in the North Atlantic of 37%, 27%, and 28%, respectively. A significant impact of black carbon and organic aerosols on cyclone frequency is found over the North Indian Ocean, with explained log likelihoods of 27%. A weaker but still significant impact is found for observed dust aerosols in the North Atlantic with an explained log likelihood of 11%. Changes in lower stratospheric temperatures explain 28% of the log likelihood in the North Atlantic. Lower stratospheric temperatures from a subset of Coupled Model Intercomparison Project Phase 5 models properly simulate the warming and subsequent cooling of the lower stratosphere that follows a volcanic eruption but underestimates the cooling by about 0.5 degrees C.