A Study of Tropical Cyclone Genesis Indices with Poisson Regression Model and Their Relationship with Various Environmental Predictors

Tropical cyclones are the most deadly and financially costly natural disasters. Surprisingly, although the basic physics of tropical cyclones is well understood, there are still many open questions about their formation. Genesis events, therefore, must be studied empirically, and only statistical conclusions can be drawn. In my thesis research, I worked with Professor Suzana Camargo at the Lamont Institute of Earth Science, in Palisades, New York, whose specialty is tropical cyclones. She provided me with historical weather data describing observed tropical cyclone frequency along with the relevant environmental variables like temperature and humidity. Global data was given on a grid with 2° resolution in both latitude and longitude, averaged monthly, for the years 1979 to 2013. I fitted the observed tropical cyclone frequency simultaneously to 4 or 5 independent environmental variables to find the relative strength of their dependence. Comparing the strength of their dependence, I identified the variables with the highest correlations with tropical cyclone genesis. The correlations of environmental variables with tropical cyclone genesis can be quantified by coefficients of regression. The environmental variables, then, can be weighted by the coefficients and combined to create a rough mathematical model describing the expected number of tropical cyclones per area (2° x 2° latitude-longitude) per month. There were eight environmental variables available: one parameter related to the difference in wind speed at different altitudes (called vertical shear), two parameters related to the speed of atmospheric rotation (called vorticity), three parameters related to sea surface temperature, and two parameters related to humidity. Among these eight variables, I identified a combination that best captured the spatial and temporal distributions of tropical cyclones. The technical names of the five most important environmental variables are relative sea surface temperature, clipped vorticity, saturation deficit, potential intensity, and vertical shear. The quantitative model for tropical cyclone genesis produced by a weighted combination of these environmental variables is called the tropical cyclone genesis index. This index will help us gain an understanding of genesis events in the absence of an exact theory. In low-resolution computer models of climate, they can be used as a proxy to relate tropical cyclone number to local environmental conditions. Additionally, the index may be helpful in making projections of tropical cyclone frequency under future environmental conditions, which is especially important for considerations of global climate change.