Bivariate Splines for Hurricane Path Forecasting

Every year, hurricanes cause a lot of damage, especially, when they hit cities along the coast line. A notorious example is Hurricane Katrina in 2005 which hit New Orleans and damaged the city significantly. Human intervention cannot change the path of a hurricane. Hence the best possible way to reduce the damage is to predict the path of the hurricane and avoid direct contact. The formation, movement and strength of hurricanes is not understood very well. Although meteorologists have several methods to predict the path of hurricanes, current methods are still not able to predict the paths with enough lead time, accuracy and confidence to evacuate and prepare cities. For example, Hurricane Rita in 2005 was wrongly predicted to hit Houston which caused tens of thousands residents to evacuate for nothing. Little is known about the physical, chemical and thermodynamic explanations of the speed and direction of the movement of a hurricane. Hurricanes seem to move randomly. Nevertheless, people are determined to understand the phenomenon of hurricane. Satellites are constantly observing hurricanes from space. Airplanes track hurricanes from the sky. Buoys are placed in the oceans to collect hourly barometric information. In particular, after Hurricane Katrina, more buoys were placed in the Gulf of Mexico. Several technologies have been initiated and developed to decipher the data and information to predict the path of a hurricane. See Figures 1 and 2 for buoy locations in the Gulf of Mexico below. We approach this prediction problem by using bivariate spline functions. Although univariate splines have been applied for predictions (cf. (2), (4), (5) and (6)), bivariate splines have never been used for hurricane predictions. Bivariate spline functions have the advantage of fitting scattered data conveniently. We find a se-