Thanks to an increase in the use of devices that come with a builtin gps sensor, a large number of gps data is available. This work presents an algorithm to extract concrete patterns from GPS data based on the changes on the behaviour of the car drivers when the weather drastically changes. The data used in this project comes from a week when the Typhoon Roke struck Japan, the algorithm will try to select the most likely areas affected by an accident like floods, banned roads, etc. This data is excelent for this pourpose but there is a need to built a visual and interactive