Whirlwind prediction using cloud movement patterns on satellite image

Whirlwind is a local-scale meteorological phenomenon that occurs in a short time, destructive, and can cause loss of life and material. Until now, when and where the whirlwind will occur can not be predicted precisely. However, the signs are still recognizable from some of the symptoms before the phenomenon occurs, such as high temperatures and the formation of many Cumulus clouds which then suddenly transform into Cumulonimbus clouds. Because this incident caused a lot of damage and casualties, the whirlwind needs to be predicted, so that later people can be more vigilant and the impact can be minimized. This research aims to build a system for taking cloud movement patterns from observing cloud clusters on the satellite image. The clustering method is used to classify clouds, and then find out the pattern of movement in each cluster. This pattern of movement is a model to predict the occurrence of the whirlwind. The results obtained from the experiments in several whirlwind incidents in Indonesian territory indicate that at 24 hours before the event, there are at least Cumulus, Middle Cloud and/or Stratocumulus clouds that have a curving pattern; approaching and then away from the location where the whirlwind appears. Furthermore, the pattern of cloud movement will be collected to build a data test. The results obtained from the K-NN method show the accuracy of the data test collected from a number of the whirlwind phenomenon in 2016 is 88%. Meanwhile, when the data test tested with SVM method, the percentage of accuracy is 84%.