Comparison between LDA technique and fuzzy membership roster method for pre-monsoon weather forecasting

Abstract

In the present study, an attempt is made to propose a new operational technique for weather forecasting at Kolkata (22.53º N, 88.33º E), India, during the pre-monsoon season (March, April and May). The technique is based on fuzzy membership roster method. It can handle inherent non-linearity in a physical phenomenon. To establish the new method, a comparative study is performed between the existing multivariate technique, the linear discriminant analysis and the newly suggested technique based on fuzzy membership roster method. It is interesting to note that for the prediction of weather of next 12 hours based on Radio/Rawin Sonde observation at 1200 UTC of a day, the fuzzy membership roster method is better than the multivariate technique. Both the methods are however almost equally suitable to predict the weather of the next 12 hours based on Radio/ Rawin Sonde observation at 0000 UTC. So, the fuzzy logic based technique, adopted here, is as efficient as the linear statistical rules, but computationally simpler. The degrees of compatibility and the discriminant functions are defined using a training data set for the period 1985-1996 and validated for the period 1997-1999.