Mamdani Fuzzy Based Prediction of Rainfall Fuzzy Rainfall Index

Abstract

This study attempts to investigate the link between rainfall and climatic large-scale synoptic patterns. Here, we employed the fuzzy inference system (FIS) to forecast rainfall. Six linguistic variables were considered as an input for the proposed system, split into two FIS. Average Mean Temperature (degree celsius), Mean Wind velocity (KM-PH), and Evapotranspiration (mm/day) belong to FIS1. In contrast, Relative humidity (%), Season, and Mean sun shines (hrs/day) belong to FIS2, and each model has three triangular membership functions, excluding the season variable. It has four bi-membership functions: summer, winter, South West, and North East. The output models have four memberships function Very high, High, Normal, Low, and Very Low for FIS1 and FIS2; respectively. The IF-THEN rules were assigned based on the individual significance of linguistic variables for rainfall and from expert opinion for the model FIS1, FIS2 and FRFI, In this article, we had implemented 72 numbers of possible rules, moreover model results can be reviewed into surface 3D-Plot. The implication of each variable with output such as FIS1, FIS2, and FRFI was addressed. Finally, predicted values were compared with the actual rainfall data. Thus, the proposed model would be expected rainfall at a reasonable accuracy. All the implementation has been done in MATLAB7 Fuzzy toolbox.