Study of Atmospheric Parameters Variation using Time Series Multiplicative Decomposition Model for Accurate Forecasting

Abstract

In this study, the variations in selected atmospheric parameters (rainfall, temperature, relative humidity, and wind speed) were investigated via a time series multiplicative decomposition model to enhanceaccurate weather forecasting. The time series multiplicative decomposition model application in atmospheric studies provides a sophisticated tool to better understand, model, and predict complex weather and environmental phenomena, offering significant improvements in accuracy over simpler forecasting methods. In this study, the selected atmospheric parameters were analysed over several years to capture both short-term fluctuations and long-term trends, with a focus on improving the forecasting accuracy. The findings highlight significant variability in multiplicative model performance across the different atmospheric parameters studied. The results show that rainfall exhibited extremely poor accuracy, with a high MAPE of 275.068, indicating significant forecasting errors with the model and suggesting a need for model enhancement. Conversely, the model demonstrated strong predictive performance for temperature, with a low MAPE value of 7.65%, indicating a reliable forecast with only occasional larger deviations, as reflected by the moderate MSD value. The relative humidity model showed relatively good accuracy of forecasts, with a MAPE value of 5.11%, although the high MSD suggested occasional outliers. The result of the wind speed model, however, showed an exceptionally high MAPE value of 93.38%, indicating a high forecast error, despite lower MAD and MSD values. These results underscore the importance of multiplicative model refinement, particularly for rainfall and wind speed, to improve prediction accuracy. The findings of this study provide insights into the strengths and weaknesses of current forecasting models for atmospheric parameters, guiding future improvements in predictive modelling. Similarly, it will serve as a foundation for more accurate and reliable forecasting techniques, which can be applied in climate monitoring, agricultural planning, disaster management and control, particularly in regions experiencing variable weather patterns.