Performance comparison of filtering methods on modelling and forecasting total precipitation amount

Ünal, Ecem
The performance of condensed water vapour in the atmosphere observed as precipitation on the earth surface with the consequence of gravity. It is hard to observe and measure the amount and concentration of total precipitation with its all types changing over time. This difficulty can be explained by the association between the changing amount of precipitation and the variability in the climate with its both causes and consequences. As a result of these factors, modelling and forecasting of monthly total precipitation series is a difficult procedure because of being highly parametrized and varied nature of data. To predict and forecast total precipitation, filtering methods are suggested as an alternative in the literature. Therefore, this study focus on the comparison of modelling and forecasting performances of different types of filtering methods on monthly total precipitation series. To do this, the Kalman Filter method is preferred in order to predict and forecast the naturally uncontrollable outcomes. The Kalman Filter is an algorithm for the estimation of the unobservable true state of the system, which is conducted by incorporation with the models of the system and noisy measurements of parameters. For this purpose, we used the monthly precipitation series of Muğla, Konya and Ordu stations from 1950 to 2010. The regions have been selected in terms of the amount of precipitation as moderate, scarce and abundant regions. The results of modelling and forecasting performance comparison will be a guide for the choice of best performing method for further work related to the precipitation.
Citation Formats
E. Ünal, “Performance comparison of filtering methods on modelling and forecasting total precipitation amount,” Thesis (M.S.) -- Graduate School of Natural and Applied Sciences. Statistics., Middle East Technical University, 2019.