Abstract by Spencer Ebert
Application of Kalman Filter on El Nino Data
This paper explores the use of a Kalman filter on an AR(1) time series model. The data examined is the Southern Oscillation Index and is used to predict El Nino events (SOI below -1) that have a great impact on climate throughout the western hemisphere. Measuring SOI has a lot of room for error so this paper uses the Kalman filter to distinguish between process noise and observational noise. The goals of this analysis is to determine the false positive rate for an El Nino event and find the best parameters that fit the SOI data using expectation maximization. This analysis also examines the effectiveness of a Kalman filter on one dimensional data through a simulation study. It was found that the Kalman filter has a hard time distinguishing between process noise and observational noise when the values are far apart, which suggests the Kalman filter has greater effectiveness for multidimensional data.