All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 27 6 0
PDF Downloads 3 0 0

Prediction and Frequency Tracking of Nonstationary Data with Application to the Quasi-Biennial Oscillation

View More View Less
  • 1 National Center for Atmospheric Research, Boulder, CO 80307
Full access

Abstract

Prediction of meteorological phenomenon is an important problem in the Atmospheric Sciences. For this purpose the periodic components are usually identified first. Then, to apply well-known analytic tools, stationarity and ergodicity are often invoked. this tacitly implies fixed periodicities. However, we often come across instances where the data are nonstationary, having time-dependent periodicities. Further, some stationary noise component may also be superimposed on the data. The Quasi-Biennial Oscillation (QBD) is one such example. In such cases, only those data analysis techniques should be used which can handle both, stationary as well as nonstationary, data generating processes. the least mean square (LMS) algorithm is one such technique.

In this paper we explore the capabilities of the LMS algorithm for the prediction and frequency tacking of nonstationary processes. The technique is then applied to the QBD zonal winds to achieve a several month prediction and to highlight its “quasi” characteristic.

Abstract

Prediction of meteorological phenomenon is an important problem in the Atmospheric Sciences. For this purpose the periodic components are usually identified first. Then, to apply well-known analytic tools, stationarity and ergodicity are often invoked. this tacitly implies fixed periodicities. However, we often come across instances where the data are nonstationary, having time-dependent periodicities. Further, some stationary noise component may also be superimposed on the data. The Quasi-Biennial Oscillation (QBD) is one such example. In such cases, only those data analysis techniques should be used which can handle both, stationary as well as nonstationary, data generating processes. the least mean square (LMS) algorithm is one such technique.

In this paper we explore the capabilities of the LMS algorithm for the prediction and frequency tacking of nonstationary processes. The technique is then applied to the QBD zonal winds to achieve a several month prediction and to highlight its “quasi” characteristic.

Save