Fast Fourier Transform (FFT) for Forecasting Long-Term Droughts and Floods

News source: The Philippine Council for Agriculture, Forestry and Natural Resources Research and Development
For further information, Baldazo, P.G., Ramirez, A.M., Acebec, T.B. Bautista, M.P. Forecasting long-term droughts and series of wet years. Highlights 2001, PCARRD, Los Banos, Laguna, Philippines, 2002, pp. 2-4., 2002-03-01

Long-Term Prediction of Floods and Droughts

Using Mathcad7 Routine Program (commercial computer software), the Fast Fourier Transform (FFT) model was developed to predict long-term droughts and floods. The FFT model assumes that natural time data series (tree rings, rainfall, stream flow, runoff, and temperature) follow a normal distribution, and that climate is periodic in nature. It also assumes that the wavelength of the climate signal is universal, and that its amplitude is the reflection of the local characteristics of a particular area.

Because of FFT's ability to decompose and reconstruct any natural time data series, and identify the presence of strong signals from observed long-term data series, it can be used to forecast or model annual rainfall and/or any natural time data. This ability is more advantageous than the autoregressive moving average technique, which is specific only for univariate data series.

Tree Ring Series from Other Countries

To validate the FFT model, tree ring data series from other countries including China, Indonesia, India, Japan, Australia, New Zealand, and USA (California) were used to reconstruct and predict the annual rainfall series/pattern in the Philippines. Results revealed that some of the signals evaluated from the China tree ring data closely fit with the rainfall patterns in several parts of the Philippines, namely: Aparri, Cagayan; Baler, Aurora; Basco, Batanes; Casiguran, Quezon; Cuyo, Palawan; Laoag City; Mactan, Cebu; Malaybalay, Bukidnon; Port Area, Manila; and Romblon.

Correlation Coefficients of FFT Model

Additionally, the FFT model gave high correlation coefficients, ranging from 92 to 99%, between the observed and the predicted values from 1951 _ 1999. The Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) is now using the FFT model as one of the tools to forecast weather or come up with a climate advisory. The FFT model is being studied by PAGASA and the University of the Philippines at Los Banos for seasonal climatic forecasting to be used by farmers.

Download the PDF. of this document, 28,669 bytes (28.0 KB).