Long Lead Time Drought Forecasting Using a Wavelet and Fuzzy Logic Combination Model: A Case Study in Texas

Abstract

Drought forecasting is important for drought risk management. Considering the El Niñ o-Southern Oscillation (ENSO) variability and persistence in drought characteristics, this study developed a wavelet and fuzzy logic (WFL) combination model for long lead time drought forecasting. The idea of WFL is to separate each predictor and predictand into their frequency bands and then reconstruct the predictand series by using its predicted bands. The strongest-frequency bands of predictors and predictand were determined from the average wavelet spectra. Applying this combination model to the state of Texas, it was found that WFL had a significant improvement over the fuzzy logic model that did not use wavelet banding. Comparison with an artificial neural network (ANN) model and a coupled wavelet and ANN (WANN) model showed that WFL was more accurate for drought forecasting. Also, it should be noted that the ENSO variability is not a global precursor of drought. For this reason, prior to an application of such a data-driven model in different regions, significant work is required to identify appropriate independent predictors. Drought forecasting with longer lead times and higher accuracy is of significant value in engineering applications.

Monthly Prediction of Drought Classes Using Log-Linear Models under the Influence of NAO for Early-Warning of Drought and Water Management

“…Examples of those hybrid approaches which conjugate statistics and physical-based techniques are the works by Mishra et al [54] and Kim and Valdes [55], which use respectively a hybrid stochastic and neural network model and the later conjugates wavelet transforms and neural networks for a nonlinear model. However, other examples are the model combination of the wavelet and fuzzy logic models used in Ozger et al [56] and the adaptive neuro-fuzzy inference used in Bacanli et al [57]. Among the techniques used for drought forecasting, statistical models are chosen many times, since they are simple to implement, do not have a high computational burden, and produce useful predictions [58].…”

Drought Forecasting Using Stochastic Models in a Hyper-Arid Climate

“…Mishra and Singh developed a new hybrid wavelet-Bayesian regression model for simulating the hydrological drought time series [30]. Özger et al created a combination model based on wavelet and fuzzy logic for long lead-time drought forecasting [31]. Özger et al investigated the use of a wavelet fuzzy logic model to estimate the PDSI based on meteorological variables [32].…”

Standard Precipitation Index Drought Forecasting Using Neural Networks, Wavelet Neural Networks, and Support Vector Regression

“…Wavelet analysis, an effective tool to deal with nonstationary data, has recently been applied in hydrological forecasting to examine the rainfall-runoff relationship in a Karstic watershed [17], to characterize daily streamflow [18,19] and monthly reservoir inflow [20], to evaluate rainfallrunoff models [21], to forecast river flow [22][23][24], to forecast future precipitation values [25], and for the purposes of drought forecasting [26]. The study conducted by Kim and Valdes [26] is the only study that has explored the ability of a wavelet-neural network conjunction model (WN) to forecast a given drought index.…”