Drought monitoring based on SPI with varying timescales using wavelet based hybrid intelligence methods

Drought monitoring based on SPI with varying timescales using wavelet based hybrid intelligence methods

Roghayeh Ghasempour1 Kiyoumars Roushangar2 Hassan Sani3 Farhad Alizadeh Afshar4

1) 1. Department of Water Resources Engineering, Faculty of Civil Engineering, University of Tabriz
2) 1. Department of Water Resources Engineering, Faculty of Civil Engineering, University of Tabriz
3) 1. Department of Water Resources Engineering, Faculty of Civil Engineering, University of Tabriz
4) M.Sc., Civil Engineering

IOI Code : XDCB-FEECB
Publication : 7th.International Congress On Civil Engineering, Architecture & Urban Development(7icsau.com)
Abstract :
    Drought modeling is an important hydrology issue, and has a key role in risk management, drought readiness and alleviation. Drought time series data consists of nonlinear features and various time scales. Therefore, this study mixed the strengths of the wavelet transformation and several intelligence methods to develop a new method of a hybrid model for their ability to accurately predict future Standardized Precipitation Index (SPI) as drought index. A 40-year precipitation data from the year 1980–2018 was used for the SPI series with 3, 9 and 24 months timescales. For improving the applied intelligence approaches efficiency, pre-processing technique was used via wavelet transform (WT). A comparison between the results of single approaches showed that Multilayer Perceptron Firefly Algorithm (MLP-FFA) method led to better outcomes. The results showed that data processing with WT method enhanced the models capability up to 40%. Also, it was observed that the applied methods were more successful in long-term drought modeling. In general, it was found that the hybrid forecasting models were effective forecasting models in short- to long-term drought modeling.

Keywords : discrete wavelet transform drought multilayer perceptron firefly algorithm pre processing standardized precipitation index