Development of Multidecomposition Hybrid Model for Hydrological Time Series Analysis

Abstract

Accurate prediction of hydrological processes is key for optimal allocation of water resources. In this study, two novel hybrid models are developed to improve the prediction precision of hydrological time series data based on the principal of three stages as denoising, decomposition, and decomposed component prediction and summation. The proposed architecture is applied on daily rivers inflow time series data of Indus Basin System. The performances of the proposed models are compared with traditional single-stage model (without denoised and decomposed), the hybrid two-stage model (with denoised), and existing three-stage hybrid model (with denoised and decomposition). Three evaluation measures are used to assess the prediction accuracy of all models such as Mean Relative Error (MRE), Mean Absolute Error (MAE), and Mean Square Error (MSE). The proposed, three-stage hybrid models have shown improvement in prediction accuracy with minimum MRE, MAE, and MSE for all case studies as compared to other existing one-stage and two-stage models. In summary, the accuracy of prediction is improved by reducing the complexity of hydrological time series data by incorporating the denoising and decomposition.

Peer Review #2 of "Improving the prediction accuracy of river inflow using two data pre-processing techniques coupled with data-driven model (v0.1)"

“…The schematic view of proposed methodology is illustrated in Figure 1 Manuscript to be reviewed denoised daily river inflow series into multiple IMFs (Rezaie-Balf et al, a2019). The high irregular IMF is set as threshold with EBT to remove its sparsity and irregularity (Nazir et al, 2019). Further, in the prediction stage, SVM is applied on all IMFs to establish the prediction models and all predicted IMFs are aggregated to get a final prediction (Yaseen et al, a2016).…”

“…In EBT method, the posterior distribution is derived with the help of prior distribution to remove sparseness and noises from the coefficients derived from wavelets (To et al, 2009;Nazir et al, 2019). In this study, we used this wavelet-based denoising method to remove noises and sparseness of VMD based coefficients.…”

“…However, all data pre-processing techniques have some drawbacks with different aspects to decompose the non-linear, multi-scale and noisiest data. The most widely used WA depends heavily on the selection of wavelet basis function (Wang et al, 2018), the application of EMD is limited by its own mathematical mode mixing and sensitivity to denoise property (Nazir et al, 2019), and EEMD suffers a strict mathematical theory (Qian et al, 2019;Wu et al, 2019).…”

Dependence structure analysis of multisite river inflow data using vine copula-CEEMDAN based hybrid model

“…In the past few decades, reliable prediction of rivers inflow has gained popularity in all water-related departments because of their crucial role in the reservoir, irrigation management, water planning, risk evaluation and flood controlling (Porporato & Ridolfi, 2001;Jandhyala, Liu & Fotopoulos, 2009;Di, Yang & Wang, 2014;Tiwari et al, 2017;Nazir et al, 2019). Johnston & Smakhtin (2014) reviewed the importance of river data and concluded that river inflow data is an indispensable component of water resources.…”

“…Wu & Chau (2010) explained in their review that the data-driven approaches performed better than the traditional statistical models to predict the non-linear data. However, data-driven models may suffer an overfitting problem and are sensitive to parameter selection (Nazir et al, 2019). Moreover, data-driven models ignored the time-varying or multi-scale characteristics of time series data.…”