Search

In recent years, deep learning (DL) approaches have been proven effective in addressing high nonlinear relationships within complex systems. Although various scientific studies have primarily focused on optimizing model architecture and enhancing its computational efficiency through hybridization, they have neglected the crucial aspect of selecting appropriate predictor variables (PV) and their influence on the model's predictions. In this paper, a new multi-phase stochastic DL model is designed with a feature selection framework for monthly streamflow prediction. The multi-phase hybrid MCMC–BC–BiLSTM–BiGRU model is based on bidirectional long short-term memory (BiLSTM) and bidirectional gated recurrent unit (BiGRU) integrated with Markov Chain Monte Carlo (MCMC) based bivariate copulas (BC). First, Ant Colony Optimization (ACO) algorithm was employed for screening optimal PV, and then significant antecedent streamflow was coupled with it to generate optimized input combinations. In addition, 25 different MCMC–BC models were adopted for each input combination to determine the dependency of previous month's streamflow on current and future streamflow. Finally, best-fitted MCMC–BC model was integrated with hybrid BILSTM–BiGRU model to predict streamflow at nine different catchments in the Victoria region of the Upper Murray Basin (UMB), Australia. The experimental results show that the multi-phase hybrid MCMC–BC–BiLSTM–BiGRU model showed remarkable performance in comparison to its benchmark counterparts with respect to different robust statistical error metrics. Therefore, this study demonstrates that stochastic DL techniques can be effectively employed as a promising alternative predictive tool for upstream discharge predictions with high consistency and accuracy, specifically when a statistically significant relationship with the antecedent streamflow exists. [ABSTRACT FROM AUTHOR]

The intensity, duration, and frequency (IDF) of low streamflow events are often considered stationary in time. Given droughts are likely to become more frequent and extreme, here those IDFs are analyzed for their non‐stationarity in relation to three climate drivers, namely, the El Niño‐Southern Oscillation, the Indian Ocean Dipole, and the Southern Annular Mode. For this, low flow events were identified for 161 unregulated catchments in Victoria, Australia. The relationship between the climate drivers and the IDFs were developed using covariate‐based non‐stationary statistical models. We found that several of these catchments exhibit strong non‐stationary influences from the climate modes. Such catchments have a higher probability of experiencing more intense, prolonged, and/or frequent low flows when a dry phase of these climate drivers occurs. We found that catchments having strong evidence of non‐stationary frequency (n = 77) were more compared to those having non‐stationary intensity (n = 19) or duration (n = 34). Further, we derived projected indices of these climate drivers from CMIP5 models for the RCP8.5 scenario. These indicate an intensification in the dry phase of the climate drivers. Using information of the projected indices in the IDF models, we found that future dry phases of these climate drivers are likely to further increase the probability of experiencing higher magnitudes of low flow IDFs in catchments non‐stationary in these characteristics in relation to one or more of these climate drivers. Frequency shows greater response to future dry phases of these drivers than intensity or duration of low flows. Plain Language Summary: The interaction of the ocean and atmosphere can lead to conditions that can control how dry or wet a season might be for a region. These conditions are called climate modes or climate drivers. Drought periods in many regions of the world are known to be impacted by these climate modes. The activity of these climate modes can make an ongoing drought more severe compared to a situation when these modes are not active. Drought spells often entail abnormally low flows in streams. In this study, we show that low flows in southeast Australia have a significant link with the activity of the climate drivers. In fact, when a drought coincides with a dry phase of these climate modes, low flow events can become more intense, frequent and/or can last longer. Unabated climate change has the potential to make these climate modes more intense in future. Our study shows that if this happens, low flow events can become more extreme. Thus future dry phases of these climate drivers may likely cause shortage of freshwater availability, disruption of ecosystems vitally dependent on flows, uncertainty in dam operations, all of which increases the risk associated with future droughts for such regions. Key Points: Low flow characteristics of catchments in southeast Australia exhibit significant non‐stationary influences from large‐scale climate modesDry phase of these climate drivers is expected to intensify under climate change, leading to reduced water availability when they occurFrequency of low flows shows greater response to non‐stationary influences from the climate modes than intensity or duration [ABSTRACT FROM AUTHOR]