Exploring a Long Short-Term Memory based Encoder-Decoder framework for multi-step-ahead flood forecasting

A gated recurrent unit (GRU) network, which is a kind of artificial neural network (ANN), has been increasingly applied to runoff forecasting. However, knowledge about the impact of different input data filtering strategies and the implications of different architectures on the GRU runoff forecasting model’s performance is still insufficient. This study has selected the daily rainfall and runoff data from 2007 to 2014 in the Wei River basin in Shaanxi, China, and assessed six different scenarios to explore the patterns of that impact. In the scenarios, four manually-selected rainfall or runoff data combinations and principal component analysis (PCA) denoised input have been considered along with single directional and bi-directional GRU network architectures. The performance has been evaluated from the aspect of robustness to 48 various hypermeter combinations, also, optimized accuracy in one-day-ahead (T + 1) and two-day-ahead (T + 2) forecasting for the overall forecasting process and the flood peak forecasts. The results suggest that the rainfall data can enhance the robustness of the model, especially in T + 2 forecasting. Additionally, it slightly introduces noise and affects the optimized prediction accuracy in T + 1 forecasting, but significantly improves the accuracy in T + 2 forecasting. Though with relevance (R = 0.409~0.763, Grey correlation grade >0.99), the runoff data at the adjacent tributary has an adverse effect on the robustness, but can enhance the accuracy of the flood peak forecasts with a short lead time. The models with PCA denoised input has an equivalent, even better performance on the robustness and accuracy compared with the models with the well manually filtered data; though slightly reduces the time-step robustness, the bi-directional architecture can enhance the prediction accuracy. All the scenarios provide acceptable forecasting results (NSE of 0.927~0.951 for T + 1 forecasting and 0.745~0.836 for T + 2 forecasting) when the hyperparameters have already been optimized. Based on the results, recommendations have been provided for the construction of the GRU runoff forecasting model.

Section: Introductionmentioning

confidence: 99%

Impact of Input Filtering and Architecture Selection Strategies on GRU Runoff Forecasting: A Case Study in the Wei River Basin, Shaanxi, China

Wang

Liu

Yue

et al. 2020

“…where t is the current time, ∆t is the lead time,Ĥ t+∆t is the forecasted river stage at time t + ∆t, L denotes the lag length of the input variables, R t−L is the antecedent rainfall at time t − L, H t−L is the antecedent river stage at time t-L, and S t−L is the antecedent tidal level at time t − L. Following the approach adopted by Wang et al [45], the lag length was set as 6 h in the present study; this lag length takes into consideration the concentration time of a watershed. To investigate the lead time, the lead time commonly applied in hydrology modeling of 1-6 h was used in this study [45][46][47].…”

Section: Methodology 21 Data-driven Model For River Stage Forecastingmentioning

confidence: 99%

Prediction of River Stage Using Multistep-Ahead Machine Learning Techniques for a Tidal River of Taiwan

Guo

Chen

Yeh

et al. 2021

Time-series prediction of a river stage during typhoons or storms is essential for flood control or flood disaster prevention. Data-driven models using machine learning (ML) techniques have become an attractive and effective approach to modeling and analyzing river stage dynamics. However, relatively new ML techniques, such as the light gradient boosting machine regression (LGBMR), have rarely been applied to predict the river stage in a tidal river. In this study, data-driven ML models were developed under a multistep-ahead prediction framework and evaluated for river stage modeling. Four ML techniques, namely support vector regression (SVR), random forest regression (RFR), multilayer perceptron regression (MLPR), and LGBMR, were employed to establish data-driven ML models with Bayesian optimization. The models were applied to simulate river stage hydrographs of the tidal reach of the Lan-Yang River Basin in Northeastern Taiwan. Historical measurements of rainfall, river stages, and tidal levels were collected from 2004 to 2017 and used for training and validation of the four models. Four scenarios were used to investigate the effect of the combinations of input variables on river stage predictions. The results indicated that (1) the tidal level at a previous stage significantly affected the prediction results; (2) the LGBMR model achieves more favorable prediction performance than the SVR, RFR, and MLPR models; and (3) the LGBMR model could efficiently and accurately predict the 1–6-h river stage in the tidal river. This study provides an extensive and insightful comparison of four data-driven ML models for river stage forecasting that can be helpful for model selection and flood mitigation.

“…In the last decades, because the artificial neural network model can deal with the nonlinearity problem of the inflow data, some scholars have applied it to inflow flood forecasting of the reservoir [24][25][26][27][28]. At the same time, with the continuous development of machine learning technology, some new algorithms have also been widely used in flood forecasting, such as deep learning neural networks [29][30][31][32]. However, these datadriven models mainly focus on the accuracy of the simulation results based on the given datasets and ignore the physical causality between input and output; thus, they face some classical opposition due to reasons inherent in machine learning techniques (e.g., lack of transparency and difficulty of reproducing the results).…”

Section: Introductionmentioning

confidence: 99%

Baipenzhu Reservoir Inflow Flood Forecasting Based on a Distributed Hydrological Model

Chen

Xing

et al. 2021

For reservoir basins, complex underlying surface conditions, short flood confluence times, and concentrated water volumes make inflow flood forecasting difficult and cause forecast accuracies to be low. Conventional flood forecasting models can no longer meet the required forecast accuracy values for flood control operations. To give full play to the role of reservoirs in flood control and to maximize the use of reservoir flood resources, high-precision inflow flood forecasting is urgently needed as a support mechanism. In this study, the Baipenzhu Reservoir in Guangdong Province was selected as the study case, and an inflow flood forecast scheme was designed for the reservoir by a physically based distributed hydrological model, the Liuxihe model. The results show that the Liuxihe model has strong applicability for flood forecasting in the studied reservoir basin and that the simulation results are very accurate. This study also found that the use of different Digital Elevation Model (DEM) data sources has a certain impact on the structure of the Liuxihe model, but the constructed models can both simulate the inflow flood process of the Baipenzhu Reservoir well. At the same time, the Liuxihe model can reflect the spatial variation in rainfall well, and using the Particle swarm optimization (PSO) algorithm to optimize the initial model parameters can greatly reduce the uncertainty of the model forecasts. According to China’s hydrological information forecast standards, the Liuxihe model forecast schemes constructed by the two data sources are rated as Grade A and can be used for real-time flood forecasting in the Baipenzhu Reservoir basin.