Capabilities of deep learning models on learning physical relationships: Case of rainfall-runoff modeling with LSTM

Yokoo, Kazuki; Ishida, Kei; Ercan, Ali; Tu, Tongbi; Nagasato, Takeyoshi; Kiyama, Masato; Amagasaki, Motoki

doi:10.1016/j.scitotenv.2021.149876

Cited by 32 publications

(9 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, the positive effect of longer sliding windows is limited in improving the simulation of streamflow, consistent with findings of a previous research (Gao et al, 2020). However, since NSE and R 2 were almost in the identical format of equations in simulation, a R 2 around 0.6 was concluded to be effective in some cases (Bai et al, 2021;Moriasi et al, 2007;Yokoo et al, 2022), which is contrary to our study, due to differing accuracy requirements. In addition to such hyper-parameters, input data has a major impact on the performance of the proposed scheme at different stations.…”

Section: 1epcecmcsce Etfgsefllng Es Ce Nmtlcsnfgeffe Seccmflfisupporting

confidence: 81%

“…However, in an hourly prediction of shortterm runoff, the optimal time step of LSTM was detected to be unnecessary for the gradually stable accuracy, as time steps increase (Gao et al, 2020). Concerning the input data, imbalanced mass conservation was recognized in LSTM in the relationship of precipitation and flow discharge in the snow melting period (Yokoo et al, 2022), indicating its incapability besides energy conservation (Jia et al, 2019). Thus, concerning streamflow, samples are mainly generated on short-term runoff data in hours, with rare consideration on other units, like days.…”

Section: Kcye Ifeimentioning

confidence: 99%

See 1 more Smart Citation

Effects of stacking LSTM with different patterns and input schemes on streamflow and water quality simulation

Hu,

Jiang,

Yao

et al. 2024

Preprint

View full text Add to dashboard Cite

Streamflow and water quality parameters (WQs) are commonly forecasted by mechanism models and statistics models. However, these models are challenged due to computational complexity, redundant parameters, etc. Therefore, a stacking Long short-term memory networks (LSTM) model with two patterns and different input schemes was applied to simulate streamflow and eight WQs in this study. The results showed that sliding windows was detected as the more stable pattern for both forecasts. The accuracy of predicting streamflow using only meteorological inputs was limited especially with low-volume flow. Whereas, the prediction of WQs with three input variables (i.e., meteorological factors, streamflow, other influential WQs) was reliable reaching an average relative error (RE) below 17%. When adding historical data into the input dataset, both accuracies could be increased close to benchmarks of the Delft 3D model. Our study documents that the LSTM model is an effective method for streamflow and water quality forecasts.

show abstract

Section: 1epcecmcsce Etfgsefllng Es Ce Nmtlcsnfgeffe Seccmflfisupporting

confidence: 81%

Section: Kcye Ifeimentioning

confidence: 99%

Effects of stacking LSTM with different patterns and input schemes on streamflow and water quality simulation

Hu,

Jiang,

Yao

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…The Long Short-Term Memory (LSTM) model is a variant of the Recurrent Neural Network (RNN) model and is widely used for single-variable time series forecasting. The LSTM model was designed to address the vanishing gradient problem that traditional RNNs encounter [34]. It features an internal state called the cell state that can retain information over long periods, making it well-suited for capturing long-term dependencies in time series data.…”

Section: Lstm Modelmentioning

confidence: 99%

Exploring Time Series Models for Wind Speed Forecasting: A Comparative Analysis

Li,

Shen

et al. 2023

Energies

View full text Add to dashboard Cite

The sustainability and efficiency of the wind energy industry rely significantly on the accuracy and reliability of wind speed forecasting, a crucial concern for optimal planning and operation of wind power generation. In this study, we comprehensively evaluate the performance of eight wind speed prediction models, spanning statistical, traditional machine learning, and deep learning methods, to provide insights into the field of wind energy forecasting. These models include statistical models such as ARIMA (AutoRegressive Integrated Moving Average) and GM (Grey Model), traditional machine learning models like LR (Linear Regression), RF (random forest), and SVR (Support Vector Regression), as well as deep learning models comprising ANN (Artificial Neural Network), LSTM (Long Short-Term Memory), and CNN (Convolutional Neural Network). Utilizing five common model evaluation metrics, we derive valuable conclusions regarding their effectiveness. Our findings highlight the exceptional performance of deep learning models, particularly the Convolutional Neural Network (CNN) model, in wind speed prediction. The CNN model stands out for its remarkable accuracy and stability, achieving the lowest mean squared error (MSE), root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and the higher coefficient of determination (R2). This underscores the CNN model’s outstanding capability to capture complex wind speed patterns, thereby enhancing the sustainability and reliability of the renewable energy industry. Furthermore, we emphasized the impact of model parameter tuning and external factors, highlighting their potential to further improve wind speed prediction accuracy. These findings hold significant implications for the future development of the wind energy industry.

show abstract

“…With the development of computational hardware resources, deep learning methods have been applied in runoff forecasting [34] and have made progress compared with traditional machine learning methods [35]. Li et al applied convolutional deep belief networks (CDBNs) for rainfall-runoff simulation, and the CDBNs outperformed the traditional Xinanjiang model [36].…”

Section: Introductionmentioning

confidence: 99%

A New Rainfall-Runoff Model Using Improved LSTM with Attentive Long and Short Lag-Time

Chen

Huang

Wang

et al. 2022

Water

View full text Add to dashboard Cite

It is important to improve the forecasting performance of rainfall-runoff models due to the high complexity of basin response and frequent data limitations. Recently, many studies have been carried out based on deep learning and have achieved significant performance improvements. However, their intrinsic characteristics remain unclear and have not been explored. In this paper, we pioneered the exploitation of short lag-times in rainfall-runoff modeling and measured its influence on model performance. The proposed model, long short-term memory with attentive long and short lag-time (LSTM-ALSL), simultaneously and explicitly uses new data structures, i.e., long and short lag-times, to enhance rainfall-runoff forecasting accuracy by jointly extracting better features. In addition, self-attention is employed to model the temporal dependencies within long and short lag-times to further enhance the model performance. The results indicate that LSTM-ALSL yielded superior performance at four mesoscale stations (1846~9208 km2) with humid climates (aridity index 0.77~1.16) in the U.S.A., for both peak flow and base flow, with respect to state-of-the-art counterparts.

show abstract

Capabilities of deep learning models on learning physical relationships: Case of rainfall-runoff modeling with LSTM

Cited by 32 publications

References 30 publications

Effects of stacking LSTM with different patterns and input schemes on streamflow and water quality simulation

Effects of stacking LSTM with different patterns and input schemes on streamflow and water quality simulation

Exploring Time Series Models for Wind Speed Forecasting: A Comparative Analysis

A New Rainfall-Runoff Model Using Improved LSTM with Attentive Long and Short Lag-Time

Contact Info

Product

Resources

About