Abstract
In this study, a novel temporal convolutional neural network model is developed for long-term streamflow projection in California within the CAMELS watershed regions. The ensemble performance of the model is first compared with other machine learning models for streamflow prediction. The model is further assessed through comparison with reduced models and using different hyperparameters, with results suggesting that this model correctly ascertains the physical relationship between input variables and streamflow. The stability of the model and its behavior in the extrapolated regime is assessed through an idealized extreme test with quadruple precipitation and 5 C higher temperature. Future streamflow projections are then developed using daily high-resolution LOCA statistically downscaled input products. To understand the importance of the nonlinear machine learning approach, we estimate the degree of nonlinearity in the streamflow response among input variables. Our work shows the ability and potential for TCNNs to perform future hydrology projections.
Highlights
Streamflow is an undeniably important hydrologic quantity for agriculture, society and ecosystems
The Artificial Neural Networks (ANNs) model tends to achieve a higher Nash-Sutcliffe model efficiency (NSE) value than the linear regression model for almost all basins, but in terms of NSE the ANN is still inferior to the recurrent networks and the Temporal Convolutional Neural Network (TCNN), especially for basins where the NSE values for the recurrent networks and the TCNN are over 0.6, such as 11475560(NC) and 11522500(NC)
The relatively low NSE scores from the ANN indicate that there are some temporal features that the ANN cannot capture but which are important for streamflow prediction
Summary
Streamflow is an undeniably important hydrologic quantity for agriculture, society and ecosystems. Models for streamflow prediction and projection can be generally divided into two categories: physically-based models and data-driven models (Shen, 2018). Since physically-based hydrological models typically require significant computational expense and extensive calibration of land surface characteristics, machine learning (ML) models are being increasingly employed for streamflow prediction, especially Artificial Neural Networks (ANNs) (Gao et al, 2010; Noori and Kalin, 2016; Atieh et al, 2017; Peng et al, 2017), Support Vector Machines (SVMs) (Kisi and Cimen, 2011; Huang et al, 2014), and recurrent networks like Long-Short Term Memory (LSTM) (Feng et al, 2019; Kratzert et al, 2019; Le et al, 2019; Yan et al, 2019)
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
