Long Short-Term Memory models of water quality in freshwater environments

Authors: PYO, JONG CHEOL - Pusan National University; Pachepsky, Yakov; KIM, SOOBIN - Ulsan National Institute Of Science And Technology (UNIST); ABBAS, ATHER - Ulsan National Institute Of Science And Technology (UNIST); KIM, MINJEONG - Korea Atomic Energy Research Institute (KAERI); KWON, JONGSUN - Korea Atomic Energy Research Institute (KAERI); LIGARAY, MAYZONEE - University Of The Philippines; CHO, KYUNGHWA - Ulsan National Institute Of Science And Technology (UNIST)

Submitted to: Water Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/14/2023
Publication Date: 11/16/2023
Citation: Pyo, J., Pachepsky, Y.A., Kim, S., Abbas, A., Kim, M., Kwon, J., Ligaray, M., Cho, K. 2023. Long Short-Term Memory models of water quality in freshwater environments. Water Research. https://doi.org/10.1016/j.wroa.2023.100207.
DOI: https://doi.org/10.1016/j.wroa.2023.100207

Interpretive Summary: Progress in environmental monitoring technologies has resulted in accumulating long-term observations from permanently installed sensors. Using such data for predictions has long been viewed as the essential goal. Predicted measurements often depended on recent observations and observations made in the distant past. Whereas standard statistical regression methods were not helpful for predictions in such situations, one artificial intelligence method, long short-term memory modeling (LSTM), appeared to be efficient. LSTM recently became popular in water quality research and applications. Our review showed that LSTM becomes more accurate when combined with other machine learning methods, such as convolutional neural networks and attention networks. LSTM is useful in estimating mission data and may benefit from data preprocessing. Utilizing site specific static information about the environmental settings holds promises. This work will be helpful to researchers and practitioners creating and using long sequences of ecological measurements, including water quality parameters.

Technical Abstract: Long short-term memory (LSTM) is a type of recurrent neural network that can account for longer-term traits of time-dependent data. It is the most important network used to predict the time series of water quality variables. First, we reviewed applications of a standalone LSTM, which demonstrated that it has the advantages of a short calculation time, high prediction accuracy, and good robustness compared with other machine learning models. This review was expanded into the LSTM model with data preprocessing techniques, including the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise method and Synchrosqueezed Wavelet Transform. The review then focused on the coupling of LSTM with a convolutional neural network, attention network, and transfer learning. The coupled networks demonstrated their performance over the standalone LSTM model. We also emphasize the influence of the static variables in the model and used the transformation method on the dataset. Outlook and further challenges were addressed. The outlook for research and application of LSTM in hydrology is concludes the review.