In-stream Escherichia coli modeling using high-temporal-resolution data with deep learning and process-based models

Authors: ABBAS, ATHER - Ulsan National Institute Of Science And Technology (UNIST); BAEK, SANGSOO - Ulsan National Institute Of Science And Technology (UNIST); SILVERA, NORBERT - Sorbonne Universities, Paris; SOULILEUTH, BOUNSAMAY - National Agriculture And Forestry Research Institute (NAFRI); Pachepsky, Yakov; RIBOLZI, OLIVIER - University Of Toulouse; BOITHIAS, LAURIE - University Of Toulouse; CHO, KYUNG HWA - Ulsan National Institute Of Science And Technology (UNIST)

Submitted to: Hydrology and Earth System Sciences
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/15/2021
Publication Date: 12/6/2021
Citation: Abbas, A., Baek, S., Silvera, N., Soulileuth, B., Pachepsky, Y.A., Ribolzi, O., Boithias, L., Cho, K. 2021. In-stream Escherichia coli modeling using high-temporal-resolution data with deep learning and process-based models. Hydrology and Earth System Sciences. 25(12):6185-6202. https://doi.org/10.5194/hess-25-6185-2021.
DOI: https://doi.org/10.5194/hess-25-6185-2021

Interpretive Summary: Water quality predictions in catchments are commonly done with so called mechanistic water quality models that compute rates of water flow, chemical and biological degradation, and die-off, and concentrations of dissolved and suspended constituents of importance for water quality. These models cannot account for many site-specific intricacies of water quality development. Therefore, their accuracy is understandably limited. The alternative is to use a machine learning method to create a statistical model that t will learn specifics of water quality development for this site from the long-term monitoring data. We compared the above two methods of modeling with data for the catchment in the mountainous area, using the HSPF (Hydrological Simulation Program – Fortran) as the mechanistic model and the LSTM (long short term memory model) as the machine learning model. The LSTM substantially outperformed the HSPF in predicting E. coli transport in the small tropical catchment. The machine learning model will not have applications outside of the site where it was trained, but appears to provide a much better use of long term monitoring datasets. Results of this work will enable water quality management professionals to substantially increase the usefulness of the microbial water quality monitoring.

Technical Abstract: Correct estimation of fecal indicator bacteria in stream waters is critical for measures pertaining to public health. In this study, we modeled the transport of Escherichia coli in a small tropical catchment located in Lao PDR using a deep learning and a process-based model. The deep learning model was built using the long short-term memory (LSTM) technique, while the process-based model was constructed using the Hydrological Simulation Program–FORTRAN (HSPF). First, we calibrated both models with data on surface runoff as well as sub-surface flow. Then, simulated the E. coli transport with 6-min time step with both the HSPF and LSTM models. We show that the LSTM could be trained using the same input data as used by HSPF. The LSTM produced more realistic E. coli predictions. The LSTM model’s performance improved significantly for the mean absolute error (31%) as compared to that of HSPF. The coefficient of determination (R2) also improved from 0.18 to 0.31 during validation for LSTM model. Our method shows the incorporation of land use change information into LSTM model and the consequence of this new information on E. coli prediction. This study shows the application of deep learning-based models as an efficient alternative to process-based models for water quality modeling in catchments.