Comparative efficiency of the SWAT model and a deep learning model in estimating nitrate loads at the Tuckahoe creek watershed, Maryland

Authors: LEE, JIYE - University Of Maryland; KIM, DONGHO - University Of Seoul; HONG, SEOKMIN - Ulsan National Institute Of Science And Technology (UNIST); YUN, DAEUN - Ulsan National Institute Of Science And Technology (UNIST); KWON, DOHUYCK - Ulsan National Institute Of Science And Technology (UNIST); HILL, ROBERT - University Of Maryland; Gao, Feng; Zhang, Xuesong; CHO, KYUNG HWA - Korea University; LEE, SANGCHUL - University Of Seoul; Pachepsky, Yakov

Submitted to: Water Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/11/2024
Publication Date: 9/20/2024
Citation: Lee, J., Kim, D., Yun, D., Kwon, D., Hill, R., Pachepsky, Y.A., Gao, F.N., Zhang, X., Lee, S., Cho, K. 2024. Comparative efficiency of the SWAT model and a deep learning model in estimating nitrate loads at the Tuckahoe creek watershed, Maryland. Water Research. 954. Article e176256. https://doi.org/10.1016/j.scitotenv.2024.176256.
DOI: https://doi.org/10.1016/j.scitotenv.2024.176256

Interpretive Summary: Excess nitrogen in surface waters leads to water quality degradation In particular, aquatic environmental health of the Chesapeake Bay in the U.S. Mid-Atlantic region has been threatened. To understand the nitrogenous fate in the Chesapeake Bay, modelling is necessary to simulate it and evaluate management practices. Both mechanistic Soil and Water Assessment Tool and data-driven deep learning model were used to simulate nitrate loads. The deep learning model showed higher accuracy and flexibility in estimating nitrate loads compared to the conventional modeling approach. Results of this work can be useful to provide tools for water quality planning and regulations decisions in that value of monitoring data can be increased by applying the machine learning modeling.

Technical Abstract: Modeling nitrate fate and transport in water sources is an essential component of predictive water quality management. Both mechanistic addata-driven models are currently in use. Mechanistic models, such as SWAT, simulate daily nitrate loads based on the results of simulating water flow. Data driven models are more flexible and allow one to simulate of nitrate loads and water flow independently. evaluate the performance of SWAT and deep learning model in terms of nutrient loads in cases when deep learning model is used in (a) independent simulations of flow series and nitrate concentration series, and (b) in both flow and concentration simulations to obtain nitrate load values. The data sere collected at the Tuckahoe Creek watershed in Maryland, USA.The data-driven deep learning model was built using long-short-term-memory (LSTM) and three-dimensional convolutional networks (3D Convolutional Networks) to simulate flow rate and nitrate concentration using weather data and imagery to derive leaf area index according to land use. SWAT was calibrated with data over training period 2014-2017 and validated with data over testing period qo19 to estimate flow rate and nitrate load. He Nash-Sutcliffe efficiency (NSE) was 0.31 and 0.40 for flow rate and -0.26 and -0.18 for the nitrate load rate over training and testing periods, respectively. Three data-driven modeling scenario were implemented. Scenario 1 used the observed flow rate and simulated nitrate concentration, scenario 2 used the simulated flow rate and observed nitrate concentration, and scenario 3 used the simulated flow rate and nitrate concentration. The deep learning model performed better than SWAT in all three scenarios with NSE from 0.49 too 0.58 for training and from 0.28 to 0.80 for testing periods with scenario 1 showing the best results. The difference in performance was most pronounced in fall and winter seasons. The deep learning modeling can be an efficient alternative to mechanistic watershed -scale water quality models provided the regular high frequency data collection is implemented.