Predicting Dynamic Riverine Nitrogen Export in Unmonitored Watersheds: Leveraging Insights of AI from Data-Rich Regions

Abstract

Terrestrial export of nitrogen is a critical Earth system process, but its global dynamics remain difficult to predict at a high spatiotemporal resolution. Here, we use deep learning (DL) to model daily riverine nitrogen export in response to hydrometeorological and anthropogenic drivers. Long short-term memory (LSTM) models for the daily concentration and flux of dissolved inorganic nitrogen (DIN) were built in a coastal watershed in southeastern China with a typical subtropical monsoon climate. The DL models exhibited excellent accuracy for both DIN concentration and flux, with Nash-Sutcliffe efficiency coefficients (NSEs) up to 0.67 and 0.92, respectively, a performance unlikely to be achieved by generic process-based models with comparable data quality. The flux model ensemble, without retraining, performed well (mean NSE = 0.32–0.84) in seven distinct watersheds in Asia, Europe, and North America, and retraining with multi-watershed data further improved the lowest NSE from 0.32 to 0.68. DL interpretation confirmed that interbasin consistency of riverine nitrogen export exists across different continents, which stems from the similarities in rainfall–runoff relationships. The multi-watershed flux model projects 0.60–12.4% increases in the nitrogen export to oceans from the studied watersheds under a 20% increase in fertilizer consumption, which rises to 6.7–20.1% with a 10% increase in runoff, indicating the synergistic effect of human activities and climate change. The DL-based method represents a successful case of explainable artificial intelligence in environmental science, providing a potential shortcut to a consistent understanding of the global daily-resolution dynamics of riverine nitrogen export under the currently limited data conditions.

Enhanced Insights into Effluent Prediction in Wastewater Treatment Plants: Comprehensive Deep Learning Model Explanation Based on SHAP

“…Unknown reasons behind predictions have been a shortcoming limiting DL driven by artificial intelligence (AI) to be fully harnessed in application fields. , To compensate for the shortcoming, explainable AI or explainable DL has become an emerging research area as it makes the black-box models more comprehensible and strengthens users’ confidence in the prediction . Many feature importance methods have been adopted to enhance the model explainability such as SHAP. ,− SHAP is based on the unification of game theory and local explanations and can be used for global and local explainability analysis of models .…”

“…Nevertheless, the structure also makes its black-box nature obvious, and its internal principles and mechanisms are unknown, which has led to controversy. 28 And explainable DL is now an emerging research front in many fields, and there have been attempts in the field of wastewater treatment process modeling. Hwangbo et al developed a deep neural network for process modeling and considered global sensitivity analysis based on variance decomposition to identify the key parameters.…”

“…The recurrent and gate structure determines the power of the LSTM in capturing long-term dependencies from a dynamic time series, − which matches the characteristics of WWTPs. Nevertheless, the structure also makes its black-box nature obvious, and its internal principles and mechanisms are unknown, which has led to controversy . And explainable DL is now an emerging research front in many fields, and there have been attempts in the field of wastewater treatment process modeling.…”

Contaminant Transport Modeling and Source Attribution With Attention‐Based Graph Neural Network

“…The Shapley value method (SHAP) quantifies the contribution of each participant in a cooperative game when their actions result in joint outcomes (Shapley, 1953). It can be used to measure feature importance in DL methods (Lundberg et al., 2020; Lundberg & Lee, 2017; Xiong et al., 2022). In this study, we evaluate the influence of each pollutant source via SHAP based on the groundwater contamination at spatial locations of interest predicted by our proposed aGNN.…”

Explaining the Mechanism of Multiscale Groundwater Drought Events: A New Perspective From Interpretable Deep Learning Model

“…With the advancement of explanation techniques, it has become increasingly possible to obtain generalizable, machine‐captured patterns. In this study, we apply two DL explanation techniques that have already shown promising results in the field of surface water research (Jiang et al., 2022; Xiong et al., 2022). The EG technique relies on the gradient of a model's output with respect to its input features to identify the contributions of each input (Erion et al., 2021).…”