Can machine learning accelerate process understanding and decision‐relevant predictions of river water quality?

Abstract

The global decline of water quality in rivers and streams has resulted in a pressing need to design new watershed management strategies. Water quality can be affected by multiple stressors including population growth, land use change, global warming, and extreme events, with repercussions on human and ecosystem health. A scientific understanding of factors affecting riverine water quality and predictions at local to regional scales, and at sub‐daily to decadal timescales are needed for optimal management of watersheds and river basins. Here, we discuss how machine learning (ML) can enable development of more accurate, computationally tractable, and scalable models for analysis and predictions of river water quality. We review relevant state‐of‐the art applications of ML for water quality models and discuss opportunities to improve the use of ML with emerging computational and mathematical methods for model selection, hyperparameter optimization, incorporating process knowledge into ML models, improving explainablity, uncertainty quantification, and model‐data integration. We then present considerations for using ML to address water quality problems given their scale and complexity, available data and computational resources, and stakeholder needs. When combined with decades of process understanding, interdisciplinary advances in knowledge‐guided ML, information theory, data integration, and analytics can help address fundamental science questions and enable decision‐relevant predictions of riverine water quality.

Progress and opportunities in advancing near‐term forecasting of freshwater quality

“…Moreover, insights from marine and freshwater quantity forecasting may be particularly relevant to freshwater quality forecasting as all three disciplines involve aquatic ecosystems. For example, researchers are now applying machine learning methods long popular in freshwater quantity forecasting to water quality forecasting (reviewed by Poh Wai et al, 2022), and several challenges informed by the use of machine learning models in water quantity have been identified, including the need for knowledge‐guided machine learning, incorporation of uncertainty, transfer learning (i.e., models trained at data‐rich sites are then applied at data‐poor sites), and improved interpretability of model output (Khudhair et al, 2022; Poh Wai et al, 2022; Varadharajan et al, 2022). As another example, many of the lessons learned in the development and dissemination of predictive water quality guidance at marine beaches may readily transfer to freshwater beaches, such as the utility of three‐dimensional models for capturing diurnal fluctuations in water quality (Choi et al, 2022), methods for coordinating data collection among multiple agencies to assess urban water quality (Aznar et al, 2022), or the difficulty of developing adequate water quality predictive tools (e.g., Escherichia coli predictions) for beaches subject to frequent visits by large flocks of birds (U.S. EPA, 2016).…”

“…Freshwater quality forecasters can also apply lessons learned from marine and water quantity forecasters regarding, for example, model development (Varadharajan et al, 2022), forecast dissemination (Choi et al, 2022), and the ethical implications of providing operational forecasts (Hobday et al, 2019;Record & Pershing, 2021).…”

Progress and opportunities in advancing near-term forecasting of freshwater quality

“…Freshwater quality forecasters can also apply lessons learned from marine and water quantity forecasters regarding, e.g., model development (Varadharajan et al, 2022), forecast dissemination (Choi et al, 2022), and the ethical implications of providing operational forecasts (Hobday et al, 2019;Record & Pershing, 2021). Moreover, insights from marine and freshwater quantity forecasting may be particularly relevant to freshwater quality forecasting as all three disciplines involve aquatic ecosystems.…”

“…Moreover, insights from marine and freshwater quantity forecasting may be particularly relevant to freshwater quality forecasting as all three disciplines involve aquatic ecosystems. For example, researchers are now applying machine learning methods long popular in freshwater quantity forecasting to water quality forecasting (reviewed by Poh Wai et al, 2022), and several challenges informed by use of machine learning models in water quantity have been identified, including the need for knowledge-guided machine learning, incorporation of uncertainty, transfer learning (i.e., models trained at data-rich sites are then applied at data-poor sites), and improved interpretability of model output (Khudhair et al, 2022;Poh Wai et al, 2022;Varadharajan et al, 2022). As another example, many of the lessons learned in development and dissemination of predictive water quality guidance at marine beaches may readily transfer to freshwater beaches, such as the utility of three-dimensional models for capturing diurnal fluctuations in water quality (Choi et al, 2022), methods for coordinating data collection among multiple agencies to assess urban water quality (Aznar et al, 2022), or the difficulty of developing adequate water quality predictive tools (e.g., E. coli predictions) for beaches subject to frequent visits by large flocks of birds (U.S. EPA, 2016).…”

“…The acute threats to freshwater ecosystems from global change (Field et al, 2014;Maasri et al, 2022) highlight the potential of near-term freshwater forecasting for advancing water management and freshwater resource use, as well as our understanding of freshwater ecosystems (Bradford et al, 2018(Bradford et al, , 2020Coreau et al, 2009). Recent advances in next-generation technology for environmental monitoring of a broad range of freshwater ecosystem variables via in situ sensors, satellites, and internet of things (IoT) networks (Hestir et al, 2015;Marcé et al, 2016;Singh & Ahmed, 2021); development of diverse modeling, data assimilation, and uncertainty propagation methods in ecological studies (e.g., Chen et al, 2021;Heilman et al, 2022;Varadharajan et al, 2022); and a growing community of practice around ecological forecasting (Dietze & Lynch, 2019) are synergistically facilitating the increased production of near-term freshwater forecasts (Fig. 2).…”