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.
Reliable and fine resolution estimates of surface net-radiation are required for estimating latent and sensible heat fluxes between the land surface and the atmosphere. However, currently, fine resolution estimates of net-radiation are not available and consequently it is challenging to develop multi-year estimates of evapotranspiration at scales that can capture land surface heterogeneity and are relevant for policy and decision-making. We developed and evaluated a global net-radiation product at 5 km and 8-day resolution by combining mutually consistent atmosphere and land data from the Moderate Resolution Imaging Spectroradiometer (MODIS) on board Terra. Comparison with net-radiation measurements from 154 globally distributed sites (414 site-years) from the FLUXNET and Surface Radiation budget network (SURFRAD) showed that the net-radiation product agreed well with measurements across seasons and climate types in the extratropics (Wilmott’s index ranged from 0.74 for boreal to 0.63 for Mediterranean sites). Mean absolute deviation between the MODIS and measured net-radiation ranged from 38.0 ± 1.8 W∙m−2 in boreal to 72.0 ± 4.1 W∙m−2 in the tropical climates. The mean bias was small and constituted only 11%, 0.7%, 8.4%, 4.2%, 13.3%, and 5.4% of the mean absolute error in daytime net-radiation in boreal, Mediterranean, temperate-continental, temperate, semi-arid, and tropical climate, respectively. To assess the accuracy of the broader spatiotemporal patterns, we upscaled error-quantified MODIS net-radiation and compared it with the net-radiation estimates from the coarse spatial (1° × 1°) but high temporal resolution gridded net-radiation product from the Clouds and Earth’s Radiant Energy System (CERES). Our estimates agreed closely with the net-radiation estimates from the CERES. Difference between the two was less than 10 W·m−2 in 94% of the total land area. MODIS net-radiation product will be a valuable resource for the science community studying turbulent fluxes and energy budget at the Earth’s surface.
Research can be more transparent and collaborative by using Findable, Accessible, Interoperable, and Reusable (FAIR) principles to publish Earth and environmental science data. Reporting formats—instructions, templates, and tools for consistently formatting data within a discipline—can help make data more accessible and reusable. However, the immense diversity of data types across Earth science disciplines makes development and adoption challenging. Here, we describe 11 community reporting formats for a diverse set of Earth science (meta)data including cross-domain metadata (dataset metadata, location metadata, sample metadata), file-formatting guidelines (file-level metadata, CSV files, terrestrial model data archiving), and domain-specific reporting formats for some biological, geochemical, and hydrological data (amplicon abundance tables, leaf-level gas exchange, soil respiration, water and sediment chemistry, sensor-based hydrologic measurements). More broadly, we provide guidelines that communities can use to create new (meta)data formats that integrate with their scientific workflows. Such reporting formats have the potential to accelerate scientific discovery and predictions by making it easier for data contributors to provide (meta)data that are more interoperable and reusable.
Stream temperature () is an important water quality parameter that affects ecosystem health and human water use for beneficial purposes. Accurate predictions at different spatial and temporal scales can inform water management decisions that account for the effects of changing climate and extreme events. In particular, widespread predictions of in unmonitored stream reaches can enable decision makers to be responsive to changes caused by unforeseen disturbances. In this study, we demonstrate the use of classical machine learning (ML) models, support vector regression and gradient boosted trees (XGBoost), for monthly predictions in 78 pristine and human-impacted catchments of the Mid-Atlantic and Pacific Northwest hydrologic regions spanning different geologies, climate, and land use. The ML models were trained using long-term monitoring data from 1980–2020 for three scenarios: (1) temporal predictions at a single site, (2) temporal predictions for multiple sites within a region, and (3) spatiotemporal predictions in unmonitored basins (PUB). In the first two scenarios, the ML models predicted with median root mean squared errors (RMSE) of 0.69–0.84 °C and 0.92–1.02 °C across different model types for the temporal predictions at single and multiple sites respectively. For the PUB scenario, we used a bootstrap aggregation approach using models trained with different subsets of data, for which an ensemble XGBoost implementation outperformed all other modeling configurations (median RMSE 0.62 °C).The ML models improved median monthly estimates compared to baseline statistical multi-linear regression models by 15–48% depending on the site and scenario. Air temperature was found to be the primary driver of monthly for all sites, with secondary influence of month of the year (seasonality) and solar radiation, while discharge was a significant predictor at only 10 sites. The predictive performance of the ML models was robust to configuration changes in model setup and inputs, but was influenced by the distance to the nearest dam with RMSE <1 °C at sites situated greater than 16 and 44 km from a dam for the temporal single site and regional scenarios, and over 1.4 km from a dam for the PUB scenario. Our results show that classical ML models with solely meteorological inputs can be used for spatial and temporal predictions of monthly in pristine and managed basins with reasonable (<1 °C) accuracy for most locations.