Sensors in the Stream: The High-Frequency Wave of the Present

Evidence of deviations between experimental and empirical mixing lengths: Multi-discharge field tests in an arid river system

Despite advances in wastewater treatment plant (WWTP) efficiencies, multiple contaminants of concern, such as microplastics, pharmaceuticals, and per- and poly-fluoroalkyl substances (PFAS) remain largely untreated near discharge points and can be highly concentrated before they are fully mixed within the receiving river. Environmental agencies enforce mixing zone permits for the temporary exceedance of water quality parameters beyond targeted control levels under the assumption that contaminants are well-mixed and diluted downstream of mixing lengths, which are typically quantified using empirical equations derived from one-dimensional transport models. Most of these equations were developed in the 1970s and have been assumed to be standard practice since then. However, their development and validation lacked the technological advances required to test them in the field and under changing flow conditions. While new monitoring techniques such as remote sensing and infrared imaging have been employed to visualize mixing lengths and test the validity of empirical equations, those methods cannot be easily repeated due to high costs or flight restrictions. We investigated the application of Lagrangian and Eulerian monitoring approaches to experimentally quantify mixing lengths downstream of a WWTP discharging into the Rio Grande near Albuquerque, New Mexico (USA). Our data spans river to WWTP discharges ranging between 2-22x, thus providing a unique dataset to test long-standing empirical equations in the field. Our results consistently show empirical equations could not describe our experimental mixing lengths. Specifically, while our experimental data revealed "bell-shaped" mixing lengths as a function of increasing river discharges, all empirical equations predicted monotonically increasing mixing lengths. Those mismatches between experimental and empirical mixing lengths are likely due to the existence of threshold processes defining mixing at different flow regimes, i.e., jet diffusion at low flows, the Coanda effect at intermediate flows, and turbulent mixing at higher flows, which are unaccounted for by the one-dimensional empirical formulas. Our results call for a review of the use of empirical mixing lengths in streams and rivers to avoid widespread exposures to emerging contaminants.

Meta-Analysis of Optical Surrogates for the Characterization of Dissolved Organic Matter

Optical surrogates, derived from absorbance and fluorescence spectra, are widely used to infer dissolved organic matter (DOM) composition (molecular weight, aromaticity) and genesis (autochthonous vs allochthonous). Despite the broad adoption of optical surrogates, several limitations exist, such as context- and sample-specific factors. These limitations create uncertainty about how compositional interpretations based on optical surrogates are generalized across contexts, specifically if there is duplicative or contradictory information in those interpretations. To explore these limitations, we performed a meta-analysis of optical surrogates for DOM from diverse sources, both from natural systems and after water treatment processes (n = 762). Prior to analysis, data were screened using a newly developed, standardized methodology that applies systematic quality control criteria before reporting surrogates. There was substantial overlap in surrogate values from natural and treated samples, suggesting that the gradients governing the surrogate variability can be generated in both contexts. This overlap provides justification for using optical surrogates originally developed in the context of natural systems to describe DOM changes in engineered systems, although the interpretations may change. Absorbance-based surrogates that describe the amount of spectral tailing (e.g., E2:E3 and S275-295) had a high frequency of strong correlations with one another but not to specific absorbance (SUVA254) or absorbance slope ratio (SR). The fluorescence index (FI) and biological index (β/α) were strongly correlated with one another and to the peak emission wavelength but not to the humification index (HIX). Although SUVA254 and FI have both been correlated to DOM aromaticity in prior research, there was a lack of reciprocity between these optical surrogates across this data set. Additionally, there were patterns of deviations in the wastewater subset, suggesting that effluent organic matter may not follow conventional interpretations, urging caution in the use of optical surrogates to track DOM in water reuse applications. Finally, the meta-analysis highlights that three aspects should be captured when optical spectra are used for DOM interpretation: specific absorbance, absorbance tailing, and the extent of red-shifted fluorescence. We recommend that SUVA254, E2:E3, and FI or β/α be prioritized in future DOM studies to capture these aspects, respectively.

Multi-year time series of daily solute and isotope measurements from three Swiss pre-Alpine catchments

Time series analyses of solute concentrations in streamwater and precipitation are powerful tools for unraveling the interplay of hydrological and biogeochemical processes at the catchment scale. While such datasets are available for many sites around the world, they often lack the necessary temporal resolution or are limited in the number of solutes they encompass. Here we present a multi-year dataset encompassing daily records of major ions and a range of trace metals in both streamwater and precipitation in three catchments in the northern Swiss Pre-Alps. These time series capture the temporal variability observed in solute concentrations in response to storm events, snow melt, and dry summer conditions. This dataset additionally includes stable water isotope data as an extension of a publicly available isotope dataset collected concurrently at the same locations, and together these data can provide insights into a range of ecohydrological processes and enable a suite of analyses into hydrologic and biogeochemical catchment functioning.

Macroscale controls determine the recovery of river ecosystem productivity following flood disturbances

River ecosystem function depends on flow regimes that are increasingly modified by changes in climate, land use, water extraction, and flow regulation. Given the wide range of variation in flow regime modifications and autotrophic communities in rivers, it has been challenging to predict which rivers will be more resilient to flow disturbances. To better understand how river productivity is disturbed by and recovers from high-flow disturbance events, we used a continental-scale dataset of daily gross primary production time series from 143 rivers to estimate growth of autotrophic biomass and ecologically relevant flow disturbance thresholds using a modified population model. We compared biomass recovery rates across hydroclimatic gradients and catchment characteristics to evaluate macroscale controls on ecosystem recovery. Estimated biomass accrual (i.e., recovery) was fastest in wider rivers with less regulated flow regimes and more frequent instances of biomass removal during high flows. Although disturbance flow thresholds routinely fell below the estimated bankfull flood (i.e., the 2-y flood), a direct comparison of disturbance flows estimated by our biomass model and a geomorphic model revealed that biomass disturbance thresholds were usually greater than bed disturbance thresholds. We suggest that primary producers in rivers vary widely in their capacity to recover following flow disturbances, and multiple, interacting macroscale factors control productivity recovery rates, although river width had the strongest overall effect. Biomass disturbance flow thresholds varied as a function of geomorphology, highlighting the need for data such as bed slope and grain size to predict how river ecosystems will respond to changing flow regimes.

MAKING WAVES: Effluent to estuary: Does sunshine or shade reduce downstream footprints of cities?

Riparian tree canopies are key components of river systems, and influence the provision of many essential ecosystem services. Their management provides the potential for substantial control of the downstream persistence of pollutants. The recent advent of new advances in mass spectrometry to detect a large suite of emerging contaminants, high-frequency observations of water quality and gas exchange (e.g., aquatic eddy covariance), and improved spatial resolution in remote sensing (e.g., hyperspectral measurements and high-resolution imagery), presents new opportunities to understand and more comprehensively quantify the role of riparian canopies as Nature-based Solutions. The paper outlines how we may now couple these advances in observational technologies with developments in water quality modelling to integrate simulation of eutrophication impacts with organic matter dynamics and fate of synthetic toxic compounds. In particular regarding solar radiation drivers, this enables us to scale-up new knowledge of canopy-mediated photodegradation processes at a basin level, and integrate it with ongoing improvements in understanding of thermal control, eutrophication, and ecosystem metabolism.

Machine Learning-Assisted, Process-Based Quality Control for Detecting Compromised Environmental Sensors

Machine learning (ML) techniques promise to revolutionize environmental research and management, but collecting the necessary volumes of high-quality data remains challenging. Environmental sensors are often deployed under harsh conditions, requiring labor-intensive quality assurance and control (QAQC) processes. The need for manual QAQC is a major impediment to the scalability of these sensor networks. Existing techniques for automated QAQC make strong assumptions about noise profiles in the data they filter that do not necessarily hold for broadly deployed environmental sensors, however. Toward the goal of increasing the volume of high-quality environmental data, we introduce an ML-assisted QAQC methodology that is robust to low signal-to-noise ratio data. Our approach embeds sensor measurements into a dynamical feature space and trains a binary classification algorithm (Support Vector Machine) to detect deviation from expected process dynamics, indicating whether a sensor has become compromised and requires maintenance. This strategy enables the automated detection of a wide variety of nonphysical signals. We apply the methodology to three novel data sets produced by 136 low-cost environmental sensors (stream level, drinking water pH, and drinking water electroconductivity), deployed by our group across 250,000 km2 in Michigan, USA. The proposed methodology achieved accuracy scores of up to 0.97 and consistently outperformed state-of-the-art anomaly detection techniques.

Development of The Navigator: A Lagrangian sensing system to characterize surface freshwater ecosystems

Most freshwater aquatic studies rely on Eulerian monitoring, i.e., water quality and quantity are monitored using grab samples or semi-continuous sensors deployed at fixed cross-sections. While Eulerian monitoring is practical, it provides a limited understanding of spatial and temporal heterogeneity. We designed and built The Navigator, a Lagrangian (i.e., along a flow path) monitoring system that offers cost-effective solutions for in-situ, real-time data collection in surface freshwater ecosystems. The Navigator features a suite of technologies, including an autonomous surface vehicle with GPS and LTE connectivity, water quality sensors, a depth sonar, a camera, and a webpage dashboard to visualize real-time data. With these technologies, The Navigator provides insight into where, how, and why water quality and quantity change over time and space as it moves with the current or follows user-specified pathways. We tested The Navigator monitoring water quality parameters at high spatial-temporal resolution in multiple surface water bodies in New Mexico (USA) to: (1) identify water quality changes associated with land use changes along a 7th-order reach in the Rio Grande, (2) identify the fate of wildfire disturbances ∼175 km downstream of a burned watershed affected by the largest wildfire ever recorded in the state, (3) monitor the water quality of a recreational fishing pond in the City of Albuquerque. Our three successful tests confirm that The Navigator is an affordable (USD 5,101 in 2023) monitoring system that can be used to address questions involving mass and energy balances in surface waters.

Large-stream nitrate retention patterns shift during droughts: Seasonal to sub-daily insights from high-frequency data-model fusion

High-frequency nitrate-N (NO3--N) data are increasingly available, while accurate assessments of in-stream NO3--N retention in large streams and rivers require a better capture of complex river hydrodynamic conditions. This study demonstrates a fusion framework between high-frequency water quality data and hydrological transport models, that (1) captures river hydraulics and their impacts on solute signal propagation through river hydrodynamic modeling, and (2) infers in-stream retention as the differences between conservatively traced and reactively observed NO3--N signals. Using this framework, continuous 15-min estimates of NO3--N retention were derived in a 6th-order reach of the lower Bode River (27.4 km, central Germany), using long-term sensor monitoring data during a period of normal flow from 2015 to 2017 and a period of drought from 2018 to 2020. The unique NO3--N retention estimates, together with metabolic characteristics, revealed insightful seasonal patterns (from high net autotrophic removal in late-spring to lower rates, to net heterotrophic release during autumn) and drought-induced variations of those patterns (reduced levels of net removal and autotrophic nitrate removal largely buffered by heterotrophic release processes, including organic matter mineralization). Four clusters of diel removal patterns were identified, potentially representing changes in dominant NO3--N retention processes according to seasonal and hydrological conditions. For example, dominance of autotrophic NO3--N retention extended more widely across seasons during the drought years. Such cross-scale patterns and changes under droughts are likely co-determined by catchment and river environments (e.g., river primary production, dissolved organic carbon availability and its quality), which resulted in more complex responses to the sequential droughts. Inferences derived from this novel data-model fusion provide new insights into NO3- dynamics and ecosystem function of large streams, as well as their responses to climate variability. Moreover, this framework can be flexibly transferred across sites and scales, thereby complementing high-frequency monitoring to identify in-stream retention processes and to inform river management.

Challenges in studying water fluxes within the soil-plant-atmosphere continuum: A tracer-based perspective on pathways to progress

Tracing and quantifying water fluxes in the hydrological cycle is crucial for understanding the current state of ecohydrological systems and their vulnerability to environmental change. Especially the interface between ecosystems and the atmosphere that is strongly mediated by plants is important to meaningfully describe ecohydrological system functioning. Many of the dynamic interactions generated by water fluxes between soil, plant and the atmosphere are not well understood, which is partly due to a lack of interdisciplinary research. This opinion paper reflects the outcome of a discussion among hydrologists, plant ecophysiologists and soil scientists on open questions and new opportunities for collaborative research on the topic "water fluxes in the soil-plant-atmosphere continuum" especially focusing on environmental and artificial tracers. We emphasize the need for a multi-scale experimental approach, where a hypothesis is tested at multiple spatial scales and under diverse environmental conditions to better describe the small-scale processes (i.e., causes) that lead to large-scale patterns of ecosystem functioning (i.e., consequences). Novel in-situ, high-frequency measurement techniques offer the opportunity to sample data at a high spatial and temporal resolution needed to understand the underlying processes. We advocate for a combination of long-term natural abundance measurements and event-based approaches. Multiple environmental and artificial tracers, such as stable isotopes, and a suite of experimental and analytical approaches should be combined to complement information gained by different methods. Virtual experiments using process-based models should be used to inform sampling campaigns and field experiments, e.g., to improve experimental designs and to simulate experimental outcomes. On the other hand, experimental data are a pre-requisite to improve our currently incomplete models. Interdisciplinary collaboration will help to overcome research gaps that overlap across different earth system science fields and help to generate a more holistic view of water fluxes between soil, plant and atmosphere in diverse ecosystems.

Determination of Optimal Predictors and Sampling Frequency to Develop Nutrient Soft Sensors Using Random Forest

Despite advancements in sensor technology, monitoring nutrients in situ and in real-time is still challenging and expensive. Soft sensors, based on data-driven models, offer an alternative to direct nutrient measurements. However, the high demand for data required for their development poses logistical issues with data handling. To address this, the study aimed to determine the optimal subset of predictors and the sampling frequency for developing nutrient soft sensors using random forest. The study used water quality data at 15-min intervals from 2 automatic stations on the Main River, Germany, and included dissolved oxygen, temperature, conductivity, pH, streamflow, and cyclical time features as predictors. The optimal subset of predictors was identified using forward subset selection, and the models fitted with the optimal predictors produced R² values above 0.95 for nitrate, orthophosphate, and ammonium for both stations. The study then trained the models on 40 sampling frequencies, ranging from monthly to 15-min intervals. The results showed that as the sampling frequency increased, the model's performance, measured by RMSE, improved. The optimal balance between sampling frequency and model performance was identified using a knee-point determination algorithm. The optimal sampling frequency for nitrate was 3.6 and 2.8 h for the 2 stations, respectively. For orthophosphate, it was 2.4 and 1.8 h. For ammonium, it was 2.2 h for 1 station. The study highlights the utility of surrogate models for monitoring nutrient levels and demonstrates that nutrient soft sensors can function with fewer predictors at lower frequencies without significantly decreasing performance.

Value and limitations of machine learning in high-frequency nutrient data for gap-filling, forecasting, and transport process interpretation

High-frequency monitoring of water quality in catchments brings along the challenge of post-processing large amounts of data. Moreover, monitoring stations are often remote and technical issues resulting in data gaps are common. Machine learning algorithms can be applied to fill these gaps, and to a certain extent, for predictions and interpretation. The objectives of this study were (1) to evaluate six different machine learning models for gap-filling in a high-frequency nitrate and total phosphorus concentration time series, (2) to showcase the potential added value (and limitations) of machine learning to interpret underlying processes, and (3) to study the limits of machine learning algorithms for predictions outside the training period. We used a 4-year high-frequency dataset from a ditch draining one intensive dairy farm in the east of The Netherlands. Continuous time series of precipitation, evapotranspiration, groundwater levels, discharge, turbidity, and nitrate or total phosphorus were used as predictors for total phosphorus and nitrate concentrations respectively. Our results showed that the random forest algorithm had the best performance to fill in data-gaps, with R² higher than 0.92 and short computation times. The feature importance helped understanding the changes in transport processes linked to water conservation measures and rain variability. Applying the machine learning model outside the training period resulted in a low performance, largely due to system changes (manure surplus and water conservation) which were not included as predictors. This study offers a valuable and novel example of how to use and interpret machine learning models for post-processing high-frequency water quality data.

High-frequency sensor data capture short-term variability in Fe and Mn concentrations due to hypolimnetic oxygenation and seasonal dynamics in a drinking water reservoir

The biogeochemical cycles of iron (Fe) and manganese (Mn) in lakes and reservoirs have predictable seasonal trends, largely governed by stratification dynamics and redox conditions in the hypolimnion. However, short-term (i.e., sub-weekly) trends in Fe and Mn cycling are less well-understood, as most monitoring efforts focus on longer-term (i.e., monthly to yearly) time scales. The potential for elevated Fe and Mn to degrade water quality and impact ecosystem functioning, coupled with increasing evidence for high spatiotemporal variability in other biogeochemical cycles, necessitates a closer evaluation of the short-term Fe and Mn dynamics in lakes and reservoirs. We adapted a UV-visible spectrophotometer coupled with a multiplexor pumping system and partial least squares regression (PLSR) modeling to generate high spatiotemporal resolution predictions of Fe and Mn concentrations in a drinking water reservoir (Falling Creek Reservoir, Vinton, VA, USA) equipped with a hypolimnetic oxygenation (HOx) system. We quantified hourly Fe and Mn concentrations during two transitional periods: reservoir turnover (Fall 2020) and HOx initiation (Summer 2021). Our sensor system successfully predicted mean Fe and Mn concentrations and trends, ground-truthed by grab sampling and laboratory analysis. During fall turnover, hypolimnetic Fe and Mn concentrations began to decrease more than two weeks before complete mixing of the reservoir, with rapid equalization of epilimnetic and hypolimnetic Fe and Mn concentrations in less than 48 h after full water column mixing. During the initiation of HOx in Summer 2021, Fe and Mn displayed distinctly different responses to oxygenation, as indicated by the rapid oxidation of soluble Fe but not soluble Mn. This study demonstrates that Fe and Mn concentrations are sensitive to changes in redox conditions induced by stratification and oxygenation, although their responses to these changes differ. We also show that high spatio-temporal resolution predictions of Fe and Mn can improve drinking water monitoring programs and reservoir management practices.

Advances in Catchment Science, Hydrochemistry, and Aquatic Ecology Enabled by High-Frequency Water Quality Measurements

High-frequency water quality measurements in streams and rivers have expanded in scope and sophistication during the last two decades. Existing technology allows in situ automated measurements of water quality constituents, including both solutes and particulates, at unprecedented frequencies from seconds to subdaily sampling intervals. This detailed chemical information can be combined with measurements of hydrological and biogeochemical processes, bringing new insights into the sources, transport pathways, and transformation processes of solutes and particulates in complex catchments and along the aquatic continuum. Here, we summarize established and emerging high-frequency water quality technologies, outline key high-frequency hydrochemical data sets, and review scientific advances in key focus areas enabled by the rapid development of high-frequency water quality measurements in streams and rivers. Finally, we discuss future directions and challenges for using high-frequency water quality measurements to bridge scientific and management gaps by promoting a holistic understanding of freshwater systems and catchment status, health, and function.

Laboratory evaluation of open source and commercial electrical conductivity sensor precision and accuracy: How do they compare?

Variation in the electrical conductivity (EC) of water can reveal environmental disturbance and natural dynamics, including factors such as anthropogenic salinization. Broader application of open source (OS) EC sensors could provide an inexpensive method to measure water quality. While studies show that other water quality parameters can be robustly measured with sensors, a similar effort is needed to evaluate the performance of OS EC sensors. To address this need, we evaluated the accuracy (mean error, %) and precision (sample standard deviation) of OS EC sensors in the laboratory via comparison to EC calibration standards using three different OS and OS/commercial-hybrid (OS/C) EC sensors and data logger configurations and two commercial (C) EC sensors and data logger configurations. We also evaluated the effect of cable length (7.5 m and 30 m) and sensor calibration on OS sensor accuracy and precision. We found a significant difference between OS sensor mean accuracy (3.08%) and all other sensors combined (9.23%). Our study also found that EC sensor precision decreased across all sensor configurations with increasing calibration standard EC. There was also a significant difference between OS sensor mean precision (2.85 μS/cm) and the mean precision of all other sensors combined (9.12 μS/cm). Cable length did not affect OS sensor precision. Furthermore, our results suggest that future research should include evaluating how performance is impacted by combining OS sensors with commercial data loggers as this study found significantly decreased performance in OS/commercial-hybrid sensor configurations. To increase confidence in the reliability of OS sensor data, more studies such as ours are needed to further quantify OS sensor performance in terms of accuracy and precision across different settings and OS sensor and data collection platform configurations.

Evaluation of Chemcatcher® passive samplers for pesticide monitoring using high-frequency catchment scale data

Passive samplers (PS) have been proposed as an enhanced water quality monitoring solution in rivers, but their performance against high-frequency data over the longer term has not been widely explored. This study compared the performance of Chemcatcher® passive sampling (PS) devices with high-frequency sampling (HFS: 7-hourly to daily) in two dynamic rivers over 16 months. The evaluation was based on the acid herbicides MCPA (2-methyl-4-chlorophenoxyacetic acid), mecoprop-P, fluroxypyr and triclopyr. The impact of river discharge parameters on Chemcatcher® device performance was also explored. Mixed effects modelling showed that time-weighted mean concentration (TWMC) and flow-weighted mean concentration (FWMC) values obtained by the HFS approach were both significantly higher (p < 0.001) than TWMC values determined from PS regardless of river or pesticide. Modelling also showed that TWMC_PS values were more similar to TWMC_HFS than FWMC_HFS values. However, further testing revealed that MCPA TWMC values from HFS and PS were not significantly different (p > 0.05). There was little indication that river flow parameters altered PS performance-some minor effects were not significant or consistent. Despite this, the PS recovery of very low concentrations indicated that Chemcatcher® devices may be used to evaluate the presence/absence and magnitude of acid herbicides in hydrologically dynamic rivers in synoptic type surveys where space and time coverage is required. However, a period of calibration of the devices in each river would be necessary if they were intended to provide a quantitative review of pesticide concentration as compared with HFS approaches.

River metabolic fingerprints and regimes reveal ecosystem responses to enhanced wastewater treatment

Although many studies have examined how improvements in wastewater treatment impact river nutrient concentrations and loads, there has been much less focus on measuring river metabolism to evaluate the wider aquatic ecosystem benefits of reducing nutrient inputs to rivers. The objectives of this study were to evaluate the effects of enhanced wastewater treatment (nitrification) on river metabolism in the Grand River, Canada's largest river draining into Lake Erie. Metabolic fingerprints and regimes (calculated from high-frequency dissolved oxygen [DO] measurements) were used to visualize whole-river ecosystem functional responses to these wastewater treatment upgrades. There was a 60% reduction in ecosystem respiration during summer, in response to reductions in effluent total ammonia inputs, causing a shift from net heterotrophy to net autotrophy, and contraction of river metabolic fingerprints. This resulted in major improvements in summer DO concentrations, with reductions in the percentage of days during summer that DO minima fell below water-quality guidelines for protection of aquatic early life stages, from 88% to ≤16%. The results also point to potential cascading impacts on coupled phosphorus and nitrogen cycles, which may generate further improvements in river water quality. During the summer, high rates of river metabolism and nutrient retention may result in measured water-column nutrient concentrations potentially underestimating nutrient pressures. This study also demonstrates the value of combining river metabolism with nutrient monitoring for a more holistic understanding of the role of nutrients in river ecosystem health and function.

Optimising punctual water sampling with an on-the-fly algorithm based on multiparameter high-frequency measurements

The way in which aquatic systems is sampled has a strong influence on our understanding of them, especially when they are highly dynamic. High frequency sampling has the advantage over spot sampling for representativeness but leads to a high amount of analysis. This study proposes a new methodology to choose when sampling accurately with an automated sampler coupled with a high frequency (HF) multiparameter probe. After each HF measurement, an optimised sampling algorithm (OSA) determines on-the-fly the relevance of taking a new sample in relation to previous waters already collected. Once the OSA was optimised, considering the number of HF parameters and their variabilities, it was demonstrated through a study case that the number of samples could be significantly reduced, while still covering periods of low and high variabilities. The comparison between the total HF dataset and the sampled subdataset shows that physicochemical parameter variability is preserved (Pearson correlations > 0.96) as well as the multiparameter variability (PCA axes remained similar with Tucker congruence > 0.99). This algorithm simplifies HF studies by making it easier to take samples during brief phenomena such as storms or accidental spills that are often poorly monitored. In addition, it optimises the number of samples to be taken to correctly describe a system and thus reduce the human and financial costs of these environmental studies.

Reservoir water quality deterioration due to deforestation emphasizes the indirect effects of global change

Deforestation is currently a widespread phenomenon and a growing environmental concern in the era of rapid climate change. In temperate regions, it is challenging to quantify the impacts of deforestation on the catchment dynamics and downstream aquatic ecosystems such as reservoirs and disentangle these from direct climate change impacts, let alone project future changes to inform management. Here, we tackled this issue by investigating a unique catchment-reservoir system with two reservoirs in distinct trophic states (meso‑ and eutrophic), both of which drain into the largest drinking water reservoir in Germany. Due to the prolonged droughts in 2015-2018, the catchment of the mesotrophic reservoir lost an unprecedented area of forest (exponential increase since 2015 and ca. 17.1% loss in 2020 alone). We coupled catchment nutrient exports (HYPE) and reservoir ecosystem dynamics (GOTM-WET) models using a process-based modeling approach. The coupled model was validated with datasets spanning periods of rapid deforestation, which makes our future projections highly robust. Results show that in a short-term time scale (by 2035), increasing nutrient flux from the catchment due to vast deforestation (80% loss) can turn the mesotrophic reservoir into a eutrophic state as its counterpart. Our results emphasize the more prominent impacts of deforestation than the direct impact of climate warming in impairment of water quality and ecological services to downstream aquatic ecosystems. Therefore, we propose to evaluate the impact of climate change on temperate reservoirs by incorporating a time scale-dependent context, highlighting the indirect impact of deforestation in the short-term scale. In the long-term scale (e.g. to 2100), a guiding hypothesis for future research may be that indirect effects (e.g., as mediated by catchment dynamics) are as important as the direct effects of climate warming on aquatic ecosystems.

Machine learning-based estimation of riverine nutrient concentrations and associated uncertainties caused by sampling frequencies

Accurate and sufficient water quality data is essential for watershed management and sustainability. Machine learning models have shown great potentials for estimating water quality with the development of online sensors. However, accurate estimation is challenging because of uncertainties related to models used and data input. In this study, random forest (RF), support vector machine (SVM), and back-propagation neural network (BPNN) models are developed with three sampling frequency datasets (i.e., 4-hourly, daily, and weekly) and five conventional indicators (i.e., water temperature (WT), hydrogen ion concentration (pH), electrical conductivity (EC), dissolved oxygen (DO), and turbidity (TUR)) as surrogates to individually estimate riverine total phosphorus (TP), total nitrogen (TN), and ammonia nitrogen (NH₄⁺-N) in a small-scale coastal watershed. The results show that the RF model outperforms the SVM and BPNN machine learning models in terms of estimative performance, which explains much of the variation in TP (79 ± 1.3%), TN (84 ± 0.9%), and NH₄⁺-N (75 ± 1.3%), when using the 4-hourly sampling frequency dataset. The higher sampling frequency would help the RF obtain a significantly better performance for the three nutrient estimation measures (4-hourly > daily > weekly) for R² and NSE values. WT, EC, and TUR were the three key input indicators for nutrient estimations in RF. Our study highlights the importance of high-frequency data as input to machine learning model development. The RF model is shown to be viable for riverine nutrient estimation in small-scale watersheds of important local water security.

Chlorophyll and phycocyanin in-situ fluorescence in mixed cyanobacterial species assemblages: Effects of morphology, cell size and growth phase

Cyanobacteria harmful blooms can represent a major risk for public health due to potential release of toxins and other noxious compounds in the water. A continuous and high-resolution monitoring of the cyanobacteria population is required due to their rapid dynamics, which has been increasingly done using in-situ fluorescence of phycocyanin (f-PC) and chlorophyll a (f-Chl a). Appropriate in-situ fluorometers calibration is essential because f-PC and f-Chl a are affected by biotic and abiotic factors, including species composition. Measurement of f-PC and f-Chl a in mixed species assemblages during different growth phases - representative of most field conditions - has received little attention. We hypothesized that f-PC and f-Chl a of mixed assemblages of cyanobacteria may be accurately estimated if taxa composition and fluorescence characteristics are known. We also hypothesized that species with different morphologies would have different fluorescence per unit cell and biomass. We tested these hypotheses in a controlled culture experiment in which photosynthetic pigment fluorescence, chemical pigment extraction, optical density and microscopic enumeration of four common cyanobacteria species (Aphanocapsa sp, Microcystis aeruginosa, Dolichospermum circinale and Raphidiopsis raciborskii) were quantified. Both monocultures and mixed cultures were monitored from exponential to late stationary growth phases. The sum of fluorescence of individual species calculated for mixed samples was not significantly different than measured fluorescence of mixed cultures. Estimated and measured f-PC and f-Chl a of mixed cultures had higher correlations and smaller absolute median errors when estimations were based on fluorescence per biomass instead of fluorescence per cell. Largest errors were overestimations of measured fluorescence for species with different morphologies. Fluorescence per cell was significantly different among most species, while fluorescence per unit biomass was not, indicating that conversion of fluorescence to biomass reduces species-specific bias. This study presents new information on the effect of species composition on cyanobacteria fluorescence. Best practices of deployment and operation of fluorometers, and data-driven models supporting in-situ fluorometers calibration are discussed as suitable solutions to minimize taxa-specific bias in fluorescence estimates.

The role of terrestrial productivity and hydrology in regulating aquatic dissolved organic carbon concentrations in boreal catchments

The past decades have witnessed an increase in dissolved organic carbon (DOC) concentrations in the catchments of the Northern Hemisphere. Increasing terrestrial productivity and changing hydrology may be reasons for the increases in DOC concentration. The aim of this study is to investigate the impacts of increased terrestrial productivity and changed hydrology following climate change on DOC concentrations. We tested and quantified the effects of gross primary production (GPP), ecosystem respiration (RE) and discharge on DOC concentrations in boreal catchments over 3 years. As catchment characteristics can regulate the extent of rising DOC concentrations caused by the regional or global environmental changes, we selected four catchments with different sizes (small, medium and large) and landscapes (forest, mire and forest-mire mixed). We applied multiple models: Wavelet coherence analysis detected the delay-effects of terrestrial productivity and discharge on aquatic DOC variations of boreal catchments; thereafter, the distributed-lag linear models quantified the contributions of each factor on DOC variations. Our results showed that the combined impacts of terrestrial productivity and discharge explained 62% of aquatic DOC variations on average across all sites, whereas discharge, gross primary production (GPP) and RE accounted for 26%, 22% and 3%, respectively. The impact of GPP and discharge on DOC changes was directly related to catchment size: GPP dominated DOC fluctuations in small catchments (<1 km² ), whereas discharge controlled DOC variations in big catchments (>1 km² ). The direction of the relation between GPP and discharge on DOC varied. Increasing RE always made a positive contribution to DOC concentration. This study reveals that climate change-induced terrestrial greening and shifting hydrology change the DOC export from terrestrial to aquatic ecosystems. The work improves our mechanistic understanding of surface water DOC regulation in boreal catchments and confirms the importance of DOC fluxes in regulating ecosystem C budgets.

Four years of daily stable water isotope data in stream water and precipitation from three Swiss catchments

Time series of the natural isotopic composition (²H, ¹⁸O) of precipitation and streamwater can provide important insights into ecohydrological phenomena at the catchment scale. However, multi-year, high-frequency isotope datasets are generally scarce, limiting our ability to study highly dynamic short-term ecohydrological processes. Here we present four years of daily isotope measurements in streamwater and precipitation at the Alp catchment (area 47 km²) in Central Switzerland and two of its tributaries (0.7 km² and 1.6 km²). This data set reveals short-term responses of streamflow isotopes to precipitation events, which otherwise remain obscured when isotopes are sampled weekly or monthly. The observations span the period June 2015 through May 2019, during which several hydrometeorologic extreme events occurred, including a very dry summer in 2018 and below-average snow accumulation in winter 2016/2017. In addition, we provide daily time series of key hydrometeorological variables that, in combination with the isotope data, can be useful for assessing the robustness of ecohydrological models.

Measurement(s)	((18)O)water • deuterium(.) • waterflux • Relative Humidity • Air temperature • Snow depth
Technology Type(s)	wavelength scanned cavity ring-down spectrometer • off-axis integrated cavity output spectrometry • stream gauge, rain gauge • weather station
Factor Type(s)	catchment • location_ID • source • waterflux_measured • precipitation_interpolated • analysis_method • delta_2H • delta_18O • rel_humidity • air_temperature • snow_depth • notes_sampling • notes_other
Sample Characteristic - Environment	continental temperate

Machine-accessible metadata file describing the reported data: 10.6084/m9.figshare.17013848

DeepSense: A Physics-Guided Deep Learning Paradigm for Anomaly Detection in Soil Gas Data at Geologic CO2 Storage Sites

Driven by the collection of enormous amounts of streaming data from sensors, and with the emergence of the internet of things, the need for developing robust detection techniques to identify data anomalies has increased recently. The algorithms for anomaly detection are required to be selected based on the type of data. In this study, we propose a predictive anomaly detection technique, DeepSense, which is applied to soil gas concentration data acquired from sensors being used for environmental characterization at a prospective CO₂ storage site in Queensland, Australia. DeepSense takes advantage of deep-learning algorithms as its predictor module and uses a process-based soil gas method as the basis of its anomaly detector module. The proposed predictor framework leverages the power of convolutional neural network algorithms for feature extraction and simultaneously captures the long-term temporal dependency through long short-term memory algorithms. The proposed process-based anomaly detection method is a cost-effective alternative to the conventional concentration-based soil gas methodologies which rely on long-term baseline surveys for defining the threshold level. The results indicate that the proposed framework performs well in diagnosing anomalous data in soil gas concentration data streams. The robustness and efficacy of the DeepSense were verified against data sets acquired from different monitoring stations of the storage site.

Degree of Anthropogenic Land Disturbance Controls Fluvial Sediment Hysteresis

Storm events dominate sediment delivery to stream corridors, but the effects of anthropogenic disturbances on altering the sources, pathways, and timing of delivery remain uncertain. To address this knowledge gap, we analyzed 849 events from over a decade of high-frequency turbidity data across five watersheds in an urbanization gradient. Sensing results suggested that hysteresis patterns evolved with land use from clockwise (low-rural) to figure-eight (high-rural) to counter-clockwise (high-urban), indicating a disturbance-driven shift of sediment provenance from proximal to distal. Sediment loading pathways in the lowest-disturbance rural watershed were dominated by a single hysteresis shape (>90% of export by clockwise events), whereas the most-disturbed urban basin had the greatest variability in loading pathways (∼25% of export by clockwise, counter-clockwise, figure-eight, and complex events, respectively). Finally, wastewater treatment facilities modulated the release of "hungry-water" baseflow, causing more-rapid periods of high streamflow variance in catchments with a treatment facility (∼4 h period) than in those without (∼6 h period). Together, our results indicate that anthropogenic disturbances, including tile drainage, impervious surfaces, and roadway density, increase the connectivity of distally located sediment that would-in undisturbed basins-deposit along the sediment cascade. This information is important to watershed managers as they mitigate erosion in developing basins.

A Multiscale Approach to Timescale Analysis: Isolating Diel Signals from Solute Concentration Time Series

Solute concentration time series reflect hydrological and biological drivers through various frequencies, phases, and amplitudes of change. Untangling these signals facilitates the understanding of dynamic ecosystem conditions and transient water quality issues. A case in point is the inference of biogeochemical processes from diel solute concentration variations. This analysis requires approaches capable of isolating subtle diel signals from background variability at other scales. Conventional time series analyses typically assume stationary or deterministic background variability; however, most rivers do not respect such niceties. We developed a time-series filtering method that uses empirical mode decomposition to decompose a measured solute concentration time series into intrinsic mode frequencies. Based on externally supplied mechanistic knowledge, we then filter these modes by periodicity, phase, and coherence with neighboring days. This method is tested on three synthetic series that incorporate environmental variability and sensor noise and on a year of 15 min sampled concentration time series from three hydrologically and ecologically distinct rivers in the eastern United States. The proposed method successfully isolated signals in the measured data sets that corresponded with variability in gross primary productivity. The strength the diel signal isolated through this method was smaller compared to the true signal in the synthetic series; however, uncertainty analysis showed that the process-model-based estimates derived from these signals were similar to other inference methods. This signal decomposition method retains information that can be used for further process modeling while making different assumptions about the data than Fourier and wavelet analyses.

Evaluating the Performance of Machine Learning Approaches to Predict the Microbial Quality of Surface Waters and to Optimize the Sampling Effort

Exposure to contaminated water during aquatic recreational activities can lead to gastrointestinal diseases. In order to decrease the exposure risk, the fecal indicator bacteria Escherichia coli is routinely monitored, which is time-consuming, labor-intensive, and costly. To assist the stakeholders in the daily management of bathing sites, models have been developed to predict the microbiological quality. However, model performances are highly dependent on the quality of the input data which are usually scarce. In our study, we proposed a conceptual framework for optimizing the selection of the most adapted model, and to enrich the training dataset. This frameword was successfully applied to the prediction of Escherichia coli concentrations in the Marne River (Paris Area, France). We compared the performance of six machine learning (ML)-based models: K-nearest neighbors, Decision Tree, Support Vector Machines, Bagging, Random Forest, and Adaptive boosting. Based on several statistical metrics, the Random Forest model presented the best accuracy compared to the other models. However, 53.2 ± 3.5% of the predicted E. coli densities were inaccurately estimated according to the mean absolute percentage error (MAPE). Four parameters (temperature, conductivity, 24 h cumulative rainfall of the previous day the sampling, and the river flow) were identified as key variables to be monitored for optimization of the ML model. The set of values to be optimized will feed an alert system for monitoring the microbiological quality of the water through combined strategy of in situ manual sampling and the deployment of a network of sensors. Based on these results, we propose a guideline for ML model selection and sampling optimization.

Large spatiotemporal variability in metabolic regimes for an urban stream draining four wastewater treatment plants with implications for dissolved oxygen monitoring

Urbanization and subsequent expansion of wastewater treatment plant (WWTP) capacity has the potential to alter stream metabolic regimes, but the magnitude of this change remains unknown. Indeed, our understanding of downstream WWTP effects on stream metabolism is spatially and temporally limited, and monitoring designs with upstream-downstream comparison sites are rare. Despite this, and despite observed spatiotemporal variability in stream metabolic regimes, regulators typically use snapshot monitoring to assess ecosystem function in receiving streams, potentially leading to biased conclusions about stream health. To address these important practical issues, we assessed the spatiotemporal variability in stream metabolism at nine sites upstream and downstream of four WWTPs in a suburban stream. We used one year (2017–2018) of high-frequency dissolved oxygen (DO) data to model daily gross primary productivity (GPP) and ecosystem respiration (ER). We found that GPP was 1.7–4.0 times higher and ER was 1.2–7.2 times higher downstream of the WWTPs, especially in spring when light was not limited by canopy shading. Critically, we observed that these effects were spatially limited to the kilometer or so just downstream of the plant. These effects were also temporally limited, and metabolic rates upstream of WWTPs were not different from sites downstream of the plant after leaf-out at some sites. Across sites, regardless of their relation to WWTPs, GPP was positively correlated with potential incident light suggesting that light is the dominant control on GPP in this system. Temporal windowing of DO to proposed regulatory monitoring lengths revealed that the violation frequency of water quality criteria depended on both the monitoring interval and start date. We conclude that spatiotemporal variability in metabolism and DO are crucial considerations when developing monitoring programs to assess ecosystem function, and that evidence of WWTP effects may only arise during high light conditions and at limited scales.

The impacts of biofouling on automated phosphorus analysers during long-term deployment

In-situ nutrient analysers are a promising tool for improving the temporal resolution of data and filling knowledge gaps in drivers of harmful algal blooms. There are significant challenges however regarding instrument biofouling and data drift, which remain largely unquantified and unresolved. In this study the effects of biofouling on data consistency and accuracy is quantified on automated wet chemical analysers during long-term monitoring. In 2019 three fractions of phosphorus (P); total phosphorus (TP), total dissolved phosphorus (TDP) and soluble reactive phosphorus (SRP), were measured in-situ at four sites in Southern Ontario, Canada. The analysers were exposed to a wide range of P concentrations and biofouling extremes over an 8-month period. They were calibrated using chemical standards both in the field and the lab, and validated with fortnightly grab samples, and the representativeness of real-time data under a range of biofouling conditions were analysed. Results show that analysers biofouling during long-term deployment can desensitize instrument measurements, with greatest impacts on instruments operating in highly turbid environments. Temporal changes in calibration curves suggest that equilibrium P concentrations (EPC₀) of sediments accumulating inside filters can elicit a rapid exchange of dissolved P (SRP, TDP) with the water sample. Data drift increases the further from the EPC₀ an instrument is required to analyse, and thus this study demonstrates that for in-situ P monitoring, unless filters are frequently replaced or renovated, in-situ probes should ideally be dedicated to a specific waterbody type defined by similar EPC₀ values. It is recommended that in order to ensure accuracy in in-situ monitoring of TP, TDP and SRP during long-term deployment, preliminary site trials should be conducted to ascertain sediment EPC₀; the extent of biofouling should be monitored; and/or frequent grab samples taken for post-deployment validation. The findings apply to any in-situ phosphorus monitoring techniques for SRP or TDP.

Development of statistical models for estimating daily nitrate load in Iowa

There is an ongoing need to increase our understanding of the sources and timing of stream nitrate loads across agricultural watersheds in Iowa as water quality improvement strategies are implemented. The goal of this study was to model the relationship between nitrate load and the two components of streamflow (i.e., baseflow and stormflow) to quantify in-stream nitrate patterns and develop a new method for estimating loads on days when monitoring data are not available. We analyzed eight watersheds in Iowa that had long-term water quality data where grab samples have been collected from 1987 to 2019. Four regression models were developed that related daily nitrate load to daily baseflow, stormflow, and streamflow discharge. The first model considered baseflow as a predictor, the second model used stormflow, the third model included both baseflow and stormflow as two different covariates, and the final model used total streamflow (unseparated). For all eight watersheds, the baseflowstormflow models had the highest correlation coefficients, which indicates that both components are necessary and together improve nitrate load estimates. While baseflow models estimated lower nitrate loads better, stormflow models captured the variability associated with larger loads. In addition, streamflow models tended to overestimate large nitrate loads. This simple modeling framework can be used to calculate daily, monthly and annual nitrate loads. Delineating nitrate loads between stormflow and baseflow can help identify differences in nitrate sources for nutrient reduction and remediation.

Costs and benefits of automated high-frequency environmental monitoring - The case of lake water management

Freshwater lakes are dynamic ecosystems and provide multiple ecosystem services to humans. Sudden changes in lake environmental conditions such as cyanobacterial blooms can negatively impact lake usage. Automated high-frequency monitoring (AHFM) systems allow the detection of short-lived extreme and unpredictable events and enable lake managers to take mitigation actions earlier than if basing decisions on conventional monitoring programmes. In this study a cost-benefit approach was used to compare the costs of implementing and running an AHFM system with its potential benefits for three case study lakes. It was shown that AHFM can help avoid human health impacts, lost recreation opportunities, and revenue losses for livestock, aquaculture and agriculture as well as reputational damages for drinking water treatment. Our results showed that the largest benefits of AHFM can be expected in prevention of human health impacts and reputational damages. The potential benefits of AHFM, however, do not always outweigh installation and operation costs. While for Lake Kinneret (Israel) over a 10-year period, the depreciated total benefits are higher than the depreciated total costs, this is not the case for Lough Gara (Ireland). For Lake Mälaren in Sweden it would depend on the configuration of the AHFM system, as well as on how the benefits are calculated. In general, the higher the frequency and severity of changes in lake environmental conditions associated with detrimental consequences for humans and the higher the number of lake users, the more likely it is that the application of an AHFM system is financially viable.

Prediction of stream nitrogen and phosphorus concentrations from high-frequency sensors using Random Forests Regression

Stream nutrient concentrations exhibit marked temporal variation due to hydrology and other factors such as the seasonality of biological processes. Many water quality monitoring programs sample too infrequently (i.e., weekly or monthly) to fully characterize lotic nutrient conditions and to accurately estimate nutrient loadings. A popular solution to this problem is the surrogate-regression approach, a method by which nutrient concentrations are estimated from related parameters (e.g., conductivity or turbidity) that can easily be measured in situ at high frequency using sensors. However, stream water quality data often exhibit skewed distributions, nonlinear relationships, and multicollinearity, all of which can be problematic for linear-regression models. Here, we use a flexible and robust machine learning technique, Random Forests Regression (RFR), to estimate stream nitrogen (N) and phosphorus (P) concentrations from sensor data within a forested, mountainous drainage area in upstate New York. When compared to actual nutrient data from samples tested in the laboratory, this approach explained much of the variation in nitrate (89%), total N (85%), particulate P (76%), and total P (74%). The models were less accurate for total soluble P (47%) and soluble reactive P (32%), though concentrations of these latter parameters were in a relatively low range. Although soil moisture and fluorescent dissolved organic matter are not commonly used as surrogates in nutrient-regression models, they were important predictors in this study. We conclude that RFR shows great promise as a tool for modeling instantaneous stream nutrient concentrations from high-frequency sensor data, and encourage others to evaluate this approach for supplementing traditional (laboratory-determined) nutrient datasets.

Assessment of a portable UV-Vis spectrophotometer's performance in remote areas: Stream water DOC, Fe content and spectral data

This paper presents data for the assessment of a portable UV-Vis spectrophotometer's performance on predicting stream water DOC and Fe content. The dataset contains DOC and Fe concentrations by laboratory methods, in-situ and ex-situ spectral absorbances, monitoring environmental indexes such as water depth, temperature, turbidity and voltage. The records in Yli-Nuortti river (Cold station, Finland) took place during the hydrological year 2018-2019 and in Krycklan (C4 and C5, Sweden) during the hydrological years 2016-2019. The data analyses were conducted with 'pls' and 'caret' package in R. The correlation coefficient (R), root-mean-square deviation (RMSD), standard deviation (STD) and bias were used to check the performance of the models. This dataset can be combined with datasets from other regions around the world to build more universal models. For discussion and more information of the dataset creation, please refer to the full-length article "Assessment of a portable UV-Vis spectrophotometer's performance for stream water DOC and Fe content monitoring in remote areas" [1].

Combining in-situ fluorometry and distributed rainfall data provides new insights into natural organic matter transport dynamics in an urban river

Urbanization alters the quality and quantity of Dissolved Organic Matter (DOM) fluxes to rivers potentially leading to water quality problems and impaired ecosystem function. Traditional synoptic and point sampling approaches are generally inadequate for monitoring DOM source dynamics. To identify links between spatial heterogeneity in precipitation and DOM dynamics, we used a unique approach combining high spatial and temporal resolution precipitation datasets featuring point, catchment, and land-cover weighted precipitation to characterise catchment transport dynamics. These datasets were linked to fluorescence records from an urban stream (Bourn Brook, Birmingham, UK). Humic-like fluorescence (HLF: Ex. 365 nm, Em. 490 nm) and Tryptophan-like fluorescence (TLF: Ex. 285 nm, Em. 340 nm) were measured, (plus river flow and turbidity) at 5 min intervals for 10 weeks during Autumn 2017. The relationship between discharge (Q) and concentration (C) for TLF and HLF were strongly chemodynamic at low Q (<Q50) but TLF was chemostatic when Q exceeded this threshold. Figure of eight hysteresis was the most common response type for both HLF and TLF, indicating that DOM sources shift within and between events. Key drivers of DOM dynamics were identified using regression analysis and model outputs using point, catchment-averaged, and land-use weighted precipitation were compared. Antecedent rainfall was identified as the most important predictor (negative relationship) of TLF and HLF change suggesting DOM source exhaustion. Precipitation weighted by land cover showed that urbanization metrics were linked to increased TLF:HLF ratios and changes in hysteresis index. This study presents a novel approach of using land-cover weighted rainfall to enhance mechanistic understanding of DOM controls and sources. In contrast, catchment-average rainfall data have the potential to yield stronger understanding of TLF dynamics. This technique could be integrated with existing high resolution in-situ datasets to enhance our understanding of DOM dynamics in urban rivers.

Approaches to herbicide (MCPA) pollution mitigation in drinking water source catchments using enhanced space and time monitoring

Freshwater occurrences of the selective acid herbicide 2-methyl-4-chloro-phenoxyacetic acid (MCPA) are an ongoing regulatory and financial issue for water utility industries as the number and magnitude of detections increase, particularly in surface water catchments. Assessments for mitigating pesticide pollution in catchments used as drinking water sources require a combination of catchment-based and water treatment solutions, but approaches are limited by a lack of empirical data. In this study, an enhanced spatial (11 locations) and temporal (7-hourly to daily sampling) monitoring approach was employed to address these issues in an exemplar surface water source catchment (384 km²). The spatial sampling revealed that MCPA was widespread, with occurrences above the 0.1 μg L^-1 threshold for a single pesticide being highly positively correlated to sub-catchments with higher proportions of 'Improved Grassland' land use (r = 0.84). These data provide a strong foundation for targeting catchment-based mitigation solutions and also add to the debate on the ecosystems services provided by such catchments. Additionally, of the 999 temporal samples taken over 12 months from the catchment outlet, 25% were above the drinking water threshold of 0.1 μg L^-1. This prevalence of high concentrations presents costly problems for source water treatment. Using these data, abstraction shutdowns were simulated for five scenarios using hydrometeorological data to explore the potential to avoid intake of high MCPA concentrations. The scenarios stopped abstraction for 4.2-9.3% of the April-October period and reduced intake of water containing over 0.1 μg L^-1 of MCPA by 16-31%. This represents an important development for real-time proxy assessments for water abstraction in the absence of more direct pesticide monitoring data.

Nutrient Load Mitigation with Wintertime Cover as Estimated by the INCA Model

Increased nutrient loading causes deterioration of receiving surface waters in areas of intensive agriculture. While nitrate and particulate phosphorus load can be efficiently controlled by reducing tillage frequency and increasing vegetation cover, many field studies have shown simultaneously increased loading of bioavailable phosphorus. In the latest phase of the Rural Programme of EU agri-environmental measures, the highest potential to reduce the nutrient loading to receiving waters were the maximum limits for fertilization of arable crops and retaining plant cover on fields with, e.g., no-till methods and uncultivated nature management fields. Due to the latter two measures, the area of vegetation cover has increased since 1995, suggesting clear effects on nutrient loading in the catchment scale as well. We modeled the effectiveness of agri-environmental measures to reduce phosphorus and nitrogen loads to waters and additionally tested the performance of the dynamic, process-based INCA-P (Integrated Nutrients in Catchments—Phosphorus) model to simulate P dynamics in an agricultural catchment. We concluded that INCA-P was able to simulate both fast (immediate) and slow (non-immediate) processes that influence P loading from catchments. Based on our model simulations, it was also evident that no-till methods had increased bioavailable P load to receiving waters, even though total P and total N loading were reduced.

Explainable AI Framework for Multivariate Hydrochemical Time Series

The understanding of water quality and its underlying processes is important for the protection of aquatic environments. With the rare opportunity of access to a domain expert, an explainable AI (XAI) framework is proposed that is applicable to multivariate time series. The XAI provides explanations that are interpretable by domain experts. In three steps, it combines a data-driven choice of a distance measure with supervised decision trees guided by projection-based clustering. The multivariate time series consists of water quality measurements, including nitrate, electrical conductivity, and twelve other environmental parameters. The relationships between water quality and the environmental parameters are investigated by identifying similar days within a cluster and dissimilar days between clusters. The framework, called DDS-XAI, does not depend on prior knowledge about data structure, and its explanations are tendentially contrastive. The relationships in the data can be visualized by a topographic map representing high-dimensional structures. Two state of the art XAIs called eUD3.5 and iterative mistake minimization (IMM) were unable to provide meaningful and relevant explanations from the three multivariate time series data. The DDS-XAI framework can be swiftly applied to new data. Open-source code in R for all steps of the XAI framework is provided and the steps are structured application-oriented.

High-frequency data reveal differential dissolved and suspended solids behavior from a mixed restored prairie and agricultural catchment

Quantifying temporal variability and fluxes within hydrologic catchments is critical to understanding the underlying chemical and physical processes leading to material transport. Measuring variability and fluxes requires sampling at time scales similar to the time scale of process occurrence. This demand has led to the development of automated sampling systems designed to sample at high frequencies, on the order of minutes. While widely deployed in a variety of systems, we installed two high-frequency sampling devices in a single drainage comprised of restored prairie and agricultural land uses in temperate Eastern Nebraska. The sampling systems determined flow rate, conductivity, and turbidity at 15-minute intervals for a twelve-month period. Conductivity was used as a proxy for total dissolved solids (TDS) concentrations and turbidity was used as a proxy for total suspended solids (TSS) concentrations. Using the high-frequency data, estimates of solids flux were calculated, error on the estimates was constrained, the effects of sample timing were considered, and conductivity and turbidity changes during precipitation events were examined. Overall, TDS fluxes were about three times higher than TSS fluxes from the catchment as a whole. However, the TSS fluxes were higher in the agricultural section of the catchment than from the restored prairie. Sheet and rill soil loss estimates from both the restored prairie and agricultural settings were low (<0.060 mm/yr). For TDS flux calculations, sampling at a monthly frequency gave a value that was only 11% lower than sampling every 15 min. For TSS flux calculations, sampling only during precipitation events (0.7% of the time) would capture 67% of the annual flux. Thus, minimizing error in sampling strategies depends on the constituent being analyzed.

Assessment of a portable UV-Vis spectrophotometer's performance for stream water DOC and Fe content monitoring in remote areas

Quantification of dissolved organic carbon (DOC) and iron (Fe) in surface waters is critical for understanding the water quality dynamics, brownification and carbon balance in the northern hemisphere. Especially in the remote areas, sampling and laboratory analysis of DOC and Fe content at a sufficient temporal frequency is difficult. Ultraviolet-visible (UV-Vis) spectrophotometry is a promising tool for water quality monitoring to increase the sampling frequency and applications in remote regions. The aim of this study was (1) to investigate the performance of an in-situ UV-Vis spectrophotometer for detecting spectral absorbances in comparison with a laboratory benchtop instrument; (2) to analyse the stability of DOC and Fe estimates from UV-Vis spectrophotometers among different rivers using multivariate methods; (3) to compare site-specific calibration of models to pooled models and investigate the extrapolation of DOC and Fe predictions from one catchment to another. This study indicates that absorbances that were measured by UV-Vis sensor explained 96% of the absorbance data from the laboratory benchtop instrument. Among the three tested multivariate methods, multiple stepwise regression (MSR) was the best model for both DOC and Fe predictions. Accurate and unbiased models for multiple watersheds for DOC were built successfully, and these models could be extrapolated from one watershed to another even without site-specific calibration for DOC. However, for Fe the combination of different datasets was not possible.

Hydrological and Chemical Controls on Nutrient and Contaminant Loss to Water in Agricultural Landscapes

Nutrient and contaminant losses in agricultural landscapes are directly controlled by hydrological (flow pathways), chemical (sorption, speciation and transformations), biological processes (fixation, uptake) and indirectly by demographic (growing population), economic (food production) and societal drivers (individual attitudes, farming tradition) that control how agricultural landscapes are managed [...]

Detecting Technical Anomalies in High-Frequency Water-Quality Data Using Artificial Neural Networks

Anomaly detection (AD) in high-volume environmental data requires one to tackle a series of challenges associated with the typical low frequency of anomalous events, the broad-range of possible anomaly types, and local nonstationary environmental conditions, suggesting the need for flexible statistical methods that are able to cope with unbalanced high-volume data problems. Here, we aimed to detect anomalies caused by technical errors in water-quality (turbidity and conductivity) data collected by automated in situ sensors deployed in contrasting riverine and estuarine environments. We first applied a range of artificial neural networks that differed in both learning method and hyperparameter values, then calibrated models using a Bayesian multiobjective optimization procedure, and selected and evaluated the "best" model for each water-quality variable, environment, and anomaly type. We found that semi-supervised classification was better able to detect sudden spikes, sudden shifts, and small sudden spikes, whereas supervised classification had higher accuracy for predicting long-term anomalies associated with drifts and periods of otherwise unexplained high variability.

Modeling Water Quality in Watersheds: From Here to the Next Generation

In this synthesis, we assess present research and anticipate future development needs in modeling water quality in watersheds. We first discuss areas of potential improvement in the representation of freshwater systems pertaining to water quality, including representation of environmental interfaces, in-stream water quality and process interactions, soil health and land management, and (peri-)urban areas. In addition, we provide insights into the contemporary challenges in the practices of watershed water quality modeling, including quality control of monitoring data, model parameterization and calibration, uncertainty management, scale mismatches, and provisioning of modeling tools. Finally, we make three recommendations to provide a path forward for improving watershed water quality modeling science, infrastructure, and practices. These include building stronger collaborations between experimentalists and modelers, bridging gaps between modelers and stakeholders, and cultivating and applying procedural knowledge to better govern and support water quality modeling processes within organizations.

Catchment effects of a future Nordic bioeconomy: From land use to water resources

In the future, the world is expected to rely increasingly on renewable biomass resources for food, fodder, fibre and fuel. The sustainability of this transition to bioeconomy for our water systems depends to a large extent on how we manage our land resources. Changes in land use together with climate change will affect water quantity and quality, which again will have implications for the ecosystem services provided by water resources. These are the main topics of this Ambio special issue on "Environmental effects of a green bio-economy". This paper offers a summary of the eleven papers included in this issue and, at the same time, outlines an approach to quantify and mitigate the impacts of bioeconomy on water resources and their ecosystem services, with indications of useful tools and knowledge needs.

Developing Sensor Proxies for “Chemical Cocktails” of Trace Metals in Urban Streams

Understanding transport mechanisms and temporal patterns in the context of metal concentrations in urban streams is important for developing best management practices and restoration strategies to improve water quality. In some cases, in-situ sensors can be used to estimate unknown concentrations of trace metals or to interpolate between sampling events. Continuous sensor data from the United States Geological Survey were analyzed to determine statistically significant relationships between lead, copper, zinc, cadmium, and mercury with turbidity, specific conductance, dissolved oxygen, and discharge for the Hickey Run, Watts Branch, and Rock Creek watersheds in the Washington, D.C. region. We observed a significant negative linear relationship between concentrations of Cu and dissolved oxygen at Rock Creek (p < 0.05). Sometimes, turbidity had significant positive linear relationships with Pb and Hg concentrations. There were negative or positive linear relationships between Pb, Cd, Zn, and Hg and specific conductance. There also appeared to be relationships between watershed areal fluxes of Pb, Cu, Zn, and Cd in streams with turbidity. Watershed monitoring approaches using continuous sensor data have the potential to characterize the frequency, magnitude, and composition of pulses in concentrations and loads of trace metals, which could improve the management and restoration of urban streams.

Combining punctual and high frequency data for the spatiotemporal assessment of main geochemical processes and dissolved exports in an urban river catchment

The assessment of dissolved loadings and the sources of these elements in urban catchments' rivers is usually measured by punctual sampling or through high frequency sensors. Nevertheless, the combination of both methodologies has been less common even though the information they give is complementary. Major ion (Ca²⁺, Mg²⁺, Na⁺, K⁺, Cl^-, SO₄^2-, and alkalinity), organic matter (expressed as Dissolved Organic Carbon, DOC), and nutrients (NO₃^-, and PO₄^3-) are punctually measured in the Deba river urban catchment (538 km²), in the northern part of the Iberian Peninsula (draining to the Bay of Biscay). Discharge, precipitation, and Electrical Conductivity (EC) are registered with a high frequency (10 min) in three gauging stations. The combination of both methodologies has allowed the assessment of major geochemical processes and the extent of impact of anthropogenic input on major composition of riverine water, as well as its spatial and temporal evolution. Three methodologies for loading estimation have been assessed and the error committed in the temporal aggregation is quantified. Results have shown that, even though carbonates dominate the draining area, the water major ion chemistry is governed by an evaporitic spring in the upper part of the catchment, while anthropogenic input is specially noted downstream of three wastewater treatment plants, in all nutrients and organic matter. The results of the present work illustrate how the combination of two monitoring methodologies allows for a better assessment of the spatial and temporal evolution on the major water quality in an urban catchment.

Temporal hydrochemical dynamics of the River Wensum, UK: Observations from long-term high-resolution monitoring (2011-2018)

In 2010, the UK government established the Demonstration Test Catchment (DTC) initiative to evaluate the extent to which on-farm mitigation measures can cost-effectively reduce the impacts of agricultural water pollution on river ecology whilst maintaining food production capacity. A central component of the DTC platform was the establishment of a comprehensive network of automated, web-based sensor technologies to generate high-temporal resolution (30 min) empirical datasets of surface water, groundwater and meteorological parameters over a long period (2011-2018). Utilising 8.9 million water quality measurements generated for the River Wensum, this paper demonstrates how long-term, high-resolution monitoring of hydrochemistry can improve our understanding of the complex temporal dynamics of riverine processes from 30 min to annual timescales. This paper explores the impact of groundwater-surface water interactions on instream pollutant concentrations (principally nitrogen, phosphorus and turbidity) and reveals how varying hydrochemical associations under contrasting flow regimes can elicit important information on the dominant pollution pathways. Furthermore, this paper examines the relationships between agricultural pollutants and precipitation events of varying magnitude, whilst demonstrating how high-resolution data can be utilised to develop conceptual models of hydrochemical processes for contrasting winter and summer seasons. Finally, this paper considers how high-resolution hydrochemical data can be used to increase land manager awareness of environmentally damaging farming operations and encourage the adoption of more water sensitive land management practices.

High-frequency monitoring reveals how hydrochemistry and dissolved carbon respond to rainstorms at a karstic critical zone, Southwestern China

Hydrochemical behavior and dissolved carbon dynamics are highly-sensitive to hydrological variations in the monsoon-influenced karstic critical zone which has high chemical weathering rates and experiences strong anthropogenic impact. Continuous high-frequency monitoring in the spring outlet of a karstic catchment in Southwestern China revealed that most hydrochemical variables changed distinctively in response to hydrologic variations, influenced by mixing of different sources and miscellaneous biogeochemical processes. Na⁺, K⁺ and SO₄^2- varied significantly with hydrology, showing weak chemostatic behavior controlled by dilution. The flushing effect and random behavior of NO₃^- and Cl^- likely reflect agricultural inputs from high throughflow. Soil CO₂ in infiltrated water supports carbonate weathering, enabling DIC (dissolved inorganic carbon) and weathering products (e.g., Ca²⁺ and Mg²⁺) to maintain chemostatic behavior. Biogenic DIC exhibited a stronger chemostatic response than carbonate sources and was the foremost control in DIC behavior. Carbon exchange between DIC and DOC (dissolved organic carbon) did not significantly influence DIC concentration and δ¹³C due to very low DOC concentration. More DOC was exported by flushing from increasing discharge. Hysteretic analysis indicated that the transport processes were controlled by proximal sources mixing and diverse mobilization in various periods responding to rainstorms. NO₃^- and Cl^- presented different hysteresis behavior as sourced from agricultural activities. DOC increased on the hydrograph rising limb and was controlled by a transport-limited regime. However, the hysteresis behavior of most weathering products and DIC were regulated by a process-limited regime in the karstic critical zone. Overall, biogeochemical processes, hydrogeological properties, storm intensity/magnitude and the timing of storms (antecedent conditions) are main factors influencing the response of hydrochemical variables and dissolved carbon to storm events.

High frequency fluorescence monitoring reveals new insights into organic matter dynamics of an urban river, Birmingham, UK

Natural organic matter (NOM) is fundamental to many biogeochemical processes in river ecosystems. Currently, however, we have limited knowledge of NOM dynamics across the spectrum of flow conditions as previous studies have focused largely on storm events. Field deployable fluorescence technology offers new opportunities to explore both stochastic and predictable diel NOM dynamics at finer time-steps and for longer periods than was hitherto possible, thus yielding new insight into NOM sources, processing, and pathways. Hourly fluorescence data (humic-like fluorescence [Peak C] and tryptophan-like fluorescence [Peak T]) and a suite of hydro-climatological variables were collected from an urban river (Birmingham, UK). We explored monthly concentration-discharge (C-Q) patterns using segmented regression and assessed hysteretic and flushing behaviour for Peak C, T and turbidity to infer source zone activation. Diel patterns were assessed during low flow periods. Wavelet analysis identified strong diurnal variations in Peak C with early morning peaks while no diel dynamics were apparent for Peak T. Using generalised linear modelling relationships between Peak C periodicity and both solar radiation and time since previous storm/scouring event were identified. Breakpoints and positive slopes for C-Q relationship highlighted chemodynamic behaviour for NOM over most of the monitoring period, with Peak T mobilised more relative to Peak C during high Q. Hysteresis patterns were highly variable but flushing behaviour of Peak T and C suggested exhaustion of humic compounds during long duration events and following successive rainfall events. Peak T flushing was correlated with Q magnitude highlighting the potential for combined sewer overflows to act as important NOM sources despite significant dilution potential. This research highlights the potential of real-time, field deployable fluorescence spectroscopy as a viable method for providing insight into diel and transport driven NOM dynamics.

Using high-frequency phosphorus monitoring for water quality management: a case study of the upper River Itchen, UK

Increased concentrations of phosphorus (P) in riverine systems lead to eutrophication and can contribute to other environmental effects. Chalk rivers are known to be particularly sensitive to elevated P levels. We used high-frequency (daily) automatic water sampling at five distinct locations in the upper River Itchen (Hampshire, UK) between May 2016 and June 2017 to identify the main P species (including filterable reactive phosphorus, total filterable phosphorus, total phosphorus and total particulate phosphorus) present and how these varied temporally. Our filterable reactive phosphorus (considered the biologically available fraction) data were compared with the available Environment Agency total reactive phosphorus (TRP) values over the same sampling period. Over the trial, the profiles of the P fractions were complex; the major fraction was total particulate phosphorus with the mean percentage value ranging between 69 and 82% of the total P present. Sources were likely to be attributable to wash off from agricultural activities. At all sites, the FRP and Environment Agency TRP mean concentrations over the study were comparable. However, there were a number of extended time periods (1 to 2 weeks) where the mean FRP concentration (e.g. 0.62 mg L^-1) exceeded the existing regulatory values (giving a poor ecological status) for this type of river. Often, these exceedances were missed by the limited regulatory monitoring procedures undertaken by the Environment Agency. There is evidence that these spikes of elevated concentrations of P may have a biological impact on benthic invertebrate (e.g. blue-winged olive mayfly) communities that exist in these ecologically sensitive chalk streams. Further research is required to assess the ecological impact of P and how this might have implications for the development of future environmental regulations.