High-frequency Sensor Data Research Articles

SentemQC - A novel and cost-efficient method for quality assurance and quality control of high-resolution frequency sensor data in fresh waters

The growing use of sensors in fresh waters for water quality measurements generates an increasingly large amount of data that requires quality assurance (QA)/quality control (QC) before the results can be exploited. Such a process is often resource-intensive and may not be consistent across users and sensors. SentemQC (QA-QC of high temporal resolution sensor data) is a cost-efficient, and open-source Python approach developed to ensure the quality of sensor data by performing data QA and QC on large volumes of high-frequency (HF) sensor data. The SentemQC method is computationally efficient and features a six-step user-friendly setup for anomaly detection. The method marks anomalies in data using five moving windows. These windows connect each data point to neighboring points, including those further away in the moving window. As a result, the method can mark not only individual outliers but also clusters of anomalies. Our analysis shows that the method is robust for detecting anomalies in HF sensor data from multiple water quality sensors measuring nitrate, turbidity, oxygen, and pH. The sensors were installed in three different freshwater ecosystems (two streams and one lake) and experimental lake mesocosms. Sensor data from the stream stations yielded anomaly percentages of 0.1%, 0.1%, and 0.2%, which were lower than the anomaly percentages of 0.5%, 0.6%, and 0.8% for the sensors in Lake and mesocosms, respectively. While the sensors in this study contained relatively few anomalies (<2%), they may represent a best-case scenario in terms of use and maintenance. SentemQC allows the user to include the individual sensor uncertainty/accuracy when performing QA-QC. However, SentemQC cannot function independently. Additional QA-QC steps are crucial, including calibration of the sensor data to correct for zero offsets and implementation of gap-filling methods prior to the use of the sensor data for determination of final real-time concentrations and load calculations.

Improving Lameness Detection in Cows: A Machine Learning Algorithm Application

The deployment of diverse data-generating technologies in livestock farming holds the promise of early disease detection and improved animal well-being. In this paper, we combine routinely collected dairy farm and herd data with weather and high frequency sensor data from 6 farms to predict new lameness events in various future periods, spanning from the following day to 3 weeks. A Random Forest classifier, using input features selected by the Boruta Algorithm, was used for the prediction task; effects of individual features were further assessed using partial dependence plots. We achieve precision scores of up to 93% when predicting lameness for the next 3 weeks and when using information from the last 3 weeks, combined with a balanced accuracy of 79%. Removing sensor data results have tendency to reduce the precision for predictions, especially when using information from the last one,2 or 3 weeks. Moving to a larger data set (without sensor data) of 44 farms keeps the similar balanced accuracy but reduces precision by more than 30%, revealing a substantial a trade-off in model quality between false positives (false lameness alerts) and false negatives (missed lameness events). Sensor data holds promise to further improve the precision of these models, but can be partially compensated by high resolution data from other systems, such as automated milking systems.

Solution Enables Well Surveillance in Stacked Reservoirs With Near-Critical Fluids

_ This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper IPTC 23912, “Digital Solution for Well Surveillance in Stacked Reservoirs With Near-Critical Fluid Systems,” by Kristian Mogensen, SPE, Carlos Mata, SPE, and Swarjit Samajpati, ADNOC, et al. The paper has not been peer reviewed. Copyright 2024 International Petroleum Technology Conference. _ Fully compositional integrated asset models (FC-IAMs) are being deployed for an increasing number of fields in the operator’s portfolio. The complete paper describes the development of comprehensive digital well surveillance and field-production optimization for an offshore field consisting of four stacked reservoirs, each containing near-critical fluids. Augmenting the FC-IAM with high-frequency sensor data in addition to proprietary tools to actively monitor well performance helps identify and pursue opportunities to maximize oil production from the field, subject to several system constraints. Introduction The field is approximately 120 km offshore Abu Dhabi. During the appraisal phase, eight wells penetrated the structure and confirmed that the predominant lithologies are composed of dolomite and anhydrite, with subordinate limestone. Production of oil and gas comes from four main reservoirs—A, B, C, and D. The ongoing development centers around miscible gas injection along the crestal part of each reservoir and peripheral water injection. Integrated asset models are one of many key initiatives supporting the operator’s Smart Production Growth initiative. Although the calculations identifying scope for improved production rates are firmly founded in science, substantial digitization efforts are required to automate, streamline, and sustain any such digital solution. Fig. 1 illustrates the fact that company database management plays a fundamental role in providing reliable input to keep models updated and fit for purpose. Fluid-Property Modeling Each of the four producing reservoirs contains near-critical fluids exhibiting a steep compositional depth gradient, which means that the solution gas/oil ratio (GOR) is well-specific and may change over time. The GOR trend is governed by several factors. The lightest part of the fluid column may be produced first because of the improved mobility; this would lead to a decreasing solution GOR over time. However, because of the crestal miscible gas injection, significant swelling will result in a gradual GOR increase followed by gas breakthrough. Solution GOR increases because of gas swelling. It is important to make the distinction between solution gas and free gas because the operational guidelines dictate that the reservoir pressure should be maintained above the saturation pressure and above the minimum miscibility pressure. Well-Test Validation Workflow The compositional framework adds an extra layer of complexity because fluid composition needs to be specified at the well level. A robust procedure is needed to first assign an initial reservoir fluid composition to each well. The second key task is to capture changes over time. The method of solution is referred to as a target GOR calculation. For any given fluid system, an equation of state (EOS) model must be tuned to experimental data to convert compositional information to black-oil properties at specific pressures and temperatures. A so-called anchoring fluid composition initially is assigned to the well, and the EOS can then calculate the resulting GOR and other variables from a simulated separator test. The evolved gas and liquid phases are then recombined mathematically to match a measured GOR from a production test. Matching is an iterative process but normally converges very quickly.

Concurrent warming and browning eliminate cold-water fish habitat in many temperate lakes

Cold-water species in temperate lakes face two simultaneous climate-driven ecosystem changes: warming and browning of their waters. Browning refers to reduced transparency arising from increased dissolved organic carbon (DOC), which absorbs solar energy near the surface. It is unclear whether the net effect is mitigation or amplification of climate warming impacts on suitable oxythermal habitat (<20 °C, >5 mgO/L) for cold-loving species because browning expands the vertical distribution of both cool water and oxygen depletion. We analyzed long-term trends and high-frequency sensor data from browning lakes in New York's Adirondack region to assess the contemporary status of summertime habitat for lacustrine brook trout. Across two decades, surface temperatures increased twice as fast and bottom dissolved oxygen declined >180% faster than average trends for temperate lakes. We identify four lake categories based on oxythermal habitat metrics: constrained, squeezed, overheated, and buffered. In most of our study lakes, trout face either seasonal loss (7 of 15) or dramatic restriction (12 to 21% of the water column; 5 of 15) of suitable habitat. These sobering statistics reflect rapid upward expansion of oxygen depletion in lakes with moderate or high DOC relative to compression of heat penetration. Only in very clear lakes has browning potentially mitigated climate warming. Applying our findings to extensive survey data suggests that decades of browning have reduced oxythermal refugia in most Adirondack lakes. We conclude that joint warming and browning may preclude self-sustaining cold-water fisheries in many temperate lakes; hence, oxythermal categorization is essential to guide triage strategies and management interventions.

Sensor-based behavioral patterns can identify heat-sensitive lactating dairy cows

Heat stress impairs the health and performance of dairy cows, yet only a few studies have investigated the diversity of cattle behavioral responses to heat waves. This research was conducted on an Italian Holstein dairy farm equipped with precision livestock farming sensors to assess potential different behavioral patterns of the animals. Three heat waves, defined as at least five consecutive days with mean daily temperature-humidity index higher than 72, were recorded in the farm area during the summer of 2021. Individual daily milk yield data of 102 cows were used to identify “heat-sensitive” animals, meaning the cows that, under a given heat wave, experienced a milk yield drop that was not linked with other health events (e.g., mastitis). Milk yield drops were detected as perturbations of the lactation curve estimated by iteratively using Wood’s equation. Individual daily minutes of lying, chewing, and activity were retrieved from ear-tag-based accelerometer sensors. Semi-parametric generalized estimating equations models were used to assess behavioral deviations of heat-sensitive cows from the herd means under heat stress conditions. Heat waves were associated with an overall increase in the herd’s chewing and activity times, along with an overall decrease of lying time. Heat-sensitive cows spent approximately 15 min/days more chewing and performing activities (p < 0.05). The findings of this research suggest that the information provided by high-frequency sensor data could assist farmers in identifying cows for which personalized interventions to alleviate heat stress are needed.

Intelligent anomaly detection for dynamic high-frequency sensor data of road underground structure

Intelligent anomaly detection for dynamic high-frequency sensor data of road underground structure

Understanding links between water-quality variables and nitrate concentration in freshwater streams using high frequency sensor data.

Real-time monitoring using in-situ sensors is becoming a common approach for measuring water-quality within watersheds. High-frequency measurements produce big datasets that present opportunities to conduct new analyses for improved understanding of water-quality dynamics and more effective management of rivers and streams. Of primary importance is enhancing knowledge of the relationships between nitrate, one of the most reactive forms of inorganic nitrogen in the aquatic environment, and other water-quality variables. We analysed high-frequency water-quality data from in-situ sensors deployed in three sites from different watersheds and climate zones within the National Ecological Observatory Network, USA. We used generalised additive mixed models to explain the nonlinear relationships at each site between nitrate concentration and conductivity, turbidity, dissolved oxygen, water temperature, and elevation. Temporal auto-correlation was modelled with an auto-regressive-moving-average (ARIMA) model and we examined the relative importance of the explanatory variables. Total deviance explained by the models was high for all sites (99%). Although variable importance and the smooth regression parameters differed among sites, the models explaining the most variation in nitrate contained the same explanatory variables. This study demonstrates that building a model for nitrate using the same set of explanatory water-quality variables is achievable, even for sites with vastly different environmental and climatic characteristics. Applying such models will assist managers to select cost-effective water-quality variables to monitor when the goals are to gain a spatial and temporal in-depth understanding of nitrate dynamics and adapt management plans accordingly.

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.

Nitrate Hysteresis as a Tool for Revealing Storm‐Event Dynamics and Improving Water Quality Model Performance

AbstractUnderstanding the physics of nitrate contamination in surface and subsurface water is vital for mitigating downstream water quality impairment. Though high frequency sensor data have become readily available and computational models more accessible, the integration of these two methods for improved prediction is underdeveloped. The objective of this study was to utilize high‐frequency data to advance our understanding and model representation of nitrate transport for an agricultural karst spring in Kentucky, USA. We collected 2‐years of 15‐min nitrate and specific conductance data and analyzed source‐timing dynamics across dozens of events to develop a conceptual model for nitrate hysteresis in karst. Thereafter, we used the sensing data, specifically discharge‐concentration indices, to constrain modeled nitrate prediction bounds as well as the uncertainty of hydrologic and nitrogen processes, such as soil percolation and biogeochemical transformation. Observed nitrate hysteresis behavior at the spring was complex and included clockwise (n = 11), counterclockwise (n = 13), and figure‐eight (n = 10) shapes, which contrasts with surface systems that are often dominated by a single hysteresis shape. Sensing results highlight the importance of antecedent connectivity to nitrate‐rich storages in determining the timing of nitrate delivery to the spring. After integrating hysteresis analysis into our numerical model evaluation, simulated nitrate prediction bounds were reduced by 43 ± 12% and parameter uncertainty by 36 ± 20%. Taken together, this study suggests that discharge‐concentration indices derived from high‐frequency sensor data can be successfully integrated into numerical models to improve process representation and reduce modeled uncertainty.

Continuous-Time Trajectory Estimation for Differentially Flat Systems

Continuous-time estimation using splines on Lie groups has been gaining traction in the literature due to the ability to incorporate high-frequency sensor data without introducing new optimization parameters. However, evaluating time derivatives and Jacobians of Lie group splines is computationally expensive, limiting their use mainly to offline applications. Motivated by the trajectory planning literature, we develop a new estimation technique that leverages the differential flatness property of many dynamic systems to define the spline in the system's flat output space, which is often Euclidean. Doing so has the added benefit of providing a simple and effective way to include system inputs in the estimation process. We show an example of flatness-based estimation for the unicycle dynamic model. We then show that this new method can achieve similar performance as Lie group spline estimation with significantly less computation time, and validate its use in hardware using a differential-drive robot.

Using Mobile Dual Eye-Tracking to Capture Cycles of Collaboration and Cooperation in Co-located Dyads

The goal of this paper is to bring new insights to the study of social learning processes by designing measures of collaboration using high-frequency sensor data. More specifically, we are interested in understanding the interplay between moments of collaboration and cooperation, which is an understudied area of research. We collected a multimodal dataset during a collaborative learning activity typical of makerspaces: learning how to program a robot. Pairs of participants were introduced to computational thinking concepts using a block-based environment. Mobile eye-trackers, physiological wristbands, and motion sensors captured their behavior and social interactions. In this paper, we analyze the eye-tracking data to capture participants’ tendency to synchronize their visual attention. This paper provides three contributions: (1) we use an emerging methodology (mobile dual eye-tracking) to capture joint visual attention in a co-located setting and replicate findings that show how levels of joint visual attention are positively correlated with collaboration quality; (2) we qualitatively analyzed the co-occurrence of verbal activity and joint visual attention in low and high performing groups to better understand moments of collaboration and cooperation; (3) inspired by the qualitative observations and theories of collaborative learning, we designed a new quantitative measure that captures cycles of collaborative and cooperative work. Compared to simple measures of joint visual attention, we found it to increase correlation coefficients with learning and collaboration scores. We discuss those results and describe how advances in analyzing sensor data can contribute to theories of collaboration. We conclude with implications for capturing students’ interactions in co-located spaces using Multimodal Learning Analytics (MMLA).

How do inorganic nitrogen processing pathways change quantitatively at daily, seasonal, and multiannual scales in a large agricultural stream?

Abstract. Large agricultural streams receive excessive inputs of nitrogen. However, quantifying the role of these streams in nitrogen processing remains limited because continuous direct measurements of the interacting and highly time-varying nitrogen processing pathways in larger streams and rivers are very complex. Therefore, we employed a monitoring-driven modelling approach with high-frequency in situ data and the river water quality model Water Quality Analysis Simulation Program (WASP) 7.5.2 in the 27.4 km reach of the sixth-order agricultural stream called Lower Bode (central Germany) for a 5-year period (2014–2018). Paired high-frequency sensor data (15 min interval) of discharge, nitrate, dissolved oxygen, and chlorophyll a at upstream and downstream stations were used as model boundaries and for setting model constraints. The WASP model simulated 15 min intervals of discharge, nitrate, and dissolved oxygen with Nash–Sutcliffe efficiency values higher than 0.9 for calibration and validation, enabling the calculation of gross and net dissolved inorganic nitrogen uptake and pathway rates on a daily, seasonal, and multiannual scale. Results showed daily net uptake rate of dissolved inorganic nitrogen ranged from −17.4 to 553.9 mgNm-2d-1. The highest daily net uptake could reach almost 30 % of the total input loading, which occurred at extreme low flow in summer 2018. The growing season (spring and summer) accounted for 91 % of the average net annual uptake of dissolved inorganic nitrogen in the measured period. In spring, both the DIN gross and net uptake were dominated by the phytoplankton uptake pathway. In summer, benthic algae assimilation dominated the gross uptake of dissolved inorganic nitrogen. Conversely, the reach became a net source of dissolved inorganic nitrogen with negative daily net uptake values in autumn and winter, mainly because the release from benthic algae surpassed uptake processes. Over the 5 years, average gross and net uptake rates of dissolved inorganic nitrogen were 124.1 and 56.8 mgNm-2d-1, which accounted for only 2.7 % and 1.2 % of the total loadings in the Lower Bode, respectively. The 5-year average gross DIN uptake decreased from assimilation by benthic algae through assimilation by phytoplankton to denitrification. Our study highlights the value of combining river water quality modelling with high-frequency data to obtain a reliable budget of instream dissolved inorganic nitrogen processing which facilitates our ability to manage nitrogen in aquatic systems. This study provides a methodology that can be applied to any large stream to quantify nitrogen processing pathway dynamics and complete our understanding of nitrogen cycling.

Toward capturing divergent collaboration in makerspaces using motion sensors

PurposeThis study aims to uncover divergent collaboration in makerspaces using social network analysis to examine ongoing social relations and sequential data pattern mining to invesitgate temporal changes in social activities.Design/methodology/approachWhile there is a significant body of qualitative work on makerspaces, there is a lack of quantitative research identifying productive interactions in open-ended learning environments. This study explores the use of high frequency sensor data to capture divergent collaboration in a semester-long makerspace course, where students support each other while working on different projects.FindingsThe main finding indicates that students who diversely mix with others performed better in a semester-long course. Additional results suggest that having a certain balance of working individually, collaborating with other students and interacting with instructors maximizes performance, provided that sufficient alone time is committed to develop individual technical skills.Research limitations/implicationsThese discoveries provide insight into how productive makerspace collaboration can occur within the framework of Divergent Collaboration Learning Mechanisms (Tissenbaum et al., 2017).Practical implicationsIdentifying the diversity and sequence of social interactions could also increase instructor awareness of struggling students and having this data in real-time opens new doors for identifying (un)productive behaviors.Originality/valueThe contribution of this study is to explore the use of a sensor-based, data-driven, longitudinal approach in an ecologically valid setting to understand divergent collaboration in makerspaces. Finally, this study discusses how this work represents an initial step toward quantifying and supporting productive interactions in project-based learning environments.

Dynamic models of stress-smoking responses based on high-frequency sensor data

Self-reports indicate that stress increases the risk for smoking; however, intensive data from sensors can provide a more nuanced understanding of stress in the moments leading up to and following smoking events. Identifying personalized dynamical models of stress-smoking responses can improve characterizations of smoking responses following stress, but techniques used to identify these models require intensive longitudinal data. This study leveraged advances in wearable sensing technology and digital markers of stress and smoking to identify person-specific models of stress and smoking system dynamics by considering stress immediately before, during, and after smoking events. Adult smokers (n = 45) wore the AutoSense chestband (respiration-inductive plethysmograph, electrocardiogram, accelerometer) with MotionSense (accelerometers, gyroscopes) on each wrist for three days prior to a quit attempt. The odds of minute-level smoking events were regressed on minute-level stress probabilities to identify person-specific dynamic models of smoking responses to stress. Simulated pulse responses to a continuous stress episode revealed a consistent pattern of increased odds of smoking either shortly after the beginning of the simulated stress episode or with a delay, for all participants. This pattern is followed by a dramatic reduction in the probability of smoking thereafter, for about half of the participants (49%). Sensor-detected stress probabilities indicate a vulnerability for smoking that may be used as a tailoring variable for just-in-time interventions to support quit attempts.

Macrosystems EDDIE Teaching Modules Increase Students’ Ability to Define, Interpret, and Apply Concepts in Macrosystems Ecology

Ecologists are increasingly using macrosystems approaches to understand population, community, and ecosystem dynamics across interconnected spatial and temporal scales. Consequently, integrating macrosystems skills, including simulation modeling and sensor data analysis, into undergraduate and graduate curricula is needed to train future environmental biologists. Through the Macrosystems EDDIE (Environmental Data-Driven Inquiry and Exploration) program, we developed four teaching modules to introduce macrosystems ecology to ecology and biology students. Modules combine high-frequency sensor data from GLEON (Global Lake Ecological Observatory Network) and NEON (National Ecological Observatory Network) sites with ecosystem simulation models. Pre- and post-module assessments of 319 students across 24 classrooms indicate that hands-on, inquiry-based modules increase students’ understanding of macrosystems ecology, including complex processes that occur across multiple spatial and temporal scales. Following module use, students were more likely to correctly define macrosystems concepts, interpret complex data visualizations and apply macrosystems approaches in new contexts. In addition, there was an increase in student’s self-perceived proficiency and confidence using both long-term and high-frequency data; key macrosystems ecology techniques. Our results suggest that integrating short (1–3 h) macrosystems activities into ecology courses can improve students’ ability to interpret complex and non-linear ecological processes. In addition, our study serves as one of the first documented instances for directly incorporating concepts in macrosystems ecology into undergraduate and graduate ecology and biology curricula.

How Good Is the STW Sensor? An Account from a Larger Shipping Company

This paper examines the statistical properties and the quality of the speed through water (STW) measurement based on data extracted from almost 200 container ships of Maersk Line’s fleet for 3 years of operation. The analysis uses high-frequency sensor data along with additional data sources derived from external providers. The interest of the study has its background in the accuracy of STW measurement as the most important parameter in the assessment of a ship’s performance analysis. The paper contains a thorough analysis of the measurements assumed to be related with the STW error, along with a descriptive decomposition of the main variables by sea region including sea state, vessel class, vessel IMO number and manufacturer of the speed-log installed in each ship. The paper suggests a semi-empirical method using a threshold to identify potential error in a ship’s STW measurement. The study revealed that the sea region is the most influential factor for the STW accuracy and that 26% of the ships of the dataset’s fleet warrant further investigation.

Quantifying the frequency of synchronous carbon and nitrogen export to the river network

The biogeochemical cycles of carbon (C) and nitrogen (N) are inextricably linked for a range of reactions. For coupled reactions such as denitrification to occur, however, solutes must be found together in space and time. Using the framework of concentration-discharge (c-Q) relationships, we examine the frequency of synchronous C and N export (i.e. identical c-Q behavior) across a river network using > 5 years of high-frequency sensor data. We demonstrate that across space and time the export of C and N to a river network is asynchronous 57% of the time. The probability of simultaneous export in largely forested watersheds demonstrates little temporal structure, while in more human-impacted watersheds, we observe the highest frequency of asynchronous c-Q behavior. We discuss the implications of synchronous c-Q behavior for solute flux estimation models and develop a theoretical framework for predicting where within a landscape we expect the probability of coupled C and N reactions to be greatest. By simultaneously comparing the variability in C and N c-Q relationships we develop an integrated framework for predicting synchronous export of solutes.

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.

Influence of land use and hydrologic variability on seasonal dissolved organic carbon and nitrate export: insights from a multi-year regional analysis for the northeastern USA

Land use/land cover (LULC) change has significant impacts on nutrient loading to aquatic systems and has been linked to deteriorating water quality globally. While many relationships between LULC and nutrient loading have been identified, characterization of the interaction between LULC, climate (specifically variable hydrologic forcing) and solute export across seasonal and interannual time scales is needed to understand the processes that determine nutrient loading and responses to change. Recent advances in high-frequency water quality sensors provide opportunities to assess these interannual relationships with sufficiently high temporal resolution to capture the unpredictable, short-term storm events that likely drive important export mechanisms for dissolved organic carbon (DOC) and nitrate (NO3−–N). We deployed a network of in situ sensors in forested, agricultural, and urban watersheds across the northeastern United States. Using 2 years of high-frequency sensor data, we provide a regional assessment of how LULC and hydrologic variability affected the timing and magnitude of dissolved organic carbon and nitrate export, and the status of watershed fluxes as either supply or transport controlled. Analysis of annual export dynamics revealed systematic differences in the timing and magnitude of DOC and NO3−–N delivery among different LULC classes, with distinct regional similarities in the timing of DOC and NO3−–N fluxes from forested and urban watersheds. Conversely, export dynamics at agricultural sites appeared to be highly site-specific, likely driven by local agricultural practices and regulations. Furthermore, the magnitude of solute fluxes across watersheds responded strongly to interannual variability in rainfall, suggesting a high degree of hydrologic control over nutrient loading across the region. Thus, there is strong potential for climate-driven changes in regional hydrologic cycles to drive variation in the magnitude of downstream nutrient fluxes, particularly in watersheds where solute supply and/or transport has been modified.

Hysteretic Response of Solutes and Turbidity at the Event Scale Across Forested Tropical Montane Watersheds

Concentration-discharge relationships are a key tool for understanding the sources and transport of material from watersheds to fluvial networks. Storm events in particular provide insight into variability in the sources of solutes and sediment within watersheds, and the hydrologic pathways that connect hillslope to stream channel. Here we examine high-frequency sensor-based specific conductance and turbidity data from multiple storm events across two watersheds (Quebrada Sonadora and Rio Icacos) with different lithology in the Luquillo Mountains of Puerto Rico, a forested tropical ecosystem. Our analyses include Hurricane Maria, a category 5 hurricane. To analyze hysteresis, we used a recently developed set of metrics to describe and quantify storm events including the hysteresis index (HI), which describes the directionality of hysteresis loops, and the flushing index (FI), which can be used to infer whether the mobilization of material is source or transport limited. We also examine the role of antecedent discharge to predict hysteretic behavior during storms. Overall, specific conductance and turbidity showed contrasting responses to storms. The hysteretic behavior of specific conductance was similar across sites, displaying clockwise hysteresis and a negative flushing index indicating proximal sources of solutes and consistent source limitation. In contrast, the directionality of turbidity hysteresis was significantly different between watersheds, although both had strong flushing behavior indicative of transport limitation. Overall, models that included antecedent discharge did not perform any better than models with peak discharge alone, suggesting that the magnitude and trajectory of an individual event was the strongest driver of material flux and hysteretic behavior. Hurricane Maria produced unique hysteresis metrics within both watersheds, indicating a distinctive response to this major hydrological event. The similarity in response of specific conductance to storms suggests that solute sources and pathways are similar in the two watersheds. The divergence in behavior for turbidity suggests that sources and pathways of particulate matter vary between the two watersheds. The use of high-frequency sensor data allows the quantification of storm events while index-based metrics of hysteresis allow for the direct comparison of complex storm events across a heterogeneous landscape and variable flow conditions.