Estimation of nonlinear water-quality trends in high-frequency monitoring data

Long-term water quality monitoring in agricultural catchments in Sweden: Impact of climatic drivers on diffuse nutrient loads

Water quality related to non-point source pollution continues to pose challenges in agricultural landscapes, despite two completed cycles of Water Framework Directive actions by farmers and landowners. Future climate projections will cause new challenges in landscape hydrology and subsequently, the potential responses in water quality. Investigating the nutrient trends in surface waters and studying the efficiency of mitigation measures revealed that loads and measures are highly variable both spatially and temporally in catchments with different agro-climatic and environmental conditions. In Sweden, nitrogen and phosphorus loads in eight agricultural catchments (470-3300 ha) have been intensively monitored for >20 years. This study investigated the relationship between precipitation, air temperature, and discharge patterns in relation to nitrogen (N) and phosphorus (P) loads at catchment outlets. The time series data analysis was carried out by integrating Mann-Kendall test, Pettitt break-points, and Generalized Additive Model. The results showed that the nutrient loads highly depend on water discharge, which had large variation in annual average (158-441 mm yr^-1). The annual average loads were also considerably different among the catchments with total N (TN) loads ranging from 6.76 to 35.73 kg ha^-1, and total P (TP) loads ranging from 0.11 to 1.04 kg ha^-1. The climatic drivers were highly significant indicators of nutrient loads but with varying degree of significance. Precipitation (28-962 mm yr^-1) was a significant indicator of TN loads in five catchments (loamy sand/sandy loam) while annual average temperature (6.5-8.7 °C yr^-1) was a significant driver of TN loads in six out of eight catchments. TP loads were associated with precipitation in two catchments and significantly correlated to water discharge in six catchments. Considering the more frequent occurrence of extreme weather events, it is necessary to tailor N and P mitigation measures to future climate-change features of precipitation, temperature, and discharge.

Assessing temporal variability of lake turbidity and trophic state of European lakes using open data repositories

Water turbidity is one of the more important water quality parameters that is strictly linked with the productivity of the lake and is commonly used as an indicator of the trophic state. However, limited field data availability across wide geographic gradients may hinder the conduction of large scale longitudinal studies. In this study, time series of lake turbidity and trophic state index (TSI) between 2002 and 2012 were obtained from the Copernicus Lake Water products to create a large longitudinal dataset of lake variables for 22 European lakes. The dataset was combined with estimates of nutrient concentrations and surface water temperature obtained from the Hydrological Predictions for the Environment (HYPE) and ERA5-Land data repositories, that were used as environmental predictors. Hence, the validity of the lake water quality parameters was tested by a) exploring their spatial and temporal variability and b) identifying associations with the environmental predictors. For this purpose, seasonal Mann-Kendall tests were applied to find significant inter-annual trends of turbidity and TSI for each lake, and generalized additive models (GAMs) were employed to identify the main parameters that shape their temporal dynamics. Although we did not find significant inter-annual changes, our findings highlighted the strong influence of seasonality and surface water temperature in defining the temporal variability patterns in most of the lakes. In addition, the importance of nutrients varied among lakes as several lakes exhibited narrow nutrient gradients reflecting relatively stable nutrient conditions during the examined period. Other lake intrinsic factors, such as local climate and biotic interactions, are important drivers of shaping turbidity and nutrient dynamics. This study highlighted the usefulness of combining lake data from large repositories in conducting large scale spatial studies as a valuable asset for future lake research and management purposes.

Total and dissolved phosphorus losses from agricultural headwater streams during extreme runoff events

Eutrophication continues to be a concerning global water quality issue. Managing and mitigating harmful algal blooms demands clear information on the conditions promoting large phosphorus losses from contributing watersheds. Of particular concern is the amount and form of phosphorus loading to receiving water bodies during extreme runoff events, which are expected to increase in frequency due to climate change. Five years (2015 to 2020) of water quantity and quality data from 11 agricultural watersheds in the lower Great Lakes basin were analyzed and used to model total and dissolved phosphorus losses. This study aimed to assess temporal dynamics in phosphorus concentrations and losses over runoff events covering a wide range of hydrologic conditions and to quantify their relative importance on annual phosphorus losses. Event concentration-discharge relationships for total and dissolved phosphorus were hysteretic and had contrasting dominant patterns across watersheds. The proportion of annual phosphorus losses during events was highly variable between watersheds, accounting for 47-94 %. Extreme events were particularly impactful: as few as three events per year were found to be responsible for nearly half of total phosphorus (20-50 %) and total dissolved phosphorus (14-44 %) losses. Variability in total and dissolved phosphorus losses and concentrations over a wide range of flow conditions suggests that event magnitude is an important control on the relative mobility of particulate and dissolved phosphorus fractions. This study showed that insights into nutrient dynamics and phosphorus budgets in the lower Great Lakes basin and agriculture dominated environments more broadly can be gained by assessing event nutrient losses with respect to flow conditions and patterns in concentration-discharge relationships.

Trends of lake temperature, mixing depth and ice cover thickness of European lakes during the last four decades

Lakes are particularly vulnerable ecosystems to global warming. Surface temperature of most lakes in the world has significantly increased. Here, we analysed time-series of water temperature, mixing-depth, and ice depth of 51 European lakes over the last four decades. We used data of surface temperature, total layer water temperature, mix-layer temperature, mix-layer depth, and ice cover depth obtained from the ERA5-Land reanalysis dataset. Our main objectives were a) to identify significant changes of the examined variables that have occurred from 1981 to 2019 and b) to assess the variability of changes in relation with geographical and lake morphological gradients. To this end, time series analysis was conducted using generalized additive models (GAMs). In addition, we quantified the magnitude of change by estimating the Sen's slopes for each variable and then we examined the variability of these slopes to geographical and lake morphological parameters using GAMs. Our results confirmed that water temperature parameters (surface, total-layer and mix-layer temperature) have significantly increased for all lakes during the last four decades. We also found significant changes of the mixing depth for 14 lakes. In addition, the lake ice depth has significantly decreased in all fifteen lakes of the subarctic climate region. Finally, we showed that the Sen's slopes depend on the geographic coordinates and the elevation of the lakes, whereas lake morphometry (e.g. depth) has a smaller effect on the magnitude of changes. These findings hint that lake ecosystems of Europe have substantially changed over the last forty years and urge the need to take precautionary measures to prevent future implications for the freshwater biota.

Novel predictors related to hysteresis and baseflow improve predictions of watershed nutrient loads: An example from Ontario's lower Great Lakes basin

Eutrophication has re-emerged in the lower Great Lakes basin resulting in critical water quality issues. Models that accurately predict nutrient loading from streams are needed to inform appropriate nutrient management decisions. Generalized additive models (GAMs) that use surrogate data from sensors to predict nutrient loads offer an alternative to commonly applied linear regression and may better handle relationship non-linearities and skewed water quality data. Five years (2015-2020) of water quantity and quality data from 11 agricultural watersheds in southern Ontario were used to develop GAMs to predict total phosphorus (TP) and nitrate (NO₃^-) loads. This study aimed to 1) use GAMs to predict nutrient loads using both common and novel predictors and 2) quantify and examine the variability in seasonal and annual nutrient loads. Along with routine surrogate model predictors (i.e., flow, turbidity, and seasonality), the addition of the baseflow proportion and the hydrograph position of flow observations improved model performance. Conversely, including the antecedent precipitation index minimally affected model performance, regardless of constituent. Seasonal and annual patterns in TP and NO₃^- load predictions mirrored that of the hydrologic regime. This study showed that parsimonious GAMs featuring novel model predictors can be used to predict nutrient loads while accounting for the partitioning of surface and subsurface flow paths and hysteresis between streamflow and water quality parameters that are frequently observed in a wide range of environments.

Using Machine Learning Models for Predicting the Water Quality Index in the La Buong River, Vietnam

For effective management of water quantity and quality, it is absolutely essential to estimate the pollution level of the existing surface water. This case study aims to evaluate the performance of twelve machine learning (ML) models, including five boosting-based algorithms (adaptive boosting, gradient boosting, histogram-based gradient boosting, light gradient boosting, and extreme gradient boosting), three decision tree-based algorithms (decision tree, extra trees, and random forest), and four ANN-based algorithms (multilayer perceptron, radial basis function, deep feed-forward neural network, and convolutional neural network), in estimating the surface water quality of the La Buong River in Vietnam. Water quality data at four monitoring stations alongside the La Buong River for the period 2010–2017 were utilized to calculate the water quality index (WQI). Prediction performance of the ML models was evaluated by using two efficiency statistics (i.e., R2 and RMSE). The results indicated that all twelve ML models have good performance in predicting the WQI but that extreme gradient boosting (XGBoost) has the best performance with the highest accuracy (R2 = 0.989 and RMSE = 0.107). The findings strengthen the argument that ML models, especially XGBoost, may be employed for WQI prediction with a high level of accuracy, which will further improve water quality management.

Wind drives fast changes of light climate in a large, shallow re-established lake

With ever greater frequency, wetlands and shallow lakes that had been diverted for agriculture are being re-established to reduce nutrient loss and greenhouse gas emission, as well as to increase biodiversity. Here, we investigate drivers of water column light attenuation (K_d) at multiple time scales and locations in Lake Fil, Denmark, during the first five years after its re-establishment in 2012. We found that K_d was generally high (overall mean: 3.4 m^-1), with resuspended sediment particles and colored dissolved organic matter being the main contributors. Using daily time series of light attenuation recorded at four stations, we used a generalized additive model to analyze the influence of wind speed and direction on K_d. This model explained a high proportion of the variation (R² = 0.62, RMSE = 0.74 m^-1, and MAE = 0.55 m^-1) and showed that higher wind speed increased K_d on the same day and, with smaller influence, on the next day. Furthermore, we found a significant influence of wind direction and an interaction between wind speed and wind direction, a combination that suggests that short-term variations in light climate depends on the interplay between wind direction and sources of particles. Wind from non-prevailing directions thus influence K_d more, as it can activate previously deposited particles. The maximum colonization depths of submerged vegetation occurred at ~2-6% of sub-surface light from 2014 to 2016 and peaked at 1.2 m in 2016. The fast, day-to-day variation of K_d in Lake Fil reveals the importance of wind on light climate and in turn biological elements such as phytoplankton and submerged macrophyte development in shallow lakes. The implications are essential for the prior planning and management of future lake re-establishment.

Effects of change in streamflow patterns on water quality

Streamflow patterns are closely linked with the quality of stream water, but they are often dealt separately. Due to this, the effects of change in streamflow patterns resulting from river regulation and flow diversion on stream water quality remain under-investigated. This study models change in water quality indicators including pollutants (total suspended solids and turbidity), nutrients (total nitrogen and phosphorus), dissolved oxygen, nitrogen (kjeldahl), pH, and salinity caused by the change in streamflow patterns under different scenarios of river regulation, flow diversion, and rainfall. The generalized additive model was used and the Goulburn-Broken catchment, Australia was chosen as the case study. It was found that concentrations of pollutants and nutrients increased by 38% while dissolved oxygen and nitrogen (kjeldahl) decreased by 35% during the period 1990-2018. These changes were associated with an average increase of 20% in low and medium flows, an average decline of 22% in high and overbank flows and a 15% decline in rainfall. Under the scenario of climate change, river regulation and flow diversion, the overbank flow patterns would mimic the effects of low and medium flows on the water quality indicators that would raise the concentration of pollutants, nutrients, and salinity by 19%. Restoration of high flows would decrease these concentrations by 28% relative to current concentrations, however, it would also reduce dissolved oxygen, nitrogen (kjeldahl), and pH. Effects of streamflow patterns on water quality have implications for environmental flow management, thus, this study recommends critical adjustments in low, medium, and high flows for improving water quality.

Basin-Scale Approach to Integration of Agro- and Hydroecological Monitoring for Sustainable Environmental Management: A Case Study of Belgorod Oblast, European Russia

The quantitative and qualitative depletion of water resources (both surface and groundwater) is closely related to the need to protect soils against degradation, rationalization of land use, and regulation of surface water runoff within the watershed area. Belgorod Oblast (27,100 km2), one of the administrative regions of European Russia, was chosen as the study area. It is characterized by a high activity of soil erosion (the share of eroded soils is about 48% of the total area of arable land). The development phase of the River Basin Environmental Management Projects (217 river basins from the fourth to seventh order) allowed for the proceeding of the development of an integrated monitoring system for river systems and river basin systems. The methods used to establish a geoecological network for regional monitoring include the selection and application of GIS techniques to quantify the main indicators of ecological state and predisposition of river basins to soil erosion (the share of cropland and forestland, the share of the south-oriented slopes, soil erodibility, Slope Length and Steepness (LS) factor, erosion index of precipitation, and the river network density) and the method of a hierarchical classification of cluster analysis for the grouping of river basins. An approach considering the typology of river basins is also used to expand the regional network of hydrological gauging stations to rationalize the national hydrological monitoring network. By establishing 16 additional gauging stations on rivers from the fourth to seventh order, this approach allows for an increase in the area of hydro-agroecological monitoring by 1.26 times (i.e., up to 77.5% of the total area of Belgorod Oblast). Some integrated indicators of agroecological (on the watershed surface) and hydroecological (in river water flow) monitoring are proposed to improve basin environmental management projects. Six-year monitoring showed the effectiveness of water quality control measures on an example of a decrease in the concentrations of five major pollutants in river waters.

Multi-scale trend analysis of water quality using error propagation of generalized additive models

Effective stewardship of ecosystems to sustain current ecological status or mitigate impacts requires nuanced understanding of how conditions have changed over time in response to anthropogenic pressures and natural variability. Detecting and appropriately characterizing changes requires accurate and flexible trend assessment methods that can be readily applied to environmental monitoring datasets. A key requirement is complete propagation of uncertainty through the analysis. However, this is difficult when there are mismatches between sampling frequency, period of record, and trends of interest. Here, we propose a novel application of generalized additive models (GAMs) for characterizing multi-decadal changes in water quality indicators and demonstrate its utility by analyzing a 30-year record of biweekly-to-monthly chlorophyll-a concentrations in the San Francisco Estuary. GAMs have shown promise in water quality trend analysis to separate long-term (i.e., annual or decadal) trends from seasonal variation. Our proposed methods estimate seasonal averages in a response variable with GAMs, extract uncertainty measures for the seasonal estimates, and then use the uncertainty measures with mixed-effects meta-analysis regression to quantify inter-annual trends that account for full propagation of error across methods. We first demonstrate that nearly identical descriptions of temporal changes can be obtained using different smoothing spline formulations of the original time series. We then extract seasonal averages and their standard errors for an a priori time period within each year from the GAM results. Finally, we demonstrate how across-year trends in seasonal averages can be modeled with mixed-effects meta-analysis regression that propagates uncertainties from the GAM fits to the across-year analysis. Overall, this approach leverages GAMs to smooth data with missing observations or varying sample effort across years to estimate seasonal averages and meta-analysis to estimate trends across years. Methods are provided in the wqtrends R package.

Four Decades of Surface Temperature, Precipitation, and Wind Speed Trends over Lakes of Greece

Climate change is known to affect world’s lakes in many ways. Lake warming is perhaps the most prominent impact of climate change but there is evidence that changes of precipitation and wind speed over the surface of the lakes may also have a significant effect on key limnological processes. With this study we explored the interannual trends of surface temperature, precipitation, and wind speed over 18 lakes of Greece using ERA5-Land data spanning over a period of almost four decades. We used generalized additive models (GAMs) to conduct time-series analysis in order to identify significant trends of change. Our results showed that surface temperature has significantly increased in all lakes with an average rate of change for annual temperature of 0.43 °C decade−1. With regard to precipitation, we identified significant trends for most lakes and particularly we found that precipitation decreased during the first two decades (1981–2000), but since 2000 it increased notably. Finally, wind speed did not show any significant change over the examined period with the exception for one lake. In summary, our work highlights the major climatic changes that have occurred in several freshwater bodies of Greece. Thus, it improves our understanding on how climate change may have impacted the ecology of these important ecosystems and may aid us to identify systems that are more vulnerable to future changes.

Responses of phytoplankton community structure and association to variability in environmental drivers in a tropical coastal lagoon

Spatial and seasonal heterogeneity in phytoplankton communities are governed by many biotic and abiotic drivers. However, the identification of long-term spatial and temporal trends in abiotic drivers, and their interdependencies with the phytoplankton communities' structure is understudied in tropical brackish coastal lagoons. We examined phytoplankton communities' spatiotemporal dynamics from a 5-year dataset (n = 780) collected from 13 sampling stations in Chilika Lagoon, India, where the salinity gradient defined the spatial patterns in environmental variables. Generalized additive models showed a declining trend in phytoplankton biomass, pH, and dissolved PO₄ in the lagoon. Hierarchical modelling of species communities revealed that salinity (44.48 ± 28.19%), water temperature (4.37 ± 5.65%), and season (4.27 ± 0.96%) accounted for maximum variation in the phytoplankton composition. Bacillariophyta (Indicator Value (IV): 0.74) and Dinophyta (IV: 0.72) emerged as top indicators for polyhaline regime whereas, Cyanophyta (IV: 0.81), Euglenophyta (IV: 0.79), and Chlorophyta (IV: 0.75) were strong indicators for oligohaline regime. The responses of Dinophyta and Chrysophyta to environmental drivers were much more complex as random effects accounted for ~70-75% variation in their abundances. Prorocentrum minimum (IV: 0.52), Gonyaulax sp. (IV: 0.52), and Alexandrium sp. (IV: 0.51) were potential indicators of P-limitation. Diploneis weissflogii (IV: 0.43), a marine diatom, emerged as a potential indicator of N-limitation. Hierarchical modelling revealed the positive association between Cyanophyta, Chlorophyta, and Euglenophyta whereas, Dinophyta and Chrysophyta showed a negative association with Cyanophyta, Chlorophyta, and Euglenophyta. Landsat 8-Operational Land Imager satellite models predicted the highest and lowest Cyanophyta abundances in northern and southern sectors, respectively, which were in accordance with the near-coincident field-based measurements from the lagoon. This study highlighted the dynamics of phytoplankton communities and their relationships with environmental drivers by separating the signals of habitat filtering and biotic interactions in a monsoon-regulated tropical coastal lagoon.

An approach for decomposing river water-quality trends into different flow classes

A number of statistical approaches have been developed to quantify the overall trend in river water quality, but most approaches are not intended for reporting separate trends for different flow conditions. We propose an approach called FN_2Q, which is an extension of the flow-normalization (FN) procedure of the well-established WRTDS ("Weighted Regressions on Time, Discharge, and Season") method. The FN_2Q approach provides a daily time series of low-flow and high-flow FN flux estimates that represent the lower and upper half of daily riverflow observations that occurred on each calendar day across the period of record. These daily estimates can be summarized into any time period of interest (e.g., monthly, seasonal, or annual) for quantifying trends. The proposed approach is illustrated with an application to a record of total nitrogen concentration (632 samples) collected between 1985 and 2018 from the South Fork Shenandoah River at Front Royal, Virginia (USA). Results show that the overall FN flux of total nitrogen has declined in the period of 1985-2018, which is mainly attributable to FN flux decline in the low-flow class. Furthermore, the decline in the low-flow class was highly correlated with wastewater effluent loads, indicating that the upgrades of treatment technology at wastewater treatment facilities have likely led to water-quality improvement under low-flow conditions. The high-flow FN flux showed a spike around 2007, which was likely caused by increased delivery of particulate nitrogen associated with sediment transport. The case study demonstrates the utility of the FN_2Q approach toward not only characterizing the changes in river water quality but also guiding the direction of additional analysis for capturing the underlying drivers. The FN_2Q approach (and the published code) can easily be applied to widely available river monitoring records to quantify water-quality trends under different flow conditions to enhance understanding of river water-quality dynamics.

The Long-Term and Retention Impacts of the Intervention Policy for Cage Aquaculture on the Reservoir Water Qualities in Northern China

To ensure the safety of the water supply of the Panjiakou reservoir, in 2016, the Chinese central government comprehensively banned the fishing cage culture that had lasted for almost 30 years. However, the long-term effects and retention impacts of the government’s mandatory intervention on the reservoir water quality are unknown. To determine the reservoir water quality, we employed statistical methods along with the mathematical model to investigate the internal relationship since the construction of the reservoir. We applied seasonal trend decomposition using loess (STL) to explore the long-term and seasonality trend of monthly total nitrogen (TN) and total phosphorous (TP). To separate the impact of upstream water quality changes from cage culture on reservoir water quality, we employed generalized additive models (GAMs). We created a model, the LAKE2K model, to investigate the internal sources of the sediment that accumulated during the aquaculture period and its retardant effect. The results revealed that the concentration of upstream TN was more affected by non-point sources than by TP. The long-term policy of encouraging aquaculture has greatly contributed to the increase in the reservoir TP concentration rather than an increase in TN; the prohibition of cage aquaculture has resulted in a sharp drop in TP. After the ban, the sediment became the main source of TP. We suspect that the TP concentration of the reservoir and sediment will decrease gradually until a new equilibrium is reached within 10 years. This study offers lake managers an opportunity to increase their insight into the interaction of management measures with water quality and provides valuable information for the natural recovery of the eutrophic system.

A simulation-based approach to assess the power of trend detection in high- and low-frequency water quality records

To provide more precise understanding of water quality changes, continuous sampling is being used more in surface water quality monitoring networks. However, it remains unclear how much improvement continuous monitoring provides over spot sampling, in identifying water quality changes over time. This study aims (1) to assess our ability to detect trends using water quality data of both high and low frequencies and (2) to assess the value of using high-frequency data as a surrogate to help detect trends in other constituents. Statistical regression models were used to identify temporal trends and then to assess the trend detection power of high-frequency (15 min) and low-frequency (monthly) data for turbidity and electrical conductivity (EC) data collected across Victoria, Australia. In addition, we developed surrogate models to simulate five sediment and nutrients constituents from runoff, turbidity and EC. A simulation-based statistical approach was then used to the compare the power to detect trends between the low- and high-frequency water quality records. Results show that high-frequency sampling shows clear benefits in trend detection power for turbidity, EC, as well as simulated sediment and nutrients, especially over short data periods. For detecting a 1% annual trend with 5 years of data, up to 97% and 94% improvements on the trend detection probability are offered by high-frequency data compared with monthly data, for turbidity and EC, respectively. Our results highlight the benefits of upgrading monitoring networks with wider application of high-frequency sampling.