A Framework for Assessing Concentration‐Discharge Catchment Behavior From Low‐Frequency Water Quality Data

Abstract

Effective nutrient pollution mitigation measures require in‐depth understanding of spatio‐temporal controls on water quality which can be obtained by analyzing export regime and hysteresis patterns in concentration‐discharge (c−Q) relationships. Such analyses require high‐frequency data (hourly or higher resolution), hampering the assessment of hysteresis patterns in widely available low‐frequency (monthly, biweekly) regulatory water quality data. We propose a reproducible classification of c−Q relationships considering export regime (dilution, constancy, enrichment) and long‐term average hysteresis pattern (clockwise, no hysteresis, anticlockwise) applicable to low‐frequency water quality data. The classification is based on power‐law c−Q models with separate parametrization for low and high discharge and rising and falling hydrograph limb, enabling a better representation of c−Q dynamics. The classification has been applied to a 30‐years record of daily streamflow and monthly spot samples of solute concentrations in 45 Scottish catchments with contrasting characteristics in terms of topography, climate, soil and land cover. We found that c−Q classification is solute‐ and catchment‐specific and linked to upland versus lowland catchments and streamflow variability. However as the relationship between solute behavior and catchment characteristics is variable, we propose that future typologies should integrate both water quality response, that is, c−Q classification, and catchment characteristics. The data‐driven c−Q classification allows us to increase the information content of low‐frequency water quality data and thus inform mitigation measures, monitoring strategies, and modeling approaches. Such approaches open up an ability to characterize processes and best management for a wider number of catchments, subject to regulatory surveillance and outside of research catchments.

A systems approach to modelling phosphorus pollution risk in Scottish rivers using a spatial Bayesian Belief Network helps targeting effective mitigation measures

“…While the manual calibration informed by observed data and plausible ranges of SRP losses from literature was still subject to equifinality, a major advantage of this conceptual risk-based modelling approach is that all parameters have a physical meaning and hence could be potentially constrained by observations. Identifying controlling factors on SRP pollution is known to be difficult (Glendell et al, 2019;Pohle et al, 2021) due to complex interacting processes and lag effects (Bieroza et al, 2020). In this study, sensitivity analysis highlighted the importance of catchment-specific parameterisation of key processes affecting SRP losses from different sources and SRP concentrations at the catchment outlet.…”

“…This has led to the development of models and decision support tools (Drohan et al, 2019) to inform evidence-based decision making. However, these tools often struggle to represent the site-and catchment-specific nature of P loss and the catchment-specific responses to pressures (Drohan et al, 2019;Glendell et al, 2019;Pohle et al, 2021). In addition, performance of complex models is often hampered by lack of observational data for model parameterisation (Jackson-Blake et al, 2017;Drohan et al, 2019;Fu et al, 2020).…”

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

“…Generalized additive models that extend linear models are another statistical approach used to determine water quality trends (e.g., Morton & Henderson, 2008; Yang & Moyer, 2020) and can be adapted to large datasets (Wood et al, 2015). Another common statistical technique is correlation analysis between riverine chemical concentrations and discharge (typically referred to as C‐Q relationships), which has been used for a variety of purposes such as examining constituent dynamics at short and long time‐scales (e.g., Arora et al, 2020; Evans & Davies, 1998; Godsey et al, 2009; Moatar et al, 2017; Musolff et al, 2021), identifying sources and pathways of different solutes (see Musolff et al, 2021 and references therein), and analysing water quality monitoring data for watershed management (Bieroza et al, 2018; Pohle et al, 2021; Westphal et al, 2020). However, complex C‐Q patterns, such as those resulting from variable lags between the hydrograph and chemograph, are difficult to interpret and attempts to link insights gained from C‐Q analysis to models have been limited (Liu, Birgand, et al, 2021).…”

Can Prediction and Understanding of Water Quality Variation Be Improved by Combining Phosphorus Source and Waterbody Condition Parameters?

“…For example, septic tanks, often considered diffuse sources, represent multiple small point sources ranging from diluted to concentrated effluents, discharging variously, including broken tanks flushed by rising water tables. Data on sources and water quality often differ; individual research catchments may have short-duration, intensive, sub-daily water quality characterisation to link with bespoke spatial characterisation (e.g., soil sampling; Jordan et al, 2012), whereas national studies often use country-wide spatial datasets and decadal, monthly regulatory river monitoring (Mockler et al, 2017), where utility to look at temporal aspects such as concentration vs. flow is being questioned (Pohle et al, 2021). However, approaches encompassing ecologically relevant pollution factors such as source bioavailable P are rare in individual catchment studies (McDowell et al, 2016) and very limited in national investigations.…”

“…However, these authors also concluded that such methods would be robust and best tested using the highest temporal resolution stream data that is seldom available across wide stressor gradients. Pohle et al (2021) used Multiple Factor Analysis (MFA) to integrate a range of catchment descriptors (landcovers, soils), derived hydrological indices with concentration-discharge (C-Q) hydrochemistry across macronutrients, major anions and cations in large UK catchments. The benefit of the MFA approach as a global principal components analysis type approach was to summarise multiple complex catchment datasets into an optimal lowdimensional space to facilitate interpreting data interactions; in their case catchment controls on solute exports.…”