Supporting the development of climate-resilient water safety plans through a risk-screening framework for drinking water source protection

Climate changes, through increasing temperatures and changes in precipitation patterns and amounts, are expected to lead to increased pressures on water resources and a resulting increased risk of decreasing water quality (Johnson et al. 2022). For drinking water purposes, decision-makers therefore need to anticipate and plan for potential deteriorations in the water quality of untreated abstraction sources (also called raw water quality) in order to ensure a continued public supply of high-quality potable water (Delpla et al. 2009). However, anticipatory risk management is challenged by deep uncertainty associated with climate change, the need to plan and implement for resource security over long time scales (especially due to long lead times for infrastructure implementation), and limited capital resources highlighting the need for both targeted and flexible adaptive responses (Jeuland & Whittington 2014; Hall et al. 2019).

This identifies requirements for a flexible cross-scale approach to adaptation decisions that can place insights from local-scale sampling and modelling data in the wider context of large-scale water resource management (Bartlett & Dedekorkut-Howes 2023). Such a strategic approach would enable identification of priorities for investment both in terms of geographic areas that are of higher risk from potentially decreasing water quality, as well as of risk-management options that are robust against uncertainty about future climates and environmental water-quality variations (Borgomeo et al. 2018). This paper proposes a risk-based framework to integrate relationships between climate change and raw water quality, and to build the capacity to anticipate and respond to change at the national or regional scale, where the overarching goal of ensuring continued public supply of high-quality drinking water is planned.

The management of risks to water quality and quantity is crucial to support a continuous and affordable supply with high-quality drinking water compliant with international safety standards. Guidance on methods and tools for risk assessment within the water sector is available (Pollard & Stephenson 2016), including guidelines of the World Health Organisation (WHO 2017a), which define a widely implemented framework through Water Safety Plans intended to cover risks to drinking water quality from source to potable water tap. As an extension to Water Safety Plans, climate-resilient water safety plans (CRWSP) attempt to provide preparedness to climate change-related impacts on drinking water quality and identify robust adaptation options at the individual supply level (WHO 2017b). Climate change is expected to result in changes to water quality through direct effects from rising temperatures and changing precipitation amounts and seasonality, and through indirect effects from interaction with socioeconomic drivers and changes in anthropogenic pressures such as intensified land use and management (Whitehead et al. 2009; Brown et al. 2015). There is also the prospect of new emergent risks from either biological or chemical sources (Geissen et al. 2015; Vorstius et al. 2024). For drinking water providers, deterioration in the quality of abstracted ‘raw’ (pre-treated) water can overwhelm existing treatment practices leading to increased capital and operational costs to comply with regulatory standards. Failure to anticipate significant potential changes, and over-reliance on incremental reactive responses, can result in operators being locked into unsustainable, increasingly costly, or ineffective approaches. Including a forward-looking climate-change perspective into resource monitoring, planning, and investment decisions is therefore crucial, especially while meeting obligations to reduce greenhouse gas (GHG) emissions (Garnier & Holman 2019).

The CRWSP procedure introduces climate considerations into the development steps of Water Safety Plans and should include aspects of yield reliability, historical water quality, and future climate projections and trends in land use and population that could impact water quantity and quality (WHO 2017b). This requires knowledge not only of the nature and scope of possible changes in climate, but also of their risks (direct and indirect) on the quantity and quality of water, which is inherently complex (Srivastav et al. 2021). Scientific projections of future states are hampered by the epistemic uncertainties of climate change, and render assumptions of fixed adaptation decisions unfeasible, putting iterative and adaptive risk management to the forefront (Döll & Romero-Lankao 2017). These challenges mean the creation of CRWSPs is still in its infancy with few published examples and independent reviews of their effectiveness (Rickert et al. 2019; Pham & Dam 2021; Rand et al. 2022). Furthermore, CRWSPs are developed for individual supply systems, so from the viewpoint of national or regional drinking water provision, a strategic step that assesses the risk of the large-scale supply network to water quality changes, and that provides a framework for re-evaluation and adaptive planning, is missing. This step can be provided by improved integration of CRWSPs with national/regional Water Resource Management Plans regarding recognition of key risk factors, underlying uncertainties, and appraisal of prospective responses to help ensure resource resilience and security. Ultimately, this approach can allow for systematically building the capacity to integrate climate-change preparedness into Water Safety Plans, with a focus on systems most exposed and vulnerable to changes in climate-related risks.

As an example of a generic risk-based strategy, catchment source protection is increasingly recognised as playing an important part in ensuring high-quality drinking water supply through mitigating anthropogenic impacts of land-management practices. Examples of successful applications of this approach are the New York City Watershed Agricultural Programme (Appleton 2002; Kousky 2015) and the Sustainable Catchment Management Programme of United Utilities in the Northwest of England (United Utilities 2022). Under the umbrella of ecosystem-based adaptation, which aims to develop climate responses based on knowledge of natural processes (UNEP 2012), catchment source protection can become a crucial part of creating climate-resilient supplies, by supporting healthy and resilient ecosystems that continue to supply high-quality raw water under changing conditions. Early recognition of potential changes to raw water quality from changes in catchment processes induced by climate change allows the implementation of catchment-based low-regret adaptation options that stabilise water quality and reduce necessary increases in treatment efforts, including investment into additional treatment technology. This can also support obligations to reduce GHG emissions consistent with net-zero targets as water treatment processes are traditionally energy intensive (Del Río-Gamero et al. 2020).

On a strategic level, the uncertainty of climate change and its effects on water resources, coupled with a need to invest limited capital resources where they achieve largest impacts, means that understanding which sources and their catchments are potentially at risk from direct or indirect impacts of climate change is a prerequisite for a successful application of this concept. Screening the supply network for higher risk systems based upon changing risk factors is therefore crucial for the development of overarching adaptation strategies to increase the resilience of the supply network and to maximise the outcome of targeted intervention actions (Hall et al. 2019). The framework presented in this paper provides a flexible screening approach to bridge the gap between the strategic and operational level, as demonstrated by its application to the public water supply in Scotland.

For risk screening, a framework was developed to successively build the capacity to anticipate, respond, and adapt to changes in climate through a better understanding of the supply sources, their exposure to changing risk factors, and uncertainties associated with those future changes (Figure 1).

Water Safety Plans involve a characterisation of the supply system with associated identification of risk factors and lead to an awareness of current water-quality patterns and implementation of risk controls (WHO 2017a). On a more strategic level, analysing differences in water quality in relation to catchment characteristics that vary over space and/or time establishes an understanding of overarching processes and pressures that drive water-quality outcomes. While some catchment characteristics are relatively invariant over the target time period (e.g., topography), they have important spatial variations, whereas other dynamic characteristics have notable temporal and spatial variations associated with biophysical (e.g., precipitation and land cover) and socioeconomic processes (e.g., land use).

Stage 1 of the suggested approach therefore involves identification and modelling/representation of these processes and their relationships as a basis for understanding how changes in catchment properties can influence water quality (no -). This involves techniques that enable the understanding of the relationships between raw water-quality observations and intrinsic catchment properties and how differences in, for example, land-management practices, may influence water quality. At Stage 2, projections for future dynamic changes in catchments and their properties feed into these systems’ representations, leading to possible trajectories for water quality, and enable an understanding of how risk patterns may change. In other words, this step ‘translates’ information on changes in climate, such as modelled temperature and precipitation changes, into water-quality changes using the established relationships in Stage 1. This supports robust risk management at Stage 3, as long-term effectiveness of different adaptation options can be considered for different catchment types and their sensitivities. For example, it can be better estimated if treatment work upgrades are sufficient for the duration of its lifespan. An assessment of long-term viability of the complete supply and water resource system can be achieved at Stage 4 by considering possible changes in risk profiles for a multitude of supplies and identifying general areas of concern rather than focusing on individual supplies. This supports the development of overarching adaptation strategies incorporating different catchment sensitivities at the regional/national scale.

Building on findings of Vorstius et al. (2019), using multivariate geospatial analysis to ‘explain’ patterns of raw water quality for over 150 water supply sources, the framework was followed to produce trajectories and discuss implications for two diagnostic water-quality parameters: total organic carbon (TOC) and Escherichia coli, both exhibiting different spatial patterns of contamination owing to different drivers and processes (see Box S1).

An empirical approach was chosen for Stage 1 that builds on low-frequency water-quality monitoring data from a national network of sites, as it can take advantage of the wide spatial coverage to connect differences in catchment properties to water-quality outcomes. In comparison, process-based models usually require higher frequency data for calibration and parameterisation, are developed for individual sites, and take more time to develop, limiting their potential for risk screening.

In Scotland, water is predominantly sourced from surface waters, inclusive of natural lakes, reservoirs, and rivers. We utilised a database of raw water-quality data from the national drinking water provider (Scottish Water) supplying potable water to 97% of the Scottish population extending across the mainland and major inhabited islands (Table 1).

Drinking water sources were described by catchment characteristic compiled into ‘static factors’, such as area, elevation, topography, geology and soil properties, and ‘dynamic factors’ such as climate and land use (Table S1). Multiple linear regression was used to understand the relationships between catchment characteristics and water-quality parameters. The models were developed on log-transformed catchment median TOC and E. coli values as dependent variables, and catchment properties as independent variables (Table S2 and Figure S1). Median concentrations were chosen to represent a baseline contamination levels and thus to reflect usual pressures in the catchment. Manual step-wise backward regression based on Bayesian information criterion was used to find relevant variables. Adjusted R² and normalised root mean squared error from a 10-times 10-fold cross-validations were used to evaluate goodness of fit and potential overfitting.

We used appropriately downscaled climate projections (UKCP18) to evaluate direct changes to climate factors (Box S2) and a national land-use capability modelling approach (Box S3 and Figure S2) to anticipate changes in land-cover and land-management practices that together will alter risk patterns and the need for strategic interventions.

Replacing relevant dynamic catchment property variables in the resulting final models with future projections resulted in projected water-quality outcomes. The models and their outputs were interpreted to understand underlying hydrogeochemical catchment processes, which allowed to formulate hypotheses about possible water-quality trajectories that reflect future shifts to baseline concentrations and thus the risk magnitudes for TOC and E. coli.

The statistical models based upon pooled catchment data were used to infer changing risk factors relative to climate and land-use pressures. This provided a basis to understand future risk magnitudes for the specific water-quality parameters of interest, based on changes in exposure to climate hazards and intrinsic vulnerabilities, leading to the development of overarching raw water-quality trajectories.

The raw water-quality trajectories can be used to identify potential future high-risk systems and areas, providing the basis for risk management at Stage 3, and to uncover strategic-level implications for water resource management, enabling Stage 4 of the framework.

The raw water-quality trajectories inform possible options to improve anticipatory risk management at catchment and individual water supply system levels through an evaluation of the role of different risk factors and the role of enhanced data collation and analysis. They uncover key uncertainties that require improved monitoring for subsequent analysis and modelling of individual systems. In this context, important differences can be distinguished between the two parameters.

For TOC, uncertainty relates primarily to the natural biophysical processes, which are confounded by the varying influence of different risk factors, and the analysis focused on direct responses to climate drivers and how these will change over space and time. Follow-on adaptation planning should further address key uncertainties through further monitoring and analysis, to better understand the consequences of reduced amounts of rainfall during the summer months, in combination with increasing temperatures, on TOC/DOC dynamics. For identified high-risk catchments, land-management interventions that target key risk factors (e.g., peatland restoration to reduce surface runoff rates) should be trialled and monitored as an adaptive management strategy.

Causes and processes leading to contamination of surface waters with E. coli and other faecal pathogens are relatively well studied (Vinten et al. 2004; Kay et al. 2010; Oliver et al. 2016). The main areas of uncertainty lie in forward predictions of risk factors, which were here represented by potential changes in land use and management through the proxy of land capability. Projecting changes in land capability for agriculture identifies where risk could increase unless additional adaptation measures are implemented to constrain risky land-use practices (e.g., livestock entering water bodies), or otherwise to positively incentivise practices consistent with good water-quality outcomes (e.g., manure storage). These catchments can be systematically prioritised for further engagement with land managers, including more detailed investigations of different forms of land use and practices on water quality, and prioritisation of alternative management options. To this end, scenarios that include socioeconomic factors might be particularly advantageous to investigate the range of different land-use changes, positive and negative (e.g., government incentives, regulatory regimes, consumer demands, and commodity prices). Early engagement with different stakeholders in the identified high-risk catchments would anticipate and aim to steer changes in a positive direction for water-quality outcomes, including partnership-type schemes such as payment for ecosystem services to offer land managers a broader range of economically viable options (Pynegar et al. 2018).

Consistent with an iterative risk-screening approach, further refinements can be added to support risk-management (no -) decision making. The climate projections used herein involved an ensemble approach from one climate model, albeit one with relatively high climate sensitivity compared to other models in the Intergovernmental Panel on Climate Change suite (Murphy et al. 2018). In practical terms this is consistent with a ‘reasonable worst-case scenario’ for risk assessment (Yoe 2011) and matches the primary purpose of risk screening to identify future threats to source water quality (Hrudey et al. 2006). However, further analysis could use alternative climate change-projections or timescales to assess the relative magnitude of projected water-quality outcomes under different climate scenarios. Similarly, the land-use change projections could be refined by considering additional factors and constraints influencing land use. For example, irrigation may not be available to sustain more intensive agricultural practices.

Median concentration values were used to reference against existing baseline pressures in the catchment. However, these metrics could be replaced by other indicators representative of current regulatory standards or risk tolerance (e.g., number of times raw water quality exceeds the treatment envelope, number of people served by the supply, or a combination of several factors). The choice of method and data is flexible under the framework and can be adapted to suit data and time availability in balance with desired level of detail.

The analysis is based on exploring relationships between catchment properties and water quality, for catchments with adequate data. The relationships thus generated enable extrapolation to catchments with more limited monitoring data (often a consequence of perceived risk being assumed lower for a particular water-quality parameter) and can be used to ‘flag’ potential changes in risk for some of those catchments, and therefore a need for improved local data availability. The same approach could also be applied to screen potential new drinking water sources, including for private supplies, to alert water resource managers to emerging risks and the implications for ensuring safe water supplies both in terms of treatment processes and the wider suite of catchment risk-management options. These possibilities support crossing from planning at the individual supply level to the more strategic planning at Stage 4 of the framework.

In the absence of water-quality data, alternative approaches can be used to establish baseline risk factors and apply future climate and/or land-use projections, such as risk screening based on perceived ‘risky’ catchment properties and vulnerabilities. Emerging contaminants especially lend themselves to these approaches, as monitoring is usually absent or not well established (e.g., Vorstius et al. 2024).

In this application of the framework, two indicative water-quality parameters were chosen that represented different broad types of catchments. This allows the development of hypotheses on impacts on related water-quality parameters and thus gives a more holistic picture of water-quality outcomes from changes in climate and land use. These hypotheses can be further explored through the development of additional trajectories or different modelling approaches, making it possible to integrate different water-quality aspects and study the effect of catchment-based mitigation and adaptation measures on multiple parameters. Taking a more holistic view is important to ensure that measures to improve one parameter do not inadvertently deteriorate others, especially regarding the need for additional treatment processes.

The raw water-quality trajectories identify overarching changes in spatial patterns of risk, and enable catchment comparisons based on intrinsic catchment properties and water-quality outcomes. Knowledge gained from increased monitoring, investigation, and risk management in individual systems (at Stage 3) can thus provide insight into similar supply systems and enable evaluation of the viability of the network under different future conditions.

The analysis highlights key areas of uncertainty and limits to quantitative predictions. Mechanical and chemical treatment procedures to address water-quality issues generally follow a ‘predict and provide’ approach to risk management, which strongly relies on a deterministic predictive capability that may be undermined by uncertainties from climate change (Dessai et al. 2009). The presented framework using water-quality trajectories identifies potential future high-risk areas based on projections, but rather than relying on those to plan future treatment investments, it uses them as guidance for prioritising enhanced monitoring, investigation, and engagement, as integral components of a resource-based adaptive management strategy. It thereby supports targeted knowledge generation that feeds back into regular re-evaluation of the supply system, thus building the capacity to support resilient ecosystems that can supply stable raw water quality under a range of possible futures, including those of which we may not yet be aware. This can also facilitate further development of multiple adaptation pathways to manage both specific and collective water resource assets contingent on evolving knowledge on the rate and magnitude of climate-change parameters and associated risks (cf., Muccione et al. 2024).

The challenges uncovered through the analysis highlight problems of over-reliance on treatment plants to resolve changing raw water-quality risks. For example, rising baseline concentrations could mean more frequent occurrences of water quality issues that lie beyond the originally designed treatment envelope of the treatment work, requiring costly upgrades. As it may be difficult for a government or water utility to justify investment into treatment facilities without greater evidence of prospective changes to water quality, the results emphasise a broader, more holistic, and precautionary approach to increase the resilience of catchments as functioning water purification systems (Falkenmark et al. 2019), and minimise impacts of treatment works.

Catchments are complex systems from biophysical and socioeconomic perspectives, therefore addressing water-quality outcomes through catchment interventions rather than at treatment plants can be generally more complicated (especially when involving multiple land managers). It will require a different implementation strategy for the water resource manager, based upon targeted anticipatory catchment management to ensure water quality alongside a broader suite of ecosystem services, but healthy and well-managed ecosystems provide further co-benefits such as climate-change mitigation, flood mitigation, enhanced biodiversity, and amenity value. For example, the protection and restoration of peatland is not only a measure to reduce organic carbon release into surface waters but safeguard important carbon stores and contribute to net-zero GHG strategies (Motelica-Wagenaar et al. 2020; Evans et al. 2021). Similarly, a holistic approach to livestock management is not only water sensitive but also minimises airborne GHG emissions, for example, through effective manure management (Hyland et al. 2016; Lötjönen et al. 2020). The proposed screening approach in combination with improved risk management at the catchment level is therefore also more consistent with the net-zero agenda, which is now a key strategic planning priority for water utilities.

A framework for catchment-based risk screening has been presented to facilitate a systematic, national-scale identification of direct and indirect climate-change factors that affect water resources and their catchments, and thus pre-treated water quality used for drinking water supply. The framework provides a flexible approach that builds on analysis and awareness of current water-quality outcomes associated with catchment characteristics and processes (Stage 1) and combines this knowledge with projections of future changes to understand possible trajectories for water-quality parameters (Stage 2). This approach recognises the limits of attempts to predict water-quality outcomes for specific systems based on past trends and relationships, and emphasises the insights gained from analysing spatial variabilities to understand underlying processes und catchment vulnerabilities to pressures. In this way, the screening assesses the national or regional water resource to identify areas of comparatively high risk for further monitoring, analysis, and intervention. Moving along the framework, risk-management options for identified individual high-risk supply systems can be evaluated that take into account changing drivers of and pressures on raw water quality, thereby enabling anticipatory decision making focusing on options that are beneficial under a range of possible futures (Stage 3). Furthermore, due to the national-scale assessment, overarching strategic decisions about the viability of the supply network and accompanying investment needs are also possible (Stage 4).

Application of this approach for Scotland has shown how it can help distinguish spatial and temporal variations in water quality and associated risk factors, and how projection of changes in risk factors enables anticipatory adaptation responses to further understand and manage risks, especially in areas of highest concern. In particular, risk screening highlights different areas of uncertainty, which guides strategies to reduce those if and where possible and/or implement adaptive responses that evolve with different unfolding futures. This points particularly towards low-regret options to enhance adaptive capacity through targeted monitoring, early stakeholder engagement, and catchment source protection. The limits of predictions (exemplified e.g., by modest explanatory power of modelling approaches; or cascaded from climate model uncertainties) emphasise the danger of over-reliance on treatment and projecting treatment requirements to meet future water-quality outcomes. Strengthening ecosystems to stabilise water quality, and continue the supply of multiple benefits, means that this strategy also delivers important co-benefits (notably for the net-zero emissions agenda), and can reduce costs for the drinking water provider and, ultimately, customers. Such approaches have international applications, most recently illustrated by the publication of the European climate risk assessment (EEA 2024) highlighting the climate impacts on freshwater ecosystems, inclusive of drinking water supplies.

In summary, the outlined framework constitutes a tiered assessment based on risk screening that supports a step-wise increase in capacity, to first understand facets of climate change and the ways it may impact water quality, then to respond to those, and finally enable necessary changes to increase overall resilience of the drinking water supply network. Following this approach facilitates increased integration of climate-change aspects into established risk assessments and can thus guide and support the targeted development of CRWSPs.

We acknowledge the Scottish Government for funding this research through the Hydro Nation Scholars Programme and Scottish Water for providing data. We would also like to thank Professor Stephen Hubbard for his advice on statistical methods.

This work was supported by the Scottish Government under the Hydro Nation Scholars Programme.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

R code is available in the associated PhD thesis: https://dx.doi.org/10.15132/20000343

C.V. conceptualized the article, developed the methodology, conducted the analysis, and wrote the article. J.S.R. supported conceptualisation of the article, and reviewed and edited the article. I.B. supported conceptualisation of the article, developed part of the methodology and analysis, and reviewed and edited the article. F.L. supported resource preparation, and reviewed and edited the article. Z.F. supported visualisation of the work, and reviewed and edited the article.

The dataset generated and analysed during this study is available in Discovery, the University of Dundee Research Portal: https://doi.org/10.15132/10000198.

The authors declare there is no conflict.