Impacts of climate change on drinking water quality in Norway

In Norway, climate change will lead to higher temperatures, increased precipitation and runoff, as well as more frequent extreme weather events (Hanssen-Bauer et al. 2015). In addition, with higher temperatures, rainfall patterns are expected to shift towards more frequent or more intense extreme precipitation (Meehl et al. 2007). From 1900 to 2014, the annual mean temperature has already increased by approximately 1 °C (Hanssen-Bauer et al. 2015). This observation is associated with increased runoff in winter and spring, and earlier snowmelt. The annual precipitation in Norway has increased by approximately 18% during the same period, with most of the increase occurring after the late 1970s. Heavy short-duration rainfall has increased both in intensity and frequency in recent years (Sorteberg et al. 2018).

The observed changes in the Norwegian climate are estimated to continue into the future (Hanssen-Bauer et al. 2015). Climate model projections based on a high scenario indicate an increase in the annual mean temperature for Norway of 4.5 °C and an increase in the annual precipitation of 18% by the end of this century. Following the increasing precipitation, the intensity of rainfall-induced floods is also expected to increase (Lawrence 2020). Higher temperatures are likely to lead to a shift towards earlier spring floods and increased possibility for late autumn and winter floods (Beldring et al. 2008). These changes in weather and flood patterns may threaten the quality of the Norwegian drinking water.

Extreme weather conditions have long been associated with water quality impacts and associated waterborne diseases (Whitehead et al. 2009; Delpla et al. 2009; Khan et al. 2015; Young et al. 2015). Heavy rainfall and runoff, floods, cyclones, droughts, heatwaves, extreme cold and wildfires may all affect water catchments, storage reservoirs, water treatment processes or distribution systems, with potential impacts on the quality of drinking water (Khan et al. 2015). Source waters of drinking water may be impacted by increased concentrations of suspended material, organic matter, nutrients, inorganic substances and pathogenic microorganisms (Whitehead et al. 2009; Khan et al. 2015; Young et al. 2015).

The surface waters are particularly vulnerable to changes in temperature, heavy rainfall and floods. Extreme precipitation and runoff events as well as high temperatures may lead to a higher density of microorganisms in water bodies and increase the risk of waterborne disease (Kistemann et al. 2002; Hunter 2003; Guzman Herrador et al. 2016). The increase in rainfall-induced flooding may also increase human exposure to waterborne pathogens (Kelman 2011). As most of the Norwegian drinking water is obtained from surface water, the Norwegian drinking water may be especially sensitive to the increasing events of extreme precipitation and floods. Ageing drinking water distribution systems and sewage systems will be particularly vulnerable to flooding. Many infectious microorganisms are sensitive to climatic conditions. These factors together will increase the risk and burden of waterborne illnesses.

A recent study shows evidence of associations between weather events and raw water quality in Norway (Guzman Herrador et al. 2021). Increased amounts of rain and runoff were associated with increased levels of Escherichia coli (E. coli), coliform bacteria, intestinal enterococci, colour and turbidity in raw water throughout the entire year. In raw water, increased maximum temperature was not significantly associated with increased or decreased levels of any of the outcomes throughout the entire year, however, was associated with an increase in coliform bacteria, E. coli, intestinal enterococci, and turbidity in winter, and turbidity in spring.

Norway, with its elongated shape and highly diverse topography, has a climate that varies substantially from region to region. In order to evaluate potential upgrades of the treatment capacity at the waterworks, there is a need for regional information about the risk of drinking water deterioration in the future. Therefore, to investigate potential future changes in drinking water quality, known historical associations between weather/runoff and water quality from a geographically wide distribution of Norwegian waterworks must be combined with future climate scenarios that are downscaled to local levels. This has so far not been done.

In this study, we combine the recent associations found in Guzman Herrador et al. (2021) with locally downscaled future climate projections in order to estimate regional changes in raw water quality in Norway by the end of the century.

We used the associations between weather and raw water quality which were calculated by Guzman Herrador et al. (2021). In their study, the water quality indicators coliform bacteria, E. coli, intestinal enterococci, colour and turbidity were first collected from 26 of the largest Norwegian waterworks for the period 2006–2014. These were received as weekly averages. For each water work intake point, four weekly averaged observed meteorological and hydrological data (daily maximum temperature (°C), accumulated daily rainfall (mm) and runoff (mm)) were combined with the succeeding week of water quality data for raw water. Finally, mixed-effects linear regression models were run for all water work intake points, as described in Guzman Herrador et al. (2021). The resulting associations were stratified by four seasons (winter: December–February, spring: March–May, summer: June–August and autumn: September–November).

For future values of daily maximum temperature, daily accumulated rainfall and runoff, we used post-processed climate projections derived from global climate models (GCMs). The future projections are based on results from the Fifth Assessment Report by the Intergovernmental Panel on Climate Change (Cubasch et al. 2013). In this study, one of the IPCC emission scenarios, the so-called ‘Representative Concentration Pathways (RCPs)’, is used: RCP8.5 (van Vuuren et al. 2011). RCP8.5 represents the highest scenario, referring to a future with continuously increasing greenhouse gases. As a precaution, the government demands the work on climate adaptation to consider high alternatives from the national climate projections when consequences of climate change are assessed (Miljøverndepartementet 2013).

GCMs generally operate on a horizontal grid size of 100×100 km², which is too coarse to resolve mesoscale phenomena that are important for Norway and its climate. For regional climate impact studies, the model projections are further downscaled by regional climate models (RCMs). The Euro-CORDEX initiative (https://www.euro-cordex.net/) offers a comprehensive database of climate projections with different downscaled GCM/RCM combinations. An ensemble of 10 Euro-CORDEX model runs with a spatial resolution of 12×12 km² was used in this study (Jacob et al. 2014). Despite the finer spatial resolution, the climate projections for Norway generally show significant temperature and precipitation biases. A post-processing of the 12×12 km² RCM outputs is therefore necessary. The daily temperature and precipitation values were first interpolated to a 1×1 km² grid. The observation-based datasets, which also have a 1 km resolution, were used for the bias-adjustment procedure which removed systematic biases in simulated values relative to the observed data. For further details of this procedure and the overview of the model ensemble used, see Wong et al. (2016). Runoff was then simulated by a precipitation-runoff model with bias-adjusted temperature and precipitation as input data.

Our model ensemble consists of 10 datasets of future projections of daily maximum temperature, daily accumulated rainfall and runoff under the RCP8.5 scenario. These simulation data were collapsed to the yearly median of each of the variables (daily maximum temperature, daily accumulated rainfall and runoff) and were then used as described below in the section ‘Analyses of future changes in water quality’. Control runs from the models cover the period 2006–2014, whereas for the future analyses, we used projections for the period 2071–2100.

After fitting the mixed-effects linear regression models using the real (observed) data to obtain the associations between meteorological/hydrological data (daily maximum temperature, daily accumulated rainfall and runoff) and water quality data (Guzman Herrador et al. 2021), we used the associations ‘rainfall affecting water quality’, ‘runoff affecting water quality’, ‘maximum temperature affecting water quality’ as well as future predicted exposures to estimate the future changes.

The modelled meteorological and hydrological climate data from 2006–2014 to 2071–2100 were used as exposures to the linear regression models from Guzman Herrador et al. (2021) to obtain predictions of the raw water quality in 2006–2014 and 2017–2100. These predictions were then aggregated to the yearly level. For each waterwork and climate projection, the predicted multiplicative increase in year X from 2006 to 2014 was calculated as: predicted value in year X/predicted mean in 2006–2014.

For each waterwork, climate projection and year from 2071 to 2100, 1,000 data points were resampled from the original data of water quality indicators, maximum daily temperature, daily accumulated rainfall and runoff used to fit the respective linear regression model and multiplied by the multiplicative increase from 2006 to 2014. We then estimated the 99th percentile in this new simulated dataset for each of the waterworks in order to study the changes in extreme concentrations of water quality indicators.

For this study, Norway was divided into three geographical climate regions: West, East and North, and we allocated each waterwork to its region. Then to report regional values, we used the mean of the 99th percentiles of the waterworks allocated to each region.

Due to future changes in rainfall, we estimated markedly higher future levels of water quality indicators with statistical significance (p<0.001) throughout the whole year in the future period (2071–2100) as compared to the reference period (2006–2014) (Figures 1–6; Table 1). The largest future concentrations of indicators were found for the three indicator bacteria E. coli, coliforms and intestinal enterococci in the western part of Norway (Figure 1; Table 1). The largest percentage increases in indicator concentrations were found in the western and northern regions, for all indicators (Figures 2–6). The 99th percentile of E. coli concentration is estimated to increase by 410% in the north (corresponding to a five-fold increase, from 1 colony-forming unit (CFU)/100 mL raw water to 5.1 CFU/100 mL), throughout the whole year. In the West, its concentration is estimated to double. The largest increase throughout the whole year (in percentages) is found for the 99th percentile of intestinal enterococci bacteria concentration in the North with an estimated future increase of 467% (from 0.3 to 1.7 CFU/100 mL) (Table 1). Seasonally, the largest increases in indicator levels were found in summer and autumn in the western and northern parts of Norway.

Using the effect from future changes in runoff, the results varied, with some indicators increasing and others decreasing (Figures 1–6; Table 1). Throughout the whole year, we found only very small changes in most of the water quality indicators and regions due to changes in runoff (mostly <2%). When stratifying by season, the estimated changes were also small and there were few results with statistical significance. We registered, however, a small, but statistically significant increase in all indicator bacteria concentrations as well as colour in the winter season in the West and a decrease in the colour in summer in all regions.

With future changes in maximum temperature, we found no changes in the water quality indicators that were statistically significant when the whole year was considered (Table 1). However, for the winter season, we estimated clear effects from the increasing maximum temperature. We found an increase in all regions for E. coli, coliform bacteria, turbidity and for intestinal enterococci in the East and the West (no concentration was measured in the North). The largest estimated increase is estimated for the E. coli concentration in the north, by 114.3% (from 1.4 to 3.0 CFU/100 mL). We also found an increase in turbidity in winter and spring in all regions, as a result of the increased maximum temperature.

We found increasing future levels of raw water quality indicators throughout the whole year by the end of the century due to future changes in rainfall, with the largest effects in the North and the West. In the East, small changes were seen. This follows from the finding that increased rain was associated with increased levels of coliform bacteria, E. coli, intestinal enterococci, colour and turbidity in Norway throughout the whole year (Guzman Herrador et al. 2021). Furthermore, the annual precipitation in Norway is projected to increase by about 18% from 1971–2000 to 2071–2100 (Hanssen-Bauer et al. 2015), with the largest increases estimated for the northern and western regions of the country.

Seasonally, the results due to future changes in rainfall vary more and depend on both regions, seasons and water quality indicators. While the variations between the indicators can be understood from the differences in the associations found by Guzman Herrador et al. (2021), the seasonal and regional variability may be explained by climatological differences. In the North and the West, most of the indicators show a substantial increase in their future concentrations during the summer and autumn seasons. These are the seasons when the future precipitation is projected to increase the most in the North and West by the end of the century (Hanssen-Bauer et al. 2015).

The fact that we estimate small changes in raw water quality when taking into account the effect from future changes in runoff may be explained by the future projections of runoff. Here, the results are inconsistent from model to model. On average for the whole country, the median of 10 models gives an increase of 7% in annual runoff for Norway from 1971–2000 to 2071–2100, but due to the disagreement between models, the estimate spans from −3 to +11% change (Hanssen-Bauer et al. 2015). The runoff is also generally estimated to decrease along the coast and in lower-lying areas, and to increase in mountains and inner parts of Nordland and Finnmark, which makes it difficult to capture consistent regional signals in the estimates.

The projected runoff is estimated to change much more consistently with season than annually, especially in winter and summer (Hanssen-Bauer et al. 2015). In winter, the runoff is estimated to increase, whereas in summer, it is estimated to decrease. This is shown in all regions. This supports the estimated future increase in all indicator concentrations as well as colour in the winter season in the West, and the decrease in the colour in summer in all regions, when the effect from changing runoff is included. A model study of a Norwegian catchment indicates lower future concentrations of E. coli in summer due to reduced runoff (Mohammed et al. 2019).

Several studies support the link between future changes in rainfall and runoff and decreasing drinking water quality (Delpla et al. 2009; Leveque et al. 2021). In the USA, half the waterborne disease outbreaks during the last half-century followed a period of extreme rainfall (Curriero et al. 2001). Seasonal and interannual variability in local rainfall are found to explain nearly 70% of the variability in the coliform records in the UK (Pednekar et al. 2005).

With future changes in the maximum temperature, small changes are seen in the water quality indicators when the whole year is considered (Figures 1–6; Table 1). The annual mean temperature for Norway is estimated to increase by about 4.5 °C from 1971–2000 to 2071–2100 (Hanssen-Bauer et al. 2015); however, no significant associations were found between temperature and water quality indicators throughout the whole year. Statistically significant changes due to changing temperature were mainly found in winter, and in winter and spring for turbidity. The reason for this may be that warmer winters with shorter periods of ice cover on lakes may reduce the hygienic barrier efficiency against pathogen transport from the surface to the deep-water intake in deep lakes used as drinking water sources (Tryland et al. 2011; Guzman Herrador et al. 2021). The main water sources in Norway are lakes, covered with ice during winter. The ice cover is normally a natural protection against microbes and particles flushing into the water source and prevents an increase in both microbial content and turbidity. However, in recent years, the period of ice cover has decreased. The winters are now milder and are more often associated with air temperatures above 0 °C. Observations of daily maximum temperature show the largest increase in winter. In the future, the largest increase in temperature is also projected for the winter season (Hanssen-Bauer et al. 2015).

In order to get a good estimate of the climate and hydrology at a location, a series with at least 30 years of data should be used. In this study, the water quality data from the waterworks only covered the period from 2006 to 2014. We also have a relatively limited number of waterworks included in our study, with especially few waterworks located in the northern part of Norway. Therefore, we divided the country into only three simplified geographical climate regions. Norway is a country with large variations in temperature and precipitation depending on the region and height above sea level (Førland et al. 2000). As a result, we may not have captured the entire climate and hydrological variability that the waterwork locations represent.

Our analyses were based on associations between weekly averages of raw water quality and meteorological and hydrological variables (Guzman Herrador et al. 2021). Water quality data for raw water were received as weekly averages from the waterworks. A quantitative analysis of changes in extreme precipitation or floods over a shorter time resolution was thus not made. Perhaps this may explain the small effects on water quality found in the Eastern part of the country, where short-duration extreme precipitation is more common than in the North and West. Several studies demonstrate impacts of short-term rainfall events on water quality (Roig et al. 2011; Zhu et al. 2021). A follow-up study using higher time resolutions of both water quality and meteorological and hydrological data would allow further investigations of the effects from changes in short-duration extreme rainfall and subsequent runoff events on drinking water quality.

When it comes to the future climate projections, emission scenarios, climate models and the downscaling procedures include several uncertainties (Hanssen-Bauer et al. 2015; Wong et al. 2016). Uncertainties in climate models are reduced by using several (here 10) models for the analyses. It should, however, be noted that the results get more uncertain with higher resolution in time and space. That means that seasonal estimates are more uncertain than annual estimates, and regional/local estimates are more uncertain than national estimates.

With climate change, other factors may also possibly affect the raw water quality. For instance, land-use change and longer growing seasons could increase the use of fertilisers with subsequent leaching to watercourses, rivers and lakes, causing higher levels of contaminants in the water (Arheimer et al. 2005; Whitehead et al. 2009). Small lakes with short residence times will also be more sensitive to changes in rainfall and runoff than large ones (George et al. 2007).

Increased levels of water quality indicators in raw water may lead to public health issues if the waterworks do not have treatment processes that are capable of cleaning the water properly before distributing it to the public (Khan et al. 2015). Few associations between weather/hydrology and water quality indicators were found in treated water (Guzman Herrador et al. 2021), meaning that the waterworks included were able to cope with the present levels of quality indicators. However, only large waterworks were included. Smaller waterworks may not have the resources and capacities needed for treating the raw water sufficiently. In addition, private supplies would be at risk (Hunter 2003).

The results of our study suggest that with continued climate change, the raw water quality will deteriorate by the end of the century. The degradation of water quality follows from estimated higher levels of water quality indicators in the future, especially due to increasing amounts of rainfall. Regionally, the largest effects were seen in the Western and Northern parts of Norway, with very small changes found in the East. The concentrations of bacteria, turbidity and colour predicted in raw water for the end of this century are, however, relatively small. Thus, it is likely that the large waterworks will adapt to future conditions. Any designs for new treatment systems will have to include effects from climate change, and new operational procedures may be required in the future. Furthermore, the estimated deterioration of raw water quality may cause future challenges for the treatment processes at smaller waterworks and for private supplies.

Our results may be used for assessing the need for an upgrade in treatment capacity at the waterworks in association with climate change. Combining the results with further studies of treatment effects and microbial risk assessments is needed in order to ensure sufficient treatment capacities of the raw water in the future. Whereas this study includes effects from changing climate on the raw water of the waterworks, well-functioning water distribution systems are also necessary in order to ensure the distribution of high-quality drinking water from the waterworks to the public.

We would like to thank Linda Selje Sunde and Zuzana Nordeng, Norwegian Institute of Public Health, for their highly valued coordination and support of our work. We would also like to thank two anonymous reviewers for their comments that helped to improve our manuscript.

All relevant data are included in the paper or its Supplementary Information.