Assessment of future water availability under climate change, considering scenarios for population growth and ageing infrastructure

Climate change is likely to cause higher temperatures and alterations in precipitation patterns, with potential impacts on water resources. One important issue in this respect is inflow to drinking water reservoirs. Moreover, deteriorating infrastructures cause leakage in water distribution systems and urbanization augments water demand in cities. In this paper, a framework for assessing the combined impacts of multiple trends on water availability is proposed. The approach is focused on treating uncertainty in local climate projections in order to be of practical use to water suppliers and decision makers. An index for water availability (WAI) is introduced to quantify impacts of climate change, population growth, and ageing infrastructure, as well as the effects of implementing counteractive measures, and has been applied to the city of Bergen, Norway. Results of the study emphasize the importance of considering a range of climate scenarios due to the wide spread in global projections. For the specific case of Bergen, substantial alterations in the hydrological cycle were projected, leading to stronger seasonal variations and a more unpredictable water availability. By sensitivity analysis of the WAI, it was demonstrated how two adaptive measures, increased storage capacity and leakage reduction, can help counteract the impacts of climate change. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/wcc.2018.096 s://iwaponline.com/jwcc/article-pdf/10/1/1/533154/jwc0100001.pdf Erle Kristvik (corresponding author) Tone M. Muthanna Knut Alfredsen Department of Civil and Environmental Engineering, NTNU, Trondheim 7491, Norway E-mail: erle.kristvik@ntnu.no


INTRODUCTION
A safe and steady drinking water supply is one of the most important public goods there is. As awareness of climate change increases, there is rising concern for the future reliability of drinking water supplies. Climate change is likely to cause higher temperatures and alter precipitation patterns (IPCC ) and the understanding of the local impacts of this on the hydrological cycle is highly relevant for planning a provident water supply. At the same time, more and more people live in cities, yielding more strain on existing water supply systems as the water demand increases in pace with population growth. In addition, many cities experience high levels of water losses due to ageing infrastructure and deteriorating pipes. Responsible water suppliers need to assess both the potential negative effects of climate change to supply and the trends towards increased water demand if they wish to secure reliable water supply services in the future.
There exist numerous studies of the impacts of climate change on the hydrological cycle, water resources, and availability, see for instance Barnett et al. () Benestad et al. () as 'the process of making the link between the state of some variable representing a large space and the state of some variable representing a much smaller space'. Compared with the statistical approach, a strength of dynamical downscaling is that it is based on physics and resolving of atmospheric processes at the local level (Wilby et al. ). However, the application of dynamical downscaling requires significant computing resources compared with statistical models, which are also more flexible because they can be adapted to other regions other than the ones for which they are built. Some of the statistical downscaling techniques have resulted in practical tools, which contributes to making climate scenarios more available to impact assessors. Examples of such are the statistical downscaling software SDSM (Wilby et al. ) and the R-package 'esd' by Benestad et al. ().
The availability of climate projections for impact studies are improving (CMIP ), but there are still challenges related to handling the uncertainty of the projections. Ekström et al. () categorized the uncertainty in climate projections into the uncertainty related to external forces (Type I), the uncertainty related to the climate system's response to these forces (Type II), and uncertainty due to natural variability (Type III). The paper argues that Type I uncertainty is handled using different emissions scenarios.
Furthermore, using multi-model ensembles (ensembles of different GCMs) and perturbed physics ensembles (ensemble of one GCM with differing initial conditions and parameter schemes) should account for Type II and Type III uncertainty, respectively. Giorgi & Mearns () proposed the 'Reliability Ensemble Averaging' method for assessing the reliability of simulated changes in multi-model GCM runs. The method involves quantifying the reliability of regional GCM simulations by combining two reliability criteria that accounts for: (1) the models' ability to reproduce historical and present day climate (the model performance criterion); and (2) the convergence of the models' simulated climate to the ensemble mean (the model convergence criterion). By following this framework it is possible to assess the probability of climate projections exceeding given thresholds and reduce predictive uncertainty in hydrological impacts studies (Giorgi & Mearns ).
GCM ensembles, downscaling, and reliability-weighted projections add valuable information that enables a better understanding of the future climate. However, the intrinsic uncertainty that accompanies the climate scenarios and projections makes it complicated to use them as a basis for decision making. Local water managers and stakeholders are still in need of easy-to-use tools that facilitate the assess- Recent studies of water resources availability under climate change in the city of Bergen, Norway, suggests a potential conflict between water supply and demand unless water losses in the distribution network are reduced (Kristvik & Riisnes ). Accordingly, water supply security could improve by making changes at the demand side of water management (i.e. reduce leakages). However, the study also highlights the need for practical tools that both reveal a supply system's vulnerability to external factors, such as climate change and population growth, and shows the system response and sensitivity to changes in conditions that decision makers can control, such as leakage rates (demand side) and levels of installed storage capacity (supply side). This paper suggests a framework for assessing future water availability in cities with the aim of resolving some of the issues described in this section. These are, specifically: (1) high levels of uncertainty in local climate projections; and (2) lack of easy-to-use tools to facilitate water availability assessments. To address the first issue, a large ensemble of climate data is statistically downscaled and the site-specific projections are prepared. Furthermore, an index for water availability (WAI) is introduced. This index accounts for climate change as well as other straining factors that cities may experience, such as population growth and deteriorating infrastructure for water supply. Finally, a demonstration of the WAI is presented through scenario and sensitivity analyses where the effects of counteractive measures that reduce negative impacts on water availability are investigated.

Study area
Bergen is the second largest city in Norway and located on the west coast of the country. The climate is wet and mild with an annual normal precipitation of 2250 mm and mean annual temperature of 7.6 C (monthly normal values for 50540 Florida Weather Station, http://www. eklima.no/). Bergen is a particularly rainy city due to its exposure to westerly winds and the pronounced topography characterizing the city. Statistically, the spring and summer months represent the driest period (see Figure 1 in Results and Discussion section). Usually, this does not conflict with water supply as snowmelt in this period makes up for lower precipitation amounts. However, the city has experienced substantial dry periods that have challenged water supply. The latest incident was in winter 2009-2010 when the climate was unusually dry and cold. At the turning point, water levels had dropped to half of their usual levels (Kristvik & Riisnes ).
The raw water serving the water supply system in the city is drawn from several reservoirs located close to the city centre and the water is treated at five major treatment plants: Svartediket, Jordalsvatnet, Espeland, Kismul, and Saedalen. Water from these plants is supplied to the inhabitants of Bergen through a distribution system comprising 900 km of pipe network. The network is complemented by transfer tunnels between treatment plants, securing a steady supply even if one plant is out of service (Bergen Municipality ).
Most (97%) of the total population of 278,000 inhabitants in Bergen are connected to the municipal water supply. In 2014 the estimated domestic consumption amounted to 45% of the produced drinking water, 21% was consumed by industry and 31% was ascribed to leakages in the distribution network (3% unspecified) (Bergen Municipality ; Statistics Norway a). The municipality is continuously working on reducing the high level of leakages and the objective is to achieve a leakage level that equals 20% of produced water by 2024 (Bergen Municipality ). However, regional centres in Norway, such as Bergen, are expected to experience high population growth due to urbanization (Tønnessen & Leknes ). Thus, although the municipality is working on reducing water production by rehabilitating leaking pipes, the overall consumption is expected to increase as there are strong indications of continued population growth throughout the 21st century.

Projections of future climate
Output from GCMs is available through the Coupled Model Intercomparison Project phase 5 (CMIP5). The projections of temperature and large-scale precipitation for all available emissions scenarios (RCP2.6, RCP4.5, RCP6.0, and RCP8.5) and from a selection of GCMs (Table 1) were statistically downscaled. The GCMs were selected based on a criterion that the results had to be comparable across emissions scenarios. Thus, only models that were run with all RCPs were selected. In addition, only GCM output from simulations with the same realization ID were selected. Based on this, the total number of common GCMs was 19. The downscaling was performed following the statistical approach as

Water availability index (WAI)
An index for water availability was defined to facilitate the analyses on the effects of different drivers on water availability in the future. Herein, water availability is defined as the total amount of water that is available for water supply when requirements to minimum storage reserves are accounted for. Minimum storage reserves (RR) refer to the volume of water that is always required in the reservoirs.
The municipality in Bergen has set this threshold to a volume that corresponds to 50 days of consumption. The water availability index (WAI) is defined as the ratio between the available water and the capacity of the system to store water (Equation (1)): where WAI(t) is the WAI at time t, SW(t) is the stored water at time t, RR(t) is the required storage reserves at time t, and SC is the installed storage capacity. Stored water, SW, is a reservoir balance considering all the water that enters the reservoirs and all that is withdrawn, such that: where Q in represents the inflow from surrounding catchments to the drinking water reservoirs. As there are transfer tunnels in the distribution network of Bergen that connect the treatment plants, the water balance is treated as a one-reservoir model where Q in is the sum of all inflows to the various reservoirs. Q out covers consumption, water lost to overflow when reservoirs are full, and a regulated flow of 12 m 3 s À1 that is released from the reservoir connected to the Espeland treatment plant during the period 1 April to 30 September. The consumption is defined as: where C tot (m 3 /timestep) is the total consumption, a is the leakage rate, C sp (m 3 /people/timestep) is the specific con-

RESULTS AND DISCUSSION
The climatology for the reference period 1975-2005 simulated by the downscaled GCMs is presented in Figure 1 along with observed climatology for the same period. All The results are, to some degree, in agreement with other  Furthermore, the level of leakages, population growth and storage capacity were changed one by one (while the others were kept at base level) as given in Table 2.  (5) increased storage capacity (SC H). The results are presented in Figure 5. In all emission scenarios, the WAI is most sensitive to, and negatively affected by, population growth.
There are two main reasons for this: increased population causes increased water consumption putting more strain on stored water (SW); and the WAI is constrained by required storage reserves (RR), which are directly influenced by the population as the required volume equal to 50 days of consumption will increase with population growth. Moreover, the scenario for low population growth (Pop L) has the most positive impact on the WAI for each emission scenario. Furthermore, Figure 5 shows the effect of the counteractive measures (leakage reduction and increased storage capacity), as well as the effects of letting the leakage level exacerbate to higher levels (40%

CONCLUSIONS
This paper has presented a framework for assessing future water availability in cities. The suggested framework is tailored to account for not only climatic changes at the local level, but also other factors that might put strain on future water supply, such as population growth and leakages in the distribution network. These driving forces are summarized in an index for water availability which has been demonstrated for use in scenario and sensitivity analyses. By applying the proposed framework, three main conclusions regarding future water availability in the city of Bergen, Norway, can be drawn. Firstly, the results of downscaling suggest higher seasonal variations in inflow and thus an increased potential for storage such that more water can be preserved for dryer seasons. Secondly, in a 'business-asusual' scenario-analysis of the WAI indicated a more vulnerable water supply due to decreased and more unpredictable water availability. Finally, it was shown that the city's policy of reducing leakages to a level of 20% would have approximately the same effect on water availability as a 10%