Climatic Change DOI 10.1007/s10584-013-0871-8
Quantification of uncertainty sources in a probabilistic climate change assessment of future water shortages C. N. P. Harris & A. D. Quinn & J. Bridgeman
Received: 7 December 2012 / Accepted: 10 August 2013 # Springer Science+Business Media Dordrecht 2013
Abstract As the incorporation of probabilistic climate change information into UK water resource management gathers apace, understanding the relative scales of the uncertainty sources in projections of future water shortage metrics is necessary for the resultant information to be understood and used effectively. Utilising modified UKCP09 weather generator data and a multi-model approach, this paper represents a first attempt at extending an uncertainty assessment of future stream flows under forced climates to consider metrics of water shortage based on the triggering of reservoir control curves. It is found that the perturbed physics ensemble uncertainty, which describes climate model parameter error uncertainty, is the cause of a far greater proportion of both the overall flow and water shortage per year probability uncertainty than that caused by SRES emissions scenario choice in the 2080s. The methodology for producing metrics of future water shortage risk from UKCP09 weather generator information described here acts as the basis of a robustness analysis of the North Staffordshire WRZ to climate change, which provides an alternative approach for making decisions despite large uncertainties, which will follow.
1 Introduction Maintaining sustainable water resource use throughout the 21st century in the face of climate change represents a significant challenge to the UK water industry, with legislative and ethical pressure to both maintain constant supply to customers whilst protecting the ecosystem services that freshwater provides (Arnell 2011). Although the scientific basis for the anthropogenic warming of Earth’s climate is well established, the exact extent of change to key climatic variables and their effect on water resources in the future is impossible to quantify, with considerable uncertainties in our ability to accurately model the Earth system, downscale those models to regional scales relevant to impact analyses and define the greenhouse gas emissions pathways that human behaviour will dictate. Future climate change uncertainty, then, is a crucial consideration in any assessment of future water resource supply stress (Prudhomme and Davies 2009)
C. N. P. Harris (*) : A. D. Quinn : J. Bridgeman School of Civil Engineering, University of Birmingham, Birmingham B15 2TT, UK e-mail:
[email protected]
Climatic Change
It is therefore clear that assuming stationarity is no longer valid when making decisions on future water resource planning (Milly et al. 2008), and nor is utilising precise yet potentially highly inaccurate deterministic projections of climate change that lead to overly-confident predictions of future hydrological conditions (Dessai et al. 2009; Gosling et al. 2011; Harris et al. 2012). Probabilistic projections of climate change from perturbed physics ensembles (hereafter referred to as PPEs) such as the United Kingdom Climate Projections (UKCP09) (Murphy et al. 2009) have emerged as an attractive means of projecting future conditions for use in impact assessments as a result of enabling a thorough exploration of the uncertainties involved, which is not possible with climate model ensemble approaches (Knutti et al. 2010). Many studies have been carried out to assess sources of uncertainty in climate change impacts on flows, notably Wilby and Harris (2006), Prudhomme and Davies (2009), Kay et al. (2009) and Todd et al. (2010). From this body of research, the disagreement between climate models and the methods used to downscale that climate model information are routinely found to represent the largest source of future flow uncertainty, although differences between emissions scenarios and other sources such as hydrological model error and statistical post-processing are often also found to account for significant uncertainty also and should not be ignored in practical applications (Bosshard et al. 2013). This paper extends such work to compare climate uncertainty within the UKCP09 PPE to emissions scenario selection uncertainty in terms of future water shortage probability for the first time. This is useful to water resource planners who are interested in communicating water resource vulnerability in terms of the future probability of unwanted outcomes such as water restrictions for customers, rather than more abstract terms such as Deployable Output (DO). In a robustness analysis of a water resource zone (WRZ), the probability of triggering a control line of a given severity can act as the metric against which the effectiveness of interventions to the system are judged (Groves and Lempert 2007; Hall et al. 2012), so the uncertainty in terms of future flows is less important and therefore demands less attention. Furthermore, the general need for a better understanding of PPE uncertainty comes as legislative pressure on water companies from the UK water sector regulator, Ofwat, to use probabilistic UKCP09 climate change projections in their adaptation plans comes into force. The production of practical and replicable frameworks for the effective use of UKCP09 in the water industry, which explicitly requires a better understanding of uncertainty, is now crucial to the long-term sustainability of water resource supply (Arnell 2011). Conveying the range of uncertainty involved with a climate change assessment of future water shortage can facilitate the establishment of policies and strategies that are statistically robust to the range of plausible futures, therefore increasing resilience and reducing the possibility of maladaptation (Groves et al. 2008; Hall et al. 2012).
2 Data and methods 2.1 Research area: the North Staffordshire water resource zone This research is carried out at the North Staffordshire WRZ in central England, within which Tittesworth Reservoir serves as the main surface water resource (Fig. 1). The region is managed by Severn Trent Water (STW) and includes the urban centre Stoke-on-Trent, multiple groundwater resources and river abstractions sites. The key drought management tools in the WRZ are based on the crossing of various control lines at Tittesworth Reservoir.
Climatic Change
Fig. 1 Map of the North Staffordshire WRZ in the context of the UK (Harris 2013). The overall upper River Churnet catchment is shown in pink, whilst the UC, SOL and DHY sub-catchments can be approximated from the tributaries. As the River Churnet flows away to the SSE, water is piped from Tittesworth Reservoir to the major demand centre of Stoke-on-Trent
These include the Storage Alert Line (SAL), falling below which represents the first indication of dry conditions, the Drought Warning Trigger (DWT) which catalyses a variety of potential responses to the threat of water shortage but rarely creates disruption to supply, and the more severe Temporary Use Ban (TUB) which imposes restrictions on water use by customers. Output from the Tittesworth Water Treatment Works (WTW) to the surrounding demand centres can be shut off completely provided sufficient groundwater is available, significantly reducing water stress at the reservoir. Three small sub-catchments influence the reservoir; Upper Churnet (30 km2) provides all of the inflow whilst Deep Hayes (10 km2) and Solomon’s Hollow (6 km2) flow into the River Churnet downstream (at the area marked * on Fig. 1), reducing the compensation flow needed from the reservoir. Upper Churnet produces the greatest flow of the sub-catchments (56.5 Ml/d), and is an upland area with greater average precipitation than elsewhere in the region. Groundwater resources are considered stable and largely robust to drought events by STW, although more rigorous assessment of the climate change impact on groundwater models would be useful for further studies in the area. 2.2 Instrumental and climate data Flows sequences for the sub-catchments are simulated in Hysim, a physically-based lumped conceptual rainfall runoff model (Manley 1978) which has been used extensively in climate change impact assessments in the UK (Murphy et al. 2004; Severn Trent Water 2010; 2011; Hall and Murphy 2010). The hydrological parameters used in this project are based on an extensive survey carried out by Severn Trent Water (2011), in which a full validation and calibration of the hydrological models used in this project can be found. It is assumed that relationships between variables in the catchment will remain constant in the future as the climate changes. A pre-constructed Aquator water resource model of the North Staffordshire WRZ is used to simulate the water infrastructure systems used by STW in the area. Hysim
Climatic Change
flows, direct reservoir daily rainfall and open water evaporation rates drive the Aquator model (Oxford Scientific Software 2008), which simulates surface water resources at Tittesworth Reservoir and the ability to supply water to the demand centres given the operational procedures in place. The UKCP09 weather generator (UKCP09WG) (Jones et al. 2009) is used as the core climate change data source. Weather generator approaches have the advantage of allowing for changes to the sequencing and timing of rainfall events in the future by creating entirely synthetic weather sequences. However, considerable limitations still exist, particularly with regards to the production of extreme climatic events (including multi-seasonal droughts). This limits the extent to which the approach outlined here can be used as a tool to adapt to changes in the most extreme conditions and thus some of the outcomes are ‘semi-quantitative’ (Harris et al. 2012). 1000 simulations, each of 98 years at a daily time-step are created for the low (B2), medium (A1B) and high (A1FI) IPCC SRES emissions scenarios in the 2071–2100 timeslice, as well as a set of 100 control simulations for the 1961–1990 period for validation purposes. Amongst the range of daily weather variables produced by the weather generator, precipitation and potential evapotranspiration (PET) simulations are isolated and used as the inputs for the Hysim model. PET is calculated in the UKCP09WG using the FAO-modified version of the Penman equation (Jones et al. 2009) Differences between methodologies for producing PET have been shown to increase uncertainty in simulated future flows, although the effect is less than that due to differences between climate models (Kay and Davies 2008). A sensitivity analysis of the effect of using different approaches for estimating PET can be found in Bormann (2011). The A1B emissions scenario dataset is sub-sampled to produce 20 simulations using the UNEP aridity index (annual precipitation/annual sum PET (UNEP 1992)), the spread of which represents the PPE uncertainty. This relatively simple and fast technique for subsampling UKCP09 data is found to adequately describe the range of water shortage probabilities in the North Staffordshire WRZ (not shown). The median simulations in terms of UNEP aridity index are selected for each of the B2, A1B and A1FI datasets, the spread of which represents the emissions scenario uncertainty. By assessing the extent of these ranges for a given variable (such as water shortage probability), the scale of uncertainty created by each can be estimated. As the aim here is to gain an understanding of climate-related uncertainties, all other variables that would affect water shortage risk in the future are assumed to remain unchanged, allowing for the explicit investigation of climate risks (following Donaldson et al. 2001; Gosling et al. 2012). It should be noted that in many cases changes to other factors such as irrigation demands, groundwater infiltration and demand changes would affect a WRZ significantly on top of raw surface water availability reductions (Groves et al. 2008). 2.3 Scaling approach for producing pseudo-spatial weather generator information and validation UKCP09WG works on a single site basis, meaning a computationally-inexpensive technique that creates artificial rainfall sequences for Deep Hayes and Solomon’s Hollow based on the weather generator output at Upper Churnet is required. To produce daily time-step simulations at Solomon’s Hollow and Deep Hayes, the information from Upper Churnet is scaled using a z-transform, a quick and replicable method for using UKCP09WG information in areas where its conventional use would result in unmanageable errors, and full-scale spatial
Climatic Change
140 120
Flow (Ml/d)
100 80 60 40 20 0
1
2
3
4
5
6
7
8
9
10
11
12
Month (calendar year) Instrumental Baseline Lower 2080 Median 2080 Lower 2080 Low Scenario 2080 Medium Scenario
Baseline Median Baseline Upper 2080 Upper 2080 High Scenario 25
Flow (Ml/d)
20
15
10
5
0
1
2
3
Baseline Median Baseline Upper 2080 Upper 2080 High Scenario
4
5
6
7
8
Month (calendar year) Instrumental 2080 Median 2080 Low Scenario
9
10
11
12
Baseline Lower 2080 Lower 2080 Medium Scenario
Fig. 2 UKCP09WG validation statistics and 2080s simulations for Upper Churnet (left) and Solomon’s Hollow (right) for average daily flow. For each of the 1961–1990 baseline and 2080s data, the outer dashed lines represent the lower and upper boundaries of the sub-sampled dataset. The median 2080s B2, A1B and A1FI simulations, used to describe emissions scenario uncertainty, are also shown. Instrumental data is modelled flows based on the gridded UK Met Office precipitation dataset and MORECS PET data
Climatic Change
weather generation would mean formidable computational expense. The process for carrying this procedure out is as such: PUC − μUC δSOL þ μSOL δUC Where, Precipitation at Upper Churnet on a given day Monthly mean of Upper Churnet rainfall in the simulation Monthly standard deviation of Upper Churnet rainfall in the simulation Monthly standard deviation of Solomon’s Hollow, calculated by change factor from Upper Churnet Monthly mean of Solomon’s Hollow rainfall, calculated by change factor from Upper Churnet
PUC μUC δUC δSOL μSOL
Figure 2 shows that the resultant average monthly mean flow statistics for the derived sequences at Solomon’s Hollow are in line with the original Upper Churnet sequence in terms of reproducing the observed record, which can be deemed adequate. The slight over/underestimation of summer/winter flow is carried over from Upper Churnet to the derived catchments. Precipitation monthly average, variance and dry days are reproduced adequately by UKCP09WG and the scaled data when validated against the gridded UK Met Office dataset (not shown). Flow duration curves (FDCs) are used to show the extent of time a certain flow is equalled or exceeded within a dataset. Figure 3 is an FDC that describes the extent of the 1000.0
Flow (Ml/D)
100.0
10.0
1.0 0
10
20
30
40
50
60
70
80
90
100
Percentage of time flow equalled or exceeded Instrumental
Baseline upper
Baseline lower
Baseline Median
Fig. 3 Baseline (1961–1990) flow duration curve for the Upper Churnet catchment compared to the 1961–1990 instrumental record (simulated as in Fig. 1). The vertical bars split the plot into high flows (exceeded 0–20 % of the time), average flows (20–70 %), and low flows (70–100 %)
Climatic Change
Probability of DWT trigger in a calendar year
0.3
Rex= 50% 0.25
0.2
0.15
0.1
0.05
0
1 0.9
Rex= 60%
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
Probability of SAL trigger in a calendar year
Fig. 4 Rex plots showing water shortage trigger probabilities at three different severity levels; TUB (left), DWT (right) and SAL (bottom). The Rex value quantifies the amount of the simulated A1B range (blue rhombi) that lay outside the range of the three emissions scenario median simulations (red squares)
Probability of TUB trigger in a calendar year
underestimation of low flow events at Upper Churnet in the simulated dataset compared to the combined observed record. Whilst high and medium flows are reproduced well, the observed record deviates from the simulated range at around the 65 % point. This inaccuracy stems from the inability of the UKCP09WG to produce the most extreme dry events (Jones et al. 2009).
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
Rex= 45%
Climatic Change
Extent of Simulation Range (%)
100
y = TUB
90 80
t = 2080s
70 60 50 40 30
(i)
20 10 0
0
(x)
0.05
0.1
0.15
0.2
0.25
0.3
Water Shortage Probability
2080m range High Scenario Median
Low Scenario Median Level of Service
Med Scenario Median
Extent of Simulation Range (%)
100 90
y = DWT
80 70
t = 2080s
60 50 40 30
(i)
20 10 0
0
0.1
0.2
0.3
(x)
0.4
0.5
0.6
0.7
0.8
Water Shortage Probability
2080m range High Scenario Median
Low Scenario Median Level of Service (Inferred)
Med Scenario Median
Extent of Simulation Range (%)
100 90
y = SAL
80
t = 2080s
70 60 50 40 30
(i)
20 10
(x)
0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Water Shortage Probability 2080m range High Scenario Median
Low Scenario Median Level of Service (Inferred)
Med Scenario Median
Fig. 5 CDFs of water shortage probability for three severities within the A1B (medium) emissions scenario simulation range in the 2080s (purple lines). The water shortage probabilities of the B1 (low), A1B and A1FI (high) scenario median simulations are also shown (green lines). The LoS for each water shortage severity and the arbitrary level of acceptable risk are shown as in Fig. 3
Climatic Change
2.4 Water shortage risk Increasing the resilience of UK water supply management into the future can be achieved by moving towards a risk-based approach when making decisions on resources. Following Hall et al. (2012), passing a trigger condition at a key water resource component at more than a pre-determined frequency represents a failure to meet a particular Level of Service (LoS) is deemed a suitable metric of risk and is used in this study. Stressed water supply conditions in the North Staffs WRZ are indicated by falling below water levels of varying severity at Tittesworth Reservoir within a calendar year. The SAL, DWT and TUB curves quantify the breaches of key drought action triggers in the region, and water levels below those lines are taken as a signal that the WRZ is under stress to various extents. The process for incorporating uncertainty into water shortage risk assessment and the ensuing decision-making on adapting to that changing envelope of risk is based on a modified version of that detailed in Hall et al. (2012). Assuming a target frequency x of water shortage severity event y…n occurring in a time horizon t…n, an ‘acceptable risk’ i of z% of the modelled uncertainty range can be used to assess the robustness of the water supply system. By organizing the water stress dataset into a cumulative distribution function (CDF) with x and i set, the extent to which i is satisfied can be seen by comparing it to the actual percentage of the model range that lies beyond x (these denominations are labelled in Fig. 5). This approach lends itself to a robustness analysis of the WRZ, where various interventions to the system are applied to a range of potential future simulations in order to assess how successful they are at satisfying an acceptable risk of an unwanted outcome, such as breaching a LoS (Groves and Lempert 2007).
3 Results 3.1 Climate model uncertainty and emissions scenario uncertainty Rex plots (Fig. 4) describe the extent of water shortage uncertainty within the PPE range and the emissions scenario selection. The Rex value in the corner of each plot relates to the amount of the PPE range that lies beyond the emissions scenario range. The Rex values at three different levels of severity are 50 %, 60 % and 45 % for TUB, DWT and SAL, respectively, showing that around half of the PPE range from the A1B scenario is beyond the boundaries of the emissions scenario range. This indicates that a large proportion of the feasible future water shortage range in the North Staffs WRZ is as a result of climate model uncertainty rather than the emissions scenario that is chosen. This is in-line with the results for flows in the catchments shown in Fig. 1, where the emissions scenario medians are clustered in the centre of the uncertainty range in the low-flow section of the FDC. 3.2 The usefulness of the central estimate and comparing uncertainty in terms of robustness Figure 5 shows how the PPE uncertainty relates to the emissions scenario uncertainty in terms of water shortage frequency across different severities. The simulation ranges are organised into cumulative distribution functions (CDFs) that communicate the probabilistic range of water shortage probability at the three severity levels, as described in Section 2.4. The probability of water shortage per year in the median simulations is also shown for comparison. LoS for TUBs is taken from STW policy, whilst LoS for DWTs and SALs are estimated based on a maximum DO simulation for the instrumental record.
Climatic Change
It can be seen that the median simulation for the A1B scenario projects that the probability of water shortage satisfies current LoS for each severity level in the 2080s. However, much of the wider A1B scenario range does not conform to LoS in each case (38 %, 52 % and 40 % for TUB, DWT and SAL, respectively), all of which would constitute being outside of acceptable risk if the robustness criteria is for the system to remain within current LoS across 80 % of the range of feasible futures. All of the emissions scenario median simulations sit within the A1B range for each severity level, again showing that PPE uncertainty substantially outweighs emissions scenario selection uncertainty in all cases at the North Staffordshire WRZ. Current UK water industry practice is to use a sub-sampled range from the A1B scenario only, and this research shows that doing so gives a reasonably large uncertainty range and would overlap the median low and high emissions scenario projections. However, the process for selecting emissions scenarios for climate change assessments in the water industry is rather ad-hoc and requires further justification. Figure 5 makes it clear that assuming the mean water shortage frequency value to be the ‘most likely’ realisation of the future (i.e. at 50 % of the simulation range) is not good practice in water shortage assessment, and would lead to unjustified assumptions regarding the impact of climate change on the frequency with which control curves denoting water scarcity conditions are crossed. In this case, assuming a central estimate of the A1B scenario would lead to over-confidence that the system would remain within LoS in the 2080s and would provide only an extremely limited assessment of future conditions to a decision-maker. As Figs. 4 and 5 have indicated, the PPE uncertainty creates a wider range of water shortage probability for each level of severity than the emissions scenario uncertainty source. The probability ranges for PPE uncertainty and emissions scenario uncertainty are:
& & &
Severity 1 (TUB): 0 to 0.24 (PPE uncertainty) and 0.01 to 0.06 (emissions scenario) Severity 2 (DWT): 0.01 to 0.72 (PPE uncertainty) and 0.09 to 0.27 (emissions scenario) Severity 3 (SAL): 0.11 to 0.95 (PPE uncertainty) and 0.24 to 0.56 (emissions scenario)
It is therefore clearly shown that climate model physics uncertainty accounts for a vast majority of future projections of water shortage uncertainty in UKCP09, regardless of which event severity is scrutinised. This understanding of the scales of the two key uncertainty sources is important as the movement towards water shortage risk approaches to climate change assessment continues to increase in prominence in the UK water sector. Although this information does not explicitly answer the question of how many and which emissions scenarios should be used by water resource decision-makers, it does highlight the greater importance of including the range of PPE uncertainty within the UKCP09 projections when assessing future water shortage.
4 Discussion A number of outcomes relevant to increasing the resilience of water resources to climate change are identified from this research, although it should be noted that the work presented here is catchment-specific and would not necessarily be representative of conditions elsewhere:
Climatic Change
1) Although only one of a suite of factors that affect water supply in the future, climate change alone is likely to significantly increase the threat of water scarcity at the North Staffordshire WRZ. 2) Assuming the mean future water shortage probability represents the most likely outcome is poor practice and can lead to maladaptive reactions to the threat of climate change. 3) Disagreement between climate models is a larger source of uncertainty in an assessment of climate change impacts on water shortage than emissions scenario uncertainty. The same is also true for flows, in-line with previous research. 4) The organisation of uncertain climate change information into probabilities of avoiding unwanted outcomes enables robust decision-making approaches to be employed. In such an approach, the performance of each option or strategy is measured against the range of potential climate futures offered by sources such as UKCP09 and strategies can be selected based on the amount of a simulated range that lies within a pre-determined ‘acceptable risk’ (Hall et al. 2012). This allows those strategies that are effective in terms of water supply, environmental and/or financial cost to be identified and implemented. In a practical application of such a study, estimations of LoS that are scaled based on the under-representation of drought events by the UKCP09WG can be used and/or a change factor method (CFM) can be employed, where the statistics from the weather generator are used to perturb the instrumental record. Either of these approaches produces usable probabilistic water shortage risk information despite the limitations of the weather generator that could be used to drive investment in the water industry. The ‘unwanted outcome’ that forms the metric against which the robustness of a system is analysed can be changed from the control curves used here to others such as cost, probability of switching operational processes by a certain time in the year or percentage of a license used in a year. This means a water company can assess a potential strategy for a WRZ against whatever metric they deem important from a single set of future simulations. Combinations of such metrics can be put together to form an optimised system in a way proposed by Lempert and Groves (2010). This approach to using UKCP09 data to drive decision-making is not limited to the water sector, and is of relevance to any area with multiple criteria against which to judge the success, or otherwise, of an adaptation scheme. 5) Further uncertainties, such as hydrological modelling, PET calculation and statistical post-processing are not taken into account here, but are assumed to no larger than the PPE uncertainty (Kay and Davies 2008; Kay et al. 2009; Prudhomme and Davies 2009; Bosshard et al. 2013). This is justified by the research being based on gaining a better understanding of the bespoke UKCP09 tools available to water resource managers in the UK.
5 Conclusions This paper quantifies the relative uncertainties from the PPE used to construct the probabilistic range of UKCP09 climate projections and emissions scenario selection on water shortage probability in the future (2080s) using a multi-model approach. PPE uncertainty from the UKCP09WG is found to be much larger than uncertainty sourced from emissions scenario selection, with 45–60 % of the water shortage described
Climatic Change
by the climate model range outside the water shortage range described by the median projections for each emissions scenario, dependent on which water severity metric is considered. This confirms that climate model uncertainty significantly outweighs emissions scenario uncertainty in terms of water shortage. Uncertainties sourced from hydrological models are not searched for, but are expected to not outweigh those for climate model selections and emissions scenario selection based on previous research. A simple technique to sub-sample the UKCP09 range using the UNEP aridity index is deemed adequate for climate change-based water shortage assessments. The spatial limitations of the UKCP09WG are reduced by the introduction of a z-transform scaling approach, although this technique is not universally viable and depends on the presence of the same large-scale weather systems affecting all of the sub-catchments under consideration. It is found that using median projections from the range of water shortage probabilities cannot be assumed to relate to the most ‘likely’ outcome in terms of water shortage. The importance of introducing probabilistic climate change projections into assessments of water security in the water industry in order to avoid costly maladaptation is highlighted by the substantial range of future water scarcity projections. Significant challenges remain in using UKCP09 effectively in water resource decision-making, but organising data in the way described here lends itself to tackling climate change threats to water shortage through robust decision-making approaches. A further paper will continue this work to describe a robustness assessment for the same WRZ that both assesses the extent to which the system avoids unacceptable risks in the future as a result of climate change and provides a platform for selecting adaptation options based on probabilities of avoiding unwanted outcomes such as crossing control curves. Finally, it is found that whilst the weather generator approach is a useful tool for describing ‘normal’ operating conditions in the future, the incapability to describe the most extreme events lessens the extent to which it can be used for providing fully quantitative information on extreme future drought events. Methods for coping with this limitation are suggested here and put into practice in the follow-up paper. Acknowledgments The authors would like to thank Oxford Scientific Software for access to the Aquator Water Resource Model, Ron Manley for the use of Hysim, and Mott MacDonald Ltd and Severn Trent Water Ltd for releasing instrumental data and supplying modelling parameters for the North Staffordshire subcatchments. The authors are also grateful for the helpful insights of two anonymous reviewers.
References Arnell NW (2011) Incorporating climate change into water resources planning in England and Wales1. JAWRA J Am Water Res Assoc 47:541–549 Bormann H (2011) Sensitivity analysis of 18 different potential evapotranspiration models to observed climatic change at German climate stations. Clim Chang 104:729–753 Bosshard T, Carambia M, Goergen K, Kotlarski S, Krahe P, Zappa M, Schär C (2013) Quantifying uncertainty sources in an ensemble of hydrological climate-impact projections. Water Resour Res 49:1523–1536 Dessai S, Hulme M, Lempert R, Pielke R Jr (2009) Climate prediction: A limit to adaptation. In: Adger N, Lorenzoni I, O’Brien K (eds) Adapting to climate change: Thresholds, values, governance. Cambridge University Press, Cambridge Donaldson GC, Kovats RS, Keatinge WR, McMicheal AJ (2001) Heat- and cold related mortality and morbidity and climate change. In: Maynard RL (ed) Health effects of climate change in the UK. Department of Health, London, pp 70–80 Gosling SN, Warren R, Arnell NW, Good P, Caesar J, Bernie D, Lowe JA, Linden P, van der O’Hanley JR, Smith SM (2011) A review of recent developments in climate change science. Part II: the global-scale impacts of climate change. Prog Phys Geogr 35:443–464
Climatic Change Gosling S, McGregor G, Lowe J (2012) The benefits of quantifying climate model uncertainty in climate change impacts assessment: an example with heat-related mortality change estimates. Clim Chang 112:217–231 Groves DG, Lempert RJ (2007) A new analytic method for finding policy-relevant scenarios. Glob Environ Chang 17:73–85 Groves DG, Yates D, Tebaldi C (2008) Developing and applying uncertain global climate change projections for regional water management planning. Water Resour Res 44, 16 PP Hall J, Murphy C (2010) Vulnerability analysis of future public water supply under changing climate conditions: a study of the Moy catchment, western Ireland. Water Resour Manag 24:3527–3545 Hall JW, Watts G, Keil M, de Vial L, Street R, Conlan K, O’Connell PE, Beven KJ, Kilsby CG (2012) Towards risk-based water resources planning in England and Wales under a changing climate. Water Environ J 26:118–129 Harris CNP, Quinn AD, Bridgeman J (2012) The use of probabilistic weather generator information for climate change adaptation in the UK water sector. Meteorol Appl. doi:10.1002/met.1335 Harris CNP (2013) “North Staffordshire WRZ” [PDF map], Strategi [SHAPE geospatial data], Scale 1:250000, Tiles: GB, Updated: January 2013, Ordnance Survey (GB), Using: EDINA Digimap Ordnance Survey Service, , Downloaded: Tue Jul 09 10:30:41 GMT 2013 Jones PD, Kilsby CG, Harpham C, Glenis V, Burton A (2009) UK climate projections science report: Projections of future daily climate for the UK from the weather generator. University of Newcastle, UK Kay AL, Davies HN (2008) Calculating potential evaporation from climate model data: a source of uncertainty for hydrological climate change impacts. J Hydrol 358:221–239 Kay A, Davies H, Bell V, Jones R (2009) Comparison of uncertainty sources for climate change impacts: flood frequency in England. Clim Chang 92:41–63 Knutti R, Furrer R, Tebaldi C, Cermak J, Meehl GA (2010) Challenges in combining projections from multiple climate models. J Clim 23:2739–2758 Lempert RJ, Groves DG (2010) Identifying and evaluating robust adaptive policy responses to climate change for water management agencies in the American west. Technol Forecast Soc Chang 77:960–974 Manley RE (1978) Simulation of flows in ungauged basins. Hydrol Sci J 23:85–101 Milly PCD, Betancourt J, Falkenmark M, Hirsch RM, Kundzewicz ZW, Lettenmaier DP, Stouffer RJ (2008) Stationarity is dead: whither water management? Science 319:573–574 Murphy C, Fealy R, Charlton R, Sweeney J (2004) Changing precipitation scenarios: preliminary implications for groundwater flow systems and planning. Presented at the 25th Anniversary Conference on Groundwater in Ireland, International Association of Hydrogeologists (Irish Group), Tullamore, pp. 49–56 Murphy JM, Sexton DMH, Jenkins GJ, Boorman PM, Booth BBB, Brown CC, Clark RT, Collins M, Harris GR, Kendon EJ, Betts RA, Brown SJ, Howard TP, Humphrey KA, McCarthy MP, McDonald RE, Stephens A, Wallace C, Warren R, Wilby R, Wood RA (2009) UK climate projections science report: Climate change projections. Met Office Hadley Centre, Exeter Oxford Scientific Software (2008) A guide to aquator, 1. Application, version 3.0. Oxford Scientific Software, Oxford Prudhomme C, Davies H (2009) Assessing uncertainties in climate change impact analyses on the river flow regimes in the UK. Part 2: future climate. Clim Chang 93:197–222 Severn Trent Water (2010) Water resource management plan. Final Version. June 2010. Severn Trent Water, Coventry Severn Trent Water (2011) Aquator Flow Database Extension. December 2011. Severn Trent Water, Coventry Todd MC, Taylor RG, Osborne T, Kingston D, Arnell NW, Gosling SN (2010) Quantifying the impact of climate change on water resources at the basin scale on five continents—a unified approach. Hydrol Earth Syst Sci Discuss 7:7485–7519 UNEP (1992) World Atlas of Desertification. Middleton, N., Thomas, D. (eds) Edward Arnold, London Wilby RL, Harris I (2006) A framework for assessing uncertainties in climate change impacts: low-flow scenarios for the river Thames, UK. Water Resour Res 42, W02419
