Prioritising climate adaptation options to minimise financial and distributional impacts of water supply disruptions

Climate-related disruptions to water supply infrastructure services incur direct financial losses to utilities (e.g. to repair damaged assets) and externalise a societal cost to domestic customers due to additional costs that they may incur (e.g. to acquire water from alternative sources). The latter often represents an uncompensated social burden, which should be properly accounted for in investment planning. Here we present a new framework for quantifying direct financial risks burdened by utilities and alternative water purchase losses incurred by domestic customers, including those in low-income groups, during flood- and drought-induced utility water supply disruptions. This framework enables the comparison of benefit-cost ratios of a portfolio of flood protection and leakage reduction for water supply systems across the island of Jamaica. A system-level optioneering analysis allows the identification of the optimal adaptation option per system. We estimate that 34% of systems would benefit from flood defences and 53% would benefit from leakage reduction to adaptation to droughts. The benefit that could be achieved by implementing all system optimised adaptation options is estimated to be 720 million Jamaican dollars per year on average, representing a substantial saving for the utility and its customers, including low-income customers. We identify options that offer strong synergies between economic and equity objectives for both types of adaptation option. The proposed framework is established to support the business case for climate adaptation in the water supply sector and to prioritise across flood and drought mitigation options. We take a first step towards mainstreaming equity considerations in water supply sector optioneering frameworks by estimating the contribution of adaptation options towards reducing household costs for low-income customers.


Introduction
Formally operated and regulated water supply utility services offer a level of quality and affordability assurance that is not necessarily guaranteed by informal markets or other unregulated supplies (WHO 2014).The safety and affordability of water supplies are key components of Sustainable Development Goal (SDG) for clean water and sanitation (SDG6), highlighting the importance of utility operated water infrastructure for overall sustainable development (UN 2021).However, many utilities in low-and middle-income countries are caught in a vicious cycle of disrupted services, non-revenue water and a low rate of reinvestment in infrastructure resilience (Rouse 2013, Libey et al 2020).This has contributed towards a large, and widening, service gap in many countries (Hutton and Varughese 2016, Rozenberg and Fay 2019, OECD 2022).Water supply systems in several low-and middle-income countries are also exposed and vulnerable to climate-related disruptions and climate change (Mirza 2003).Utilities in these countries have smaller budgets for emergency water supply deployments to affected communities during disruptions and lower creditworthiness, hindering their ability to attract loans for long-term infrastructure investments (Winpenny 2015).Further, low income water utility customers suffer the most from disrupted services (Robak and previously integrated with the optioneering literature.(1) Literature that explores the economic, financial and welfare impacts of household water supply disruptions (Altaf 1994, Baisa et al 2010, Freire-González et al 2017, Roibas et al 2019, Jenkins et al 2021), including through the costly replacement from private tanker trucks (Bharti et al 2020, Yoon et al 2021).Economic and welfare impacts to customers can be captured through interviews and relatively small-scale surveys (Heflin et al 2014, Selelo et al 2017, Sjöstrand et al 2021).Such research is typically removed from the literature on climate risk to water supply systems, but rather focusses on non-climate related service intermittency (Grasham et al 2022a).National scale studies have focussed on GDP losses across a range of economic sectors caused by water supply disruptions (Jenkins et al 2021); there is limited research that estimates the costs of climate-related water supply disruptions to customers at the national scale.(2) Literature dedicated to the distributional impacts of climate extremes (Hallegatte et al 2016, Hallegatte andRozenberg 2017).This literature places focus on the disproportionate welfare impacts burdened by low income households, as they have a relatively low financial capacity to cope and recover from shocks, resulting in a so-called poverty trap (Borgomeo et al 2018a).This can be obscured by research that focusses only on the economic costs of extremes, as higher income households typically have higher asset values at risk to shocks (Hallegatte et al 2017).Although a major channel through which climate extremes impact households, including in vulnerable communities, is via disruptions to critical infrastructure services (e.g.transport, electricity and water supply), this has received far less attention (Verschuur et al 2023a).There is scope to approach the traditional adaptation infrastructure optimisation problem from the perspective of the distributional costs of water supply disruptions to households.
Here, we present a framework that combines the multiple channels through which climate-induced water supply disruptions incur direct financial losses to utilities and to their customers.By stress-testing a model representation of Jamaica's potable water supply network to a large synthetic event set, comprising of both flood and drought events, the annual expected risk in terms of direct losses per water supply system is calculated.Direct losses considered include the losses incurred by utilities and the alternative water purchase losses incurred by domestic customers.Here, we define utility losses as flood-induced damages to the utility's asset-base, and alternative water purchase losses as the costs of replacing disrupted supply from tanker trucks.By repeating this process with and without the integration of a range of infrastructure interventions, comprising flood defence walls and leakage reduction strategies, we quantify the effect that each intervention has on the annual expected losses.The benefit of a range of adaptation options is, thus, quantified in terms of the reduced direct cost of disruptions.This is compared against each option's implementation cost to compare options based on their benefit-cost ratios (BCRs).By estimating the reduction in alternative water purchase losses incurred by customers in poverty provided by each intervention, we identify 'pro-poor' adaptation strategies.The proposed framework quantifies the climate-related risk mitigated by adaptation projects to promote the business case for climate adaptation and enables optioneering analyses with respect to both economic objectives and 'pro-poor' agendas.
The proposed optioneering framework is applied to the national water utility in Jamaica, although it is applicable to any national infrastructure setting with adequate data availability.The National Water Commission (NWC) of Jamaica is a state-owned utility that supplies drinking water to 90% of Jamaica's total potable water supply to residential and non-residential customers (National Water Commission 2020).Jamaica's water infrastructure is intensely vulnerable to damage from flooding (Mandal and Maharaj 2013, Burgess et al 2015, Glas et al 2015, Mandal et al 2016, Nandi et al 2016) and droughts (Setegn et al 2014, Goyal et al 2015).During climate-induced disruptions, customers typically turn to other sources, such as tanker trucks or untreated well or spring supply (Lester 2015).There are several health risks associated with water stored in household tanks (ibid) and unregulated sources, such as wells, which are becoming increasing contaminated by untreated sewage (Mandal et al 2011, Jamaica Water Resources Authority 2020).Therefore, tanker trucked water offers an essential alternative replacement service during utility water supply disruptions, to protect the health of disrupted communities.
The NWC aims to significantly improve system reliability over the coming decade (for example, by reducing intermittent supply) (NWC 2015) and is an essential pillar of national development and poverty alleviation by providing high quality, affordable potable water services to customers (JIS 2008, MEGJC 2019).However, the capacity for the utility to invest in better services is hindered by a high level of debt, with the highest debt to equity (leverage) ratio in the Caribbean (Burdescu et al 2020).In this context, it is essential that the adaptation strategy adopted is as cost-efficient as possible.This situation is comparable to utilities in many other low and middle-income countries that are trapped in a cycle of disrupted services, low-cost recovery and low adaptive capacity.An important but often unrecognised driver of this downward spiral is utility's vulnerability to various climatic hazards, that will likely be further intensified under future climate change conditions (Mirza 2003, Srinivasan et al 2013, Becher et al 2023).

Overview
In a previous study, a spatial model of the national water supply infrastructure network in Jamaica was established and stress-tested with a large set of (synthetic) spatially coherent, climate conditioned drought and pluvial and fluvial flood events to conduct a climate risk analysis (Becher et al 2023).The flood events were derived from synthetic daily rainfall-runoff simulations which generated a set of 14 000 events.Drought events were extracted from a synthetic monthly rainfall-runoff simulation.Both event sets were used to provide risk outputs up to a maximum return period of 1000 years.Assets are grouped into their respective systems, each serving a particular population or demand zone across the island (National Spatial Management Branch 2010).In total ∼240 systems in the potable water systems representation of Jamaica, with relatively weak linkages and redundancies between each other, were identified.The water systems model represents both asset-level physical vulnerability as well as domestic water usage, leakage losses and major storage dynamics in the supply network to compute the number of domestic customers impacted (hereafter CDD) per hazard event.Note that only surface water-fed systems were included in the drought risk analysis.The steps undertaken in the previous study (Becher et al 2023) in calculating the physical vulnerability and CDD are summarised in section 2.2 and relevant datasets are described in supplementary table 1.
Here, we employ the outputs from this probabilistic multi-hazard risk framework (Becher et al 2023) to quantify the effectiveness of a set of multi-hazard adaptation options for mitigating drought-induced water shortages and flood-induced asset damages and disruptions.Figure 1 provides a conceptual overview of the proposed framework.First, we identify the ways in which climate-induced disruptions impose financial losses (1) to the utility through asset damages and (2) to the utility's water customers in Jamaica.Our previous analysis demonstrated that a significant threat of damage to the water utility's assets comes from flooding, so we focus upon the financial consequences of these losses for the utility.Domestic water users incur financial losses through the purchase of water from tanker truck suppliers when their potable water supplies fail.The value of these financial losses for people is estimated by applying the unit cost of acquiring water from utility vendors per person per day to the CDD risk output provided by Becher et al (2023).We collect a set of adaptation options comprising flood defence walls, proposed to protect all exposed assets against flood-induced damages and disruptions, and leakage reduction strategies, proposed to protect systems against drought-induced water shortage disruptions.The costs and benefits of each flood protection and drought mitigation option are quantified on a system scale across range of flood and drought hazard intensity, respectively.Thus 10 options for 240 systems, provide a longlist of 2400 options under consideration under this framework.By testing the system performance with and without each adaptation option, we investigate the benefit of each option, in terms of avoided utility damages and alternative water purchase losses.These benefits are divided by each option's cost to compare options based on their BCRs.We go on to estimate the reduction in alternative water purchase losses burdened by customers in poverty provided by each intervention.This allows us to assess each climate adaptation option in terms its business case and its contribution towards 'pro-poor' , or equity-oriented development agendas.
Our methodological approach draws from two studies undertaken in Bangladesh that approached the problem of planning critical infrastructure networks through the lens of the distributional benefits of service provision among diverse populations.Roman et al (2021) explore the trade-offs between purely utilitarian (maximising the number of people with access) and poverty-orientated (maximising the number of people in poverty with access) objectives when planning the expansion of water supply infrastructure.Verschuur et al (2023a) quantify the impacts of cyclonic flooding on a range of critical infrastructure networks, including water supply, to understand distributional impacts on multi-dimensional poverty (MDB), a description of poverty defined partially in relation to access to critical infrastructure services, such as water, sanitation and electricity.Thus, the permanent spatial distribution of MDB, extracted from census data (Bangladesh Bureau of Statistics 2020), is superimposed with temporary climate-induced critical service disruption to evaluate the benefit of coastal flood protection measures in vulnerable communities.

Application of multi-hazard risk approach for optioneering
The framework outlined in Becher et al (2023) involves quantifying the impact of present-day and future climate extremes on the national water supply utility in Jamaica, in terms of CDD.Below, we summarise the overarching steps of this framework: 1. Generate a large set of spatially coherent drought and pluvial and fluvial flood events, representing both historic climate conditions, as well as future climate change conditions.We consider climate change projections up to the year 2080 under RCP2.6,RCP4.5 and RCP8.5 emissions scenarios.2. Combine the water utility's assets (1208 pumping stations, treatment works, etc) in a system model, which also represents water supply to municipal water users, leakage and major storage capacities in the supply network.Assets are grouped into their respective systems, each serving a particular population or demand zone across the island (National Spatial Management Branch 2010).3.For flood events, intersect each hazard event with the water utility asset and identify the set of directly damaged (and hence failed) assets that will lead to disruptions across different water utility sub-systems.4. For drought events, estimate the risk of system-level disruptions by estimating the unmet water demand due to loss of water supply resulting from lower stream flows, reservoir levels, and system leakages.5. Quantify, for each disruptive event, the number of water users impacted per day during the event, multiplied by the duration of the event (in days), i.e. the CDD. 6. Compute the annual expected CDD per utility supply system by integrating the risk distribution using quadrature, or the trapezoidal rule.
The water system model and risk analysis methods are published and described in detail in Becher et al (2023).Here we employ the CDD risk outputs, projected up to the year 2080 under RCP2.6,RCP4.5 and RCP8.5 emissions scenarios in this aforementioned study (Becher et al 2023), to quantify the effect of a portfolio of climate adaptation options on mitigating future water shortage-and flood-induced asset damages and disruptions.

Cost of damages incurred by the utility
Here we consider the costs required to repair assets following flood inundation as an important climate-related financial risk transmission pathway for the Jamaican water utility.Climate-related financial losses may also include lost tariff revenue during disruptions, however, these losses are not considered as part of the current framework.The level of damage to the utility's potable asset-base is modelled by applying damage curves, which quantify the ratio of the expected repair cost to the replacement cost of an asset against a given flood depth (Porter 2021).Asset-specific damage curves for flooding for pumping stations (applied to wells, intakes and pumping station) and treatment plants were obtained from (FEMA 2020), see supplementary table 2. Reservoirs were assumed to be invulnerable to flooding at the return periods considered here, given that design standards for reservoirs typically specify a higher protection level.We apply the fraction damage curve value for a given flood event to the cost of reconstruction for the different types of assets to estimate the cost of flood-induced damage, as shown in equation ( 1).
Where PD j is the physical damage cost estimate for an asset (j) for the flood hazard event (i), F j denotes the function that describes the damage curve ratio (Miyamoto 2019, FEMA 2020) for the asset from which we estimate the damage ratio value at given flood depth (d i ) for the flood hazard event i, R j denotes the full reconstruction cost of that particular asset j (National Water Commission 2011).Equation ( 1) was applied to estimate damages to wells, treatment plants, intakes and pumping stations.Replacement costs for these assets are assumed to be equivalent to their construction costs, collected from the NWC's Parish water supply plans (National Water Commission 2011) and are summarised in supplementary table 3. Asset level direct damage risk across several events is estimated, as the annual average damage (PD j ), by numerically integrating the inverse cumulative probability function of the direct damage costs Q PD (p), obtained by ranking the damages for all events to estimate annual exceedance probabilities of damages (p).The annual average damage is estimated for given flood events set under a given climate scenario. (2) The asset level annual average damages are aggregated at the system level, where each system serves a set of customers.For a given system s, the annual average damage (PD s ) is estimated by adding up the annual average damages of all assets j within s (i.e.j ∈ s), as we can assume that direct physical damages to assets are independent of each other (3)

Cost of disruptions externalised to society
There is a strong reliance on tanker trucks to replace disrupted water utility supplies in Jamaica, although most households are fitted with one or multiple water tanks which can be filled during periods of non-disrupted utility supply (Lester 2015), providing a limited buffer against disruptions.Further, some households also turn to unregulated spring and well water sources during disruptions.However, in the absence of detailed data on the dynamics of tanker truck vendors and household storage tank capacity, we simplify complex, household-level responses to utility water supply disruptions, and assume that the only viable alternative form of water supply comes from tanker trucks.Thus, the alternative water purchase losses (AWL is ) are assumed to be equivalent to the CDD associated with system (s) due to event (i) (CDD is ) multiplied by the unit cost (c) ($J per person per day) of tankering water for the utility (Jamaica Observer 2018) Note that, the framework can accommodate contexts where tariff revenue is elastic to the rate of disruption: disruption-induced lost tariff revenue can be estimated by multiplying the CDD metric by the daily lost tariff revenue per customer.
Similar to estimating an annual average damage (from equation ( 2)), the annual average loss at the system level (CRL s ) is estimated by integrating the inverse cumulative probability function of the alternative water purchase losses Q SL (p), obtained by ranking the alternative water purchase losses for all events to estimate losses versus annual exceedance probabilities of losses (p).The annual average loss at the system level is estimated for a given hazard type (floods or drought) and for a given climate scenario, as shown in equation ( 5) (5)

Relative burden to customers in poverty
To estimate the relative burden of public water supply disruptions to customers in poverty, we estimate the fraction of each NWC system's customer base that is below the Jamaican poverty line, defined as an income of 1.9 US dollars per day.Those in poverty are less likely to be able to afford a household connection to the utility networks and would likely source water from other unregulated (and potentially contaminated) sources such as springs and wells (Velleman 2009).Indeed, approximately a third of Jamaica's poorest households obtain water from untreated wells and rivers (Haughton 2015).In the absence of detailed system-level data on the income distribution of customers, our approach estimates the poverty-incidence of each NWC system as follows.The population residing in communities that intersect with the infrastructure network boundaries of each NWC system, are assumed to encompass the population served by each system.Therefore, to find this representative population, we spatially overlay the system boundaries layer with a community level population (Statistical Institute of Jamaica 2010) and poverty incidence layer (Planning Institute of Jamaica 2011).All communities that intersect the NWC system boundaries were allocated to that system.Where a community intersects multiple NWC systems, the population within that community is split across systems in proportion to the population served by each system.Because the NWC water supply network covers ∼70% of the full Jamaican population (National Water Commission 2020), the population allocated to each system based on the census data, exceeds the population that is served by the utility, according to the NWC Parish Water Supply Plans.Thus, we assume that each NWC system's customer-base would first be made up of the full population allocated to that system that is not in poverty.Where the above-poverty-threshold population allocated to a NWC system is less than the NWC customers served by that system, the remaining customers are assumed to be made up of the population is in poverty.
The fraction of each NWC system's customer base in poverty can be estimated as follows: where f s denotes the fraction of people in poverty, TA s denotes the number of customers estimated to be above the poverty line within the service area of NWC system (s) and TB s denotes the number of customers estimated to be below the poverty line within the service area of NWC system.Our results, are sensitive to input data on poverty incidence (Planning Institute of Jamaica 2011), as illustrated in supplementary figure 1, which is outdated by over 10 years.However, the consistent use of the input datasets across systems provides an indication of the relative measure of how climate-related water infrastructure disruptions affect utility customers in poverty.Household survey data (e.g.World Bank 2012), that captures information on whether households have access to piped water supply connection and their rate of income, could be usefully applied in future applications of this framework.
To summarise the risk per NWC water system s due to each hazard (flooding and drought) in total (TR s ) and the total poverty risk (TPR s ) affecting the fraction of people in poverty (f s ) are estimated as following:

Multi-hazard, multi-objective optioneering framework
Flood defence walls and leakage reduction are selected as the two most relevant types of intervention for our optioneering study, as they are targeted specifically for climate adaptation interventions.We test different levels of flood wall and leakage reduction designs meant to prevent floods and drought events of return periods of 20, 50, 100, 200 and 500 years at each NWC water system level.
The optioneering involves doing a cost benefit analysis (CBA), where the costs of adaptation options are weighed against the benefits, or avoided risks, that would be accumulated (i.e. the net present value) over the lifetime of a project.We assume the lifetime of a project would be 50 years and simulate the time window from project implementation in 2030 (the date of the Capital Improvement Plan) to the year 2080.We assume a linear rate of change in annual expected risk between the baseline TPR s (2010) and future TPR s under each climate change scenario, where future risk is averaged across three RCP scenarios (RCP 2.6, 4.5 and 8.5) projected to the end of century epoch (year 2080).A discount rate of 10% was applied, following the approach that is taken by the utility in their Water Supply Parish plans (National Water Commission 2011).
For a selected adaptation option of flood defence walls or leakage reduction the planning of the chosen option is done over an annual timescale denoted by t = 0, . . ., T, where t = 0 is the start of the implementation of the option and t = T = 50 years is the option lifetime.Here we assume different options (flood defence walls or leakage reduction) would provide a level of protection up to given (flood or drought) event severity.For example, a flood defence wall designed to a 1/100 return period standard will provide protection against flood events up to 1/100 year return periods.Then we apply the aforementioned CBA approach to assess the effectiveness of each option.
As noted above we apply different adaptation options to present floods or drought risks for the set of design return periods RP = {20, 50, 100, 200, 500} (in years) or design annual exceedance probabilities P = {1/20, 1/50, 1/100, 1/200, 1/500}.Assuming that the discount rate of r% is applied over the adaptation planning horizon, the discounted cost (DC s (p d )) for applying the adaptation option of a given design annual exceedance probability (p d ∈ P) to a water system s is estimated in terms of the sum of initial investment costs of all assets j within system s (CI j0 (p d )) at t = 0 and the sum of the subsequent asset level discounted annual maintenance costs CM jt (p d ) over the planning horizon.This is formulated in equation ( 8) Using equations ( 2)-( 6) these reduced risks are obtained by rerunning the risks analysis with the hazard event sets, where now the risks to a system are only estimated for more severe events with exceedance probabilities ⩽ p d .This is formulated as following: The risk before and after the adaptation option are estimated at annual timescales over the planning horizon.As noted above, future risks are driven by climate change driven hazard events.The discounted benefit (DB s (p d )) of risk reduction due to the adaptation option is estimated by discounting the annual timeseries of marginal risk reduction (TR st − TR st (p d )) over the planning horizon The effectiveness of the option is assessed in terms of the benefit-cost ratio (BCR s (p d )), given by equation ( 11) The optioneering solution (for a given type of hazard protection) involves finding the optimal intervention (OI s ) per system, identified as the option where the ratio between benefit and cost (BCR) is greatest (13)

Choice of flood defences
For flood defences adaptation, we look across all assets exposed to flooding.First, we find the flood depths to which each asset exposed for return periods equivalent to 20, 50, 100, 200 and 500 years.The costs of flood walls built up to a range of heights to protect a number of water treatment plants and pumping stations in the UK are extracted from a report provided by a Mott MacDonald Report (2018) which were converted to Jamaican dollars.Extracted costs are used to produce cost curves of flood walls built to a range of heights, with upper and lower uncertainty bounds.This curve is provided in the supplementary table 4. Thus, we derive a cost curve for the initial investment costs (CI j0 (p d )) of flood defence walls to protect against the aforementioned range of return periods for all exposed assets, which are estimated at 0.5% of the original value of the flood wall (Aerts 2018), also provided in supplementary table 4.
As water systems are made up of a network of assets with varying levels of redundancy, implementing flood defences at an individual site may still result in the same risk of customer disruptions as not implementing any defences, if multiple assets in a system are exposed to the same events.Here, we adopt a systems perspective and calculate the costs, and the risk reduction potential of, protecting each system from events across the return periods of 20-, 50-, 100-, 200-and 500 years.Note that this requires assuming that a given flood event would cause disruptions at all exposed (have a probability of flooding > 0) and vulnerable (experience damage given a flood inundation depth above some level) assets within in a given system simultaneously, which is a reasonable assumption given their spatial proximity to each other.This assumption has also been adopted in other studies (e.g.Koks et al 2019b).As such, for a given system to be protected from a 1 in 100 year flood, we sum the costs of defences at all exposed assets that would be required to mitigate damage and disruptions in a system equivalent to a 1 in 100-year event.

Choice of leakage reduction options
Leakage reduction is proposed to improve the reliability of services to the currently served population.Further, the NWC considers leakage reduction a priority intervention, due to the significant losses of up to 70% that current affect the distribution networks (National Water Commission 2011).In the capital improvement plan, the budget proposed for leakage reduction is provided as a national total volume leakage loss mitigated (NWC 2015).We use this to derive a unit cost per m 3 of leakage loss reduction.In reality, the marginal cost of reducing leakage increases as the target loss rate becomes more ambitious: for example, it is cheaper to reduce leakage from 50% to 40% than from 20 to 10%.In higher income countries such as in the UK, 15% leakage is considered to be an economic loss rate, beyond which costs become unjustifiable (GWI 2022).In cases where water infrastructure is thought to be most advanced globally, such as Singapore and Israel, leakage rates can be as low as ∼5% (ibid).Here, we assume a linear cost curve and adopt 30% as the economic loss rate for Jamaica, based on the utility's target as outlined in their Parish water supply planning documents (National Water Commission 2011).
To determine the effect of reducing leakage on water shortage risk per system, we run iterations of the water balances established in (Becher et al 2023), with a range of leakage reduction scenarios from the original loss rate to a minimum rate of 30% (National Water Commission 2011).The water balances involve simulating monthly supply inputs for all surface water systems across the NWC network against a system specific drought threshold.The drought threshold per surface water abstraction asset is the sum of the volume of demand, leakage losses and an environmental flow threshold.The sensitivity of each system's risk of drought-induced water shortages with respect to changes in leakage loss rates was derived.This was used to find the reduction in leakage that would be required to achieve a protection level equivalent to each of the five customer disruption events associated with return periods of 20, 50, 100, 200 and 500 years.The cost of each leakage reduction option CI j0 (p d ), is found by applying the unit cost (m 3 ) to the volume of leakage loss mitigated by each option.In some cases, reducing leakage by the minimum amount (i.e. to 30%) would not be enough to protect systems against a very extreme event; in which case the option was removed from consideration.
Following the same approach as that defined by equations ( 8)-( 12), we derive investment costs and mitigated alternative water purchase losses associated with each leakage reduction option for each system to identify the most cost-beneficial option per system.

Mitigated alternative water purchase losses for customers in poverty
We use the estimated alternative water purchase losses that would be burdened by those in poverty (TPR s ) to simulate the mitigated poverty losses as the Discounted Poverty Benefit DPB s (p d ) achieved by alternative interventions.

Uncertainty analysis
The uncertainty range is derived firstly by determining whether results are more sensitive to (1) the variation in cost and benefit of each option provided by each of the considered RCP scenarios or (2) perturbations of the combined vulnerability (damage fraction and leakage).For flood defences, the latter source of uncertainty dominates, whereas for leakage reduction, the former source of uncertainty dominates.We also factor cost estimates by ±25% for both flood defences and leakage reduction options.

Results and discussion
In this section, we present the overall distribution of costs and benefits for each option type, across all considered protection levels across all systems.We go on to explore results at the system level.We explore the BCRs of system-optimised options, to identify the optimal options tailored for each individual system and to explore to what extent economic (in terms of BCR) and equity (in terms of costs alleviated for those in poverty) objectives align or diverge for different systems.We also provide the spatial distribution in utility damages induced by flooding per climate change scenario in supplementary figure 2.

Benefit cost ratio of all options
Figure 2 provides an overview of the distribution of discounted benefits versus discounted costs of all considered leakage reduction (top) and flood defence (bottom) options, while also showing the discounted benefits associated with poverty reduction.For leakage reduction, we see that for many options, the benefit is zero or close to zero.In these cases, the sensitivity of a system's water shortage risk to changes in leakage losses is low.We also observe a strong relationship between an option's overall benefit and its benefit to those in poverty.This is because the BCRs of leakage reduction options are defined as the mitigated alternative water purchase losses they offer, of which their benefit to those in poverty is a fraction.However, this relationship is not always clear: some options offer a high BCR but a low benefit to those in poverty, which signifies a relatively low poverty incidence among these options' beneficiaries (i.e. the customer base of the relevant NWC system).
The differences between costs and benefits for different options is lower for leakage reduction compared to those for flood defences.This is because the contribution of leakage towards water shortage risks across NWC systems is relatively pervasive in comparison to relatively localised flooding impacts.In the lower subplot of figure 2, we see the costs and benefits of flood defences.In absolute terms, the costs and benefits of flood defences are both much higher than for leakage reduction options, given the contribution of avoided asset damages and that interventions require construction of protective walls around all exposed assets in a system.We also note that, although there is still an observable relationship between overall option benefit and the benefit for those in poverty (indicated by the shade of blue of each point), it is not as clear as for leakage reduction options.This is because the contribution of avoided utility damages vastly outweighs that of avoided alternative water purchase losses in the calculation of overall benefits of flood protection, and the riskiest assets from a damage perspective are not necessarily the same from a disruption perspective.This mismatch between damage and disruption risk hotspots aligns with prior studies on flood risk to critical infrastructure services (e.g.Koks et al 2019b).The range of uncertainty associated with each point is indicated by error bars.

System level optioneering
Adopting a systems perspective, figure 3 compares the BCRs across all protection levels considered, for drought (left) and flood (right) options.Please note that we only show the 44 systems which have a BCR > 1 for either leakage reduction or flood defences.We see that the BCRs for leakage reduction options generally increase with an increased protection level, whereas the variation between BCRs for flood defences do not show this trend.This suggests that the increase in flood defence costs co-varies with the increase in avoided damages as the protection level of an option increases, such that the ratio between benefit and cost does not necessary increase.Whereas the change in drought risk, and therefore the additional required investment in leakage reduction, is relatively small for a given increase in protection level, which means that incremental investments in leakage reduction tend to pay-off more than for flood defences.This means that the losses incurred by high frequency drought events increase to a lesser extent relative to flood events, such that protecting against a 1-in-500 year protection level would not cost significantly more than protecting up to a lower protection level.System 137 is the only system where both types of option with a BCR > 1, highlighting the benefit of adopting a multi-hazard perspective for optioneering to capture all climate vulnerable systems that would benefit from adaptation (Koks et al 2019a, Dong et al 2020, Luo et al 2023, Verschuur et al 2023b).
Figure 4 provides the BCR per option, broken down between leakage reduction options (red, left) and flood defences (blue, right).Note that here we plot the optimal option per system with a BCR greater than 0. High leakage losses are pervasive across NWC systems, driving the risk of water shortage-induced alternative water purchase losses (Becher et al 2023).In contrast, flood impacts are relatively localised (ibid).Thus, as presented in figure 4, we find a greater number of leakage reduction options that offer BCRs > 1 compared to flood defence options, and a greater variation between BCRs asset protection options than for drought mitigation.
In figure 5, we represent the system-optimised option's BCR spatially (across leakage reduction and flood defences).Leakage reduction options are indicated in red, with darker red highlighting options with the greatest BCRs.Flood defence options are indicated in blue, with darker blue highlighting flood protection options with the greatest BCRs.Notable system numbers are provided for ease of comparison between plots (e.g.figures 3 and 5).We can see that the options with the greatest BCRs for leakage reduction target systems 137, 80 and 26 systems.Likewise, the flood defence strategies with the greatest BCRs target systems 9, 84, 81.Two hotspots are identified in systems serving areas near Ocho Rios and Kingston, where adaptation options for protection against both drought and flood hazards would be particularly efficient from a cost-benefit perspective.There is not a distinct spatial clustering between flood and drought optimal BCRs, which one would expect if BCR was only influence by hazard risk.This is not surprising, given the influence of local system characteristics on risk and, in turn, on the cost-effectiveness of risk mitigation strategies (Becher et al 2023).Flood risk is driven by the spatial pattern of flood inundation depth for a range of return periods in relation to asset location, vulnerability (defined by asset-specific flood depth-damage curves) and criticality (population served per asset) or reconstruction costs for disruption and damage risk, respectively.Drought risk is driven by population served by a system and the availability of water supply relative to demand (ibid), where demand reflects household consumption after leakage losses across the distribution network, and supply inputs capture the dynamics major storage reservoirs (ibid).Note that we do not consider the difference in water consumption rates for customers across income groups (Hussien et al 2016), but rather assume a constant rate of demand, following the approach taken in the utility's Parish Water Supply plans (National Water Commission 2011).We find that system-optimised flood defence and leakage reduction options with BCRs > 0 are 83 and 128, respectively, and the number of options >1 amount to 24 and 26, respectively.The total annual benefit that could be achieved by implementing all system-optimised flood defences with BCRs > 0 is estimated to be approximately 0.72 billion Jamaican dollars per year.This saving represents a significant portion of annual total repair costs, which ranges from 1 to 3 billion Jamaica dollars per year (NWC 2019).However, flood defence walls are relatively capital intensive and can have unintended consequences (Jongman 2018), such as exacerbating flooding in surrounding areas.Further, they would not address turbidity risks, one of the most common forms of flood-related disruption for the utility (NWC 2021).A broader range of option types, such as integrated, nature-based approaches (including rainwater harvesting and flood retention), could be included within the framework in future applications (FAO 2019).The difficulty in quantifying the effectiveness of nature-based solutions for flood risk reduction at scale prevented their inclusion here (Turkelboom et al 2021).Whereas, for leakage reduction, the total benefit would be 1.5 million Jamaican dollars per year.We note that reducing leakage, e.g. through pressure management and network rehabilitation, could also help to reduce energy consumption and strengthen network resilience to drought and operational risks, such as pipe bursts (Ahopelto and Vahala 2020).This is particularly challenging in the hard-to-reach, rural communities in the centre of the island, which have a higher poverty incidence (see figure S3): extreme pressure variations due to steep topography pose a challenge to the functioning of distribution networks in these systems, increasing the risk of leaks and mechanical failures (Lambert 2000).Indeed, J$ 828.8 million project is being channelled into reducing the level of Non-Revenue Water by 30% in the next 5 years and reduce maintenance by correcting for leaks (NWC 2015).
In figure 6 we provide the spatial contribution of the most economically efficient options towards equity-related objectives, defined in terms of absolute annual alleviated expenditure for those in poverty.Drought options are indicated in orange, with darker orange highlighting drought mitigation options with the greatest benefit to those in poverty.Flood defence options are indicated in purple, with darker purple highlighting flood protection options with the greatest benefit to those in poverty.Systems with the most economically efficient options are annotated.The total annual benefit to those in poverty that would be achieved by implementing these options is estimated at 0.08 million Jamaican dollars per year, with optimal flood defences and leakage reduction options offering a comparable contribution towards this annual benefits.Although this is relatively small number in absolute terms, from the perspective of a low-income household, could represent a significant saving relative to income (Hallegatte and Rozenberg 2017).
However, the options with the greatest BCRs do not always align with the options that provide the greatest overall benefit to those in poverty.Figure 7 explores the interaction between economic-and equity-orientated objectives spatially, offering an integrated perspective for decision-makers.Options were ranked (or scored) based on their economic performance (BCR) and equity-related benfit.Options with a high score based on both economic and equity performance, offer a strong synergy between economic and equity objectives and are represented in red in figure 7. System-optimised options with a high score for only one objective, are more subject to trade-offs: options with a high score for equity and low for economy, appear in yellow; options with a high score for economy and low for equity, appear in blue.Options with a low score with respect to both equity and economic performance, are represeted in grey.

Assumptions, limitations, uncertainty and merits
A new optioneering framework for potable sector planning has been presented to quantify direct social and utility losses incurred during climate-induced water supply disruptions.This framework allows us to directly compare the benefit cost ratios of a portfolio of protection works, proposed to reduce asset-level vulnerability to flooding, and leakage reduction options, proposed to reduce system vulnerability to drought impacts.It also allows us to estimate the alternative water purchase losses burdened by customers in poverty that would be alleviated due to potable water sector investments.
In the application of the framework to the water supply utility in Jamaica, we find that system-optimised flood defence and leakage reduction options with BCRs > 0 are 83 and 128, respectively, and the number of options >1 amount to 24 and 26, respectively.The total annual benefit that could be achieved by implementing all system-optimised flood defences with BCRs > 0 is estimated to be between ∼0.72 billion Jamaican dollars per year.Whereas implementing all leakage reduction options would provide a total benefit of ∼1.5 million Jamaican dollars per year.The total annual benefit to those in poverty that would be achieved by implementing these options is estimated at 0.08 million Jamaican dollars per year.Overall, asset protection interventions are cost effective for mitigating utility damages, whereas drought mitigation options are more effective at mitigating alternative water purchase losses.Drought mitigation options are more cost effective than flood defences, however, investment-worthy flood defences provide greater financial benefits in absolute terms.Both option types have a comparable benefit to those in poverty in absolute terms.
We acknowledge that, in such a budget constrained environment, non-climate-related factors would likely dominate investment decision-making process of utilities, such as the drive to expand networks to unconnected communities.Nonetheless, an optioneering framework, such as that presented here, can help utilities to more efficiently allocate limited budgets.The framework proposed also serves to demonstrate the positive impacts of adaptation investments, which can be useful for attracting concessional financing.
The framework presented here introduces an alternative quantified measure of the direct financial returns that could be realised by investing in more climate resilient municipal water supply infrastructure.Typically, cost benefit analysis on new water supply infrastructure is done based on the projected tariff revenues that would be gained by increasing the volumetric sale of water, which are limited due to intense political pressures to keep tariffs low.Further, benefits defined as such would be zero for projects that improve the climate resilience of supplies to existing customers and, therefore, do not necessarily increase tariff revenue.Our approach supports both the economic case for climate adaptation projects in the water sector at the same time as demonstrating the benefit of adaptation infrastructures for society.In the context of an infrastructure investment landscape that is increasingly motivated to enable climate resilience and equitable social outcomes, such analytical frameworks could help water supply utilities access financing sources that meet their unique requirements and constraints.
The wider applicability of certain aspects of the approach should be re-evaluated before employing this framework in other utility contexts.Firstly, here, we simplify the definitions of utility and alternative water purchase losses: in applications of the framework in the other contexts, the components of these losses must be considered.For example, many utilities lose tariff revenue during drought-induced water shortages, due to a lower overall volumetric sale of water; in such cases, lost tariff revenue should be incorporated into the definition of utility losses.Further, we focus on direct financial losses, omitting more intangible, indirect losses such as the time spent on acquiring alternative water sources and associated losses in productivity or wellbeing impacts.
Another aspect of our approach to consider is that utilities typically consider a much broader portfolio of infrastructure investments than those considered in this paper, such as piped network construction or supply capacity enhancement projects.We do not consider options that are proposed to increase the number of household connections to the network as, although they may also improve the overall resilience of the network, it is not possible to disentangle this from the effects of increasing demands on the relevant system and heightening societal exposure to disruption.
Third, due to data constraints, here we do not account for the uneven quality of service provision, or rate of intermittency, across systems.This information could, in subsequent applications, be harnessed to better understand the vulnerability of households to disruptions of different durations.For example, the less frequently households receive water supply, the less likely they would be able to store a sufficient amount to buffer against disruptions.Given that many low-income communities are often burdened with greater levels of supply intermittency, the uneven levels of vulnerability that that exist across water supply networks, could be incorporated into the underlying CDD risk metric in future applications.
Despite the limitations of the framework, our approach highlights the importance for stakeholders to capture multiple hazards and financial risk transmission pathways when evaluating what measures would offer the most cost-beneficial climate resilience benefits.The framework also takes an early step towards mainstreaming equity considerations in potable sector decision making, to identify co-benefits and trade-offs between economic-and equity-related objectives.

Concluding remarks
Through the delivery of potable water supply services, water utilities are responsible for among the most vital services for humanity.Climate-related disruptions to water supply services incur direct financial losses to utilities (e.g. to repair damaged assets) and temporarily deprive households, imposing a financial burden in acquiring water from alternative water sources.This uncompensated social burden, which may disproportionately impact lower-income households that have less disposable income to spend on alternative supply (e.g. from tanker trucks), should be properly accounted for in investment planning.
Here we show that adapting potable water supply networks to climate change would not only mitigate annual asset damage losses for utilities, but also improve the affordability of safe water services and protect the most vulnerable domestic customers from climate extremes.This speaks to a wider discussion on the interlinkages between climate adaptation and overall development and demonstrates the synergy between the two in the case of potable water supply investments.
We draw three main policy recommendations from this work.(1) It is necessary to capture the costs of customer disruptions, and distributional effects across income groups, in water supply optioneering frameworks to promote the multi-dimensional benefits of climate adaptation options.(2) Utilities that are exposed to multiple climate extremes require an integrated approach for directly comparing the risk mitigation potential of options that adapt to different hazards.(3) Leakage reduction is a low-regret, climate adaptation option for water supply utilities to improve overall operational efficiency and reduce the risk of customer disruptions.

Figure 1 .
Figure 1.Methodological framework for multi-hazard optioneering with respect to economic and equity objectives.

Figure 2 .
Figure 2. Cost (DCs) and benefit (DBs) per leakage reduction option (top) and flood defence option (bottom), where points are coloured by the tankering costs saved by customers in poverty.All costs are provided in million Jamaican dollars.Vertical lines associated with each scatter point represent the uncertainty associated with the benefit of each option.The black line illustrated the 1-1 relationship between benefit and cost; therefore, options are associated with BCRs that are >1.

Figure 4 .
Figure 4. Benefit Cost Ratio (BCRs) per option, with flood defence options highlighted in blue and leakage reduction options in red.Options are ranked from those offering the highest BCR to the lowest.BCRs, plotted on the y axis using a log scale for plot clarity.

Figure 5 .
Figure5.Option with the greatest BCR per system (OIs).Drought options are indicated in red, with darker red highlighting drought mitigation options with the greatest BCRs.Flood defence options are indicated in blue, with darker blue highlighting flood protection options with the greatest BCRs.Systems with the most economically efficient options are annotated.Parish outlines are plotted for greater spatial context.

Figure 6 .
Figure 6.The contribution of the most efficient option (OIs) of each type towards reducing costs for those in poverty (DPBs) in absolute terms in Jamaican dollars.Drought options are indicated in orange, with darker orange highlighting drought mitigation options with the greatest benefit to those in poverty.Flood defence options are indicated in purple, with darker purple highlighting flood protection options with the greatest benefit to those in poverty.Systems with the most economically efficient options are annotated.Parish outlines are plotted for greater spatial context.

Figure 7 .
Figure7.Bivariate plot showing the interaction between economic-and equity-orientated objectives for each system, spatially.Strong option performance solely with respect to equity appear in yellow.Strong option performance solely with respect to economic objectives appear in blue.Strong performance with respect to both objectives appear in red.Systems with optimal options that have a strong synergy between both objectives are annotated.Five quantiles are applied to the score of each option relative to each objective.Parish outlines are plotted for greater spatial context.