Global analysis of urban surface water supply vulnerability

Water vulnerability assessments were performed for all basins supporting cities with populations in excess of 750 000 that obtain water supplies solely from either (a) river withdrawals or (b) reservoirs (natural or man-made) that are not shared with another large city. In total, 71 cities in 39 countries were identified that meet these criteria. Information on the type and location of urban water sources was obtained from the City Water Map Initiative (McDonald et al 2014). Hydrologic data, including mean annual reliable basin discharge (GWSP 2008a), environmental flow requirements (GWSP 2008b), mean annual reservoir storage capacity (Lehner et al 2011, NID 2009), global potential evapotranspiration rates (Zomer et al 2007, 2008) and urban demand (FAO 2013) were obtained from global databases (see supplemental material). Available environmental flow requirements and discharge statistics for all supply basins were based on time series of monthly climate variables over the climate normal period 1961–1990. Estimates of future vulnerability are based on urban growth projections developed in the UN World Urbanization Prospects (United Nations 2012). Basin boundaries for each urban water source were delineated using HydroSHEDS (Lehner et al 2008), yielding 107 surface-water basins, with 49 cities relying on a single supply basin and 22 cities relying on more than one basin. Where a city uses more than one supply basin, that city's supply basin refers to the combination of discharge from all of the supply basins used.

The focus of this study is on large cities using surface water supplies that lack source diversity, relying on either direct river withdrawals or releases from reservoir storage. Three vulnerability metrics, discussed below, were used to identify those cities that fail to meet simultaneous thresholds of freshwater supplies for urban, agricultural, and environmental purposes. Urban water vulnerability was calculated for baseline conditions (2010) and for a future scenario (2040) in which increasing water demands are proportional to predicted population growth and projected irrigated agricultural demand. Although projected demand estimates account for population increases, surface water supply changes resulting from expanding urban boundaries and additional future supplies a city might tap by 2040 are not considered. These estimates also do not include any impacts of changes in land use or climate, which could potentially increase or decrease the amount of water available to an urban area (Velpuri and Senay 2013). Rather, this analysis establishes a baseline estimate of vulnerability assuming stationarity in urban hydrologic systems over the next 25 years. For the future scenario, the reduction in urban water vulnerability that could be achieved is examined in two end-member demand scenarios: (1) demands based on projected water-sharing allocations, and (2) demands where excess water from agriculture or the environment is redirected for the other sectoral uses.

2.3.1. Environmental water requirements (${{Q}_{E}}$)

Environmental flows are a critical water resources component (Poff et al 1997, Smakhtin et al 2004). Ideally, a supply basin should be capable of fostering urban well-being while supporting a viable ecosystem. The environmental water metric reflects the vulnerability of freshwater-dependent ecosystems to urban water use, and is defined as a ratio of critical environmental water demand to available basin flow. Environmental water demand (${{D}_{E}}$) is based on low-flow conditions derived from the WaterGap model in which global estimates of environmental water flow were determined at the watershed scale (Döll et al 2003, Smakhtin et al 2004).

This vulnerability metric compares the water required to meet environmental demands to the mean low-flow monthly water discharge within each basin (${{q}_{90}}$). Although it has been shown that urbanization can cause changes in runoff in some US urban watersheds (Velpuri and Senay 2013), global data of changes in runoff trends are currently not available and may not be applicable in cities that collect water from basins external to their urban footprint. Rather, within each basin, the ${{q}_{90}}$ value is used to represent the 'reliable monthly discharge,' which is equal to the minimum flow that is exceeded on average 90% of the time (based on monthly discharge data). We define the environmental water requirement (${{Q}_{E}}$) for each basin as the ratio of ${{D}_{E}}$ to the low-flow runoff $({{q}_{90}})$ after agricultural (${{D}_{A}}$) and urban demands (${{D}_{H}}$) are met (see 2.3.2 for agricultural and urban demand calculations) (equation (1)).

$$\begin{eqnarray}&&{{Q}_{E}}={{D}_{E}}/\left( {{q}_{90}}-{{D}_{H}}-{{D}_{A}} \right).\end{eqnarray} \tag{ 1 }$$

Any basins for which environmental water demands $({{D}_{E}})$ are large relative to net reliable flow are considered threatened (${{Q}_{E}}$ > 0.4). The threshold of 0.4 adopted here builds on a similar scarcity threshold applied in other well-established contexts (Falkenmark et al 1989, Vörösmarty et al 2005).

2.3.2. Human uses: urban water and agricultural irrigation

Urban water requirements were assessed using an analogous metric to that used for environmental flow requirements. Specifically, the urban water requirement represents a ratio of urban demand during low-flow conditions to renewable surface water in basins when environmental needs are met first. Since urban water demand (${{D}_{H}}$) is not a commonly reported metric, estimates of demand for each city were computed by multiplying the population by the mean annual national municipal and industrial per capita demands, excluding withdrawals for thermoelectric cooling (GWSP 2008c). The discharge required to meet urban needs was calculated as the ratio of ${{D}_{H}}$ to the mean 'reliable monthly discharge' for a basin $({{q}_{90}})$ after agricultural (${{D}_{A}}$) and environmental demands (${{D}_{E}}$) are met (equation (2)).

$$\begin{eqnarray}&&{{Q}_{H}}={{D}_{H}}/\left( {{q}_{90}}-{{D}_{E}}-{{D}_{A}} \right).\end{eqnarray} \tag{ 2 }$$

As with ${{Q}_{E}}$, a threshold is applied such that all basins for which urban demands D_H are large relative to net available low flow are deemed threatened (${{Q}_{H}}$ > 0.4).

Agricultural demands (${{D}_{A}}$) were estimated from the most recent annual national water use statistics (FAO 2013) for urban supply basins with >1% of the total area occupied by land under irrigation (Siebert et al 2013). The relative percent of water withdrawn for agriculture was used as a proxy for the flow in each basin that is allocated for irrigation purposes. In equation (2), systems supported by river withdrawals, have ${{D}_{A}}$ subtracted directly from the ${{q}_{90}}$ while basins supporting urban supplies through reservoirs, versus directly from streamflows, have a ${{D}_{A}}=0$. For these reservoir-supplied systems, values of ${{D}_{A}}$ are instead subtracted from reservoir storage (equation (3)). Projections of future agricultural demands for the 2040 scenario were made based on available national water use statistics from 1965–2010.

2.3.3. Storage requirements (${{Q}_{S}}$)

The third measure of vulnerability provides insight into a basin's hydraulic capacity to supply water. This measure is based on the ratio of total annual urban demand volume to reservoir normal storage. Reservoir storage is a key component in many urban supply systems, increasing hydraulic system stability by helping to buffer variability in available supply (Padowski and Jawitz 2012). In this study, 28 of the 71 cities relied on releases from surface reservoirs. Here, urban supply basin reservoirs are assumed to be managed such that urban demands receive top priority. It is also assumed that no additional significant reservoir capacity will be added before 2040. Understanding that these assumptions likely provide just one assessment of water available for human consumption, storage threat is measured as the ratio of urban water demand (${{D}_{H}}$) to the reported reservoir annual storage capacity (${{Q}_{{\rm cap}}}$) minus evaporative losses estimated using potential evapotranspiration $({{D}_{V}})$ (Ward and Trimble 2003) and agricultural demands (${{D}_{A}}$) for a year (equation (3)), where basins are considered threatened when ${{Q}_{S}}\gt 0.4$ (Lane et al 1999).

$$\begin{eqnarray}&&{{Q}_{S}}={{D}_{H}}/({{Q}_{{\rm cap}}}-({{D}_{V}}+{{D}_{A}})).\end{eqnarray} \tag{ 3 }$$

Reservoir-supplied cities were evaluated using each of the three metrics described above, however river-supplied cities have no major storage infrastructure and were assessed using only the two flow-based metrics (${{Q}_{E}}$ and ${{Q}_{H}}$).

For both categories—reservoir and river supplied—cities were considered 'vulnerable' if all demand thresholds were violated. If a city had one or two but not all three metrics exceeding a threshold, then it was considered 'threatened,' as these supplies may no longer be able to support urban demand without causing damage to agriculture, the environment, or both. We refer to a city as 'susceptible' if it has either threatened or vulnerable status; the number of susceptible cities is the sum of those that are threatened or vulnerable. Finally, cities were considered 'non-threatened' if no demand thresholds were violated.

Each city was assessed according to the three vulnerability metrics defined in section 2. The status of each city, either vulnerable (red), threatened (yellow) or non-threatened (light blue) is shown on the global map in figure 1. The total number of cities in each category is given in table 1. Of the 71 cities included in this study, 25 are classified as vulnerable in 2010, of which 23 are cities supplied solely by direct river withdrawals. These vulnerable river-supplied cities have significantly lower mean per capita flows (2700 l/p/d) than those cities with sufficient water (196 100 l/p/d), and fall within the range of water availability that Falkenmark et al (1989) identify as having water management problems due to water stress (2700–4700 l/p/d).

Zoom In Zoom Out Reset image size

In addition to supporting urban demands, urban supply basins often provide water for agricultural and environmental needs as well. To better understand why vulnerability occurs, the primary source of water demand by sector was determined for each supply basin. Considering only the 25 vulnerable cities, demand from the agricultural sector is dominating for the supply basins of 16 cities (64%). There are five vulnerable cities for which environmental needs are most prominent within supply basins (20%), with demand from the urban sector dominating in the supply basins of four cities (16%) (table 2).

Also examined were 'threatened' cities, where 'threatened' means that one or two of the three metrics exceed their demand threshold. Of the 71 cities considered, 34 are threatened and represent the majority of reservoir-supplied systems (71%), but only 33% of river-supplied cities are threatened (table 1). Environmental demand was the only stressor for river-supplied cities, however reservoir-supplied cities revealed more varied sources of stress, suffering from a variety of environmental, human and storage-related threats (table 2). The number of threatened reservoir-supplied cities is ten times greater than those considered vulnerable, whereas the number of threatened river-supplied cities is approximately half of those found to be vulnerable (table 1).

In 2040, the number of vulnerable cities rises from 25 to 32, representing a 28% increase in urban areas facing serious water issues (table 1). Of these 32 cities, those that are river-supplied continue to make up the majority of those that are vulnerable. Our analysis provides a baseline estimate of vulnerability assuming near-term stationarity in urban surface water hydrologic systems. Even without accounting for potential changes in local hydrology due to climate change, as populations increase, mean per capita flows for vulnerable river-supplied cities drop over this 30 year period to 1200 l/p/d, beyond 'chronic water scarcity', 1370 l/p/d (Falkenmark et al 1989). During this same period, the total number of vulnerable reservoir-supplied cities more than doubles, from 7% to 18%, as storage infrastructure loses its advantage as a supply buffer. When considering only the 32 vulnerable cities in 2040 (table 2), 18 (56%) cities had supply basins dominated by demand from the agricultural sector. The number of cities with vulnerable supply basins dominated by demand from the urban sector increases from four out of 25 vulnerable cities (16%) in 2010 to eight out of 32 vulnerable cities (25%) in 2040.

Although more cities become vulnerable in 2040, the percent of 'threatened' cities decreases from 48% in 2010 to 42% in 2040 (table 1). Similar to 2010, the majority of reservoir-supplied cities are those considered threatened (71%), however the relative difference between the numbers considered threatened versus vulnerable decreases. Environmental demands continue to be the only stressor for river-supplied cities, however, stress from storage over-allocation becomes the leading threat in reservoir-supplied cities, occurring in 15 of the 28 cities depending on this supply type (table 2). Although the ratio of reservoir- to river-supplied threatened cities increases, by 2040 all but three reservoir-supplied and six river-supplied cities remain unthreatened. Access to large sources of water combined with relatively low urban demand and/or environmental demands keeps these nine cities protected from water-supply compromise.

Between 2010 and 2040, 10 'cities-of-concern' were identified as becoming more susceptible, with conditions worsening from a non-threatened to threatened status, or from threatened to vulnerable. In total, three of the cities identified as non-threatened in 2010 become threatened in 2040 (Accra, Ghana; Freetown, Sierra Leone; Panama City, Panama). Seven cities-of-concern change from threatened to vulnerable status (Dublin, Ireland; Salem and Charlotte, USA; Ouagadougou, Burkina Faso; Guangzhou, Wuhan, and Nanjing, China; figure 1 insets). Three of these cities-of-concern (Charlotte, Ouagadougou, and Panama City), each transition from a system dominated by the environmental sector demand to one dominated by the urban sector demand, as none of their supply basins support significant irrigated agriculture. Of these 10 total cities that show increased susceptibility by 2040, 6 are reservoir-supplied cities and 4 are river-supplied. This suggests that increased demand by 2040 puts at risk a significant number of both river-supplied and reservoir-supplied cities.

Two water reallocation strategies are examined to assess their potential usefulness of reducing the threat of failing to meet urban demands. In these alternative assessments, urban water threat predictions operate under the assumption that growing urban demands may either completely consume water required for environmental flows (reallocation through environmental neglect), or be tempered through an urban-rural partnership where water is reallocated from agriculture via a water market to urban uses within a supply basin (reallocation through agricultural transfers). In this scenario, surface reservoirs continue to provide irrigation water and a portion of that water is transferred to the urban sector sufficient to meet the urban water demand threshold. The remaining water is used for irrigation or to satisfy environmental needs. Similarly, irrigation water in excess of urban needs is also used to meet minimum environmental flows in river-supplied cities. This analysis represents an end-member strategy in which enough irrigation water is transferred to satisfy urban demands plus other sectoral demands to the degree possible.

Table 3 shows the number of cities for which reallocation provides sufficient water under three scenarios: (1) satisfying urban demands only, (2) satisfying both urban and environmental demands, and (3) securing adequate reservoir storage. Here, 'sufficient water' means that enough water exists within a city's supply basin to satisfy sectoral demands in each scenario.

Growing demand for freshwater will likely force human needs to be placed ahead of environmental requirements. As such, the first strategy is reallocation through environmental neglect in which cities divert all water available for maintaining minimum environmental flows to meet urban needs. Under this strategy agricultural demands are also met, such that neither the urban or agricultural sectors are threatened. Of the 71 cities included in this study, the number of cities that could support urban demands in 2040 increases from 35 to 52 when exercising environmental neglect reallocation (table 3). Urban sector threats alone are reduced in 14 otherwise vulnerable and 29 threatened cities, where 9 cities remain non-threatened. Those aided by reallocation through environmental neglect include 5 of the 7 cities-of-concern likely to become vulnerable by 2040, with only Ouagadougou and Charlotte failing to benefit.

The second strategy reallocates water from agriculture to urban use and shows similar benefits to reallocation by environmental neglect for 51 versus 52 cities. However, agricultural reallocation preserves environmental flows, and shows that roughly the same benefit can be gained without further damage to freshwater-dependent ecosystems. Under this strategy, urban sector threats are reduced in 16 otherwise vulnerable and 26 threatened cities, whereas 9 cities remain non-threatened. Reallocation from agriculture leaves 28% of the 71 cities (including the cities-of-concern Ouagadougou, Charlotte, and Dublin) unable to meet urban demands. Additionally, under reallocation of irrigation water, the number of cities for which both urban and environmental demands are met increases from 18 to 28, and the number able to achieve reservoir storage security increases from 10 to 13 cities (table 3). Water available only from agricultural transfers could only support both urban and environmental demands in 3 of the 10 cities-of-concern (Guangzhou, Nanjing, Wuhan) that become more vulnerable between 2010–2040. A combination of environmental neglect and agricultural transfers would leave only Ouagadougou vulnerable in 2040.