Population growth will in many regions increase the pressure on water resources and
likely increase the number of people affected by water scarcity. In parallel, global
warming causes hydrological changes which will affect freshwater supply for human use in
many regions. This study estimates the exposure of future population to severe
hydrological changes relevant from a freshwater resource perspective at different levels
of global mean temperature rise above pre-industrial level (
Within the 2030 Agenda for Sustainable Development of the United Nations (United Nations, 2015), “access to clean water and sanitation” is one of the 17 sustainable development goals (SDGs). For other SDGs, such as “zero hunger” and “affordable and clean energy”, access to sufficient water resources is a precondition (International Council for Science, 2017). Already today, more than 2 billion people live in countries where total freshwater withdrawals exceed 25 % of the total renewable freshwater resource (United Nations, 2017). Population increase and economic development are expected to further increase pressure on water resources leading to enormous challenges for water resource management to maintain or increase water supply. Climate change potentially aggravates this challenge in some regions by altering precipitation patterns in time and space, increasing atmospheric demand, or accelerating glacial melt, to name just a few. Such changes can lead to a reduction in total physical water availability, but also a change in the flow regime, which may lead to more frequent or more severe drought events or an increased risk of flooding (Döll and Schmied, 2012). All these changes affect water supply management and will make meeting the demand and achieving SDGs more costly or impossible.
As of April 2017, 194 countries responsible for > 99 % of
global greenhouse gas emissions have signed the Paris climate agreement that
aims at “holding the increase in the global average temperature to well
below 2
Unlike most global assessments of climate change impacts on water resources,
which have employed a measure of water stress like the water crowding index
(WCI; Falkenmark, 1989) or the withdrawal-to-availability ratio
(WTA; Raskin et al., 1996), here we analyse
hydrological changes relevant from a water resource perspective directly.
This allows us to focus on climate-induced hydrological change alone
(unobscured by the effects of population change) and to include aspects of
hydrological change important from a water resource perspective other than mean annual discharge (MAD), on which both WCI
and WTA are based. In order to gain a detailed and comprehensive
understanding of changes in the water sector, this study analyses climate
impacts with respect to a decrease in mean water availability, growing
prevalence of hydrological droughts, and an increase of flooding hazards. To
estimate these hydrological changes, three key metrics are used to assess
flow regime changes: (i) MAD, (ii) the average number of drought months per
year (ND), and (iii) the 10-year flood peak (Q10). Severe hydrological change
is defined as crossing a critical threshold (defined below) for at least one
of these key metrics. By combining these changes with spatially explicit
population projections consistent with shared socio-economic pathways (SSPs; Jones and O'Neill, 2016),
the number of people exposed to severe hydrologic changes is estimated for
each level of
However, looking at the total number of people affected by severe hydrological change provides only limited insights into the consequences of severe hydrological change and the challenges for adaptation. These are greatly determined by the underlying population-driven water-scarcity level; that is, when options for supply-side management are exhausted or become too costly under water-scarcity conditions, the focus of water management has to shift towards demand management (Falkenmark, 1989; Ohlsson and Turton, 1999). Thus, adaptation to severe hydrological change under already water-scarce conditions will also have to involve demand-side management strategies to prevent negative social and economic consequences. Because demand-side options are complex and their implementation is faced with behavioural, economic, political, and institutional obstacles (Kampragou et al., 2011; Russell and Fielding, 2010), adaptation to severe hydrological change is more challenging under already water-scarce conditions. To account for this aspect, we apply the WCI to estimate the future population pressure on water resources under the assumption of no climate change and jointly analyse climate-induced severe hydrological change and population-driven water scarcity.
For the estimation of future population affected by severe hydrological change and to calculate WCI, we use spatially explicit population projections from Jones and O'Neill (2016, 2017). These are based on the SSP national population projections (KC and Lutz, 2017) and have been downscaled making additional assumptions on urbanization consistent with the respective SSP storyline. The five SSP storylines are designed to cover a broad range of future socio-economic development pathways with plausible future changes in demographics, human development, economy, institutions, technology, and environment (O'Neill et al., 2017). However, they do not account for the impact of climate change on development pathways. For this study we only use the population projections of the SSPs. The analysis focuses on the middle-of-the-road scenario SSP2 (with a total population of 9.0 billion in 2100), but we use the other scenarios (with a total population between 6.9 and 12.6 billion in 2100) to test the sensitivity of our findings to different population scenarios.
In order to systematically assess climate change impacts on freshwater
resources, we use the PanClim climate scenarios described in
Heinke et al. (2013a). The dataset consists of 8 different scenarios of
A total of 152 climate scenarios (8
For assessing the impacts of climate change on the hydrological cycle, we employ the LPJmL Dynamic Global Vegetation Model version 4 (LPJmL4) that simulates the growth of natural vegetation and managed land coupled with the global carbon and hydrological cycles (Schaphoff et al., 2018a, c). The model has been extensively evaluated showing good performance in representing the global hydrological cycle (Rost et al., 2008; Schaphoff et al., 2018b). LPJmL has been widely applied in water resource assessments (Gerten et al., 2011; Jägermeyr et al., 2016; Rockström et al., 2014; Steffen et al., 2015).
For the simulations conducted here, the model is first run without land use for a spin-up
period of
In addition to the 152
The focus of this study is on hydrological changes due to climate change that are relevant from a water resource perspective. “Water resources” refers to “blue” water – the water that can be withdrawn from rivers, lakes, and aquifers, and which can be directly managed by humans – as opposed to “green” water, i.e. the soil moisture in the root zone from local precipitation that can only be used by locally growing plants (Rockström et al., 2014).
Here we use river discharge as an approximation of the blue water resource. River discharge is simulated in LPJmL by means of a linear storage cascade (Schaphoff et al., 2018a) along a river network defined by the Simulated Topological Network (STN-30p) flow direction map (Vörösmarty et al., 2000, 2011). The simulated discharge of a grid cell includes all the water that enters the cell from upstream areas and all surface and subsurface runoff generated within the cell. Although water is often withdrawn from lakes and aquifers, no more than the possible recharge to these storages can be withdrawn over a prolonged period. Therefore, river discharge as computed with LPJmL represents a good approximation of the total renewable blue water resource (excluding non-renewable fossil groundwater from aquifers with very long recharge times).
Three metrics relevant from a water resource perspective, i.e. mean annual discharge
(MAD), the number of drought months per year (ND), and the 10-year flood peak (Q10), are
calculated for each grid cell for the 8 levels of
Changes in MAD are used as a measure for changes in mean water availability, assuming that a substantial decline in MAD will make it difficult to satisfy existing and future societal water demands with the existing water supply infrastructure. We define a decrease in MAD by 20 % or more as a severe hydrological change that requires some form of management intervention (either on the supply or the demand side). The same threshold was also used by Schewe et al. (2014) to define severe decrease in annual discharge.
The occurrence of prolonged periods of below-average discharge, mostly initiated by inter-annual climate variability, is referred to as hydrological drought. To provide stable water supply to society, water supply systems are adjusted to seasonal variability and drought regimes. A substantial increase in drought periods thus impairs the capability of existing water management infrastructure.
We apply a drought identification method proposed by van Huijgevoort et al. (2012) to determine which months of a monthly time series of river discharge are in drought condition. The method is based on a combination of the threshold level method (TLM) and the consecutive dry month method (CDM). The TLM method classifies a month as drought-stricken if discharge falls below a given threshold (here the month-specific discharge that is exceeded 80 % of the time). However, in ephemeral rivers a method that accounts for the duration of dry periods is more appropriate since the TLM would classify all months with zero flow as drought. We adopt this combination of TLM and CDM from van Huijgevoort et al. (2012) but make some modifications to obtain a more robust and plausible algorithm. First, a month-specific discharge threshold is applied to identify drought months according to the TLM method. Then, if the TLM threshold is zero and the number of drought months in a given calendar month (e.g. January) exceeds 20 %, the CDM is used to determine which of the months with zero discharge can be classified as drought months. To this end, the number of preceding consecutive TLM droughts is determined for each month with zero discharge in the given calendar month. Finally, a threshold is selected that retains only the months with the longest preceding dry period so that the total number of drought months in that calendar month is 20 %. The TLM and CDM thresholds are determined from the reference simulation representing present-day climate conditions. These thresholds are then used to estimate the number of drought months for all climate scenarios. Note that the thresholds are derived from and applied to the continuous 30-year time series, which allows for the detection of multi-year droughts.
We define an increase in the average number of drought months per year (ND) by 50 % (i.e. from 20 % to 30 %) as a severe hydrological change that will require an upgrade of existing water management systems to maintain a reliable water supply.
All water supply infrastructure should be designed to withstand typical flooding events. A flood with a return time of 50–100 years (Q100) is typically used as a reference case (Coles, 2001). However, spillways of critical infrastructure such as dams and reservoirs are designed for even more severe flood events, with a return time of 1000 years or more (Dyck and Peschke, 1995). An increase in the magnitude of floods poses a serious threat to water management systems with potentially disastrous consequences.
The magnitude of extreme events with long return periods is usually derived from much
shorter observed time series of annual maximum floods by fitting a suitable extreme value
distribution (e.g. a Gumbel or generalized extreme value, GEV, distribution; Coles 2001).
The obtained extreme value distribution is then used to extrapolate the magnitude of
flood events with long return periods. This procedure can also be used to detect changes
in the magnitude or the return time of such events from two fitted extreme value
distributions (Dankers et al., 2013). However, fitting a GEV distribution to 5-day
average peak flow estimates form LPJmL using
We use the cases where a good fit of the GEV to data was achieved to assess how well the estimated changes in directly derived Q10 can be used as a proxy for changes in events with a higher return time (Q100 or Q1000) derived from GEVs. Because the overall goal is to detect a severe increase in Q100 or Q1000, we estimate how many false positives and false negatives occur when a threshold of 20 % or 30 % increase, respectively, in Q10 is used. False positives are defined as increases in Q10 by more than 20 % or 30 %, which does not coincide with an increase in Q100 or Q1000 by at least 10 %; false negatives are defined as an increase in Q100 or Q1000 by more than 50 %, which do not coincide with an increase in Q10 by at least 20 % or 30 %. For Q100, we find that a threshold of 20 % for Q10 produces 6.3 % and 4.7 % of false positives and negatives, respectively; a threshold of 30 % produces 2.6 % and 11.0 % of false positives and negatives, respectively. For Q1000 the figures are much higher with 15.9 % (10.7 %) of false positives and 33.8 % (47.0 %) of false negatives for a threshold of 20 % (30 %) for Q10. This demonstrates that Q10 can be used as proxy to detect severe changes in Q100 with reasonable accuracy but not to detect severe changes in Q1000.
We give the avoidance of false positives a higher priority to obtain conservative estimates of flood hazard increase. Therefore, we choose an increase in Q10 by 30 % as a threshold to detect a severe increase in flooding hazard that needs to be addressed by investment in enhancing flood resistance of water supply infrastructure or by changing reservoir operation schemes to increase the safety buffer for flood protection (at the cost of storage capacity for water supply). However, it needs to be kept in mind that this indicator only detects about half of the increases in Q1000 by more than 50 %, which can be particularly harmful to water management infrastructure.
In order to determine where transgressions of severe hydrological change thresholds in the three metrics matter most, we estimate which part of the global population is experiencing water stress in the absence of additional climate change. We use the WCI originally proposed by Falkenmark (1989) to assess different levels of population pressure on water resources. Originally, the water crowding index was applied at the country scale, which may hide important within-country variations (Arnell, 2004). With improved spatial resolution of population data and a desire to use natural hydrological units, instead of administrative boundaries, it has become more common to calculate WCI at the basin scale (Arnell and Lloyd-Hughes, 2014; Falkenmark and Lannerstad, 2004; Gerten et al., 2013; Gosling and Arnell, 2016). In this paper, we develop a new calculation procedure to obtain a measure of water crowding that can be calculated and interpreted at the grid-cell scale. This can then be combined with the simulated hydrological changes at the grid-cell scale to estimate hydrological change for different levels of water crowding.
To calculate the effective population pressure on the total available water
within each grid cell, we treat local (within grid cell) runoff and the
inflow from each upstream cell
Five different WCI levels can be distinguished, each characterized by a
different degree of water scarcity (Falkenmark, 1989). WCI below 100 people
per flow unit (p/fu; 1 fu
Spatial
pattern of water crowding in 2010
Between 1950 and 2010 the number of people that live with
This trend is projected to continue in the future under all five SSP
population scenarios (Fig. 1c and d). The total number of people living
under
Under the majority of climate change patterns within the range of
If global warming was limited to 2
To get an indication of the adaptation challenges associated with the exposure to severe
hydrological change, we use the assessment of future water scarcity due to population
change to distinguish two principal adaptation domains. Coping with water scarcity
conditions (WCI > 1000 p/fu) even without further aggravation by climate
change requires a combination of supply-side and demand-side management measures
(Falkenmark, 1989; Ohlsson and Turton, 1999). Therefore, water
demand management interventions will also have to play a role in the adaptation to severe
hydrological change under already water-scarce conditions. In contrast, adaptation to
severe hydrological change under comparatively abundant water availability conditions
(WCI
Under the assumption of no climate change, as much as 3.30 billion people (36.8 % of
global population) are estimated to live under absolute water scarcity by 2100 in the
SSP2 scenario. For all aspects of severe hydrological change and across the whole range
of
Fraction of SSP2 population in 2100 exposed to severe hydrological change at
different levels of
Because of the challenges associated with the implementation of demand-side management
interventions, the population already experiencing water scarcity in the absence of
climate change is of primary concern when analysing exposure to severe hydrological
change. We estimate that 2.14 billion people (23.9 % of global population) in the
SSP2 population scenario would be affected by water scarcity due to population change and
Number of people in 2100 for the SSP2 population scenario that would
experience absolute water scarcity (> 1000 p/fu) under
present-day climate conditions and be
The remaining number of people exposed to severe hydrological change at 2
Failure of the Paris Agreement would substantially increase exposure to severe
hydrological change in many regions. In 5 out of 10 regions, the number of people
affected by
Although numbers differ among population scenarios, the overall pattern of where and how much change occurs in the different regions is consistent across all SSP population scenarios. A comprehensive overview of population under high water crowding and affected by severe hydrologic change in different world regions for all population scenarios is given in Fig. S5.
Our estimate that 26.8 % of global population today live under
A direct comparison of hydrological changes estimated here to previous studies is not
straightforward due to the unique design of this study. Only few global studies have
assessed climate change impacts on water resources as a function of
In summary, the global and regional estimates of population living under
Apart from these uncertainties in model projections, the results of a global study like ours are necessarily determined by simplifications and generalization in the data analysis. The most important generalization in this study are the choice of aspects of severe hydrological change and the corresponding critical thresholds. While not all selected aspects may be relevant in all cases (e.g. where supply is primarily fulfilled from groundwater), we believe that in the vast majority of cases they reflect important hydrological properties that are relevant from a freshwater resource perspective. The respective thresholds may also differ depending on hydrological and other local conditions, and using unique global values will always produce a number of false positives and false negatives. However, the selected thresholds are rather conservative, and thus are expected to produce more false negatives than false positives. Another aspect is the choice of the WCI to differentiate population groups in terms of adaptation challenges. This indicator is widely applied because it only requires data on mean water availability and population numbers, but it can neither account for hydrological aspects that limit the utilization of water resources nor for actual per capita water requirements. Despite these shortcomings of the WCI, it gives a rough impression of the overall population pressure on water resources, which is linked to the challenges to adapt to severe hydrological change. Last but not least, it is important to note that this study only addresses quantity aspects of freshwater resources and does not consider water quality.
Future freshwater supply will be affected by population growth and climate change, which
are both subject to uncertainty and heterogeneous distribution patterns. Under all five
SSP population projections considered here, a strong increase in the number of people
living under
If global warming would continue unabated to reach 5
Due to the heterogeneous spatial distribution of
The CRU TS3.1 historical climate data are available
from
The supplement related to this article is available online at:
WL and DG formulated the overarching research goal; JH designed the study, performed all model runs, and carried out the analysis; JH, CM, and ML discussed results; JH wrote the manuscript with contributions from all co-authors.
The authors declare that they have no conflict of interest.
This article is part of the special issue “The Earth system at
a global warming of 1.5
This work was in part supported by German Federal Ministry of Education and Research (BMBF) through the project SUSTAg (031B0170A). The authors would like to thank the two anonymous referees for helpful comments on the manuscript.The article processing charges for this open-access publication were covered by the Potsdam Institute for Climate Impact Research (PIK).Edited by: Rolf Aalto Reviewed by: two anonymous referees