Multiple stressors on water availability at global to catchment scales: understanding human impact on nutrient cycles to protect water quality and water availability in the long term


Ann Louise Heathwaite, Centre for Sustainable Water Management, Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, U.K. E-mail:


1. Freshwater systems are subject to multiple stressors that include changing climate, changing land use, changing demands on water resources and changing nutrient cycles. Global trends suggest these stressors that impact on water availability will increase over the coming decades, and – without action – will constrain opportunities to sustain ecosystem services to deliver the Millennium Development Goals.

2. Although a key ‘service’ freshwaters provide is buffering inputs from the land system, predicting ecosystem response through observation and modelling is complex because nonlinear and dynamic interactions amongst a large number of constituents operate to regulate biogeochemical transformations in freshwater systems.

3. Reductionist approaches have been successful at unravelling many of the processes and some of the interactions in freshwater systems. However, reductionist approaches cannot provide the concepts or methods to understand how system properties will emerge in response to a changing climate (particularly the changing spatial and temporal distribution of precipitation); to the consequent change in water availability and water quality in the context of social drivers on the demand for water; and to feedbacks arising from nutrient cycling across a range of scales.

4. This study presents a review of the evidence for landscape-scale filtering of nutrient delivery to receiving waters and examines the role of the internal processing of nutrients at critical interfaces such as the hyporheic in attenuating nutrient loads. Analysis of research on the spatial scales and time step appropriate for catchment and water quality observations indicates the importance of small scale and short time step data for unravelling nutrient cycling in freshwater ecosystems.

5. Thematic implications: leading work in the catchment and aquatic sciences on the importance of diffuse nutrient losses from land, and on nutrient cycling in freshwaters, in governing water quality and protecting water availability, is making an increasing contribution to mainstream environmental science. Critically, the evidence base is starting to grow to inform policy-related debates with respect to food security, to climate change adaptation and for sustaining ecosystem services in freshwater environments.

Introduction: multiple stressors on water availability

Freshwater is a heavily exploited and highly managed natural resource delivering many services and meeting multiple functions. Freshwater ecosystems, as well as housing unique and diverse biota, provide ecosystem goods (e.g. drinking water, fish, electricity) and services (e.g. detoxification and purification of water and nutrient cycling, flood mitigation, recreation). Three major global forces are driving future critical uncertainties that will impact inevitably on freshwater systems: growing populations, growing economies and growing environmental insecurity. Climate change is a key driver of environmental insecurity that is inseparable from the complex interrelationship between food security, energy security and water security (Beddington, 2008). Water security is linked to water availability, and only approximately 25% or 12 000 km3 year−1 of the global total river runoff and groundwater recharge is available for human use (after subtracting uncaptured storm runoff), and nearly 40% (or 5000 km3 year−1) of this fraction is already being withdrawn for human use from rivers, lakes and groundwater (Table 1). Water also supports the natural ecosystems on which we depend. Yet sufficient regard for ecosystem requirements for water or of the value that freshwater ecosystems provide in terms of ecosystem services is rare.

Table 1.   Stocks and flows of global water
  1. 1 km3 = 109 m3 = 1012 L.

  2. Source: Holdren (2008) The author defines ‘available’ river runoff and groundwater recharge = runoff plus groundwater recharge minus uncaptured storm runoff minus runoff in remote areas.

  3. Average withdrawals for human use are estimated for 2007. The per capita withdrawals are for 2000.

Global Stockskm3
 Oceans1 400 000 000
 Ice30 000 000
 Groundwater10 000 000
 Rivers and Lakes100 000
 Atmosphere10 000
Global flowskm3 year−1
 Precipitation on land120 000
 Evaporation from land70 000
 Total river runoff and groundwater recharge50 000
 Available river runoff and groundwater recharge12 000
 Average withdrawals for human use5000
Global flows per capitam3 per person per year
 Available river flow/groundwater recharge/global population1800
 Global average per capita withdrawals800

Climate adaptation measures linked to the forecast changes in water availability tend to focus on the demand end of the water balance. Popular measures include water metering and attempts to influence water consumption patterns and social behaviour (e.g. House of Lords Science and Technology Committee, 2006; All Party Parliamentary Water Group 2008). To deliver sustainable solutions that address the multiple stressors on water availability and protect ecosystem services, we need to get to grips both with changes in land use and the supply end of water availability at scales from catchment to regional, and with the contingent changes in terrestrial and freshwater biogeochemical processes.

This article evaluates the multiple stressors on water availability and water quality at global to catchments scales using the example of diffuse nutrient pollution to illustrate the complexity and dependencies linking land to water. The concluding sections attempt an assessment of the critical future drivers of pressures on water availability and the implications for freshwater systems.

Why water quality is important in sustaining water availability

Balancing the ecological flow requirements of rivers with water demand and ensuring that groundwater extraction does not exceed natural recharge are a major challenge (Environment Agency 2009). It is argued here that understanding human drivers of changes to biogeochemical cycles, particularly nutrient cycles, and tackling the multiple stressors that lead to diffuse pollution are important for sustaining the long-term quality and hence availability of water from rivers and groundwater. Whilst other stressors on water quality exist, notably in urban environments, the insidious nature and scale of human impact on global nutrient cycles has gone unrecorded until recently. Yet human activities have enhanced global cycles of nitrogen (N) and phosphorus (P) by on average 100 and 400%, respectively (Falkowski et al., 2000). Globally, N used in food production and in fossil fuel combustion contributes c.160 Tg N annually to the N cycle, exceeding that supplied naturally by biological N fixation on land or in the ocean (Gruber & Galloway, 2008). Phosphorus is the linchpin for metabolism in biological systems, accounting for c. 2–4% of the dry weight of most cells. Yet it is present in minute quantities in the Earth’s crust (0.09 wt %) and has no stable atmospheric gas phases. Consequently, ecosystems depend on its aqueous transfer, and because ambient pools are small in natural habitats, P flux not quantity has evolved as the critical ecological driver (Karl, 2000). Human inputs of P from sewage, industry and agriculture have hugely distorted this balance (Filippelli, 2008). Elevated N and P compromise the ecosystem services on which we depend through degradation of soil and freshwater resources (e.g. Vitousek et al., 2009) and loss of biodiversity; they affect human health through poor drinking water quality and for N, through reductions in air quality. Costanza et al. (1997) in their attempt to value of the world’s ecosystem services and natural capital calculate that nutrient cycling, relative to other ecosystem services, is the most valuable, estimated at $13 trillion per year (at 1994 prices).

Where diffuse nutrient pollution poses particular challenges for sustainable water management

Detection and attribution

Multiple stressors from both point and diffuse sources in urban, agricultural and industrial sectors compromise the quality of water resources, particularly from microbial (Kay et al., 1999; Oliver et al., 2006; Kay & Falconer, 2008), sediment (e.g. Walling, 1983; Bennett, Carpenter & Caraco, 2001) and nutrient (e.g. Paerl, 2009; Smith & Schindler, 2009) pollution, and contamination from urban runoff (Ellis, 2009) especially pesticides and heavy metals (e.g. Foerstner & Wittmann, 1981). Diffuse pollutants pose a particular problem because they are characterised by sources that are generally widespread and hard to detect and by fluxes that are highly variable in time (Cullen & O’Loughlin, 1982; Carpenter et al., 1998; Thornton et al., 1999). As a result, the delivery pathways for diffuse pollutants to receiving waters are difficult to quantify (Beven et al., 2005), and the location of impact may be some distance from the source of the problem.

To date, it has proved exceptionally difficult to demonstrate to policy makers the extent to which diffuse sources of pollution are responsible for localised water quality or habitat deterioration in parts of river reaches. The challenge is twofold: (i) obtaining adequate measurements to model appropriately the sources of water quality problems in catchments, and (ii) making predictions about the impact of investments in improvements (e.g. Stevens & Quinton, 2008). An example of the scale of the challenge for P source apportionment is highlighted in White & Hammond (2009). The uncertainties in both their estimates (and those of others) and ascription of point and diffuse source contributions to total P loads in water (and between different P fractions) are evident. This is despite the fact that in the UK, the Environmental Research Funders Forum (ERFF) estimates that we spend around £88 million on environmental monitoring each year (ERFF, 2007). Others (e.g. Slater, Mole & Waring, 2006) suggest the figure may be as high as £500 million. Because most environmental monitoring is focused on statutory requirements (c. 70% of the total monitoring spend) that are largely linked to point source inputs and/or to temporal measurements at fixed points, major knowledge gaps remain both in our understanding of the spatial variation in the pattern of delivery and the impacts of diffuse pollutants on freshwater ecosystems. Recent national and supranational legislation (e.g. Habitats Directive 92/43/EEC, Water Framework Directive 2000/60/EC) has highlighted the paucity of suitable data on diffuse pollution and ecosystem health. The launch of the ERFF UK Environmental Observation Framework (ERFF, 2008) seeks to address some of the issues surrounding environmental observations made for and by the UK. However, until a more integrated monitoring system is in place, there remains a particular policy challenge because the incremental improvements in water quality observed through measures to reduce diffuse pollution may not justify the added cost in mitigation activities. For example, the cost of tackling only the tangible aspects of diffuse pollution from agriculture in the UK is estimated to be at least £300 million per year (Pretty et al., 2003).

Nitrogen or phosphorus, or both?

Long-term and large-scale experiments point to the importance of P control to reduce water quality degradation in lake ecosystems because N-fixing cyanobacteria can meet seasonal N limitation (Jeppesen et al., 2005; Schindler et al., 2008; Schindler & Hecky, 2009), although this remains contested (e.g. Conley et al., 2009), and is less clear for river systems. In both lakes and rivers, the reduction in point source P loadings from, for example, sewage treatment works, shifted the emphasis to agricultural diffuse pollution as the perceived threat to river water quality (Haygarth et al., 2005; Heathwaite et al., 2005a,b; Edwards & Withers, 2008). Livestock-derived diffuse pollution may be particularly important (e.g. Nash et al., 2000; Sharpley, Foy & Withers, 2000; Johnes et al., 2007 and earlier work by Johnes & Butterfield cited therein; Chadwick et al., 2008). The recent emphasis for land management control measures has been on P mitigation (e.g. Cuttle et al., 2007).

It is clear that multiple stressors derived from microbial, nutrient and sediment sources operate through both point and diffuse pathways linking land to water, but source apportionment remains a challenge, with recent research (e.g. Jarvie, Neal & Withers, 2006a; Jarvie et al., 2006b; White & Hammond, 2009), suggesting that point source loads of P may be more important than previously thought; although White & Hammond relied on simple empirical modelling to estimate diffuse P delivery to water, which introduces uncertainties in the estimates (Beven et al., 2005; Wade, Jackson & Butterfield, 2008). Point sources may be relatively important at low flows when the contribution from diffuse catchment sources is lower (Bowes et al., 2005; Arnscheidt et al., 2007).

The role of landscape filtering in nutrient transfer from land to water

The landscape will filter the transfer and delivery of solutes and sediment at large scales (Dillon & Molot, 1997), possibly in a self-organised way (Kirchner, Feng & Neal, 2000). The landscape ‘spatial signature’ (Perron, Kirchner & Dietrich, 2009; Whipple, 2009) appears important in understanding how diffuse pollution is generated, then connected and finally integrated or delivered through to the drainage network (Harris & Heathwaite, 2005).

Research has shown that not all locations in a catchment, even if they have the same land use, contribute equally to the delivery of diffuse pollutants to receiving waters. This is captured in the critical source areas (CSA) concept used widely in nutrient management planning and pioneered for P by Pionke, Gburek & Sharpley (2000) and developed further for N and P by Heathwaite, Sharpley & Gburek (2000) and Heathwaite, Quinn & Hewett (2005c). Generation of a CSA requires both high source risk (i.e. fields or parts of fields) and high transport risk. The CSA thus describes where a particularly risky land use or land management activity is co-located with a high probability of connection of those risks to the river system (Sharpley et al., 2008a, 2009).

Whilst the CSA concept might be able to point to the riskiness of different parts of the landscape for pragmatic nutrient management, there is no mechanism to link this risk to delivery to water. Landscape wetness can be used as a metric of delivery potential by integrating the small-scale spatial variation in runoff generation and probability of hydrological connection linked to topography (Lane et al., 2004). The scaling properties of landscape wetness (e.g. Western, Blöschl & Grayson, 2001) appear to help explain both the variation in diffuse pollution risk between catchments and the pattern of risk within a catchment (Lane, Reaney & Heathwaite, 2009). An initial analysis linking hydrological connectivity with CSAs (Lane et al., 2006) showed that different river reaches in a catchment vary in their sensitivity to diffuse pollution inputs and that contributing areas may occur some distance from receiving waters. This suggests that focusing restoration measures on the riparian zone as described in some of the classical literature (e.g. Naiman & Décamps, 1990; Billen et al., 2006) may not deliver sustainable solutions for freshwater ecology. It may also explain in part why land management activities (e.g. spreading manures; surface compaction through overgrazing or cultivation) that occur over a large geographical area may appear to have a small local impact but when integrated, generate significant changes to the freshwater ecosystem at points in a river reach. Here, predicting and mitigating the effects of altered nutrient loading on freshwater ecosystems requires an understanding of if, where, and by how much these key nutrients limit production (Elser et al., 2007). If we can understand the spatial controls on the delivery of diffuse pollutants at the landscape scale, it may be possible to identify where in a river, restoration measures would give the best return on investment.

Observing nutrient delivery to watercourses

Figure 1 (after Jordan et al., 2007) illustrates the types of information that monitoring at different time steps reveal for a single measurement station in Northern Ireland. Both weekly and daily measurement intervals for total P fail to capture the relationship between the delivery of P to water from diffuse sources in the catchment, and river discharge: the hourly time step identified both storm-dependent and storm-independent signals in the total P time series. These data challenge prior assumptions that most P is delivered during major storm events (see also Jordan et al., 2005). Likewise, when Sharpley et al. (2008b) compared the surface runoff contributing area and stream flow and total P response for 248 storms over a 10-year period from 1997 through to 2006, they found that 93% of storm flows had a return period of <1 year and delivered 63% of the flow and 47% of the total P load. Johnes (2007) also illustrated how water quality observation at inappropriate time intervals, together with insufficient fractionation of water samples for different nutrient forms, underestimate the actual contribution to river nutrient loads, leading to a high degree of uncertainty in modelling and nutrient budgeting studies. Clearly, appropriate sampling intervals, nutrient fractionation and lengths of record are important in observing diffuse pollution trends across scales.

Figure 1.

 Time series of flow (m3 s−1) versus total phosphorus concentration (mg L−1) at different observation time steps (modified from Jordan et al., 2007).

Cummins, Cushing & Minshall (2006) suggest that the ‘distinction between pattern and noise is largely a scaling issue’. Although, as discussed by Harris & Heathwaite (2005) the ‘noise’ may contain critical signals that relate to processes and feedbacks on land and in rivers. The solution to the scaling issue is to match the scale of observation to the scale of occurrence of the observed phenomena, but this is not always possible using current technology or practicable based on cost balanced against added value. As a result, our experimental results contain contingent information from broader scales that is largely unknown (Harris 2007; Downs this issue). We need advances in sensor technologies and the development of appropriate and inclusive cyberinfrastructures to make simultaneous observations at the range of time and space scales of processes operating in freshwaters – particularly rivers.

Understanding the implications of diffuse nutrient inputs to rivers on instream processes and reactions at interfaces

The fate of diffuse pollutants entering rivers depends not only on landscape filtering of diffuse and point sources but also on internal ‘instream’ processes that may transform, immobilise or eliminate diffuse pollutants delivered from land to water. Some of the changes induced by land management stressors on landscape filtering are synchronous; for example, increased inorganic sediment together with increased nutrient loading and temperature, the combined effects of which are poorly understood (Battarbee et al., 2005). Consequently, predicting ecosystem response to diffuse pollutants is challenging because multiple factors regulate biogeochemical transformations in freshwater systems (e.g. Seitzinger, Pilson & Nixon, 1983) and through to coastal ecosystems (e.g. Sundareshwar et al., 2005), and these need to be mapped onto the multiple stressors that drive their delivery.

Attempts to quantify the relationship between nutrient cycling and stream geomorphology and the various controls on transient storage, such as hyporheic flows, macrophyte and/or river margin sediment storage, biofilms and water column turbulence, are limited (Billen et al., 2006; Ensign & Doyle, 2006; Thouvenot, Billen & Garnier, 2007). This is perhaps not surprising because internal processing in freshwater ecosystems is complex and nonlinear (e.g. Odum, Finn & Franz, 1979) and includes multiple drivers such as instream nutrient cycling (Jarvie et al., 2005; Bowes et al., 2007; Withers & Jarvie, 2008), microbial interactions (e.g. Falkowski, Fenchel & Delong, 2008) and redox processes at critical zones such as the hyporheic (Smith, 2004). These drivers are explored in later sections. All are subject to multiple stressors that include changes to nutrient and sediment loads in rivers because of diffuse and point source pollution, changes to flow regimes because of changes in patterns of precipitation and temperature, and changes in groundwater quality because of water abstraction and N pollution.

Nutrients, flow and ecological hystereses.  Nutrients and flow play a key role in the internal processing in freshwaters. Figure 2 shows some of the multi-stressors on instream cycling. Particularly important are the characteristics of the flow regime in terms of the velocity and return period and duration of disturbance (Wollheim et al., 2006). Sensitivity of river ecology to flow regime is reported by Biggs (2000), Bunn & Arthington (2002) and Riis & Biggs (2003) amongst others. For nutrients and P in particular (Reddy et al., 1999), the stores and forms of P in river sediments (e.g. Welch & Cooke, 1995; House & Denison, 2002; House, 2003; Surridge, Heathwaite & Baird, 2007; Walling, Collins & Stroud, 2008) and the bioavailability of P (Heathwaite & Dils, 2000) are important. Instream P cycling is important because elevated concentrations enhance epiphytic, epibenthic and planktonic algal growth, leading to shading and growth limitation of higher plants (Mainstone & Parr, 2002; Wade et al., 2002; Bowes et al., 2007) and potential loss of refugia for river fauna. Elser et al. (2007) showed that combined N and P enrichment in aquatic habitats produces strong synergistic effects with both N and P limitation being important in freshwater ecosystems, although N limitation is stronger in marine systems.

Figure 2.

 Key drivers of instream phosphorus processes in freshwater systems (modified from Page, 2008)

Co-dependence of flow and nutrients may drive hysteretic relationships between P status and ecological status that has been described as alternative stable states (e.g. Beisner, Haydon & Cuddington, 2003), although the appropriateness of this concept for rivers with short residence times is questioned (e.g. Hilton et al., 2006). Better knowledge of ecological hystereses will inform our assessment of ecological recovery and may offer a means to help the recovery of altered systems (Battarbee et al., 2005). To do this means addressing the challenge of scale captured by Cummins et al. (2006): ‘A challenge for lotic ecologists in the 21st century remains the integration of data rich studies at the reach (micro scale) level of restoration to entire watersheds (meso scale) and finally the coarse resolution of regional (macro scale) basin analysis.’ The development of conceptual models to explain internal nutrient processing (e.g. Meyer & Likens, 1979) addresses in part this challenge and is broadened to rivers by the nutrient spiralling concept using a set of simple rules (e.g. Newbold et al., 1981; Newbold, 1992). The nutrient spiralling concept focuses largely on longitudinal processes. Conceptual models at larger scales include the river continuum concept (RCC) originally proposed by Vannote et al. (1980); the Networks Dynamics Hypothesis (Benda et al., 2004), and the patch dynamics concept proposed by Townsend (1996). The RCC integrates landscape dynamics with instream processes and appears to capture invertebrate community functions along a linear continuum from headwaters to estuary fairly well but is perhaps less able to represent patch dynamics or the networking characteristic of most channel systems (e.g. Fisher, Sponseller & Heffernan, 2004; Montgomery, 2008). Doyle (2005) argues that the RCC concept fails to consider hydrological controls sufficiently. The bias in measurement towards first- and second-order systems may underplay the role of internal processing in larger rivers. Using network scale analysis, Ensign & Doyle (2006) suggest that larger (> third order) river systems may be equally important in recycling and buffering nutrient inputs.

Groundwater flows and hyporheic exchange.  Conceptual models of freshwater ecosystems tend to take only limited account of groundwater processes despite the importance of groundwater-fed rivers in terms of their ecology and nutrient processing (e.g. Winter, 2001; Jarvie et al., 2006a,b; Pretty, Hildrew & Trimmer, 2006), and their water resource capacity (e.g. Gooddy et al., 2006; Griffiths et al., 2006). In general, there is more focus on observing or modelling processes at longitudinal and lateral interfaces rather than vertical interfaces despite recognition of the need to integrate across all ‘ecotones’ (Ward & Wiens, 2001). Groundwater flux influences river discharge, water quality and freshwater ecology, particularly during baseflow conditions (Heathwaite et al., 2005d; Cotton et al., 2006). The riverbed may be a particularly important interface for freshwater systems. Riverbeds are transitional environments between groundwater and surface water and are both a sink and source of fine organic and inorganic sediment and associated pollutants including P through transient storage (e.g. Jarvie et al., 2005; Ensign & Doyle, 2006). Furthermore, biogeochemical hot spots in the landscape may occur at critical interfaces between terrestrial and aquatic systems (McCain et al., 2003) such as in the hyporheic or the riparian zone.

The attenuation of nutrients in the hyporheic zone (HZ) may be important for maintaining the ecological health of groundwater-fed rivers (e.g. Smith, 2004). There is considerable research on nutrient transformations in the HZ (e.g. Hill & Lymburner, 1998; Bencala, 2005). The focus in hyporheic research has tended to be on the upper riverbed sediments where biogeochemical activity may be high (e.g. Haggerty, Wondzell & Johnson, 2002). Dahm et al. (1998) showed that concentration gradients exist across the HZ that may influence the mobilisation of redox-sensitive elements such as N. Respiratory denitrification by bacteria is thought to be the main N loss mechanism in freshwater ecosystems (Downing & McCauley, 1992). Although Burgin & Hamilton (2007) suggest that other microbially mediated N loss pathways, such a dissimilatory reduction of nitrate to ammonium (DNRA) and anaerobic ammonium oxidation (anammox) may also be important (see also Revsbech, Jacobsen & Nielsen, 2005). Anammox is responsible for up to 70% of N2 formation in marine sediments (e.g. Trimmer, Nicholls & Deflandre, 2003; Voss & Montoya, 2009), yet its role in freshwater sediments is largely unknown.

Recent work (e.g. Wondzell, 2006; Krause et al., 2009) suggests the focus of hyporheic research on the upper riverbed ignores the role of groundwater flows in the chemical reactivity of the HZ. Malcolm et al. (2003) and Sophocleous (2002) also found local significant heterogeneity in the HZ described by Conant (2004) as ‘preferential discharge locations.’ Together, these findings suggest that the interaction between ground- and surface waters across the hyporheic may take place at greater depths and/or in ‘hotspots’ where groundwater flux and nutrient transformations are maximised. Davies, O’keeffe & Snaddon (2006) suggest that for parts of the year, nutrient retention in riverbeds switches nutrient cycling into ‘nodes’ similar to the hotspots proposed by Krause et al. (2009) rather than enabling downstream nutrient ‘spiralling’ (Newbold, 1992).

Future stressors on freshwater in a changing environment

The discussions above present evidence for the complexity of freshwater systems and the need for them to be interpreted in multiple dimensions and also over time. The underlying complexity arises from intertwined nonlinear and dynamic interactions amongst large numbers of constituents. The multiple stressors governing diffuse pollution inputs and instream nutrient cycling in freshwater systems are an example of some of the drivers of this complexity. Weins (2002) suggested that the actual spatial patterns of ‘connectedness’ and variations in flows and deposition that occur in river systems are far more complex than most currently available conceptual models. Consequently, although traditional reductionist approaches have been successful at unravelling many of the processes and some of the interactions in freshwater systems, they cannot provide the concepts or methods to understand how system properties will emerge in response to a changing climate (particularly the changing spatial and temporal distribution of precipitation); to the consequent change in water resource availability and water quality in the context of social drivers on the demand for water; and to feedbacks arising from nutrient cycling across a range of scales. A brief examination of the potential implications of these critical pressure points from the perspective of their impacts on freshwater systems is given below. The focus is largely on diffuse sources. For an evaluation of the pressures on urban systems, see, for example, Ellis (2009) and United Nations (2008).

Changing climate and diffuse pollution risk

Environmental systems are undergoing a period of unprecedented change. Much of the attention has been on climate and marine systems; however, there is growing evidence of changes to freshwater systems (Whitehead et al., 2006; Wilby et al., 2006). Clear signals of very large-scale human-induced perturbations of the water cycle are beginning to emerge. Barnett et al. (2008) showed that up to 60% of the climate-related trends in river flow, winter air temperature and snow pack in western United States between 1950 and 1999 are human induced. Raymond & Cole (2003) suggest that changes in weathering rates over the past 50 years as a result of changes in climate and land use are changing the chemistry of rivers. The recent IPCC Report (IPCC 2008) states: ‘Climate change challenges the traditional assumption that past hydrological experience provides a good guide to future conditions.’ Research suggests that we can no longer assume hydroclimatic stationarity as a foundation concept for models of water resource availability and management (Milly et al., 2008). Consequently, we need new non-stationary models of relevant environmental variables to predict the consequences of climate change on water availability.

Critically, for freshwater ecosystems, hydrological connectivity and, therefore, diffuse pollution risk may change under a changing climate. Over 10 years ago, Schindler (1997) was calling for greater understanding of the linkages between land and water systems: ‘Consideration of land–water interactions and interactions between climate warming and other human stresses are important for the accurate prediction of the effects of climatic change’. Science is only now starting to get to grips with this challenge. For example, to understand the implications of changes in surface hydrological connectivity under future climate scenarios for the 2080s (HadCM3 GCM medium high scenario, IPCC 2001) for diffuse pollution risk, Reaney, Lane & Heathwaite (2007) combined a physically-based hydrological model (CAS-Hydro, Conlan et al., 2005) with the treatment of hydrological connectivity described earlier and in Lane et al. (2009) and Reaney et al. (in press). The authors found the predicted changes in the volume and timing of rainfall had knock-on consequences for both the area of a catchment connected to receiving waters and the duration of connection. For freshwater ecosystems, the implication is both the timing and the magnitude of diffuse pollution input to freshwaters may increase in the future as the areal extent of critical source areas in a catchment increase. A further challenge is to understand the feedbacks from changes in land and water systems, such as increased hydrological connectivity onto climate (e.g. coupling between soil moisture and precipitation, changes in water retention capacity and soil sealing, changes in rates of microbially mediated reactions), and how these interactions change with scale.

Changing demands on water resources

By 2025, 40% world’s population could live in water scarce regions (World Water Assessment Programme 2006) as a result of rising demands for water and through degradation of available water supplies (Foster & Chilton, 2003). Globally, 18% of people have no safe drinking water, and 2.4 billion people lack access to improved sanitation. The United Nations Human Development Report ‘Beyond Scarcity’ (2006) estimates that water-related diseases kill a child every 8 s and are responsible for 80% of all illnesses and deaths in the developing world. The reason these statistics are important lies in the role freshwater systems have in the natural regulation of freshwater quality and hence water availability (Millennium Ecosystem Assessment 2005). Unfortunately, we rarely take account of the ecological requirements of freshwater ecosystems, including the ecological flow requirements of rivers, to support safe and secure water on a global scale. Ecological flow requirements define the quality, quantity and distribution of water required to maintain the components, functions and processes of freshwater ecosystems on which we depend. Although we can treat water to improve its potable quality, and we can deal with point sources – albeit at significant cost and energy use, we cannot treat the water on which freshwater and terrestrial ecosystems depend in the same way with the consequent degradation of freshwater ecosystems: the global decline in populations of freshwater species is greater than that for other species generally (WWF Living Planet Index 2004).

Water withdrawals have increased sixfold since the 1900s – twice the rate of population growth. The predicted increase in water abstraction is 50% in developing countries and 18% in developed countries by 2025, as population growth and development drive up water demand. Water use varies widely between countries and is particularly sensitive to changing patterns of climate (e.g. Economist 2008) and the global trend towards increasing urbanisation (United Nations 2008). This sensitivity has knock-on impacts on food security, because agriculture is particularly water demanding, using about 70% of the average total freshwater withdraw for human use (Beddington, 2008) or c. 3500 km3 water annually (Holdren, 2008).

Water scarcity is not just a ‘problem’ of the developing world (WWF, 2006). In Europe, 41 million people lack access to safe drinking water. In the UK, the average annual water availability per person is c. 2465 m3, which is almost the same as Sudan (2074 m3) and less than Spain (2794 m3), Italy (3325 m3) and France (3439 m3) (FAO 2005). However, the spatial distribution of rainfall in the UK mapped onto the spatial pattern of population density is where – for the UK – the real future challenge lies in terms of water availability. The areal rainfall for the South East region of England is 740 mm year−1, and the actual evaporation is estimated at 480 mm (Rodda, 2008) giving an effective rainfall of only 260 mm year−1. Combining this with the high population density of this region gives 610 m3 of water per person per year. On a global ranking, the availability of water in southeast England would be around 161st position out of 180 countries, and southeast England has less water per person than Egypt (859 m3). It is estimated that by 2020, increasing population and housing growth will increase water demand by 5% or an extra 800 million litres of water per day (Environment Agency 2009). Most measures to protect or conserve water resources remain focused on end-of-pipe solutions or measures to constrain water demand such as water metering. Too few measures tackle water availability by enhancing water conservation in catchments. This is a lost opportunity that fails to build on the ecosystem services offered by land and freshwater systems to protect and harness catchment hydrology and regulate water quality through internal nutrient processing in rivers.

For freshwater ecosystems, the pressures on groundwater resources represent perhaps the most significant threat to their sustainable use in the future (WWF, 2006). Other stressors come from land use change in the form of urbanisation, bioenergy crops and afforestation. Groundwater systems are the predominant reservoir and strategic reserve of global freshwater storage at c. 30% of the global water total and 98% of freshwater in liquid form. Globally, groundwater provides c. 50% of potable water supplies. The global withdrawal rate is 600–700 km3 year−1; making ground water, the world’s most extracted raw material. Groundwater provides approximately 70% of the piped water supply in EU. In the UK, at least 50% of the groundwater used for public supply shows significant deterioration in groundwater quality and has cost the UK water industry c. £754 million since 1975 in amelioration (Chilton et al. 2004). The costs reflect a combination of deterioration in groundwater quality and more stringent regulatory standards for drinking water. All these activities have implications for the quality of freshwater systems, because they either increase point source inputs or modify natural flow patterns. Combined with climate change impacts suggests a potential acceleration of impact on freshwater systems unless a more holistic approach to balance ecological and human needs is put into place.

Changing nutrient cycles

Profound ecosystem-level changes are now evident as a consequence of changing nutrient cycles. On a global scale, dead zones (hypoxia), for example, now cover an area of c. 500 000 km2 but predominate in coastal shelf seas important for food security (Diaz & Rosenberg, 2008). Freshwaters are important in attenuating N export to coastal waters through internal processing in wetlands (e.g. Zedler, 2003) and rivers – see previous discussions and Peterson et al. (2001) and Burgin & Hamilton (2007). Coastal systems also have an important role in retaining fluvial N inputs (Seitzinger et al., 2006), but perturbations of the N cycle may threaten this function (Duce et al., 2008). Recent research suggests elevated concentrations of reactive N derived from agricultural, vehicular and industrial sources degrade freshwaters through both acidification and base cation depletion (Galloway et al., 2008) and by distorting the balance between N and P cycling, particularly in estuarine and coastal waters (Paerl, 2009). These changes challenge the capacity of these areas to continue to act as net sinks for N and P.

Evidence is emerging to suggest the lack of integrated analysis of biogeochemical cycles – particularly nutrient cycles, is shortsighted; may lead to unforeseen consequences for land and freshwater systems through, for example, N saturation in catchments and prolonged eutrophication in receiving waters; and could result in assumptions being built into predictions of climate change impacts that are misleading. A changing climate and carbon cycle will both interact with and have implications for other biogeochemical cycles. For example, N availability will influence the capacity of the biosphere to continue to sequester C from the atmosphere (e.g. Bardgett, Freeman & Ostle, 2008), and freshwater ecosystems will have a role in this process as N cascades through the system (e.g. Gruber & Galloway, 2008). But the N cascade is not independent of changes in other cycles such as the P cycle. Highly contested is the claim that exclusive focus on P control in rivers has exacerbated N-limited downstream eutrophication in estuaries and coastal waters (e.g. Conley et al., 2009; Schindler & Hecky, 2009). Therefore, it is becoming increasingly important to understand better the co-dominant role of N and P in freshwater systems, particularly as controls of aquatic plant community structure and composition (e.g. Maberly et al., 2002; Elser et al., 2007), and to predict how this balance may change under environmental change through an integrated analysis of macronutrient cycles.

Conclusions: science for sustainable water management

The example of diffuse pollution demonstrates how multiple stressors challenge the sustainability of freshwater ecosystems under environmental change. These are complex problems that we have tended to address using reductionist rather than integrative approaches. In a changing environment, such approaches are no longer appropriate because they miss the multifaceted functions of freshwater ecosystems and their coupling with other environmental systems. To deliver sustainable water management will require a systems framework that integrates science disciplines and brings together observation and modelling at appropriate time and space scales. This is not a simple task. Meeting this challenge needs methods to understand how system properties emerge through quantitative measures of multiple components simultaneously, and by rigorous data integration with predictive models. Recent technological advances in environmental data collection, data handling and numerical and computational techniques may make this possible. Particularly important could be real-time data and high frequency measurements applied at catchment to river basin scales using distributed networks.

Amongst the many science challenges that form part of the elucidation of the multiple stressors in freshwater ecosystems, the following may be pertinent:

  • 1Can we identify where, in linking land and water systems, the main loci of control lie? Do pivot points exist in, for example, spatially defined critical areas? Alternatively or additionally, do the key controls lie in critical processes such as microbial reactions at interfaces?
  • 2River systems are complex; their nonlinear and temporally evolving properties require understanding of the way in which small-scale interactions may lead to large-scale and long-term effects. Do we have data at the right time and space scales to deliver the science to meet the challenges for the sustainable use of freshwater systems?
  • 3Do we work sufficiently well across science disciplines (e.g. biology, chemistry, hydrology, climate science, soil science and social sciences) to address the complex science and multiple stressors in freshwater ecosystems? What is the role of interdisciplinary research in advancing the research agenda?

Conflicts of interest

The authors have declared no conflicts of interest.