Data, Models and Uncertainties in the Global Water Cycle

Humanity faces many challenges in the future—not the least of these is the possibly growing scarcity of freshwater and the interrelationship between water use and use of the land surface (Gerten et al. 2013). While most of the problems and solutions to future water scarcity lie in governance and ownership of water, all assessments must be underpinned by a thorough and quantified knowledge of the global water cycle and an understanding of the interrelationships between water, climate and land surface processes (see e.g. Dadson et al. 2013).

Water is crucial to most of the services our aquatic and terrestrial ecosystems provide: agricultural production, carbon budgets (and other biogeochemical cycles), biodiversity, energy generation, industrial production and human health. Particularly hydrologic extremes play an important role—floods and droughts are pressure points on water scarcity and environmental damage (IPCC 2012). For decades there has been increasing pressures on available water in many regions of the world due to increasing water demand because of a growing population and wealth (Kummu et al. 2010), which together with the potential impacts of climate change on water availability and water demand are likely to aggravate water scarcity in the future (Schewe et al. 2014).

Increasing greenhouse gases are likely to lead to increases in temperature, a trend already observed (IPCC 2007). Higher temperatures will increase evaporation, over the oceans in particular, and hence will increase water vapour in the atmosphere (the evidence suggests the absolute humidity has increased but the relative humidity has decreased slightly, continuing the increase in evaporative demand, IPCC 2013). The increase in humidity is likely to lead to overall higher rainfall globally and the likelihood of more intense rainfall regionally (e.g. Groisman et al. 2005). The changing patterns of precipitation are far more complex than those of temperature, depending on not just the thermodynamics of the atmosphere but also on the details of the atmospheric circulations. These circulations frequently depend on the small or regional scale processes (and are difficult to simulate with climate models). It can be concluded, however, that the mean picture is likely to be one of increasing rainfall, although there will be large regions where rainfall decreases. There is a consensus that wet areas (and particularly northern latitudes) are likely to get wetter and drier areas drier (see e.g. Allan et al. 2010). The variability in space and time of rainfall makes it difficult to establish trends, though. Overall we have yet to identify a statistically robust mean increase in global precipitation, but we do see in the last few decades a trend of increasing precipitation at high latitudes and an increase in intense rainfall in some regions (IPCC 2013).

The Global Energy and Water EXchanges (GEWEX) Project is one of four core projects under the World Climate Research Programme (WCRP). GEWEX itself was set up in the early 1990s to co-ordinate international efforts to observe, understand and model the hydrological cycle and energy fluxes in the Earth’s atmosphere and at the surface. In the last three decades it has co-ordinated and encouraged the production of consistent global data sets describing the water and energy budget of the of the earth, it has sponsored a suite of extensive land surface campaigns to better characterise land-atmosphere interactions (of energy and water) and co-ordinated international modelling studies to better understand and describe these interactions and their impacts on water resources.

This paper summarizes some typical case studies undertaken by GEWEX scientists to contribute to our understanding of global water resources and presents the new GEWEX questions developed in the last 2 years to guide the future research direction of the global change research community.

To provide an unambiguous analysis of the changes and uncertainties in the global water cycle we require global data sets produced using consistent methodologies and which make use of a wide range of in situ and satellite data and modelling products. GEWEX has co-ordinated and sponsored the development of a wide range of global products. A number of consistent data sets of precipitation covering the globe (or at least the land areas) have been produced (see Dadson et al. 2013 for an overview). These gridded estimates rely on the interpolation of the network rain gauges. While many regions, such as Europe and North America, have extensive rain gauge networks, many regions have sparse networks and the estimates are often supplemented by satellite estimates and analysis from the national weather forecasting centres. Biemans et al. (2009) analyzed seven datasets and found a spread of at least 10 % for the annual global mean (see Fig. 4.1), which propagates to runoff estimates if used in hydrological models. Regionally the range in estimates can be much larger, notably in the Arctic and mountainous regions, where the spread in estimates can be greater than 100 % (see Biemans et al. 2009, Fig. 3b). Even in the larger basins precipitation uncertainty is typically 30 % of the mean. Precipitation is poorly simulated by the climate models—the mean global precipitation from the CMIP5 runs has a 22 % spread and a bias, on average, of 10 % (Trenberth pers. comm.). Again, at the basin scale this spread will be even larger. Accurate precipitation estimates are obviously fundamental to assessing water resources but are also critical in understanding flows of energy through the climate system.

Open image in new window

River discharge is monitored widely around the world. The Global Runoff Data Centre (GRDC http://www.bafg.de/) archives discharge data for almost 9,000 gauging stations worldwide, two-thirds of which have daily data. However, like the rainfall data, the spatial (and temporal) coverage is patchy with large gaps in Africa (excluding South Africa) and Southern Asia. River discharge is uniquely valuable as one of the few measures which integrate over large areas of the land surface (see e.g. Dai et al. 2009), unfortunately many discharge measurements have either been discontinued or are unavailable to the global change community. The number of stations in the GRDC data base peaks in the 1970s and many parts of Africa, south America and Asia have little new data since that time. The international community should continue strenuous effort to support countries to maintain their hydrometric networks and make their observations freely available. The interpretation of discharge measurements is complex because of the impact of human interventions: dams, extractions, transfers and land cover changes and also for some catchments ‘naturalised’ flows have been estimated.

In an attempt to provide globally consistent fields of discharge macro-scale hydrological models are often used. Global Water System Project (GWSP) and the EU funded WATCH project (e.g. Harding et al. 2011) have co-ordinated an inter-comparison of hydrological models globally (WaterMIP, Haddeland et al. 2011), making use of a new global data set of meteorological data (the WATCH Forcing Data, WFD, Weedon et al. 2011). This ensured the models used consistent driving data and a consistent terrestrial grid including a common river routing network. Eleven models were included in the intercomparison, including Global Hydrological Models (GHMs) and stand-alone versions of the land surface models commonly used in climate models (Haddeland et al. 2011). The main distinction between these two classes of models is that GHMs solve the water balance alone whereas the land surface models solve both the energy and water balances. All but one of the models (WaterGAP) was run without calibration via observed discharge data. The initial analysis was for “naturalised” conditions (Haddeland et al. 2011) i.e. excluding human influences related to land cover changes, damming, water abstraction and irrigation.

The eleven models showed a significant spread of the partitioning of precipitation into evaporation and runoff. The average annual global land surface evaporation varied between models from 415 to 586 mm/year and runoff from 290 to 457 mm/year, Fig. 4.2. There was no single cause for the spread in model outputs, although the different model treatment of snow was a major factor explaining the different shapes of the simulated annual hydrographs. Because all models use the same precipitation the range in mean annual discharge must be a reflection of the wide range in evaporation calculations. There is clearly a need to improve evaporation simulations in large scale hydrological models and provide benchmarks to validate models.

Open image in new window

Compared with observed discharge most models overestimate total annual runoff in semi arid regions (e.g. in the Oranje River, see Fig. 4.2)—probably a result of both water extractions not being included in this phase of WaterMIP, and wetland evaporation, typically not being included in these models. Interestingly the runoff for the Brahmaputra was under-estimated—this is probably a result of the underestimate of precipitation in the Himalayan region. In the Brahmaputra River basin temperature based evapotranspiration schemes resulted in less runoff than evapotranspiration schemes also taking radiation into account. In other basins, e.g. the Amazon River basin, the parameterization of the evaporation water intercepted on the canopy was found to be one of the factors causing the spread in evapotranspiration and runoff estimates.

The climate modelling community has a long history of systematic model intercomparison through the climate model intercomparison projects (CMIPs; Meehl et al. 2007). The impact modeling community has only recently started assessing future large-scale changes in land surface water fluxes and available water resources in multimodel frameworks (Hagemann et al. 2013; Schewe et al. 2014). The results show that climate change impacts do not only depend on emission scenarios and climate models, but that different impact models give considerably different results. However, the results also exhibit a number of large-scale features. In particular, discharge is projected to increase at high northern latitudes, in eastern Africa and on the Indian peninsula, and to decrease in a number of regions including the Mediterranean and parts of North and South America (Hagemann et al. 2013; Schewe et al. 2014). In other parts of the globe, however, the projections are subject to a large spread across the ensemble (Hagemann et al. 2013; Schewe et al. 2014). In addition to climate change, humans alter the water cycle by constructing dams and through water withdrawals. Multimodel analyses of climate change and direct human impacts on the terrestrial water cycle indicate that direct human impacts on the mean annual water fluxes in some regions, e.g. parts of Asia and in the western United States, are of the same order of magnitude, or even exceed impacts to be expected for moderate levels of global warming (+2° K) (Haddeland et al. 2013).

Evaporation¹ (ET) is a key component of the global hydrological and energy cycle. Together with precipitation, ET determines to a large extent the availability of soil moisture. In turn land cover and the availability of soil moisture determine the evaporation. Evaporation is difficult to measure in a consistent way and so, despite its importance and unlike rainfall and discharge, there is not extensive network of evaporation measurements (the nearest is the FLUXNET network (http://fluxnet.ornl.gov/introduction), with just over 500 sites). It is only in the last decade that consistent global estimates of evaporation have become readily available, because of the lack of an extensive network of in situ measurement such estimates rely heavily on a combination of satellite and modelling products. GEWEX has sponsored the development of these products through its LandFlux project (http://wgdma.giss.nasa.gov/landflux.html) and Regional Hydrological Projects (http://www.gewex.org/projects.html).

Current Earth System models show a large variability in ET estimates (Mueller et al. 2011) and a good benchmark data set at the global scale is still unavailable. The key components of ET (i.e. transpiration through plant’s stomata, bare soil evaporation and evaporation of vegetation-intercepted water) vary globally and in time. Interception loss (the evaporation directly from free water on vegetation following rainfall), is often not included explicitly in evaporation estimates, it will be particularly important in forest (responsible for the evaporation of approximately 13 % of the total incoming rainfall over broadleaf evergreen forests, 19 % in broadleaf deciduous forests, and 22 % in needleleaf forests—according to estimates using a novel satellite driven way of calculating the interception loss of vegetated land surfaces, Miralles et al. 2011). Thus in high rainfall regions the high interception losses from forest will lead to higher total evaporation compared with grassland or arable crops.

In the framework of the LandFlux-EVAL initiative (www.iac.ethz.ch/url/LandFlux-EVAL), several ET datasets based on observations, diagnostic datasets, land surface models and re-analyses are evaluated. We find that recent declining trends in terrestrial evaporation (Jung et al. 2010) are corroborated but suggest that these may be related to rainfall variability arising from El Nino/La Nina cycles (El Nino conditions are associated with negative anomalies of ET and soil moisture in most of the tropics and southern hemisphere), Miralles et al. (2013). Future climate scenarios suggest a possible increase in El Nino like activity, Collins (2005), further emphasising the possibilities of increased water stress in the future.

Overall the data sets evaluated suggest a mean global evaporation of 1.5 mm per day. This estimate is somehow lower than previously existing estimates. There is, however, still considerable uncertainty attached to all of these estimates. The input of reliable precipitation remains one of the key uncertainties in the current ET products. Further assimilation of soil moisture may help to reduce these uncertainties.

Droughts have profound impact not only on water supply but food production, biodiversity and human well being in general. Both GEWEX and WCRP have identified hydrological extremes as a major topic requiring research to understand better the underlying mechanisms and possibilities for future changes. Droughts take many forms: meteorological, hydrological, agricultural etc., but underlying all are a prolonged scarcity of rainfall, usually exacerbated by increased potential evaporation and water extractions (see e.g. Teuling et al. 2013). It is generally accepted that along with increasing rainfall intensity climate change will also bring increasing occurrence of drought (see e.g. IPCC 2013). Drought can be quantified in many ways, depending on a studies purpose and perspective. A simple index commonly used is the Palmer Drought Severity Index which makes a balance between precipitation and evaporation. Using this index there has been a suggestion that drought occurrence has already increased through the 20th century (Dai et al. 2004) but the magnitude of this change has been shown to depend on the methodology used to estimate evaporative losses (Sheffield et al. 2012), hence suggesting that prediction of future drought depends critically on its definition and method of calculation.

To assess the impact of climate change on hydrological droughts a multi-model experiment was undertaken, including seven Global Impact Models (GIMs) driven by climate data from five Global Climate Models (GCMs) from CMIP5 under four different Representative Concentration Pathways (RCPs), Moss et al. (2010). Drought severity was defined as the fraction of land under runoff deficit (runoff less than a drought threshold) and is a measure of the time-integrated effect of several interlinked processes and stores, including precipitation, evaporation and soil moisture storage. Results show a likely increase in the global severity of drought at the end of the 21st century, with systematically greater increases for the RCPs describing stronger radiative forcings. Under RCP8.5 (the most extreme), droughts exceeding 40 % of the non-arid parts of the land area are projected by nearly half of the simulations. This increase in drought severity has a strong signal-to-noise ratio at the global scale, but Southern Europe, Middle East, South East United States, Chile and South West Australia are identified as possible hotspots for future water scarcity issues. The uncertainty due to GIMs is greater than that from GCMs, particularly if including a GIM that accounts for the dynamic response of plants to CO₂ and climate, as this model simulates little or no increase in drought frequency. This analysis demonstrates that different representations of terrestrial water cycle processes in GIMs are responsible for a much larger uncertainty in the response of hydrological drought to climate change than previously thought. When assessing the impact of climate change on hydrology it is hence critical to consider a diverse range of GIMs to better capture the uncertainty associated with the models (Prudhomme et al. 2014).

The current era of the “Anthropocene” is characterized by multiple pressures on global freshwater resources. Especially the continuing population growth and associated growing demand for water-consumptive goods (such as food for humans and livestock) has already led to over-exploitation of surface and groundwater resources in many locations. As population growth and also lifestyle changes toward more water-demanding products are very likely to continue in the future, water resources will be exposed to even larger pressures, not least because anticipated climate change is about to reduce water availability in many regions, particularly some semi-arid regions (Mediterranean, Western USA, Southern Africa and North-eastern Brazil) where water is already scarce (Kundzewicz et al. 2007). A pressing question is whether there will be enough (both ‘blue’ and ‘green’) water resources to produce the food for a growing world population under conditions of ongoing global warming and associated precipitation changes (Rockström et al. 2007; Falkenmark and Lannerstad 2010; Gerten et al. 2011; Wada et al. 2013).

Besides data-based approaches, recent research makes use of global hydrological, vegetation and/or crop growth models to address questions of this kind (see e.g. Hoff et al. 2010; Elliott et al. 2014). Recent applications include assessments of future water resources and supply for different levels of mean global warming using either a single model (e.g. Gosling et al. 2010; Gerten et al. 2011, 2013), or comparing results from up to 12 global hydrological and land surface and ecosystem models (Davie et al. 2013; Schewe et al. 2014). Such studies show that already today there is high water scarcity in many regions (for example North Africa, the Middle—East and South Asia, Gerten et al. 2011), but that climate change—even if limited to a mean global warming of 2 °C above preindustrial levels—would increase the number of people living in water-scarce river basins or countries by several hundred millions. For example, using the dynamic global vegetation and water balance model LPJ mL (Rost et al. 2008), Gerten et al. (2013) found that solely due to climate change, an additional of 8 % of global population will live in water-scarce catchments for a +2 °C world, rising to 13 % for a +5 °C world. Projected increases in world population will increase this number strongly, which indicates the challenge to ensure water security, and food security alike. Gerten et al. (2011) found that the blue and green water resources (the latter defined as the evapotranspiration during the growing season on current cropland and partly on grazing land) of many countries is not sufficient to produce a ‘balanced’ diet of 3,000 kcal per capita per day. This was found for large areas of North Africa, the Middle East and the Indian subcontinent, where imports of food and underlying virtual water appear to be a necessity. Rising atmospheric CO₂ concentration, however, is potentially beneficial to plant water use efficiency and hence total yields, as has been shown e.g. for future projections of worldwide irrigation water demand (Konzmann et al. 2013; Elliott et al. 2014). An assessment of the potential of various options to close the emerging water-for-food gaps, among them more effective on-farm crop water management (such as harvesting runoff and suppression of bare soil evaporation) suggest that substantial increases in crop production on existing farmland is possible. However, it appears likely that further cropland expansion and virtual water trade is inevitable (Rost et al. 2009; Fader et al. 2013; Elliott et al. 2014).

See also http://www.gewex.org/pdfs/GEWEX_Science_Questions_final.pdf.