Long-term climatic change and sustainable ground water resources management

Atmospheric concentrations of greenhouse gases (GHGs), prominently carbon dioxide (CO2), methane (CH4), nitrous oxide (N2O), and halocarbons, have risen from fossil-fuel combustion, deforestation, agriculture, and industry. There is currently heated national and international debate about the consequences of such increasing concentrations of GHGs on the Earth’s climate, and, ultimately, on life and society in the world as we know it. This paper reviews (i) long-term patterns of climate change, secular climatic variability, and predicted population growth and their relation to water resources management, and, specifically, to ground water resources management, (ii) means available for mitigating and adapting to trends of climatic change and climatic variability and their impacts on ground water resources. Long-term (that is, over hundreds of millions of years), global-scale, climatic fluctuations are compared with more recent (in the Holocene) patterns of the global and regional climates to shed light on the meaning of rising mean surface temperature over the last century or so, especially in regions whose historical hydroclimatic records exhibit large inter-annual variability. One example of regional ground water resources response to global warming and population growth is presented.


Introduction
Few topics attract as much attention today as 'global warming' does. The latter term has become synonymous with the increasing trend of the Earth's mean surface temperature (defined as the average of near-surface air temperature over land and sea surface temperature) observed over the last century, and, more noticeably, post 1950. A position on the causes of global warming so defined and its probable consequences on the world's population and the global environment has been summarized in the United Nations' 2007 Intergovernmental Panel on Climate Change (IPCC) report. The IPCC (2007) report stated that human activities (mainly fossil-fuel combustion, deforestation, agriculture, industry) since the onset of the Industrial Revolution (circa 1750) have raised the atmospheric concentrations of GHGs above those that would be expected by natural causes alone. Furthermore, the IPCC (2007) report stated with very high confidence that human activities since 1750 have raised the global mean surface temperature above the level that would be expected by natural warming that followed after the last ice age. Besides the central role that mean surface temperature has on terrestrial and oceanic organisms and on essential physical and chemical environmental processes, the IPCC (2007) report cites heightened sea level rise and the melting of ice and snow as effects of global warming inferred from observations worldwide. Many other effects of global warming are identified or implied with various degrees of confidence in the IPCC (2007) report, among them changes in regional precipitation and evaporation.
Global warming driven changes in regional precipitation and evaporation have direct consequences on hydrologic processes and water resources (rivers, lakes, aquifers, wetlands, springs) through their possible shifts of basin-wide water balances. Recall the water-balance equation at the river basin or aquifer spatial scales, in which the change in water storage S during an arbitrary time interval ( t) is determined by the magnitude of the water fluxes affecting storage, namely, precipitation (P), evapotranspiration (ET), runoff (surface plus subsurface), and net human withdrawal (D): Changes on the regional water balance effected by global warming, climate-modified P and ET would trigger shifts (represented by the symbol δ) in all the terms appearing in equation (1): Equation (2) suggests that a warming climate could trigger a complex and interrelated set of changes in water storage, precipitation, evapotranspiration, runoff, and net human withdrawal that would affect all aspects of human activity and water-related environmental processes by offsetting water availability (expressed as δ P-δET, say, for plants, or δ Q for aquatic organisms, or as δ D for human consumption or agriculture). A voluminous body of literature has been produced over the last 30 years on a variety of examples of hydrologic change embodied by equation (2), see, for example, a review of global warming and the hydrologic cycle (Loáiciga et al 1996, Loáiciga 2003, global-scale predictions of water resources change driven by global warming and population change (Vörösmarty et al 2000), continental-scale predictions of hydrologic change (de Witt and Stankiewicz 2006), detection of global changes in runoff (Milliman et al 2008), and overall reviews of global warming-humanenvironment interactions (Houghton 2004, IPCC 2007. The current state of long-range climate prediction (say, 100 years hence) is beset by the complexity of the Earth-atmosphere system's response to heightened post-1750 concentrations of GHGs (see, Hansen et al 2005) and by the uncertain nature of human adaptation to global warming, which, in turn, may affect future concentrations of GHGs (Crowley 2008, Pielke 2008. The latter statement takes greater relevance in judging the accuracy of long-range predictions of regional (say, at the river basin or aquifer spatial scales, Arnell et al 2003, Pielke 2008) climatic change. Figure 1 exemplifies the sensitivity of climate evolution to small differences in initial conditions. Global mean surface temperature data were fitted by Tsonis (1991) with an autoregressive model, which was then used to simulate temperature departures from the historical mean through year 2100.
Two time series of temperature departures were simulated and are shown in figure 1, the only difference between the simulated time series being a 1% difference in the initial condition used to start the simulations. Notice that the two simulated time series in figure 1 diverge from each other in magnitude and timing of lows and highs after about 40 years of simulation. Long-term, transient, simulations of the Earth's climate with coupled atmosphere-ocean general circulation models (AOGCMs) are not immune to the sensitivity of climate prediction to initial conditions (akin to the so-called 'butterfly effect' of weather prediction, see Lorenz 1964). To compound matters, AOGCMs, by necessity, parameterize and simplify their abstract representations of actual climatic feedbacks involving coupled atmospheric, oceanic, and terrestrial processes (Loáiciga et al 1996, Grassi 2000, Forest et al 2002, Roe and Baker 2007, Pielke 2008). On the other hand, one must not overlook the fact that there have been advances over the last decade in the predictive capability of AOGCMs. The IPCC (2007) report based its climate prediction on the best climate models currently available. The uncertainty of climate predictions' notwithstanding, the IPCC (2007) report called for pronounced reduction of GHGs to avert what the IPCC (and others) consider a serious threat of irreversible damage to the global environment and society. According to this precautionary principle, lack of full scientific certainty may not be used to postpone cost-effective action to prevent deleterious impacts from global warming. Nationally and internationally, the search for means to transition to a vigorous world economy less dependent on fossil fuels continues.
The world's water resources are impacted by global warming in a complex manner associated with the changes in storage and fluxes written in equation (2). Also, since the emergence of agriculture circa 10 000-15 000 years before present, the ensuing population growth has had immense cumulative impacts on the world's fresh water resources, both quantitatively and qualitatively (see, e.g., Postel et al 1996, Falkenmark 1997, Oki and Kane 2006, Barnett et al 2008. Population growth worldwide suggests that people's pressure on natural and water resources will continue for the foreseeable future, giving rise to challenges on ways to meet growing demands for potable water and better sanitation of populated areas, and for water supply to sustain economic growth while avoiding environmental degradation (Cohen 1995, Falkenmark 1997. Figure 2 graphs forecasts of the world's population through year 2050, and similarly for the United States, China and India, the first country having the world's largest economy and the latter two featuring the fastest growing economies among developing countries. The world population is predicted to be near 10 billion people by 2050, most of it to take place in developing countries with inadequate infrastructure and institutions overseeing water supply and sanitation. Given the reality of population growth and the tepid improvement of water systems to cope with it, the United Nations Development Program's Millennium Development Goal of reducing by one half the number of people without access to potable water and basic sanitation by 2015 appears elusive. Ground water is currently and will continue to be an important part of the world's freshwater supply (Shiklomanov andRodda 2003, Gleick 2004).
The growing demands for water associated with population growth are intimately related to the increase in energy use required to support rising living standards worldwide and dynamic economies. In view of the immediate growing demands for water and energy (perhaps more aptly jointly called 'watergy'), the transition to a thriving world economy less dependent on fossil fuels poses enormous challenges (Hardin 1993).
The remainder of this paper is organized in three sections. First, a review of climate change and atmospheric CO 2 evolution over geologic time is presented alongside with measurements of hydroclimatic data made since the early 1900s in a region of special interest in this work. Section 3 describes a method available for assessing likely changes in ground water resources in a 2 × CO 2 -world, that is, one in which atmospheric CO 2 concentration is twice that of a baseline year (1 × CO 2 concentration), taken to be that of 1990 (approximately 360 ppmv) by the Kyoto Protocol for climate-change control (Houghton 2004). Section 4 shows an application of the method for climate-change prediction to develop courses of action for managing a regional ground water resource. Generalizations of the concepts and methods of this paper are made as needed to point directions along which-according to this author's opinion-rest some of the best opportunities to mitigate and adapt to continuing climate change.  Budyko 1977). The gap about 65 millions years before present coincides with the Cretaceous-Tertiary (K-T) boundary and the meteoric impact theorized to have occurred then. This paper summarizes the author's work on climate change/variability and ground water resources sustainability. Its contribution lies in the synthesis of that work with other pertinent publications by several researchers in the same or related fields.

Is the Earth warming?
2.1. The Earth's surface temperature Figure 3 shows the fluctuation of the mean sea surface temperature (used as a proxy for the lower-atmosphere temperature) in the northern mid-latitudes during the last 140 million years according to a reconstruction by Budyko (1977). There is gap in the reconstruction shown in figure 3 in the period surrounding the Cretaceous-Tertiary boundary (circa 65 million years before present), contemporaneous with the theorized meteorite impact on Earth that, among other effects, might have caused the extinction of the dinosaurs. Figure 3 shows pronounced fluctuations of the mean surface temperature over geologic time, and, overall, a longterm decline in mean surface temperature. It can be inferred from figure 3 that the lower-atmosphere temperature in the northern mid-latitudes has decreased about 10 • C in the last 140 million years.
The pattern of lower-atmosphere temperature during the second half of the Pleistocene (started 1 million years ago and ended with the start of the Holocene about 10 000 years ago) shown in figure 4 supplements the longer-term temperature history of figure 3. Figure 4 shows the minimum loweratmosphere temperature during the ice ages that occurred in the last 400 000 years, reconstructed from air-temperature deviations inferred from Antarctic ice cores (Fedorov et al 2006) and then adjusted by this author by an approximate Holocene mean surface temperature of 15 • C. In addition, figure 4 shows the maximum lower-atmosphere temperature during interglacial periods during the last 400 000 years (see Imbrie and Imbrie 1980, for a discussion of orbital variations and climatic response).   Figure 4 shows that the minimum and maximum temperatures associated with ice ages and interglacial periods have remained relatively steady during the second half of the Pleistocene. Current mean surface temperaturerepresentative of interglacial warming following the last ice age-is comparable with warm interglacial temperatures of the last 400 000 years. Figure 5 shows the smoothed global mean surface temperature (MST) anomaly since 1950 (see the solid line on figure 5). The anomaly equals the difference between the measured global surface temperature and its time average between 1890 and 1910, the period used as reference to calculate a time average of MST (Scafetta and West 2008). The anomaly-MST data were low-pass filtered to remove any volcanic signal and smoothed to highlight the influence of the 11 year solar cycle (of radiation intensity reaching Earth) on the surface temperature (Scafetta and West 2008). It is seen in figure 5 that the (smoothed) mean surface temperature increased about 0.68 • C between 1950 and 2008. Notice, however, that the mean surface temperature has decreased since 2002. Figure 5 also shows as a dotted line the reconstructed temperature anomaly between 1950 and 2008 caused by changes in the intensity of the solar radiation reaching the Earth. The graph in figure 5 shows that the smoothed

Atmospheric CO 2 over geologic time
Long-term average atmospheric CO 2 concentration is depicted in figure 6, adapted from Berner (1997). Evidently, there has been a substantial fluctuation of atmospheric CO 2 over geologic time, the overall trend over the last 100 million years being downwards. At the climax of the last ice age, some 25 000 years ago, atmospheric CO 2 was about 180 ppmv. By 1750 it had risen to 280 ppmv, and it is currently 385 ppmv (Fedorov et al 2006, Loáiciga 2006, 2007. It was a approximately 285 and 370 ppmv in 1800 and 1990, respectively. It is this increase of about 105 ppmv since the start of the Industrial Revolution (circa 1750) and the future emissions of GHGs what causes concern nowadays about possible future disruptions of the relative mild climate enjoyed during the Holocene.
The early high concentration of CO 2 , in the thousands of ppmv, was caused by degassing of the Earth's crust. Part of that CO 2 was consumed through geochemical reactions on land (silicate mineral weathering) and in the ocean (formation of carbonate minerals, see Krauskopf 1979, Berner 1997, Loaiciga 2006, 2007. Some of it was fixed as plant biomass and buried in the Earth crust as fossil hydrocarbons, some of which have been recently converted into energy and GHGs through fossil-fuel combustion. Figure 7 shows the minimum atmospheric CO 2 concentration at the climax of Pleistocene ice ages during the last 400 000 years reconstructed from ice cores (Fedorov et al 2006), as well as the maximum atmospheric CO 2 concentrations during the intervening warm interglacial periods. The minimum CO 2 concentration has stayed very close to 180 ppmv during the last 400 000 years. The maximum CO 2 concentration, on the other hand, has risen during the present warm interglacial period to 385 ppmv-or about 100 ppmv higher than those of previous interglacial periods in the Pleistocene.
The weight of the evidence from figures 3-7 is that atmospheric CO 2 concentration and mean surface temperature have exhibited anything but drastic change over geologic time. In contrast, the current warm interglacial period, the Holocene, has been characterized by a relatively mild climate globally, which has permitted the rise of humans and their industrialized society. Oddly, humans find themselves at present wondering if that advanced level of industrialization and the cumulative environmental change they have effected on Earth could modify its climate in ways that might threaten the survival of life as we know it.
As to the question of whether or not the Earth is warming, the answer is not a short phrase. The Earth has warmed between 5 and 10 • C-depending on location-since the end of the last ice age about 10 000 to 15 000 years before present (Budyko 1977, Gribbin 1979  Because of the complexity of the atmosphere-Earth system, with its feedbacks and the ever changing human environmental footprint, the future prediction of the Earth's climate evolution using the most advanced tools currently available-coupled atmospheric-ocean general circulation models-remains a challenge. An approach to create 2 × CO 2 steady-state scenario of future climate under a warming Earth for the purpose of ground water management is presented next. This approach does not rely on the long-term transient simulation of the Earth's climate, thus avoiding the difficult problem of longrange prediction beset by nonlinear and uncertain (if not unknown) feedbacks of the Earth's climate system. Instead, the 2 × CO 2 steady-state scenario provides a portrait of how a regional climate might look like under doubled CO 2 atmospheric concentration using some of the best tools for climate-scenario prediction. Pros and cons of the 2 × CO 2 steady-state approach are found in section 3. Loáiciga et al (2000) presented a method to create 2 × CO 2 climate scenarios-in terms of regional precipitation, runoff, mean surface temperature, etc-and using these regionally averaged hydroclimatic variables to simulate ground water recharge that drives ground water flow, hydraulic head, and spring flow in a regional karst aquifer. The creation of a 2 × CO 2 climate scenario-as proposed by the National Center for Atmospheric Research's Vegetation Ecosystem Modeling and Analysis Project (NCAR-VEMAP, see, e.g., Kittel et al 1995)-starts with a 2 × CO 2 prediction of climate variables simulated in steady-state mode, and represented by the average outputs of six GCMs (for general circulation model, see Loáiciga et al 2000). The variables correspond to a doubled concentration of atmospheric CO 2 relative to a baseline, or 1 × CO 2 , concentration (equal to the 1990 concentration of 360 ppmv). The 2 × CO 2 simulated variables represent averages within the cells of the spatial grid of the GCM, typically made up of coarse cells with sides of a few hundreds of kilometers long. The 2 × CO 2 variables simulated with the GCM serve as boundary conditions for a regional climate model (RCM) whose grid consists of cells with sides of a few tens of kilometers long. The use of the coarse-scale GCM-simulated variables as boundary conditions for a finerresolution RCM, a type of downscaling technique, is illustrated in figure 8. The RCM simulates 2 × CO 2 hydroclimatic variables at its finer spatial scale and in steady-state mode, as well. The use of a reference 2 × CO 2 concentration is reasonable when the purpose is to create a scenario of climatic change relative to that corresponding to a known (historical) 1 × CO 2 baseline condition. It also bypasses the difficulty of guessing what the CO 2 atmospheric evolution would be over the 21st century. Recent data presented by Crowley (2008) shows that the CO 2 emissions for years 2005-2007 do not conform to any of the IPCC (2007) emission scenarios, and, in fact, exceed the worst-case scenario A1F1 entertained by the IPCC (2007).

An approach to developing warmer-climate scenarios
An average 1 × CO 2 baseline, steady-state, simulation of hydroclimatic variables is also obtained, which serve as boundary conditions for the RCM-based 1 ×CO 2 , steady-state, simulation of climatic variables at its finer spatial resolution. The 1 × CO 2 is a reference concentration, like that which prevailed in 1990 (equal to 360 ppmv) and that is used by the Kyoto Protocol to set goals to reduce GHGs concentrations. With the 1 × CO 2 and 2 × CO 2 -based RCM simulations of, say, precipitation, denoted by P 1×CO 2 and P 2×CO 2 , respectively, a ratio of precipitation P 2×CO 2 /P 1×CO 2 is formed. This ratio varies from cell to cell in the RCM spatial grid. The purpose of the precipitation ratio so formed is to use it to scale historical regional precipitation (that, is from modern instrumental records), say, during wet, or dry, or normal climatic conditions. Suppose that a time series of historical precipitation, P H , recorded during a dry period is available. P H is scaled by the precipitation ratio to produce a scenario of what the precipitation would be in a 2 × CO 2 , warmer, world with dry conditions prevailing in the region of interest: Assuming that (i) P 1×CO 2 and P 2×CO 2 are statistically independent between themselves and independent of P H as well, and (ii) that P 1×CO 2 is an unbiased estimator of P H , taking the expected value on both sides of equation (3) shows that the mean of P 2×CO 2 scenario equals the mean of P 2×CO 2 , this being a desirable property of the scaled estimator of future precipitation in a warmer climate. Ratios can be so constructed for other regionally averaged variables such as runoff, evapotranspiration, etc. The scaled regional variables-akin to precipitation in equation (3)-are then used to drive basin-or aquifer-scale simulation models placed at the finest level of spatial resolution in the downscaling technique depicted in figure 8. The output from aquifer models would be ground water flow rates, hydraulic heads, spring flows, and the like, hypothesized to be representative of those to be expected in a 2 × CO 2 world. Figure 9 shows the effect of scaling a precipitation time series encompassing a dry period of duration D. In this example, the scaled time series is seen to exhibit a drought of longer duration and greater severity than the historical time series. It is customary to use differences of temperature, instead of temperature ratios, to create 2 × CO 2 surface temperature scenarios: in which T 1×CO 2 and T 2×CO 2 are the baseline CO 2 and double-CO 2 GCM-simulated time series, and T H denotes the historical time series of temperature. Assuming that T 1×CO 2 is an unbiased estimator of T H , taking the expected value in equation (4) shows that the mean of T 2×CO 2 scenario equals the mean of T 2×CO 2 , this being a desirable property of the estimator of future temperature. The scaling ratios and differences shown in equations (3) and (4), respectively, represent averages computed from several GCM/RCM outputs for regions within the conterminous United States. Loáiciga et al (2000) and Loáiciga (2003) used precipitation ratios and temperature differences to drive a hydrologic model of runoff production from which recharge (by stream seepage) was calculated. A similar approach underlies the results presented in a section 4.2 of this paper dealing with global warming scenarios and spring flow response. The reader is referred to Wilby and Wigley (1997) for a review of climatic downscaling methods. There are strong assumptions underlying the accuracy of future climate scenarios using the scaling ratios and temperature differences embodied in equations (3) and (4), respectively. It is assumed, for example, that historical climatic variables scale proportionally or shift linearly with respect to the ratios or differences, respectively. Also, the timing of events such as droughts and floods follows the same pattern observed historically. That is why this method of scaling and shifting is best suited to attempt to predict climate scenarios under below-normal precipitation (droughts), or normal precipitation, or above-normal precipitation, where the differences among extremes and averages are highlighted in the climate scenarios so created. It cannot be overlooked either that the use of ratios and differences of 1 × CO 2 and 2 × CO 2 neglects the transient rise of atmospheric CO 2 over a span of time 100 years long. In spite of the limitations of the scaling/shifting approach herein used, the alternative, or transient GCM simulations of the future climate, are fraught with their own limitations (IPCC 2007). Both steady-state and transient simulations of the future climate are beset by the current imperfect knowledge of factors governing climate change (such as modifications in planetary albedo, atmospheric transmissivity, air moisture, oceanic circulation, ice melting) and related nonlinear feedbacks (Loáiciga et al 1996, Loáiciga 2003, Roe and Baker 2007. In spite of the cited limitations, one must acknowledge the steady progress made over the last two decades in our capacity to simulate numerically the world's climate. The IPCC (2007) report summarizes several stateof-art models for climate prediction currently available. In the realm of climate prediction and water resources impacts, the papers by Scibek and Allen (2006), Serrat-Capdevila et al (2007), Fowler et al (2007 are exemplary.

Historical hydrologic records: inter-annual variability and stationarity
One of the most productive aquifers in the United States is the Edwards aquifer of south-central Texas, which provides water for economic, urban, and environmental needs in greater San Antonio, Texas and its environs. The approximate geographical layout of the Edwards aquifer is shown in figure 10, where it is shown with other river basins deemed vulnerable to current climate change (Loáiciga et al 2000, Loáiciga 2002). Vulnerable in this context is synonymous to highly probable deleterious changes in water quantity and quality that threaten society, ecosystems or water-dependent habitats. Extending over 6109 km 2 of recharge and discharge area, the Edwards aquifer is made up of karstified carbonate rocks of high transmissivity. A hydrogeologic description of the Edwards aquifer is found in Loáiciga et al (2000), Loáiciga (2003Loáiciga ( ), (2008, which also examined the vulnerability of its spring flow to global warming. The storage capacity of the Edwards aquifer is illustrated by figure 11, which shows the annual change in aquifer storage from 1934 though 2007 (adapted from Loáiciga 2008). Notice the persistence of low storage between 1947 and 1959, the period of critical drought. The difference between maximum and minimum change of storage suggests that the aquifer storage capacity is at least 5.5 × 10 9 m 3 . Figure 12 shows time series of annual precipitation (P) and recharge (R) into the Edwards aquifer from 1934 through 2007. Five year moving averages for these time series (P-5, R-5) are also shown on figure 12. The recharge data show low levels between 1947 and 1956, this being within the critical drought period (1947)(1948)(1949)(1950)(1951)(1952)(1953)(1954)(1955)(1956)(1957)(1958)(1959). Recharge is calculated by water balancing in the recharge area of the aquifer according to a  1930 1940 1950 1960 1970 1980 1990 2000 2010 Year method by Puente (1978). The precipitation and recharge time series do not exhibit any long-term trends. Their inter-annual variability is pronounced. These high-quality data establish that global warming trends do not appear to have modified either recharge or precipitation in the Edwards aquifer during the last seven decades. Figure 13 shows the annual time series of recharge and ground water pumping in the Edwards aquifer from 1934 to 2007.
Unlike recharge, the ground water pumping data in figure 13 exhibit an overall rising trend from 1934 through 1985. Lawsuits filed by environmental advocates prevented further increase of ground water pumping, which by 1980 had diminished spring flow to the point where several aquatic species had become threatened or endangered (Loáiciga et al 2000). Figure 14 displays annual recharge data and spring flow data from 1934 to 2007.
It is seen in figure 14 there that spring flow was lowest during the drought period, and also in the period 1978-1991, which, although having had average recharge, exhibited high ground water pumping. The spring flow and ground water pumping data show that the human environmental footprint has grown over time.

Global warming scenarios and spring flow response
The scaling-ratio approach embodied by equations (3) and (4) was used by Loáiciga et al (2000), Loáiciga (2003) to create recharge inputs to a ground water flow numerical simulation model. The ratios were obtained from the NCAR-VEMAP database cited above. These ratios represented averages over a spatial grid of a Regional Climate Model with cells 0.5 • × 0.5 • latitude/longitude in size within the Edward aquifer region, shown in figure 15.
The input recharge data was used to drive the regional ground water flow model GWSIMIV for the Edwards aquifer (Thorkildsen and McElhaney 1992). GWSIMIV calculates hydraulic head in each of its numerical grid cells and spring flow at selected cells where springs are located. The scaling procedure was implemented to create 2 × CO 2 recharge scenarios corresponding to average recharge (historical period 1978-1989) or drought (historical period 1947-1959) with which to drive the GWSIMIV model.    Loáiciga et al 2000) shows the results for minimum spring flow in the two major springs in the Edwards aquifer, namely, the Comal and San Marcos Springs for 2 × CO 2 recharge constructed by scaling average historical recharge (period 1978-1989). The spring flows were calculated with the GWSIMIV model for a range of annual ground water pumping. The two springs exhibit drastically different responses to ground water pumping.
Comal spring flow crosses a recommended minimum spring flow threshold at about 6.8 × 10 8 m 3 year −1 of ground water pumping. San Marcos spring flow response is relatively insensitive to ground water pumping and decreases slowly with increasing ground water pumping. At 2 × 10 8 m 3 year −1 of ground water pumping the spring flow is slightly below the recommended minimum threshold. A plausible range of recommended ground water pumping for future average recharge could be 2.0×10 8 -6.0×10 8 m 3 year −1 , given that the spring flows are close to the minimum threshold for this range of pumping. Figure 17 (adjusted from Loáiciga et al 2000) shows similar information as that in figure 17 except that the 2 × CO 2 recharge was constructed by scaling drought historical recharge (period 1947-1959).
In this case Comal spring flow is below the minimum recommended threshold at 2 × 10 8 m 3 year −1 , while San Marcos spring flow is below the minimum threshold even for nil ground water pumping. The reduced spring flow and its impacts on aquatic ecosystems calls in this case for reduced ground water pumping compared to average recharge conditions, perhaps pumping in the range 0-2.0 × 10 8 m 3 year −1 . This reduced ground water pumping would require substitute water sources to make up for the shortfall in municipal, industrial, commercial, agricultural, and military supplies. A diversified system of water sources supplemented with conservation, water recycling (reclamation), and water pricing must be developed to provide adequate water during drought periods, and, with population growth, even during average recharge years.
The importance of plentiful spring flow to support aquatic species is documented empirically. Figure 18 shows a plot of salamander counts (Eurycea sosorum) in the Eliza spring near the city of Austin, Texas, which is underlain by the Edwards aquifer. The salamander population viability was threatened by low spring flow in 2003. Spring flow recovery allowed the salamander population to rebound by mid 2005, after which its fluctuating numbers became a matter of natural seasonal variability, rather than human-induced reduction of spring flow.

Relative contributions of global warming and population growth to spring flow
The effect of global warming or climate change on spring flow is felt through changes in recharge. The effect of population growth is felt through increased ground water pumping. Let the function f (Q, R) represent the response of spring flow to annual ground water pumping Q and annual recharge R. The total change in spring flow is then: The first term on the right-hand side of equation (5) represents the change in spring flow caused by a change in ground water pumping while the recharge is constant. The second term of the same right-hand side represents the change in hydrologic response (i.e., spring flow) caused by a change in recharge (a measure of a change in climate). The function f (Q, R) is not known analytically. Therefore, one must approximate the right-hand side terms of equation (5) numerically. Loáiciga (2003) proposed an approximation whereby spring flows in the Edwards aquifer (using GWSIMIV numerical model) were calculated for four scenarios. Scenario 1, the baseline: use average historical recharge (during period 1978-1989) and baseline ground water use (period 1978-1989).
Scenario 2, approximate climate-change impact alone: use average historical recharge scaled to 2 × CO 2 conditions using scaling-ratio approach, and baseline ground water use.
Scenario 3, approximate ground water use impact alone: use average historical recharge and year-2050 forecasted ground water use.
Scenario 4, approximate climate change and ground water use impacts combined: use average historical recharge scaled to 2 × CO 2 conditions and year-2050 forecasted ground water use.
The spring flow simulations for scenarios 1-4 revealed that: (1) The climate-change scenario (scenario 2) increased spring flows at Comal and San Marcos springs by approximately 160% and 17%, respectively, relative to the baseline scenario (scenario 1).
(2) The ground water use impact scenario (scenario 3) decreased spring flows by 100% and 22% at Comal and San Marcos springs, respectively, relative to the baseline scenario.
(3) The ground water use impact and climate-change scenario (scenario 4) decreased spring flows by 73% and 1% at Comal and San Marcos springs, respectively, relative to the baseline scenario.
These results for scenarios 1-4 suggest the preponderance of the anthropogenic impact over the global warming impact (or climate-change impact) in the Edwards aquifer. This is a warning for those in charge for water supply and environmental planning in the Edwards aquifer region that a diversified water supply and conservation system must be developed to cope with increasing human water use and competing aquatic ecosystem requirements.

Summary
The heightened worldwide awareness concerning fossil-fuel use, the increase of GHGs in the atmosphere and their possible adverse climatic impacts, and about water resources impacts of global warming and population growth is timely and meritorious. This paper reviewed some of the difficulties clouding the prediction of regional effects of future climate change on water resources. To mitigate a somewhat daunting outlook in this respect, it also presented one approach to weigh the relative impacts of global warming and population growth in a very important regional aquifer of the United States. The results obtained concerning impacts of climate change and population growth in the Edwards aquifer in Texas are noteworthy and consistent with findings by this author in previous studies for various regions of the world (see Loáiciga et al 1996), and with global-scale studies by other authors (see Vörösmarty et al 2000), namely, that population growth will exert as great or greater an impact on the world's water resources as global warming might in the 21st century. The results obtained for the Edwards aquifer in this study and for other regional studies by this author (see Loáiciga and Renehan 1997) point to the need for water supply systems with diversified sources that can be tapped on under a variety of climatic conditions (wet, normal, dry), supplemented by sound conservation and water recycling programs that discourage waste.
A good example of a diversified water supply system is that of the City of Santa Barbara, California, where normally the largest (and cheapest) sources of water are regulated reservoirs. Ground water is tapped on as a supplement to reservoir water during long droughts. Ultimately, reservoir water and ground water failing, sea water desalination (the most expensive source) is activated during grave conditions of water scarcity.
The statements by Loáiciga et al (1996) about global warming and associated hydrologic-cycle issues are very pertinent today. First, the natural variability in regional hydrologic cycles is likely to dominate additional variability brought about by climate change in many regions of the world. Second, sound water resources engineering and management, with time-tested provisions for safety, ought to render water projects adequate with or without climate change. As for specifics on how to mitigate variability, change, and uncertainty, Klemes (1991) proposed guidelines to cope with future climate change in the context of water resources planning and design: (1) adherence to high professional standards in solving water resource problems, (2) commitment to measures limiting water waste and pollution, (3) striving for robust and resilient designs, (4) documenting and taking into account known uncertainties in water supply and demand, (5) documenting the ranges of feasible operation of projects, rather than providing only nominal design parameters, (6) providing a general outline of feasible contingency measures for extreme conditions not accommodated by the project under normal operation (i.e., flexible operating rules), and (7) insistence on clear disclosure of factual information, assumptions, and conjectures behind modeling results to be considered.
Lastly, this paper points to rapid and relentless population growth as one of the factors that deserves greatest attention in order to succeed in providing, adequately and timely, suitable water for society and the environment.