| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a USDA-ARS, Fort Collins, CO 80526
b CSIRO Land and Water, Floreat, Western Australia 6014, Australia
c CSIRO Land and Water, Canberra, ACT 2601, Australia (retired)
* Corresponding author (tim.green{at}ars.usda.gov).
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.
Received 9 July 2006.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONDescription of the Study...Materials and MethodsResultsSummary and ConclusionsREFERENCES |
|---|
Abbreviations: ET, evapotranspiration • GCM, general circulation model • PET, potential evapotranspiration • RCM, regional climate model • SVAT, soil–vegetation–atmosphere transfer • VPD, vapor pressure deficit
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONDescription of the Study...Materials and MethodsResultsSummary and ConclusionsREFERENCES |
|---|
Despite the critical importance of groundwater resources in many parts of the world, there have been very few direct studies of the effect(s) of global warming on groundwater recharge.
Watson et al. (1996, p. 336)
Although the effects ... on groundwater resources are not adequately understood at present, they cannot be ignored.
Watson et al. (1998, p. 122)
Groundwater is the major source of water across much of the world, particularly in rural areas in arid and semiarid regions, but there has been very little research on the potential effects of climate change.
McCarthy et al. (2001, p. 199)
As these quotations from the Intergovernmental Panel on Climate Change illustrate, there has been a continuing need to develop methods for assessing the potential effects of CO2–altered climates on aquifer recharge. An editorial article in Ground Water (Sophocleous, 2004) confirmed the general need: "Although aquifers are generally sensitive to changes in recharge, little work has been done on the impacts of climate change on particular aquifers." Furthermore, the dynamics of groundwater systems in response to human and climatic stresses require further attention (Alley et al., 2002).
Groundwater is a valuable resource in many parts of the world and particularly in arid to subhumid regions such as much of Australia. For water supply and environmental issues, the hydrologic fluxes of interest include temporal variability in groundwater recharge rates and the long-term yield (both human-induced and natural groundwater discharge) of the aquifers.
In the last several years, researchers have begun to estimate potential impacts of climate change on regional groundwater resources around the world. Simulation studies were performed on regions in Texas (Loáiciga et al., 2000), southwest Canada and northwestern USA (Scibek and Allen, 2006), and northeastern Germany (Wessolek and Asseng, 2006). These studies vary in terms of methods of simulating climates and hydrology, results of groundwater sensitivities to climate, and interpretations of how such results could be applied. Other approaches include paleo-recharge estimation using tracer methods, including isotopic data from Jordan (Bajjali and Abu-Jaber, 2001) and from north Africa and Italy (Zuppi and Sacchi, 2004) and borehole temperatures in Japan (Taniguchi, 2002).
We present here an approach and case studies for simulating the potential effects of climate change associated with doubling historical concentrations of carbon dioxide in the atmosphere. The general method involves simulation of CO2–altered climates using a GCM, generation of daily weather variables using a stochastic daily weather generator based on historical data and the GCM results, and numerical simulation of rainfall infiltration, variably saturated flow (one-dimensional), and evapotranspiration, producing time series of daily deep drainage (a surrogate for groundwater recharge) for various soil–vegetation environments.
Details of the context for the present approach and contrast with alternative methods may be gained from recent literature reviews (Overgaard et al., 2006; Xu et al., 2005). Others have used regional climate models (RCMs) to downscale GCM results, and soil–vegetation–atmosphere transfer (SVAT) schemes have been coupled to RCMs. The prior emphasis has been on surface water or shallow soil water (e.g., Bell et al., 2007), in contrast to the present focus on deep drainage below the root zone to address groundwater recharge. To our knowledge, none of the RCM–SVAT models provide equivalent physically based simulations of water and energy including plant dynamics.
Plant water use (transpiration) is a critical component of the water budget, along with evaporation needed to compute groundwater recharge. Dooge (1992, p. 2683) noted that "meaningful scenarios of hydrologic prediction and climate prediction are not possible without an understanding of vegetation response." Thus, the present model includes effects of dynamic plant growth and respiration (including simulated changes in leaf area and root density) on transpiration and deep drainage. The approach presented here demonstrates the use of physically based climate and soil–plant–water models to explore previously unquantified sensitivities of groundwater recharge to potential climate change, including simulated plant dynamics.
|
Description of the Study Sites |
|---|
|
TOPABSTRACTINTRODUCTIONDescription of the Study...Materials and MethodsResultsSummary and ConclusionsREFERENCES |
|---|
|
Gnangara is an unconfined, sandy aquifer located on the Swan Coastal Plain north of Perth in Western Australia (WAWA, 1987). It is centered at approximately 31° 40' S, 115° 50' E and covers an area of approximately 2200 km2. The landscape is covered primarily by mixed forests, dominated by Banksia woodland and eucalyptus forest (Farrington et al., 1990). Approximately 10% of the area is covered with exotic pine plantations of various ages and densities (a basal area of 11 m2 ha–1 is the target), as well as cleared grasslands. Approximately 1 to 2% of the total land area is under intensive agriculture for vegetable production. The climate is Mediterranean with cool, wet winters and hot, dry summers. We used the continuously observed daily weather data from 1975 to 1991 at Pearce Air Force Base (31° 40' S, 116° 1' E, elevation 40 m) for generating stochastic daily weather variables. The mean annual rainfall for this period was 642 mm, and the mean annual PET was approximately 1400 mm.
In both locations, infiltration rates through the sandy soils can be very high, and much of the water used by humans in coastal areas is pumped from shallow aquifers. Furthermore, groundwater discharge to wetlands affects the health of ecosystems in these groundwater-controlled environments. Only recharge sites with relatively deep water table conditions (>5 m for grasses and >25 m for trees) were considered in this paper. Soils were assumed to be homogeneous with depth, although vertical heterogeneity could be incorporated using water, vegetation, energy, and solutes (WAVES) model, if known.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONDescription of the Study...Materials and MethodsResultsSummary and ConclusionsREFERENCES |
|---|
The flow of information is shown schematically in Fig. 2 . Each component of the modeling framework is described below. In the present simulations, information flows in only one direction, and there are no feedbacks between the land surface conditions and the generated daily weather variables.
|
General Circulation Model
McGregor et al. (1993) described the spectral GCM used in this study (CSIRO9 Mark 1). The model operates with nine vertical levels in the atmosphere and a horizontal resolution of about 300 km by 600 km. GCM simulations provide daily values for 30 climatic variables, including precipitation, temperature, and global solar radiation. These output variables provide information on possible modifications of the frequency and distribution of rainfall events and potential evapotranspiration.
We first selected CSIRO9 GCM grid cells (300 x 600 km) that either contained the Gnangara and Stradbroke sites or an adjacent cell that best matched the statistics of the historical records. The changes between GCM simulations for current (Climate 1) and double CO2 (Climate 2) conditions at these cells were used to estimate changes in MWGEN parameter values below. Thus, the simulation approach was used to generate steady-state climate "scenarios" rather than "projections" of transient climate change under different greenhouse gas emission scenarios (McCarthy et al., 2001).
Subsequent work has shown that transient GCM projections have a stronger drying signal for winter rainfall in southern Australia than the scenarios generated by CSIRO9 GCM for our simulations. The evolution of Australian winter rainfall projections can be traced through the CSIRO Climate Change Scenarios reports. In 1992 they reported a –20 to +20% change projected by 2070 (CSIRO, 1992). In 1996 they reported a change of –20 to 0%. (CSIRO, 1996). In 2001 they reported that the range of change for the southwest of western Australia based on coupled models in CSIRO (1996) was –8 to 0% for 2030 and that the current projections were –20 to +4% (CSIRO, 2001; Hennessy and Whetton, 2001).
More recently, Hennessy et al. (2006) published projections for 2030 that confirm a high degree of uncertainty associated with the magnitude of projected seasonal trends, including winter drying. Climate models producing drying trends will be used in future research for comparison but are beyond the scope of the current research, which focuses on methods for assessing dynamic sensitivities of groundwater recharge to simulated climate change scenarios.
Stochastic Point Weather Generator
The stochastic weather generator used in this study is based on the WGEN model (Richardson and Wright, 1984). Subsequently, Wilks (1992) presented a method for adapting stochastic daily weather models (fitted to historical weather series) to generate synthetic series for future climates. The model parameters were adjusted to be consistent with the changes in monthly statics from GCM runs for current and double CO2 conditions. Artificial "drizzle" in GCM results remained an issue affecting rainfall frequency at the lowest amounts, but the use of differences in GCM runs (Fig. 2) helped to cancel out such GCM artifacts.
Following this type of approach, the modified model (MWGEN) also includes higher-order harmonics for temperature and an atmospheric transmission model for solar radiation to calculate the clear day maximum at the ground surface (Fleming, 1987). Ground surfaces were assumed to be flat for these analyses. For each calendar month, a generalized beta distribution was fitted to the daily residuals (i.e., differences between the clear day maximum and observations) for wet and dry days. Occurrence of either wet or dry days was determined by a two-state, lag-1 Markov model. The rainfall amount distribution for wet days was represented by a highly skewed Gamma distribution. Two climate scenarios (1000 yr each of daily data) were generated for each study site using MWGEN. The climatic variables simulated are precipitation occurrence and amount, maximum and minimum temperature, and global solar radiation. Further details regarding the generation of daily weather sequences for each climate scenario were given by Bates et al. (1994).
Soil Water and Vegetation Model
The WAVES (water, vegetation, energy, and solutes) model (Zhang et al., 1996) simulates infiltration and vertical redistribution of water through soil columns under dynamic plant cover. The computed "deep drainage" is equated with groundwater recharge. The model is used here with literature values of parameter for various components, except as noted. No data were available for site-specific calibration, so the model was not calibrated to present conditions.
Vertical Water Movement
Richards' equation for variably saturated flow in one dimension (vertical) was solved numerically using a finite difference method in WAVES (Dawes and Short, 1993; Dawes and Zhang, 1998). The solution technique is very efficient and numerically stable, owing to the methods outlined by Ross and Bristow (1990) and Dawes and Zhang (1998) using:
The soil hydraulic properties were represented by the Broadbridge and White (1988) functions, where soil water retention was represented analytically as
 | [1] |
= 
/
c and 
= (
– o)/(s – o), is the matric potential, is the volumetric water content, c is the capillary length, s and o are the saturated and residual volumetric water contents, respectively, and C is a fitting parameter which is also related to one of two parameters in the soil–water diffussivity function of Knight and Philip (1979).
Boundary and Initial Conditions
The surface boundary condition is generally rainfall infiltration with ponding occurring only for extreme events on finer textured soils. Any simulated runoff is subtracted from the water balance without further interaction. Surface runoff was an insignificant portion of the total rainfall in all cases, so it is not discussed below. The soils drain by gravity at the base of the simulated columns (5 m for grasses and 25 m for trees), and the computed "deep drainage" is equated with groundwater recharge. Simulations are continuous for 1000 yr of daily input and output, with soil water movement solved on smaller time steps as needed for numerical accuracy and convergence. However, the results can be interpreted as multiple realizations of shorter time series, each with a random, but physically realistic, initial condition for the first day of the starting year.
Energy and Carbon Balances
An important characteristic of WAVES is the coupling of water, energy, and carbon balances. Vertessy et al. (1996) summarized the use of solar energy and assimilation of carbon to drive dynamic plant growth and associated water use. Applying the integrated rate method (Hatton et al., 1992; Wu et al., 1994), plant growth is limited by the relative availabilities of light, water, and nutrients. Carbon assimilation and respiration from roots, stems, and leaves are included in an empirical representation of the vegetation response. This allows the leaf area to change dynamically with environmental conditions such as soil moisture, solar radiation, and atmospheric temperature. Thus, leaf area values need not be prespecified.
The energy balance is solved using the Penman–Monteith equation (after Monteith and Unsworth, 1990) to partition daily solar radiation provided by the climate simulator MWGEN. Solar radiation is attenuated by the canopy (only one layer is simulated here). Canopy resistances for heat and water vapor transfer are functions of the leaf area index, CO2 concentration and air–mass vapor pressure deficit (vpd). Thus, evaporation is modeled explicitly for each climate.
Input Parameters
Input to the WAVES model includes climate variables for the two climate scenarios, soil hydraulic properties, and vegetation parameter sets. The same types of soils and vegetation are used at both sites for direct comparisons between climatic zones.
The daily climate variables simulated in MWGEN and used by WAVES are precipitation occurrence and amount, maximum and minimum temperature, and global solar radiation. In addition, WAVES requires vpd and computes vapor pressure internally. Monthly mean values of vpd were computed from humidity and temperature data at Cape Moreton near Stradbroke and at Pearce AFB for Gnangara (Bureau of Meteorology, 1988). These values were used in lieu of daily vapor pressure data for each climate scenario.
Soil parameters for four generic soil types (Table 1) cover the range of soils expected for relatively deep unsaturated soil profiles at Stradbroke and Gnangara, with most of the area composed of fine to medium sands (Ahmed et al., 1994).
|
The plant-growth model uses parameters representative of eucalyptus or pine trees and perennial grasses tested under Australian conditions (Dawes and Zhang, 1998). Limited calibration periods, typically 2 or 3 yr, have been used to determine the sets of 17 vegetation parameters shown in Table 2, while simulations here are run for long time series assuming dynamic equilibrium in climate over decades to centuries.
|
An important assumption in this study is that the plant-growth model is able to simulate water use under climatic conditions that differ from those for which the model has been calibrated. We explicitly modified the model parameters that determine the changes in plant transpiration rates and vegetative cover due to CO2 doubling (cf. Hatton et al., 1992). Two key parameters in our WAVES simulations were adjusted accordingly (Warrick Dawes, personal communication): the maximum carbon assimilation rate (Amax in Table 2) was increased by 25%, and a stomatal conductance parameter (g1) was reduced by 50%. A similar approach was taken by Eckhardt et al. (2002), who modified the Soil Water Assessment Tool (Arnold et al., 1998).
Assessing Changes in Groundwater Recharge Regimes
The following methods were used to quantify changes in the simulated groundwater recharge time series. The measures below were applied to the differences in hydrologic response for each soil-vegetation combination at each study site. Climates 1 and 2 denote current and double CO2 conditions, respectively, for a given location.
Measures of Long-Term Average Recharge
Two measures of the relative change in long-term average recharge highlight the net hydrologic response. The first measure, Q2/Q1 = (Recharge 2)/(Recharge 1), is simply the net recharge under Climate 2 (double CO2) normalized by the net recharge under Climate 1 (current conditions) for each soil–vegetation combination. Given a 14% increase in mean annual rainfall, for example, values greater than 1.14 indicate that, on average, a disproportionately high amount of rainfall in Climate 2 becomes recharge relative to the Climate 1 scenario. Values less than unity indicate a reduction in recharge, despite the increase in rainfall. The second measure is
 | [2] |
Temporal Persistence in Annual Recharge Variability
In addition to the statistics of mean annual recharge, water resource managers are concerned with interannual deficits or surpluses from a given demand. Here, we set the demand to the long-term average recharge, which equals the sustained yield including discharge to surface water bodies. Residual mass curves are used to quantify the temporal persistence of deviations for every soil–vegetation–climate combination. The residual mass is defined as
 | [3] |
is the average (net) recharge (mm yr–1) after n years, and in this case, n is 1000 yr each aggregated from daily data. Instead of an actual mass, the units of Sj are given in millimeters as a sum of annual recharge deviations (mm yr–1) over j years. The range in residual mass over a (long) time series has been used in computing a measure of the persistence in variations from the mean behavior. The Hurst coefficient H (Hurst, 1951; Salas, 1992) can indicate long-term dependence or persistence when H > 0.5 as estimated by
 | [4] |
 | [5] |
Another possible measure of persistence in annual recharge variability is the normalized range of residual mass:
 | [6] |
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONDescription of the Study...Materials and MethodsResultsSummary and ConclusionsREFERENCES |
|---|
|
Alterations in the global solar radiation [Fig. 3(ii)] are relatively small and principally reflect changes in the number of rain days. The mean radiation decreased by 5.5% at Stradbroke and 5.2% at Gnangara, while the seasonal ranges were relatively unchanged in both locations. Simulated changes in daily minimum [Fig. 3(iii)] and maximum [Fig. 3(iv)] temperatures are approximately 5°C for the median values, and the spread (variance and skewness) is similar for both climates. The average daily minimum temperature increased more than the maximum (
Tmin = 5.3°C and Tmax = 5.0°C for Stradbroke, and Tmin = 5.3°C and Tmax = 4.9°C for Gnangara).
The resulting histograms of wet- and dry-period durations at Stradbroke are shown in Fig. 4 . The semi-log scale emphasizes the projected increase of long-duration periods of both continuous rain (wet) days and no rain (dry). A frequency of 1 would be zero on the log scale, so a small positive value was added for plotting purposes. Wet-period durations of more than 6 d and dry-period durations of more than 20 d were more frequent for double CO2 conditions (Climate 2) than for present conditions (Climate 1). The maximum wet- and dry-period durations for these sample distributions increased from 40 to 56 d and 87 to 109 d, respectively. Changes in the wet- and dry-period durations at Gnangara (not shown) were less pronounced.
|
|
Daily recharge rates vary much more gradually than the intermittent rainfall that drives the unsaturated flow system. This dampening of the drainage response at depth is caused by the flow impedance and storage capacity of the vadose zone and thus depends on the depth at which deep drainage (i.e., water that will become groundwater recharge at the water table) is simulated. That depth was 5 m for the grasses and 25 m for the trees.
Annual Recharge Distributions
An annual timescale is appropriate for assessment of groundwater recharge, keeping in mind that temporal dynamics of the recharge to the water table will differ somewhat from the simulated gravity drainage at 5 or 25 m for grasses and trees, respectively. The following analyses used annual aggregates of the simulated daily drainage ("recharge").
Box Plots
Figure 6a
shows box plots of 1000 annual recharge values for each combination of vegetation and soil types at Stradbroke. Climates 1 and 2 are plotted side-by-side in each soil-type panel to illustrate the changes in median annual recharge. (The notches indicate 95% confidence intervals for the medians, showing that the changes in median values are generally significant at that level.) For each climate and with every type of vegetation, the median annual recharge decreased with finer soil texture. This is due to increased water availability in the root zone as percolating water is held more strongly by capillarity in the finer textured soils. Also, the temporal variance tends to decrease with finer soil texture due to increased transit times and diffusion of wetting fronts. Grass 1 allowed less recharge than Grass 2 under both climate scenarios at Stradbroke. This may be due to a combination of a lower stomatal conductance associated with g1 (see Table 2; g1 = 10 for Grass 1 and g1 = 0.7 for Grass 2) and the maximum plant available water potential for Grass 2 being half that of Grass 1 (Table 2).
|
Figure 6b shows box plots of 1000 annual recharge values for each combination of vegetation and soil type at Gnangara. For each canopy (vegetation type), the direction of change in recharge was the same for all four soil types. Unlike Stradbroke, however, both the magnitude and the direction of change differed among vegetation types. Recharge was reduced beneath some canopies (Grass 1 and Tree 1) despite the increase in mean rainfall. A pronounced difference between grasses and trees in the annual recharge amounts was simulated at Gnangara (Fig. 6b), particularly for fine-textured soils. This is consistent with documented effects of tree clearing on increased groundwater recharge ( Peck and Williamson, 1987) and associated increases in soil salinity (Williamson et al., 1987). As shown in Table 2, the primary differences between grasses and trees are quantified by the light extinction coefficient; IRM (integrated rate methodology), weighting for light; maximum potential rooting depth; leaf respiration coefficient; and aerodynamic resistance. Although simulated recharge may be sensitive to other vegetation parameters, such as the maximum carbon assimilation rate and optimum temperature, these parameters differ among grasses or trees as much as between grasses and trees. The present results show interesting sensitivities to vegetation parameters, but a detailed sensitivity analysis is left for further research.
The net recharge can more than double with an increase in mean annual rainfall from 642 to 732 mm (14%) and a mild increase (2%) in the average number of wet days (from 116 to 123). This result is related primarily to the increased frequency of long-duration wet periods and to the change in growth rates associated with changes in the temperature regime. Increased temperatures account for the simulated reductions in recharge with increased rainfall. The effects of such interactions could not be predicted a priori and thus demonstrate the utility of the present simulation approach including the effects of dynamic plant growth and water use.
Again, the median annual recharge for each climate scenario decreased with finer soil texture for all vegetation types. We also note the pronounced effect of soil texture on recharge beneath tree canopies for either climate scenario at Gnangara versus milder effects in the same direction at Stradbroke. This may be because of higher rainfall amounts at Stradbroke and the seasonal nature of rainfall in these different climatic zones. Within an average year, rainfall is more uniform in the subtropical climate at Stradbroke than in the Mediterranean climate at Gnangara. More winter rainfall at Gnangara escapes the root zone in the medium (or coarser) sand but is held for transpiration later in the year by finer-textured soils. It is not entirely clear, however, why grasses are not affected in a similar way, except to postulate that similar amounts of infiltrated rainfall escape their relatively shallow root zones within the wet season for all soil types.
Quantile–Quantile Plots
A direct comparison of the change in annual recharge between climate scenarios is possible using quantile–quantile (q–q) plots. If there were no change in temporal distributions of recharge, points would fall on a 1:1 line. If alterations in the recharge regimes were proportional to alterations in rainfall, all points (quantiles) would fall along a straight line passing through the origin. Figure 7
shows q–q plots for each combination of vegetation (rows) versus soil (columns), with Climate 1 on the horizontal axes and Climate 2 on the vertical axes. Dashed lines mark the 1:1 equality, and solid lines are linear regressions to the simulated data. Above each q–q plot are two numbers stating the intercept and slope of the regression line.
|
For Gnangara (Fig. 7b) the changes in recharge are generally much less pronounced. This is particularly true for the grasses, where the points lie close to the 1:1 line. The most dramatic effects occur beneath the tree canopies. The intercepts are generally near zero, and the slopes range from 0.66 to 1.80. Interestingly, these extremes are observed on the same soil type (sandy loam) with different tree types. Although the soils affected the absolute value of recharge for each climate scenario (see Fig. 6b), the ratio of change in recharge may have been affected more by the vegetation type. The consistent decreases for both Grass 1 and Tree 1 appear to be related to their relatively narrow range of optimal temperature (see Table 2) and corresponding shifts in temperature into the peak zone for Climate 2. Finally, changes in the recharge regimes tend to be more linear for Gnangara than for Stradbroke.
Long-Term Mean Recharge
Measure 1
Figure 8
shows results using the first measure of long-term net change, Q2/Q1, for various combinations of four soil types and four types of vegetation at both locations. At Stradbroke (Fig. 8a) Q2/Q1 ranges from 1.74 to 5.09, consistently increasing in value from coarse to finer textured soils and from Tree 2 (row 4) up to Grass 1 (row 1). There was an increase in sensitivity to climate alteration going from trees to grass, as well as a consistent increase from vegetation with moderate assimilation and respiration rates to those with a balance of relatively high rates. The latter balance yields greater temporal variation in plant biomass, including leaf area. That is, more dynamic plants are more susceptible to extended wet and dry periods (climate variability) associated with the climate change scenario. At Gnangara (Fig. 8b) Q2/Q1 ranges from 0.66 for Tree 1 to 2.19 for Tree 2 (both values occurring with sandy loam, soil type 3). The trees experienced the greatest changes in recharge with climate, while the effect was less pronounced with grasses.
|
Q/R (Measure 2, given in Eq. [2]). For Stradbroke (Fig. 9a) there was no trend with soil type, and the principal changes (Q/R ranging from 0.76 to 1.05) were related to vegetation. The alternating bands of dark and light indicate that it is not a distinction between grasses and trees. Rather, it appears to relate to the magnitude of carbon assimilation and plant respiration (see Table 2), and the associated dynamics of leaf area and plant available radiation. For Gnangara (Fig. 9b), the principal differences (in a range of –0.79 to 0.62) were related to vegetation also. The alternating bands of dark and light (most pronounced for the medium sand, soil type 1) again indicate that it is not a distinction between grasses and trees. Rather, it relates to the temperature ranges for optimal growth (see Table 2) and the associated dynamics of leaf area and plant available radiation. This sensitivity decreased for the finer textured soils, and most of the recharge beneath trees occurred in the coarser soils.
|
Q/R decreased consistently by approximately 0.1, indicating that vegetation dynamics are responsible for making Q/R exceed 1.0, but other factors affect the first-order difference in recharge between vegetation types. The comparison between dynamic and static vegetation is not illustrated here.
Temporal Persistence in Annual Recharge Variability
We computed estimates of H for each soil–vegetation–climate scenario as described above. For Stradbroke Climates 1 and 2, values fell in the ranges 0.585 < 
< 0.713 and 0.546 < < 0.614, respectively. For Gnangara Climates 1 and 2, values fell in the ranges 0.556 < < 0.713 and 0.566 < < 0.733, respectively. Although these ranges have large overlaps, and we do not connote statistical significance here, the value of for given soil–vegetation environments generally decreased at Stradbroke while increasing at Gnangara. An increase in indicates an increase in the persistence of deviation from the mean recharge, regardless of the direction of change in the mean. Example plots of the residual mass for each climate at Stradbroke are shown for Tree 2 and fine sand in Fig. 10
. These plots include annual recharge rates, a 10-yr running average, cumulative running average (Net) recharge, and the annual residual mass curve. Because the number of years simulated is the same for all scenarios (n = 1000), the computed Hurst coefficient is based on the ratio of the range in residual mass Rn divided by the standard deviation of annual recharge sn simulated for each scenario. In Fig. 10 the range Rn changed from approximately 5800 to 7500 mm for Climates 1 and 2, respectively, but decreased slightly due to the associated increase in sn for the scenarios shown. Thus, the Hurst coefficient can reflect a change in temporal persistence which is not captured by the change in variance alone.
|
|
General Discussion and Caveats
The methods developed and applied here provide a quantitative guide to the direction of change and the potential significance of changes in evapotranspiration, soil moisture, and groundwater recharge regimes due to simulated climate change. As with any numerical experiment, the results are only indicative of the potential responses of real-world systems. There are a number of important qualifications to this assessment of the effects of climate alteration on groundwater resources:
|
Summary and Conclusions |
|---|
|
TOPABSTRACTINTRODUCTIONDescription of the Study...Materials and MethodsResultsSummary and ConclusionsREFERENCES |
|---|
For the Stradbroke (subtropical) climate scenarios, simulated net recharge consistently increased by amounts approaching, and in some cases greater than, the change in total rainfall. Vegetation type affected the transpiration and resulting recharge more than soil type, but both played important roles. The present simulations indicated that recharge can more than double (Q2/Q1 ranging from 1.74 to 5.09) under the projected climate change of a mean annual increase in rainfall amount of 37% (R2/R1 = 1.37). These results for a subtropical climate change scenario in Australia differ from the relative insensitivity of groundwater recharge to a climate change scenario in central Europe (Eckhardt and Ulbrich, 2003). Our greatest changes in recharge occurred with perennial grasses, which may be more susceptible to the impacts of climate alteration than eucalyptus trees in subtropical climates.
For the Gnangara (Mediterranean) climate scenarios, soil texture played an important role for the trees but had a less dramatic effect for grasses. The simulations indicated that the recharge can either decrease or more than double (Q2/Q1 ranging from 0.66 to 2.19) under the projected climate change, with a mean annual increase in rainfall amount of 14% (R2/R1 = 1.14). Unlike simulations for subtropical climate scenarios at Stradbroke, the results for the Mediterranean climate scenarios at Gnangara did not show a consistent increase in recharge for all vegetation types. These results appear to be controlled primarily by the optimal temperature range for plant growth, rather than by maximum carbon assimilation and respiration rates or changes in dry period durations. The feasibility of these results appears to be confirmed by recent WAVES simulations at Gnangara (Wanneroo Station) where Pinus pinaster plantations can transpire additional sources of water beyond the annual average rainfall of 750 mm yr–1 (Warrick Dawes, personal communication).
Changes in the long-term net recharge (i.e., sustainable aquifer discharge to natural and manmade systems) and the interannual variation in recharge were shown for various soil–vegetation combinations (Fig. 6). Increases in recharge variability and decreases in persistence of deviations from the mean annual recharge (or vice versa) were related to the general change in average system wetness and nonlinear response of the soil–water system. The subtropical climate change resulted in consistent decreases in the temporal persistence of recharge associated with increased mean recharge and soil water content.
The simulation results presented here are indicative of substantial changes in recharge regimes in the moderately deep, unconfined portions of the aquifers underlying North Stradbroke Island and the Gnangara region due to climate change. More detailed estimates of the potential changes in spatially averaged recharge would be possible, given knowledge of the cross-correlated spatial distributions of soil and vegetation types from more detailed site data. Shallow water table conditions could also be simulated using WAVES for such assessments. In concert with the methods demonstrated here, improved understanding and simulation of plant growth and transpiration responses are essential for reducing the uncertainty of estimates of the effects of climate change and climate variability on groundwater resources.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONDescription of the Study...Materials and MethodsResultsSummary and ConclusionsREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  B. D Smerdon Climate Change and Groundwater Environmental and Engineering Geoscience, August 1, 2009; 15(3): 211 - 212. [Full Text] [PDF]
|
|
|
|
 |
|
 G. Schnaar and D. C. Digiulio Computational Modeling of the Geologic Sequestration of Carbon Dioxide Vadose Zone J., May 21, 2009; 8(2): 389 - 403. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. R. Green, M. Taniguchi, and H. Kooi Potential Impacts of Climate Change and Human Activity on Subsurface Water Resources Vadose Zone J., August 23, 2007; 6(3): 531 - 532. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Agronomy Journal | Crop Science | |||
| Journal of Natural Resources and Life Sciences Education |
Soil Science Society of America Journal | ||||
| Journal of Plant Registrations | Journal of Environmental Quality |
The Plant Genome | |||
