Springs on rangelands: runoff dynamics and influence of woody plant cover

HYDROLOGICAL PROCESSESHydrol. Process. 20, 3277–3288 (2006)Published online in Wiley InterScience(www.interscience.wiley.com) DOI: 10.1002/hyp.6332

Springs on rangelands: runoff dynamics and influenceof woody plant cover

Yun Huang,1* Bradford P. Wilcox,1 Libby Stern2 and Humberto Perotto-Baldivieso1

1 Department of Rangeland Ecology and Management, Texas A&M University, College Station, Texas 77843, USA2 Department of Geological Sciences, University of Texas, Austin, Texas 78712, USA

Abstract:

To date, little effort has been devoted to understanding the nature of streamflow from dryland catchments where springsare found, and little is known about how changes in vegetation may alter that streamflow. But where they do occurin drylands, springs are important—ecologically as well as hydrologically. Areas that naturally support spring flow,such as those having an underlying karst geology, hold the most promise for increasing streamflow through control ofwoody plants. For rangelands, this possibility is a particularly appealing one. However, few studies have documentedthat such a method is effective (except in Mediterranean climates). For this reason, we studied runoff generation for asite on the Edwards Plateau of Texas: a first-order (19-ha) rangeland catchment supplied with intermittent streamflowfrom a spring at its base. Using hydrometric as well as isotopic characterization, we evaluated the extent to whichrunoff changed following removal of most of the woody plant cover (predominantly Ashe juniper). After monitoringstreamflow for 4 years (two before plant removal and two following removal), we found that (1) streamflow increasedfollowing removal of juniper, by about 46 mm annually, except during the summer; (2) for the 4 years of observation,runoff made up about 22% of the water budget, with baseflow from the spring accounting for about half the total flow;and (3) hydrograph separation analysis conducted for two events following juniper removal indicated that pre-eventwater made up 25% and 50% of the runoff (these numbers represent lower bounds for the percentage of water derivedfrom the spring during storm events). These findings are important, not only because they add to our understandingof runoff generation from spring-fed catchments, but also in particular because they demonstrate that where springsare present, decreasing woody plant cover may augment runoff. Copyright 2006 John Wiley & Sons, Ltd.

KEY WORDS water yield; streamflow; woody plants; range hydrology; spring flow; ecohydrology; Edwards Plateau

Received 1 June 2005; Accepted 20 December 2005

INTRODUCTION

Springs are not a common feature on most semiarid rangelands because the soils have the capacity to storemost of the water that infiltrates—the greater part of which is eventually either evaporated or transpired.However, where the soils are shallow and the underlying geology is porous or fractured, springs may develop.For example, rangelands in karst geologic settings (where underlying rocks exhibit more solution-enlargedopenings than in other settings) often support springs and spring-fed rivers and streams. In these relatively drysettings, springs will supply intermittent baseflow; but during storm events, streamflow will be contributedby a larger percentage of the watershed.

In many drylands where springs have existed historically, their number and output have decreased. InTexas, for example, many springs have diminished or even disappeared in the last 150 years, a period duringwhich groundwater pumping has greatly accelerated (Brune, 2002). But woody plant cover has increasedsignificantly during this period as well, and may also have contributed to the decline in spring flow. There

* Correspondence to: Yun Huang, Department of Rangeland Ecology and Management, Texas A&M University, College Station, Texas77843, USA. E-mail: [email protected]

Copyright 2006 John Wiley & Sons, Ltd.

3278 Y. HUANG ET AL.

are many accounts of springs drying in the wake of woody plant encroachment, as there are of spring flowbeing revived after woody plant cover was reduced. Such accounts are consistent as well as numerous.

The linkage between woody plant cover and streamflow is well established in many Mediterraneanlandscapes, including California (Hibbert, 1983), Australia (Walker et al., 1993; Zhang et al., 2001), andSouth Africa (Van Wilgen et al., 1996). Although it seems intuitively logical that a similar correlation wouldexist in non-Mediterranean-type rangelands where springs exist, the evidence so far is mainly anecdotal.

One study that has documented an increase in spring flow following juniper removal was carried out ona 7Ð9-acre catchment in the Seco Creek watershed. Wright (1996) reported that spring flow increased from11Ð8 l/min to 14Ð3 l/min (equivalent to about 40 mm on an annual basis) following removal of Ashe juniper.Although this study is commonly cited as proof that shrub control leads to increased water yield, a limitationof the report is that the methodologies, calculations, and assumptions used are not described.

Wilcox et al. (2005) studied multiple catchments on the Annandale Ranch, in the western part of theEdwards Plateau in central Texas, where springs are not present. Finding that streamflow did not increasefollowing the removal of Ashe juniper, the authors hypothesized that shrub removal would augment streamflowonly if springs are or have been naturally present. This hypothesis is consistent with the hierarchical modeldeveloped by Huxman et al. (2005), which suggests that in semiarid landscapes, woody plant cover willinfluence streamflow only if streamflow has a baseflow component. (Runoff or streamflow can be dividedinto two components: baseflow and stormflow. Baseflow is that portion of runoff that cannot be attributedto an individual precipitation event; stormflow is that portion of runoff that can. During a runoff event, notonly current precipitation but also stored soil water/groundwater may contribute to the runoff hydrograph.The former is termed ‘event water’ and the latter ‘pre-event water.’)

The Edwards Plateau region is characterized by karst geology and boasts numerous spring-fed streams aswell as productive groundwater aquifers (Maclay, 1995; Mace et al., 2000). For this reason, the possibilitythat removal of Ashe juniper (Juniperus ashei Buchholz) will enhance streamflow and recharge in this areahas attracted great interest. In addition to favourable hydrological and geological conditions of the Plateau,the fact that Ashe juniper—which has a large capacity to intercept and transpire water—has significantlyincreased in density and coverage in the last century lends credence to the idea that its presence is limitingthe amount of available water (Owens and Ansley, 1997).

To explore this issue further, we investigated the nature of runoff from a small, spring-fed catchment onthe Edwards Plateau. We used hydrometric and isotope analysis to quantify runoff dynamics and to estimatechanges in streamflow following removal of about 60% of the Ashe juniper on the catchment.

STUDY AREA

Our study area, a small (19-ha) catchment, is located within the Honey Creek State Natural Area (run by theTexas Parks and Wildlife Division) in the eastern portion of the Edwards Plateau in west central Texas, USA(29°500N, 98°290W) (Figure 1). The catchment drains into the middle section of Honey Creek, a tributary ofthe Guadalupe River. A spring at the base of the catchment supplies intermittent baseflow; its source areahas not been identified but is presumably a seasonally saturated zone or perched groundwater on the hillslopeadjacent to the spring. Above the spring, the stream channel—which flows only during storm runoff—extendsfor about 30 m. Streamflow, therefore, is composed of intermittent baseflow from the spring and ephemeralstormflow produced by individual rainfall events.

The catchment consists of a gentle, broad valley flanked by steeper hills. The elevation of the catchmentranges from 369 to 393 m a.s.l. Long-term average precipitation (based on the analysis of records from nearbyNational Oceanic and Atmospheric Administration (NOAA) weather stations at Boerne, Spring Branch, andNew Braunfels, 1956–2002) is 909 mm per year. Precipitation in this region has bimodal distribution, thefirst peak occurring around May or June and the second around September or October. Using the Malmstrom

Copyright 2006 John Wiley & Sons, Ltd. Hydrol. Process. 20, 3277–3288 (2006)DOI: 10.1002/hyp

SPRINGS ON RANGELANDS 3279

Figure 1. Location of the Honey Creek catchment study area within the Honey Creek State natural area. The 1995 image (Texas NaturalResources Information System) shows the catchment before treatment, and the 2004 image (P-2 Energy Solutions) shows it post-treatment.The green lines on the two images indicate the location of the mainstream channel (the flume is located at the west end). Both are at a 1-m

resolution

method (Malmstrom, 1969), we estimate that potential evapotranspiration (PET) is around 1200 mm annually.In this upland catchment, the regional water table is about 50 m below ground surface (Ashworth, 1983).

The soils belong to Brackett Rock outcrop–Comfort complex, which is composed of shallow (10–15 cm)gravelly loams, clay loams, and stony clay interspersed with exposed limestone outcrops. These soils extendto depths ranging from about 15 to 50 cm (U.S. Department of Agriculture, 1984). Common woody speciesinclude Ashe juniper, live oak (Quercus virginiana), vasey shin oak (Quercus pungens var. vaseyana), andredberry juniper (Juniperus pinchotti ). The herbaceous vegetation includes a variety of forbs and a mixtureof short and mid-height grasses with indiangrass (Sorghastrum nutans), little bluestem (Schizachyrium andAndropogon spp.), and switchgrass (Panicum virgatum) dominating.



METHODS

Vegetation treatment

At the beginning of the study, the catchment was about 90% woodland, dominated by Ashe juniperinterspersed with live oak. During the summer of 2001, 60% of the catchment was cleared of Ashe juniper.Woody plants were left intact along the ridge top, on steep slopes, and along banks immediately adjacentto the spring at the base of the catchment. Selective cutting was done on the north, east, and southwesthillslopes (most of the oak trees were left in place). The trees were removed with hydraulic shears attachedto a small front-end loader—a method that results in much less soil disturbance than dozing the trees overor dragging large chains between two bulldozers. The removed trees were piled in rows parallel to theslope. Following treatment, the catchment was about 30% woodland, 20% mixed oak savanna, and 45%grassland.

Hydrometric measurements and analysis

In May 1999, the catchment had been instrumented with a 0Ð6-m (2-ft) H-flume, an automated weatherstation to record hourly streamflow and climate data, and a Druck PDCR1830 pressure transducer in thestilling well for the stage readings.

A careful evaluation of discharge records revealed occasional abrupt increases in discharge even thoughthere had been no precipitation. We attributed these increases to instrument ‘drift’ in the pressure transducerthat recorded water depth in the flume. On the assumption that the calibration slope of the pressure transducerdid not change, we corrected the discharge records by deducting the additional amount from the recordedstream-depth data set. When our corrected data set was compared with a set of manual measurements takenin the later part of the study, the match was good (Figure 2).

Baseflow was separated out by means of an automated baseflow filter (Arnold et al., 1995; Arnold andAllen, 1999).

There are several different ways to assess the influence of vegetation change on streamflow. These includedouble-mass analysis (Potts, 1984), measurement of changes in the runoff/precipitation ratio (Pitt et al., 1978),

Date

9/1999 3/2000 9/2000 3/2001 9/2001 3/2002 9/2002 3/2003

Ave

rage

Dai

ly S

tage

(m

m)

0

50

100

150

200

250

300

350

Creek Depth (mm)Manual Measurements

Figure 2. Daily mean stream depths, as corrected, compared with depths of manual measurements done later in the study



use of a hydroclimatic calibration equation (Chang, 2002), and regression analysis (Trimble et al., 1987). Weused an event-based regression model to detect changes in streamflow following the removal of juniper, andwe examined runoff/precipitation ratios as well.

Over the 4-year study period, 80 storm events were recorded—43 in the 2 years before juniper removaland 37 in the 2 years after. (‘On-and-off’ rainfall with fewer than 5 days between episodes was considered asingle storm event.) Runoff from a storm event was calculated as the sum of daily runoff amounts from thefirst day of the storm until the day before the next storm.

Both rainfall and runoff data were transformed through natural logarithmic models. Test statistics indicatedthat data so transformed met the regression assumptions, including homoscedasticity, and gave the bestdistribution of data points along the entire range of the regression. A covariance (ANCOVA) modelincorporating precipitation as a covariate was used to test for differences in runoff. The ANCOVA in theform of simple linear regressions for the pre-treatment and post-treatment periods are shown by the followingformula:

loge yi D b0i C b1 loge xi C εi �1�

where yi and xi represent runoff and rainfall on a per-event basis in the ith treatment period. All statisticaltests were performed at a significance level of 0Ð1. In retransformation to the original scale, bias was correctedby means of Duan’s smearing estimate (Duan, 1983).

Isotope analysis

Natural stable isotope tracers have been widely used to clarify runoff generation in various environ-ments (Kendall and McDonnell, 1998). One method that has proved very useful—particularly in humidlandscapes—is hydrograph separation. This method is based on the fact that individual precipitation events(event water) often exhibit a ratio of naturally occurring isotopes, such as oxygen-18 and deuterium, differentfrom that of groundwater and soil water—the sources of pre-event water (Buttle, 1994).

Monitoring of naturally occurring isotopes in precipitation and streamflow did not begin until after thevegetation treatment. In 2003 and 2004, baseflow was sampled for 18O analysis at intervals ranging fromweekly to monthly, to characterize its temporal variation—including differences in isotopic composition atdifferent flow rates. Concurrently, aggregate samples of precipitation for the same time period were sampledfrom a bucket collector (similar to that used by Newman et al. (1998) to prevent evaporation. An ISCO

automatic sampler coupled with a flow meter was used for more detailed sampling during four distinct stormevents. Samples were taken at intervals of 20, 30, and 40 min, and more frequently during the rising limb ofthe hydrograph and during peak flows. All samples were refrigerated in sealedbottles until analysed.

Hydrograph separations were done for two of the four events: one on 20 February 2003 (22 samples) andone on 27 June 2004 (26 samples). For both of these, the separations yielded clearly distinguishable υ18Osignatures for the event versus the pre-event water. The υ18O isotope values were measured by the standardCO2-water equilibrator, interfaced with a Prism II isotope ratio mass spectrometer, at the University of Texas(Austin) and are expressed in terms of Vienna Standard Mean Ocean Water (V-SMOW) (Coplen, 1988). Theanalytical precision of the υ18O measurements is 0Ð14‰ at a 95% confidence level, based on analyses oflaboratory standards and duplicate samples.

Hydrograph separation into event and pre-event components is based on the following mass balance andmixing equations (Buttle, 1994):

Qt D Qp C Qe �2�

CtQt D CpQp C CeQe �3�

f D �Ct � Ce�/�Cp � Ce� �4�



where Qt is the total streamflow during an event; Qp and Qe are contributions from pre-event and event water;Ct, Cp, and Ce are the isotopic compositions of streamflow, pre-event water, and event water, respectively;and f is the contribution of pre-event water to total streamflow.

RESULTS

Runoff dynamics

The comparison of annual precipitation and runoff shown in Table I is based on the monitoring year (lateAugust–late August) instead of the calendar year. The average annual precipitation was 832 mm, the minimumrecorded being 433 mm and the maximum 1270 mm. The average annual runoff was 180 mm (minimum of16 mm and maximum of 431 mm), or about 22% of precipitation on average (varying from 4% to 34%,depending on the year). Baseflow made up about 50% of total runoff.

Higher amounts of stormflow resulted in proportionally higher runoff. In 1999, the first year of thestudy, maximum daily runoff was only 0Ð1 mm (Table I). In contrast, on 1 July 2002, 53 mm of runoffwas measured—the largest amount recorded for a single day. It was produced by an extreme floodevent that began on 29 June 2002 and continued on and off for 20 days, ending on 15 July 2002. Totalprecipitation during this period was 546 mm, of which 482 mm fell in the first week (Figure 3). The flumewas overtopped during the flood peak, causing runoff to be underestimated. Over the 20 days, storm runoffwas at least 260 mm, which accounted for 60% of the runoff in that year and 47% of the precipitationmeasured during this flood event. The elevated baseflow lasted more than 16 days after the storm hadended.

Figure 4 demonstrates that large runoff events accounted for most of the runoff during the 4-year study.Nearly 80% of the runoff was generated by the largest 20% of the runoff-producing storms, and the fivelargest runoff events accounted for 60% of total runoff.

Components of runoff

Figure 5 illustrates the isotopic composition of the runoff, as measured from baseflow (pre-event water) andprecipitation (event water) samples collected over an 18-month period. We found that the isotopic compositionof the rainfall was highly variable and exhibited little seasonal trend. Precipitation υ18O values ranged from�1Ð41 to �7Ð14‰, with a mean of �4Ð05‰ and a coefficient of variation of 40%.

Baseflow isotopic composition was very consistent throughout the sampling period. The υ18O values rangedfrom �3Ð82 to �4Ð17‰, with a mean of �4Ð00‰ and a coefficient of variation of 2Ð7%. As shown in Figure 5,this composition approximates the mean value for precipitation. For hydrograph separation, its consistencyallows us to assume that the isotopic composition of a baseflow sample taken before a storm event will berepresentative of the isotopic composition of pre-event water throughout the event.

Table I. Summary of annual hydrologic data for the study site

Treatmentperiod

Duration Precipitation(mm)

Runoff(mm)

Baseflow(mm)

Maximumdailyrunoff(mm)

Runoffratio

T(°C)

Pre 9/1999 to 8/2000 433 16 12 0Ð1 0Ð04 20Ð59/2000 to 8/2001 852 121 67 11Ð3 0Ð14 19Ð0

Post 9/2001 to 8/2002a 1270 431 177 53Ð1 0Ð34 19Ð29/2002 to 8/2003 774 150 85 14Ð3 0Ð19 18Ð8

a When the largest event was excluded, total runoff and baseflow for this record year were 171 and 83 mm, respectively.



Hou

rly R

ainf

all (

mm

) 0

20

40

60

Date

6/27/2002 6/29/2002 7/1/2002 7/3/2002 7/5/2002

Run

off (

L/S

)

050

100150200250300350

Figure 3. Hourly hydrograph of the largest precipitation event (June–July 2002) recorded during the 4-year study

Event Rank0 20 40 60 80

Run

off (

mm

)

0

50

100

150

200

250

300

Cum

ulat

ive

Run

off (

%)

0.0

0.2

0.4

0.6

0.8

1.0

Ranked RunoffCumulative

Runoff (mm)0 50 100 150 200 250

Pro

babi

lity

0.0

0.2

0.4

0.6

0.8

1.0

Figure 4. Runoff events during the study period (August 1999–August 2003), ranked by magnitude

As mentioned above, hydrograph separations were done for two storm events. The findings were as follows:

Winter storm (20 February 2003). The baseflow sample used was collected one day before this storm.Precipitation during the 2 months preceding the storm event had totalled 55 mm—42 mm between 20December 2002 and 20 January 2003; and 13 mm between 20 January and 20 February. The total rainfall forthe storm was 48 mm over 15 h, and the highest intensity was 19 mm/h. The runoff response was rapid, andthe peak rate measured was 133 l/s (Figure 6a). Pre-event water contributed only about 41% of peak flow,but accounted for about 46% of hydrograph volume. In this region, storms of this magnitude occur with afrequency of just under one per year.



Date

1/03 5/03 9/03 1/04 5/04

O-1

8 is

otop

ic c

ompo

sitio

n (p

er m

il)

-8

-7

-6

-5

-4

-3

-2

-1

BaseflowPrecipitation

Figure 5. Time series of the 18O isotopic compositions of baseflow and precipitation at the study area. Most samples were collected roughlyat monthly intervals

Rai

nfal

l (m

m)

0

5

10

15

20

25Total P=48 mm

Time

00:00 04:00 08:00 12:00 16:00 20:00 00:00

Run

off (

L/S

)

020406080

100120140160

TotalPre-event

(a) (b)

Rai

nfal

l (m

m)

0

5

10

15

20

25Total P=45 mm

Time

12:00 16:00 20:00 0:00 4:00 8:00 12:00

Run

off (

L/S

)

020406080

100120140160

TotalPre-event

Figure 6. 18O-isotope hydrograph separation for (a) a winter event (20 February 2003) and (b) a summer event (27 June 2004)

Summer storm (27 June 2004). Precipitation during the 2 months preceding this storm event had totalled271 mm—228 mm between 27 April and 27 May 2004; and 43 mm between 27 May and 27 June. The totalrainfall for the storm was 45 mm, the bulk of which fell within 4 h. The highest intensity was 27 mm/h, andthe peak runoff rate was 118 l/s. Only 20% of total flow at the runoff peak consisted of pre-event water.Volumetrically, pre-event water formed about 16% of the event hydrograph (Figure 6b).



Rainfall (mm)

Run

off (

mm

)

Rainfall (mm)

PretreatmentPre reg.PosttreatmentPost reg.

1 3 7 20 50 150 400

Rainfall (mm)

Run

off (

mm

)

1 3 7 20 50 150 400

All Non-Summer

Summer

0.1

0.4

1.0

3

7

20

50

1 3 7 20 50 150 400

0.1

0.4

1.0

3

7

20

50

150

0.1

0.4

1.0

3

7

20

50

150(a) (b)

(c)

Figure 7. Event-based regression plot of runoff amount versus rainfall amount, based on (a) 3 1/4 years of data (the largest event of thepost-treatment period was not included); (b) non-summer events (events occurring between October and April); and (c) summer events

(events occurring between May and September)

Influence of juniper cover

In comparing rainfall–runoff relationships before and after vegetation treatment, we excluded data fromtwo periods: (1) the first 9 months of the first year (late August 1999 to late May 2000) because the first6 months of that year were extremely dry; and (2) late June–early July 2002, because an extreme rainfall andrunoff-producing event occurred during that time, the inclusion of data from which would have had an overlylarge leveraging effect. Inclusion of these data would increase the chances of seeing a treatment effect.

The event-based regression analysis for the remainder of the data indicates a significantly higher runoffefficiency (p D 0Ð04) after the juniper was removed than before (Figure 7). On a per-event basis, mean runoffwas 5Ð5 mm for the pre-treatment period and 8Ð8 mm for the post-treatment period—in other words, anincrease of 60% after removal of juniper.

To evaluate potential seasonal differences, we did event-based regression analysis separately for summerevents and non-summer events—the rationale being that high-intensity, short-duration rainfall is more commonin the summer. The results of these analyses led us to conclude that treatment effects were not detectablefor summer events, but were significant (p D 0Ð02) for non-summer events. On the basis of these regressionrelationships, we anticipate that removal of juniper will increase streamflow at this site by around 46 mmannually, representing about 5% of precipitation.

DISCUSSION AND CONCLUSIONS

We found that at our study site, runoff contributed significantly to the water budget—ranging from as littleas 4% to as much as 34% and averaging about 22% over 4 years of observation. The baseflow- to stormflow-ratio was about 50/50.



The hydrograph separation analyses provided important clues concerning the relative importance ofgroundwater during stormflow events. The analysis of the winter storm described above indicates that forstorms of longer duration and lower intensity, pre-event water makes up about 50% of the runoff. Although,Buttle (1994) emphasized that one must exercise caution in using hydrograph separation for determiningsource areas, it seems likely that at our site the pre-event water and the spring flow have the same sourcearea. This being the case, estimates of pre-event water provide at least a lower limit for contributions fromthe spring during storm runoff. For the summer (higher intensity) storm, pre-event water was lower, but stillaccounted for 25% of the runoff.

A few other hydrograph separation studies have been conducted in water-limited environments—themajority in locations having Mediterranean or otherwise winter-dominated moisture regimes. Most of thesestudies found that pre-event waters dominate storm runoff. Nolan and Hill (1990), working on a 10Ð6-km2

chaparral watershed in California where annual precipitation varied from 400 to 800 mm (depending onelevation), reported that pre-event water accounted for more than 57% of storm runoff. Neal et al. (1992),working in two oak-forest locations in Catalonia, Spain (with annual precipitation of 870 and 658 mm),reported that storm runoff was mostly pre-event water. Similarly, Taha et al. (1997) reported that stormhydrographs for an oak-forest catchment in the south of France consisted of up to 80% pre-event subsurfacewater. Sandstrom (1996), in a study of forested and degraded catchments in semiarid tropical Tanzania, foundthat contributions from a soil’s saturated zone made up only 25% of the total discharge (the main mechanismsof runoff generation at this site were overland flow and throughflow; rainfall occurs primarily during winterand averages 807 mm/year).

Our results here show both similarities to and differences from those of an earlier study by Wilcoxet al. (2005), on the Annandale Ranch in the western part of the Edwards Plateau. They examinedstreamflow–woody plant cover relationships by monitoring first-order catchments without springs. Onesimilarity is that for both types of systems—first-order catchments with springs and those without—thebulk of the runoff is generated by a few large, flood-producing precipitation events. A difference is thatrunoff/precipitation ratios from spring-fed catchments are higher than those from catchments without springs.For example, runoff at the Annandale Ranch accounted for less than 5% of the precipitation. Another differenceis that Wilcox et al. (2005) observed that in the absence of spring flow and major surface disturbance,changes in woody plant cover had a minimal effect on streamflow. In contrast, the Honey Creek catchmentstudy suggests that where springs are present, streamflow in first-order catchments can be augmented throughwoody plant removal. Following the removal of approximately 60% of the Ashe juniper in the catchment,streamflow increased—though interestingly, only for non-summer events.

Our results, augmented by the hydrograph separation analyses, are consistent with the hypothesis thatthe decrease in evaporative demand brought about by woody plant removal increases the contribution fromgroundwater or spring flow. An alternative hypothesis is that surface disturbances created during woodyplant removal facilitate greater surface runoff/overland flow, and thus higher streamflows. But if this were acontributing mechanism, we would expect to see higher streamflow during summer events, which we did not.

These findings are important because they corroborate as well as constrain the numerous but poorlydocumented accounts of enhanced spring or seep flow following reduction of Ashe juniper. The only otherwork reporting an increase in spring flow following partial removal of Ashe juniper is an unpublished studyby Wright (1996), who reported increases in flow of about 40 mm/year, or 5% of annual precipitation.However, the work of Dugas et al. (1998) on evapotranspiration yielded some relevant data: their finding thatevapotranspiration rates were about 40 mm/year less for sites cleared of Ashe juniper is consistent with ourobservation that streamflow increased by approximately 46 mm/year.

On the basis of our study results and those of related work, we estimate that in the Edwards Plateauregion, reduction of woody plant cover can increase streamflow and/or groundwater recharge at the first-ordercatchment scale by about 5% of annual precipitation for average or above average years. Many questionsremain, however; for example, how long will these increases persist following treatment? How variable are



results with climate and location? And finally, to what extent do these results scale up? These questions canonly be answered from larger-scale and longer-duration studies.

ACKNOWLEDGEMENTS

Phillip Wright and Larry Allen of the Natural Resource Conservation Service generously provided streamflowand precipitation data. This work was supported by grants from the Edwards Aquifer Authority, the Guadalupe-Blanco River Authority, the Edwards Region GCLI, the San Antonio Water System, and the National ScienceFoundation (NSF Grant award number 0233667, Collaborative Research: WCR: Ecohydrology of SemiaridWoodlands: Role of Woody Plants in the Water Cycle). Support by the Texas Parks and Wildlife Departmentis also gratefully acknowledged.

REFERENCES

Arnold JG, Allen PM. 1999. Automated methods for estimating baseflow and ground water recharge from streamflow records. Journal ofthe American Water Resources Association 35: 411–424.

Arnold JG, Allen PM, Muttiah R, Bernhardt G. 1995. Automated baseflow separation and recession analysis techniques. Ground Water 33:1010–1018.

Ashworth JB. 1983. Ground-Water Availability of the Lower Cretaceous Formations in the Hill Country of South-Central Texas . TexasDepartment of Water Resources: Austin, TX.

Brune G. 2002. Springs of Texas . Texas A&M University Press: College Station, TX; 566.Buttle JM. 1994. Isotope hydrograph separations and rapid delivery of pre-event water from drainage basins. Progress in Physical Geography

18: 16–41.Chang M. 2002. Forest Hydrology: An Introduction to Water and Forests . CRC Press: Boca Raton, FL.Coplen TB. 1988. Normalization of oxygen and hydrogen isotope data. Chemical Geology 72: 293–297.Duan N. 1983. Smearing estimate: a nonparametric retransformation method. Journal of the American Statistical Association 73: 605–610.Dugas WA, Hicks RA, Wright P. 1998. Effect of removal of Juniperus ashei on evapotranspiration and runoff in the Seco Creek watershed.

Water Resources Research 34: 1499–1506.Hibbert AR. 1983. Water yield improvement potential by vegetation management on western rangelands. Water Resources Bulletin 19:

375–381.Huxman TE, Wilcox BP, Breshears DD, Scott RL, Snyder KA, Small EE, Hultine K, Pockman WT, Jackson RB. 2005. Ecohydrological

implications of woody plant encroachment. Ecology 86: 308–319.Kendall C, McDonnell JJ. 1998. Isotope Tracers in Catchment Hydrology. Elsevier: Amsterdam; 837.Mace RE, Chowdhury AH, Anaya R, Way S-CT. 2000. Ground Water Availability of the Trinity Aquifer, Hill Country Area, Texas: Numerical

Simulations through 2050 . Texas Water Development Board: Austin, TX.Maclay RW. 1995. Geology and Hydrology of the Edwards Aquifer in the San Antonio Area, Texas . U.S. Geological Survey: Austin, TX; 64.Malmstrom VH. 1969. A new approach to the classification of climate. Journal of Geography 68: 351–357.Neal C, Neal M, Warrington A, Avila A, Pinol J, Roda F. 1992. Stable hydrogen and oxygen isotope studies of rainfall and streamwaters

for two contrasting holm oak areas of Catalonia, northeastern Spain. Journal of Hydrology 140: 163–178.Newman BD, Campbell AR, Wilcox BP. 1998. Lateral subsurface flow pathways in a semiarid ponderosa pine hillslope. Water Resources

Research 34: 3485–3496.Nolan KM, Hill BR. 1990. Storm runoff generation in the Permanente Creek drainage basin, west central California: an example of flood-wave

effects on runoff composition. Journal of Hydrology 113: 343–367.Owens KW, Ansley J. 1997. Ecophysiology and growth of Ashe and redberry juniper. In Juniper Symposium , Taylor CA (ed.). Texas A&M

University: San Angelo, TX; 19–31.Pitt MD, Burgy RH, Heady HF. 1978. Influences of brush conversion and weather patterns on runoff from a northern California watershed.

Journal of Range Management 31: 23–27.Potts DF. 1984. Hydrologic impacts of large-scale mountain pine beetle (Dendroctonus Ponderosa Hopkins) epidemic. Water Resources

Bulletin 20: 373–377.Sandstrom K. 1996. Hydrochemical deciphering of streamflow generation in semi-arid East Africa. Hydrological Processes 10: 703–720.Taha A, Gresillon JM, Clothier BE. 1997. Modelling the link between hillslope water movement and stream flow: application to a small

Mediterranean forest watershed. Journal of Hydrology 203: 11–20.Trimble SW, Weirich FH, Hoag BL. 1987. Reforestation and the reduction of water yield on the Southern Piedmont since circa 1940. Water

Resources Research 23: 425–437.U.S. Department of Agriculture. 1984. Soil Survey of Comal and Hays Counties Texas . U.S. Department of Agriculture: Washington, DC.Van Wilgen BW, Cowling RM, Burgers CJ. 1996. Valuation of ecosystem services—A case study from South African fynbos ecosystems.

Bioscience 46: 184–189.Walker J, Bullen F, Williams BG. 1993. Ecohydrological changes in the Murray-Darling Basin. I. The number of trees cleared over two

centuries. Journal of Applied Ecology 30: 265–273.



Wilcox BP, Owens MK, Knight RW, Lyons RK. 2005. Do woody plants affect streamflow on semiarid karst rangelands? EcologicalApplications 15: 127–136.

Wright PN. 1996. Spring enhancement in the Seco Creek water quality demonstration project. Seco Creek Water Quality DemonstrationProject . U.S. Department of Agriculture: Natural Resource Conservation Service, Temple, Texas.

Zhang L, Dawes WR, Walker GR. 2001. Response of mean annual evapotranspiration to vegetation changes at the catchment scale. WaterResources Research 37: 701–708.


Documents

Springs on rangelands: runoff dynamics and influence of woody plant cover