Semi-Arid Aquifer Responses to ForestRestoration Treatments and Climate Changeby Clinton J.W. Wyatt1, Frances C. O’Donnell, and Abraham E. Springer

AbstractThe purpose of this study was to develop an interpretive groundwater-flow model to assess the impacts that planned forest

restoration treatments and anticipated climate change will have on large regional, deep (>400 m), semi-arid aquifers. Simulationswere conducted to examine how tree basal area reductions impact groundwater recharge from historic conditions to 2099. Novelspatial analyses were conducted to determine areas and rates of potential increases in groundwater recharge. Changes in rechargewere applied to the model by identifying zones of basal area reduction from planned forest restoration treatments and applyingrecharge-change factors to these zones. Over a 10-year period of forest restoration treatment, a 2.8% increase in recharge toone adjacent groundwater basin (the Verde Valley sub-basin) was estimated, compared to conditions that existed from 2000 to2005. However, this increase in recharge was assumed to quickly decline after treatment due to regrowth of vegetation andforest underbrush and their associated increased evapotranspiration. Furthermore, simulated increases in groundwater rechargewere masked by decreases in water levels, stream baseflow, and groundwater storage resulting from surface water diversions andgroundwater pumping. These results indicate that there is an imbalance between water supply and demand in this regional, semi-aridaquifer. Current water management practices may not be sustainable into the far future and comprehensive action should be takento minimize this water budget imbalance.

IntroductionLandscape-scale forest restoration thinning and burn-

ing treatments are planned for approximately 240,000 ha(600,000 acres) of over-dense ponderosa pine forest withinthe Coconino and Kaibab National Forests along theMogollon Rim in Northern Arizona. This area has a semi-arid climate and receives 500 to 760 mm (20–30 inches)of precipitation per year as summer monsoon rain andwinter snow. This area also has two deep (>400 m)regional water supply aquifers: the Redwall-Muav(R-) and Coconino (C-) Aquifers. Basin-fill aquifers alsoexist in the adjacent transition zone and basin and rangeprovinces found in south-central Arizona.

Forest restoration treatments are planned to beginwithin the next decade and are expected to take approxi-mately 10 years to complete (USDA Forest Service 2012).The purpose of this land management action, called theFour Forest Restoration Initiative (4FRI), is to reducethe threat of catastrophic wildfire; restore forest health

1Corresponding author: School of Earth Sciences andEnvironmental Sustainability, Northern Arizona University, P.O. Box4099, Flagstaff, AZ 86011; cjw287@nau.edu

School of Earth Sciences and Environmental Sustainability,Northern Arizona University, P.O. Box 4099, Flagstaff, AZ 86011.

Received June 2013, accepted February 2014.© 2014, National Ground Water Association.doi: 10.1111/gwat.12184

and resiliency; and restore streams, springs, and biologicfunctions of forested watersheds. Based on previousregional and international studies that show surface wateryield increasing after reductions in tree basal area inforested lands (Bosch and Hewlett 1982), a hypothesiswas developed that groundwater recharge would increaseto the deep, regional Redwall-Muav and CoconinoAquifers and basin-fill aquifers following forest restora-tion treatments. Through groundwater modeling with esti-mations of recharge that account for changing forestcover and climate change, assessments were made of howlandscape-scale forest restoration may affect groundwaterrecharge to regional aquifers.

Many previous studies have looked at the relationshipbetween land-use and land-cover change and the hydro-logic system. Researchers have attempted to quantify theeffect that removing trees has on the water budget. Mostof this previous work (Bosch and Hewlett 1982) focusedon the relationship between tree cover and surface wateryield. Some studies (Gottfried 1991) have investigated thetreatment/yield relationship in arid and semi-arid conifer-dominated watersheds. Baker (2003) reported that wateryield increased by 15% to 40% when basal area wasreduced by 30% to 100% in ponderosa pine watersheds innorth-central Arizona. A systematic review of the litera-ture from coniferous forests worldwide found an averageof 0% to 50% increase in water yield when basal area isreduced by 5% to 100% (Wyatt 2013).

NGWA.org Groundwater 1

Fewer studies have attempted to quantify variablesin the water budget other than surface water yield,such as evapotranspiration (ET), soil water storage, andgroundwater recharge. Eddy flux-based measurementsfollowing a 35% reduction in basal area in the semi-arid ponderosa pine forest of the Centennial Forest nearFlagstaff, Arizona, reported a 17 and 15% reduction in ETin the first 2 years with an average decrease of 4% overthe 5 years following thinning (Dore et al. 2012). Otherstudies reported increases in snowpack accumulationfollowing tree removal (Veatch et al. 2009; Harpold et al.2013). Owing to a combination of increased inputs fromsnowmelt and decreased ET, soil moisture is significantlyhigher in thinned ponderosa pine forests (Zou et al. 2010),though the relative contribution of these two factors hasnot been determined.

Owing to the lack of previous studies, this mod-eling study examines the question of how changes ingroundwater recharge associated with forest restoration(mechanical thinning and burning) influence groundwaterresources. There have been even fewer studies that inves-tigated the effect that removing trees has on groundwaterrecharge and soil moisture storage in arid and semi-aridforested watersheds (Bazan et al. 2012; Scanlon et al.2005). Because of the lack of published literature on theeffects of tree removal on groundwater recharge, an inter-pretive groundwater-flow model was developed and usedto understand how landscape-scale forest restoration treat-ments might impact groundwater recharge and regionalaquifers.

The 4FRI, proposed by the U.S. Forest Service,is a collaborative effort to reduce the threat of catas-trophic wildfires and restore forest ecosystem healththroughout four national forests along the Mogollon Rim,Arizona—the Kaibab, Coconino, Tonto, and Apache-Sitgreaves National Forests (USDA Forest Service 2012).The initial treatments will include mechanical thin-ning and burning that will be applied to approximately240,000 ha (600,000 acres) of Kaibab and CoconinoNational Forest land as early as 2014, pending approvalof the environmental impact statement (EIS).

The 4FRI first analysis area EIS has four treatmentalternatives. Alternative C proposes the conservation oflarge trees and would thin 175,634 ha (434,001 acres)and burn 240,064 ha (593,211 acres; USDA Forest Ser-vice 2012). Alternative C was selected for the interpre-tive groundwater model simulations because it was oneof two (B and C) that the NFS considered likely tobe implemented. Non-4FRI forest restoration treatments,called shelf-stock, within these forests were previouslyanalyzed under separate National Environmental PolicyAct (NEPA) processes and were included in the simu-lations (USDA Forest Service 2012). Shelf-stock foresttreatment data were compiled from multiple sources on theCoconino and Kaibab National Forests. There is consider-able interest in how the proposed treatments will impactdownstream areas of high water use, such as the VerdeValley (Figure 1).

Any projections of the future availability of ground-water supplies in the southwest must also consider thepotential impact of climate change. Climate models gen-erally predict a warmer, drier future for the region (Maureret al. 2007). The changes are expected to reduce rechargeboth directly, through a reduction in precipitation, andindirectly, through increased ET under warmer conditions.

Purpose and ObjectivesTo determine the response of the regional Redwall-

Muav and Coconino Aquifers to the 4FRI treatmentsand changing climatic conditions, the following objectiveswere designed for this study:

1. simulate changes in groundwater recharge from land-scape change and changing climatic conditions; and

2. assess the impacts that these changes may have on thegroundwater budget of Northern Arizona, with a focuson the Verde Valley groundwater catchment area andassociated tributaries.

The study used an interpretive modeling approachwith a recently published regional groundwater-flowmodel. The Northern Arizona Regional Groundwater-Flow Model (Pool et al. 2011), hereafter referred to asthe NARGFM, was used to simulate changes in rechargeto aquifers of the Mogollon Rim in Northern Arizonafollowing planned forest restoration treatments and pre-dicted climate change. The NARGFM was developed tosimulate the interactions between deep regional aquifers,streams, and springs in Northern Arizona and to assessthe adequacy of groundwater resources and the effectthat increased pumping, especially in the Verde Riverbasin (Figure 1), would have on these resources, thereforeallowing it to be valuable for studying regional changesto land-use/land-cover in this study.

MethodsThe groundwater system in Northern Arizona was

modeled using the NARGFM with recharge scenariosdeveloped using novel techniques for estimating thechange in recharge due to planned land-use changes. Sce-narios were developed for a baseline of calibrated modelconditions and for recharge changes from anticipated for-est restoration and climate change. Groundwater responseto these scenarios was interpreted through baseflow trendsat major U.S. Geological Survey (USGS) stream gauges,regional monitoring wells, and regional aquifer waterbudget changes.

Northern Arizona Regional Groundwater-Flow ModelThe NARGFM uses the three-dimensional finite-

difference modular groundwater-flow model codeMODFLOW-2005 (Harbaugh 2005). The NARGFMboundary conditions include surface watershed bound-aries, groundwater basin divides, and low-permeability

2 C.J.W. Wyatt et al. Groundwater NGWA.org

Figure 1. Simulated predevelopment groundwater-flow system in the Verde Valley area. Included are the 4FRI analysis area,stream boundary conditions, drain boundaries conditions, wells, and gauges of analysis.

crystalline rocks along the southern boundary of theVerde River basin and adjacent sub-basins (Figure 1).By simulating a large region and defining knownphysical boundaries, major groundwater-flow divideswithin the model area were simulated rather than set atarbitrary locations. The entire boundary of the modelwas represented as a no-flow boundary except wheregroundwater outflow was simulated at discrete locationsalong streams. There was assumed to be no groundwaterinflow anywhere along the model boundary.

The NARGFM model grid comprises 600 rows,400 columns, and three layers totaling 720,000 1 kmby 1 km (0.62 by 0.62 mi) grid cells. The model gridis rotated counterclockwise 60 degrees west of northto better align with regional structural trends that areassumed to influence groundwater flow. The NARGFMhas three model layers that were used to representhydrostratigraphic units within Northern Arizona. Layer3, the lowest layer, extends across the entire model areaand represents the Redwall-Muav Aquifer, except in thesouthern and eastern parts of the model domain wherethe Redwall-Muav Aquifer is absent and crystalline rocksare present. Layer 2 is less extensive and represents theSupai Formation, a confining unit, which extends overmost of the Colorado Plateau, as well as sands and gravelsin the Verde and Big Chino Valley extensional basins andthe lower volcanic unit in Little Chino Valley and Upper

Agua Fria sub-basins. Layer 1 is the least extensive layerwithin the model and represents the Coconino Aquifer, thealluvial basin-fill aquifers located in Big Chino Valley,Little Chino Valley, and Agua Fria sub-basins, and theVerde Formation in the Verde Valley.

Inflows and outflows were simulated at locationsof natural and artificial recharge, ET, streams, springs,and groundwater withdrawals. Hydrostratigraphic prop-erties were distributed across the model domain, basedon literature, and calibration values were applied whereappropriate. These include variables such as hydraulicconductivity, transmissivity, anisotropy (vertical and hor-izontal), specific storage, and specific yield.

The NARGFM was calibrated to steady-state condi-tions for groundwater flow that were assumed to existin 1910 and for transient conditions between 1910 and2005 (Pool et al. 2011). Recharge for the NARGFMwas derived from the Basin Characterization Model(BCM) of Flint and Flint (2008) and isotopic analy-ses developed by Blasch and Bryson (2007). Althoughrecharge to the deep, bedrock, aquifers of the region isfocused and ephemeral, the NARGFM uses areal dis-tributed, average annual recharge values. The BCM usesrecharge values derived from PRISM that are adjustedfor baseflow interpretations from stream gauging stations,providing an improved simulation of the response ofaquifers.

NGWA.org C.J.W. Wyatt et al. Groundwater 3

Forest Restoration Recharge ScenariosA set of relatively novel methods was used for

estimating changes in recharge from forest restorationtreatments. Because the NARGFM is a groundwatermodel and therefore does not model surface vegetationor precipitation, the only hydrologic parameter in themodel that was manipulated to simulate these changes wasthe specified flux recharge property. For the interpretivesimulation period from 2006 to 2099, all other parameters,including pumping and ET of riparian areas were keptat the same values used for the last stress period(2000–2005) of the calibrated groundwater-flow model.Other regional water supply studies have simulatedchanges in pumping that may result from increased humanpopulation in the future (ADWR 2011; USDI Reclamation2012; Garner et al. 2013). However, these variables werenot changed in this study because the objective was toisolate the impact of forest treatments and climate changeon groundwater recharge.

A baseline scenario based on the period of instrumentrecord was created to simulate future precipitation andrecharge values similar to recent historical values for thestudy area (Figure 2). The baseline scenario representsrecharge-conditions-absent forest restoration and climatechange. To project historic recharge conditions in theinterpretive simulations, a synthetic annual precipitationrecord was created from 2006 (the end of the calibratedNARGFM was 2005) to 2099 by randomly samplingfrom annual precipitation values from 1971 to 2000precipitation normal values (PRISM Climate Group2012). The precipitation normal was used by the USGSto estimate annual average recharge rates in the calibratedNARGFM. A longer normal from 1940 to 2005 becameavailable during the USGS study and was used to createscaled decadal variations in recharge for the period ofrecord for the calibrated NARGFM. Recharge values foreach new future stress period were estimated from thesynthetic precipitation values.

Because of the relatively few studies that havequantified groundwater recharge following tree removal insemi-arid, conifer-dominated watersheds, it was assumedthat groundwater recharge responded in a similar wayas literature documents runoff and streamflow respondingafter tree removal in these ecosystems. Specifically, whenponderosa pine tree basal area is reduced by 30% to100%, surface water yield may increase 15% to 40%(Baker 2003). Therefore, the assumption was made thatas ponderosa pine basal area is reduced, groundwaterrecharge may increase by 15% to 40%.

There were multiple, converging lines of evidencethat aided in this assumption, including a diminishmentof recharge for the historic period of record and increasedrecharge from experimental studies. Pool et al. (2011)showed a significant decline (∼50%) in baseflow of theSalt River for the period of observation (1910–2005)when forest basal area increased, except from the1960s to the 1990s during a wet phase of the PacificDecadal Oscillation. Covington and Moore (1994) useda multiresource forest growth and yield simulation

1960 1980 2000 2020 2040 2060 2080 2100150

200

250

300

350

400

450

500

550

600Mean of Emissions

Scenarios ± 1 Std. Dev.

PRISM Data (1971−2000)& Synthetic Data (2006−2099)

Year

Pre

cipi

tatio

n (m

m/y

r)

Figure 2. Precipitation values for 1971 to 2000 PRISM Base-line and climate change scenarios. PRISM data were used toestimate recharge in the calibrated NARGFM. The baselinescenario is precipitation-absent forest restoration and climatechange. Climate change scenarios are the mean, mean + 1standard deviation, and mean − 1 standard deviation ofthe IPCC A1B, A2, and B1 emission scenario precipitationprojections plus changes from forest restoration.

model (ECOSIM) to simulate a 26% reduction instream flow from the region as tree density increasedfrom presettlement to current conditions. Baker (2003)reported on ponderosa pine watersheds in semi-aridclimates, a scenario that matches the conditions of the4FRI treatment area. There are a number of studiesthat report ET (Kolb 2009) decreases and snowpackaccumulation (Stegman 1996) increases following theremoval of trees. This allows for an increase in theavailability of water to infiltrate into the subsurfaceand percolate down to recharge the aquifers. There wasgood evidence to support the notion that afforestation,or adding trees, decreases groundwater recharge (Allenand Chapman 2000), and therefore removing trees wouldhave the opposite effect. Furthermore, a systematic reviewconducted in conjunction with this work (Wyatt 2013)showed that in 13 of 15 studies, when trees were removedfrom a watershed, there was an increase in groundwatertable elevation. All of these converging lines of evidencesupported our assumptions of increased recharge fromreduction in basal area.

Quantitative estimates of the effect of forest restora-tion on recharge are not available in the literature, so dataon other components of the water balance are used toconstrain the estimated increase due to forest restoration.The annual water balance of southwestern ponderosa pineforests is approximated as

P = FET × P + FRO × P + FR × P (1)

where P is annual precipitation and F ET, F RO, and F R

are the fractions of precipitation partitioned to ET, runoff,and recharge, respectively. For an unthinned forest, F ET isequal to 0.89 (Dore et al. 2012) and F RO is approximately0.07 (Wilcox et al. 1997; Flerchinger and Cooley 2000),

4 C.J.W. Wyatt et al. Groundwater NGWA.org

making F R equal to 0.04. For a thinned forest, theequation becomes

xs × P = xET × FET × P + xRO × FRO × P

+ xR × FR × P (2)

where the factors xi indicate the fractional change in eachcomponent of the water budget due to forest restorationwith x s representing the fractional increase in precipitationinputs due to increased snowpack. Because a reliableestimate of x s is not available in the literature, it is setequal to 1 in the interest of providing a conservativeestimate of recharge increase. The change in ET, xET, isequal to 0.96 (Dore et al. 2012) and the runoff factor, xRO,is set to a range of values (1.15–1.4) that correspond torunoff increases observed for forests following basal areareductions that are characteristic of restoration treatments(Baker 1986; Bosch and Hewlett 1982). Solving for xR,the only unknown in the equation, gives a range of 1.19 to1.63, corresponding to a 19% to 63% increase in recharge.Because this range is in magnitude similar to the change inrunoff, it is assumed that the fractional change in rechargedue to forest restoration follows the same pattern as thefractional change in runoff.

Information on the 4FRI scenario C pretreatment andposttreatment basal area projections for stands within the4FRI treatments were obtained from the GIS specialistfor the 4FRI (Mark Nigrelli, personal communication2012). These projections were simulated using ForestVegetation Simulator (FVS) modeling of tree stand data(Dixon 2002). The percent change in tree basal area (x )was calculated by finding the percent difference betweenpre- and posttreatment basal area projections. Based onthe work by Baker (2003), the percent change in wateryield (y) was described as:

y = 0.36x + 4.3 (3)

This percent change in water yield (y) fromEquation 3 was applied to the groundwater model as afactor to adjust recharge based on average percent changein basal area per forest stand.

As an example of how the basal area changerelationship was applied, areas that had a basal areareduction of approximately 30% to 39% (an averagechange of ∼35%) were given a 16% increase in recharge(y = 1.16; Table 1). Any treatments that reduced basal areaby less than 30% were assumed to produce no discerniblehydrologic effect. This is supported by evidence foundin Baker (2003) where little to no discernible increase inwater yield was observed when basal area reduction inponderosa pine forests was less than 30%, a relationshipthat was assumed to hold for groundwater. The shelf-stock forest treatments were also accounted for in thesimulations. However, estimates of changes in basalarea are not consistently available for shelf-stock areas.Therefore, it was assumed that the shelf-stock areas wouldhave similar proportions of treatment intensities (i.e., basal

Table 1Forest Restoration Treatment Recharge-Change

Factors for 4FRI Scenario C and Shelf-Stock

Zone% BA

Reduction1 % IncreaseRecharge-Change

Factor

1 >0 N/A N/A2 0 N/A N/A3 1–29 N/A N/A4 30–39 16.275 1.165 40–49 19.775 1.196 50–59 23.275 1.237 60–69 26.775 1.268 70–79 30.275 1.311 31.87 15.354 1.15

1Zones were delineated based on each 10% reduction in basal area (BA), thenaveraged before calculating percent increase in groundwater recharge. Zone11 corresponds to shelf-stock treatment areas where basal area reduction wasassumed to be equal to the 4FRI average. Factors were only applied to areaswhere 4FRI treatments are planned to occur. These factors were applied tostress period 12, or the years 2014 to 2023, when 4FRI treatments are planningto be conducted.

area changes) as the 4FRI and would produce a similarhydrologic effect.

Climate Change ScenariosThis study used bias corrected and downscaled

climate projections derived from the World ClimateResearch Programme’s (WCRP’s) Coupled Model Inter-comparison Project phase 3 (CMIP3) multimodel ensem-ble (Maurer et al. 2007) to provide estimates of futureprecipitation for the study area. The WCRP CMIP3 cli-mate projections include a multimodel ensemble of resultsproduced from 16 climate models that simulated three ofthe potential emissions scenarios identified by the Inter-governmental Panel on Climate Change (IPCC) FourthAssessment Report (IPCC 2007). The emissions scenariosinclude the A1, A2, and B1 scenario families. These sce-narios are based on demographic development, socioeco-nomic development, and technological changes that mayoccur in the future and their impacts on greenhouse gasemissions.

Three emissions scenarios from these families, theA1B, A2, and B1 scenarios, were used to simulatefuture changes in precipitation for the interpretive period(2006–2099). In general, all emissions scenarios resultin warmer, drier conditions in the southwestern UnitedStates; however, there is considerable variability amongmodel estimates as to the magnitude of precipitationchanges (Maurer et al. 2007).

Recharge-change factors within the future NARGFMinterpretations were applied to simulate likely changesin precipitation resulting from changes in climaticconditions. Downscaled climate model projections of pre-cipitation were obtained from the WCRP for the chosenscenarios. To capture the range of likely precipitationfor the region under changing climatic conditions, themultimodel mean (the mean across all IPCC scenariosimulations for emission scenarios A1B, A2, and B1)

NGWA.org C.J.W. Wyatt et al. Groundwater 5

Table 2Average Climate Change Scenario

Recharge-Change Factors by Stress Period1

StressPeriod Years

Mean ofScenarios

Mean + 1StandardDeviation

Mean − 1StandardDeviation

11 2006–2013 0.954 1.19 0.71612 2014–2023 0.954 1.20 0.70913 2024–2029 0.961 1.20 0.72414 2030–2039 0.953 1.21 0.69915 2040–2049 0.950 1.21 0.69416 2050–2059 0.924 1.17 0.67817 2060–2069 0.918 1.17 0.66718 2070–2079 0.921 1.17 0.67019 2080–2089 0.922 1.18 0.66720 2090–2099 0.914 1.17 0.656

1All stress periods are transient and all factors were applied across the entiremodel domain.

and the spread (+/− 1 standard deviation) was calculatedacross all models and all three emission scenarios. Thus,three climate change scenarios are considered that repre-sent precipitation conditions associated with: (1) the meanof climate model projections, (2) conditions slightly wet-ter than the mean climate model projections (+1 standarddeviation), and (3) conditions slightly drier than themean climate model projections (−1 standard deviation).To simulate these changes in precipitation, the residualbetween the baseline synthetic precipitation and the mean,mean + 1 standard deviation, and mean − 1 standard devi-ation were calculated, converted into a factor of change(Table 2), and applied to the NARGFM (Figure 2). It wasassumed that increases or decreases in precipitation resultin similar increases or decreases in recharge.

All spatial information was manipulated with ArcGIS10.1 (ESRI 2011). The NARGFM was simulated withMODFLOW 2005 (Harbaugh 2005) using the Groundwa-ter Vistas 6.17 user interface (Rumbaugh and Rumbaugh2011). All groundwater basins for this research are delin-eated by model simulation and may not coincide withArizona Department of Water Resources basin designa-tions (Figure 1).

ResultsTo maintain consistency, all results are compared

to the average for the calibrated simulation for theNARGFM, or the years 1910 to 2005. Some of thesenumbers deviate from reported values in Blasch et al.(2006), and results are therefore dependent on thisdifference. A comprehensive evaluation of the calibratedsimulation of groundwater flow in Northern Arizona bythe NARGFM is in Pool et al. (2011). All values arereported in acre-feet or acre-feet per year (af or afy) inaddition to SI units to be consistent with Blasch et al.(2006), Pool et al. (2011), Garner et al. (2013) and watermanagement planners in the state and region. All valuesare approximated. Variables given in the following graphs

5

6

7

8

9

10

11

12

x 104

Rec

harg

e (a

cre−

feet

/yr)

2000 2020 2040 2060 2080 2100

1

x 108

Year

Baseline Scenario Annual Average

Climate Change Scenario: PrecipitationInput Mean ± 1 St. Dev.

Rec

harg

e (m

3 /yr

)

1.6

1.5

1.4

1.3

1.2

1.1

0.9

0.8

0.7

0.6

Figure 3. Changes in recharge in the Verde Valley sub-basinfor the interpretive model scenarios. The baseline scenariois precipitation-absent forest restoration and climate change.Climate change scenarios for multimodel mean, mean + 1standard deviation, and mean − 1 standard deviation ofthe IPCC A1B, A2, and B1 emission scenario precipitationprojections plus changes from forest restoration. These arethe same scenarios given for Figures 4–6.

include the baseline and climate change scenarios, plusannual average for the initial period of simulation for thecalibrated NARGFM (1910–2005).

In the interpretive model, the 4FRI treatments from2014 to 2023 simulated approximately 2.8 × 107 m3

(23,000 af) of additional recharge to the Verde Val-ley sub-basin, or about 2.8 × 106 m3/year (2300 afy;(Figure 3). This is an estimated 2.8% increase in annualrecharge from average conditions for the Verde Valley[∼1.03 × 108 m3/year (83,600 afy)]. Because some of the4FRI and non-4FRI (shelf-stock) treatments were locatedin the Coconino Plateau and Little Colorado Plateau sub-basins, it was assumed that these two sub-basins receivedadditional recharge from forest restoration treatment, butthey were not analyzed.

Recharge values (Figure 3) for the mean ofthe climate change scenarios for the interpre-tive period (2006–2099) range from 9.4 × 107 to1.0 × 108 m3/year (76,400–84,000 afy), with an averagevalue of 9.7 × 107 m3/year (79,000 afy). This represents6.2 × 106 m3/year (5000 afy) or 5.74%/year less rechargethan annual average conditions for the simulation period1910 to 2005. Recharge values for precipitation 1 stan-dard deviation above the mean (2006–2099) range from1.2 × 108 to 1.3 × 108 m3/year (98,000–107,000 afy),with an average value of 1.23 × 108 m3/year (99,800 afy).This represents 1.99 × 107 m3/year (16,100 afy) or19.3%/year more recharge than annual average recharge(1910–2005). Recharge values for 1 standard deviationbelow the mean (2006–2099) ranged from 6.77× 107

to 7.74 × 107 m3/year (54,900–62,800 afy), with anaverage value of 7.14 × 107 m3/year (57,900 afy). Thisrepresents 3.17× 107 m3/year (25,700 afy) or 30.8%/yearless recharge than annual average recharge (1910–2005).Recharge decreased for all scenarios for the interpretive

6 C.J.W. Wyatt et al. Groundwater NGWA.org

2

x 104

Bas

eflo

w (

acre

−fe

et/y

r)

1980 2000 2020 2040 2060 2080 2100

3

x 107

Baseline Scenario

Climate Change Scenario: PrecipitationInput Mean ± 1 St. Dev.

Observed

Year

Bas

eflo

w (

m3 /y

r)3.4

3.2

2.8

2.6

2.4

2.2

2.7

2.6

2.5

2.4

2.3

2.2

2.1

1.9

1.8

Figure 4. Observed and interpretive model simulated base-flow at Oak Creek near Sedona. Observed baseflow estimatesprovided by Natalie Coston (NAU Senior Thesis, 2010).Coston data from NOAA, 2009.

period (2006–2099), with the highest values comingduring the 4FRI treatment period (2014–2023) andthe lowest values coming at the end of the simulationperiod (2090–2099). The trend for recharge is reflectedin baseflow, well water level elevation, and storagechange.

Model predictions of stream baseflow at USGSstream gaging stations in the Verde Valley sub-basinwere evaluated to discern responses of the Verde Riverand its tributaries to forest restoration treatments andclimate change (Figure 1). These gaging stations (andannual average baseflow for 1910–2005) were: VerdeRiver near Clarkdale (4.81 × 107 m3/year; 39,000 afy),Verde River near Camp Verde (1.92 × 108 m3/year;156,000 afy), Oak Creek near Sedona (2.54 × 107 m3/year;20,600 afy; Figure 4), and Wet Beaver Creek near Rim-rock (7.60 × 106 m3/year; 6160 afy).

On average, baseflow for all stream gauges decreasedfrom the annual average for all interpretive scenarios,with the least decrease coming during the first yearsof interpretive simulation (2006–2013) and the mostdecrease coming at the end of the simulation period(2090–2099) with the results strongly dependent on theprecipitation scenario assumed. Baseflow values wereaverage ranges for the interpretive period (2006–2099).Losses at the Verde River near Clarkdale ranged from9.10 × 105 m3 to 2.40 × 106 m3 (738–1950 af), or 1.9% to5.0% of annual average. Losses at the Verde River nearCamp Verde ranged from 1.28 × 107 m3 to 2.73 × 107 m3

(10,417–22,155 af), or 6.7% to 14% of annual average.From annual average, Oak Creek near Sedona gained2.80 × 105 m3 (227 af) of baseflow (1.1%) under thehighest precipitation estimate and lost 1.75 × 106 m3

(1420 af; 6.9%) under the lowest precipitation estimate;the mean scenario had an average loss in baseflow of7.46 × 105 m3 (605 af), or 2.9% (Figure 4). From annualaverage, Wet Beaver Creek near Rimrock showed gainsof 4.80 × 105 m3 (389 af; 6.3%) to losses of 2.08 × 105 m3

(169 af; 2.7%) depending on climate assumptions with amean of 1.84 × 105 m3 (149 af; 2.4%) gained.

Water level data at seven wells were selected toanalyze groundwater response to landscape change andclimate change. These wells were predefined by Pool et al.(2011) as representative of the groundwater system inthe Verde Valley sub-basin. They include five wells—(A-13-05)05BDC, (A-14-05)17AAC, (A-15-03)12ADB1, (A-15-04)04DDC1, and (A-16-03)22DCD—that are in theconfined part of the Verde Formation of the alluvial basin-fill aquifer and two wells—(A-14-10)32DBD and (A-17-06)E30BBB—that are in the unconfined part of theCoconino Aquifer. None of these wells was within the4FRI treatment area, but they were assumed to showchanges in the groundwater system that resulted fromupgradient 4FRI treatments (Figure 2). Well data fromtwo alluvial basin-fill wells—(A-13-05)05BDC and (A-14-05)17AAC—and one Coconino aquifer well—(A-14-10)32DBD—are characteristic of the interpretive changesin water level altitude (Figure 5).

Generally, for the interpretive period (2006–2099)well levels declined in all wells from the last values ofthe calibrated model (2005). Well water levels declinedfrom a minimum of 0.3 m at (A-13-05)05BDC and (A-16-03)22DCD for the highest precipitation estimate toa maximum of 12.5 m at (A-15-03)12ADB1 for thelowest precipitation estimate. The exception is (A-14-10)32DBD where well water level elevation rose atleast 8.2 m for the highest precipitation inputs. Meanstandard deviation of estimated water level elevationsin wells located in the Verde Valley sub-basin is3.2 m for alluvial aquifer wells and 5.8 m for wellslocated in the Coconino Aquifer. Results from threewells—(A-13-05)05BDC, (A-15-04)04DDC1, and (A-16-03)22DCD—fell within this standard deviation andwere not considered significant changes. However, threewells—(A-14-10)32DBD, (A-15-03)12ADB1, and (A-17-06)E30BBB—showed significant changes, estimatingwell water level changes of +8.2 to −9.4 m, −11.9to −12.5 m, and −8.8 to −10.3 m, respectively. Onewell—(A-14-05)17AAC—had values that were onlysignificant for the driest scenario (decline of 3.3 m).

All scenarios project a decline in groundwa-ter storage for the interpretive period (2006–2099;Figure 6). Cumulative change in storage for theBaseline was −2.86 × 109 m3 (−2.32 × 106 af), withan annual average decline of −2.71 × 107 m3/year(−22,000 afy), and decline under the climate changescenarios ranged from −4.14 × 109 m3 to −2.13 × 109 m3

(−3.36 × 106 to −1.73 × 106 af) with a value of−3.13 × 109 m3 (−2.54 × 106 af) for the mean sce-nario [annually −3.21 × 107 m3/year (−26,000 afy) to−1.26 × 107 m3/year (−10,200 afy), −2.71 × 107 m3/year(−22,000 afy) for the mean]. Generally, the rate ofstorage loss decreased through time while net storageloss increased. There was no estimate available forpredevelopment (pre-1910) total water in storage for theVerde Valley sub-basin and therefore, percent changesper year cannot be established. Changes were lower than

NGWA.org C.J.W. Wyatt et al. Groundwater 7

3048

3049

3050

3051

3052

3053

3054

3055

3056

3057

Wat

er L

evel

Ele

vatio

n (f

eet)

1940 1960 1980 2000 2020 2040 2060 2080 2100929

929.5

930

930.5

931

931.5

932

Year

Wat

er L

evel

Ele

vatio

n (m

)

(A−13−05)05BDC − Verde Formation

Observed

Baseline Scenario

Climate Change Scenario: Precipitation InputMean ± 1 St. Dev.

3120

3125

3130

3135

3140

3145

3150

3155

3160

Wat

er L

evel

Ele

vatio

n (f

eet)

1940 1960 1980 2000 2020 2040 2060 2080 2100950

952

954

956

958

960

962

964

Year

Wat

er L

evel

Ele

vatio

n (m

)

(A−14−05)17AAC − Verde Formation

6500

6510

6520

6530

6540

6550

6560

6570

6580

6590

Wat

er L

evel

Ele

vatio

n (f

eet)

1940 1960 1980 2000 2020 2040 2060 2080 21001980

1985

1990

1995

2000

2005

2010

Year

Wat

er L

evel

Ele

vatio

n (m

)

(A−14−10)32DBD − Coconino Sandstone

Figure 5. Observed and interpretive model simulatedgroundwater levels at wells in the Verde Valley sub-basin. (A-13-05)05BDC and (A-14-05)17AAC are located in the VerdeValley near Camp Verde. (A-14-10)32DBD is located nearthe surface water and groundwater divides near Happy Jack(Pool et al., 2011). Values clipped to the years 1940 to 2099for better resolution.

normal for stress period 12, or the years 2014 to 2023,most likely because of the additional water available forrecharge following forest restoration treatments.

Discussion and ConclusionsAn interpretive groundwater-flow model was con-

structed to evaluate the response of the groundwater-flow

−3

−2

−1

x 106

Cum

ulat

ive

Cha

nge

in S

tora

ge (

acre

−fe

et)

2000 2020 2040 2060 2080 2100

−4

−3

−2

−1x 109

Cum

ulat

ive

Cha

nge

in S

tora

ge (

m3 )

Year

Baseline Scenario

Climate Change Scenario: PrecipitationInput Mean ± 1 St. Dev.

−1.5

−2.5

−3.5

−4.5

−1.5

−2.5

−3.5

Figure 6. Changes in groundwater storage for the VerdeValley sub-basin.

systems in Northern Arizona to potential changes inrecharge due to planned upland forest restoration andanticipated climate change. All values for recharge, base-flow, water table elevation, and storage, are dependenton simulated, future precipitation trends and are there-fore highly variable. The increased recharge associatedwith the 4FRI forest restoration treatments in stress period12 is relatively small compared to estimated changes inrecharge from climate change and is largely masked inthe record. Because this is an interpretive model, simu-lated changes in recharge from the 4FRI treatments wereeasily extracted by running the groundwater-flow modelonce with and once without the applied changes, isolating,and then analyzing individual stress periods. However,increases in recharge from the 4FRI treatments may bedifficult to identify due to their small magnitude relativeto other water balance components. A paired-watershedstudy may be necessary to distinguish any changes asso-ciated with recharge.

The interpretive model did not include maintenancetreatments (vegetation control through prescribed burns)for the first EIS area. Instead, the focus of the modelwas on the initial treatments. It was unclear how effectivemaintenance treatments would be, but it may sustain thehydrological benefits of reduced ET through frequentprescribed burns that occur posttreatment. Therefore,benefits may last longer than those simulated with theinterpretive model. Furthermore, because information onthe second round of 4FRI treatments, which are expectedto be largely in the Apache-Sitgreaves and Tonto NationalForests, has yet to be fully developed, these futuretreatments, which would take place after the first EIS areawas treated, were not included. However, it is anticipatedthat if treatments for the second EIS are similar to thosefor the first EIS, similar benefits may be possible.

The model suggests that recharge will decline throughtime for all scenarios. This is because precipitation projec-tions show a decrease in the water available for rechargeover time. Because the simulated precipitation projectionswere estimated from IPCC emissions scenarios, these

8 C.J.W. Wyatt et al. Groundwater NGWA.org

declining trends are assumed to be caused by the effectthat increased greenhouse gas emissions, such as carbondioxide and methane, will have on the climate. Generally,climate models predict a climate that becomes warmerand drier through time. This is reflected in the decliningrecharge trend (Figure 3).

With less recharge, there is less water available todischarge to streams and wells. This can cause baseflowand water level elevations to decline over time. Theseimpacts are reflected in Figures 4 and 5. While awetter than average climate may produce an increase inbaseflow and water level elevation, these increases areexpected to be relatively short-lived because of decliningrecharge values throughout the interpretive simulationperiod (2006–2099). Blasch et al. (2006) reports thatdischarge upstream from the streamflow gauges on theVerde River near Clarkdale and Camp Verde, Oak Creeknear Sedona, and Wet Beaver Creek near Rimrock islikely sensitive to variations in recharge rates. Becauseof this, it is assumed that changes in baseflow will beaffected by both forest restoration treatments and changesin precipitation from climate change.

Blasch et al. (2006) estimated a change in storage of−4.81 × 107 m3 (−39,000 af) for the Verde Valley sub-basin for the year 2005, the last year simulated by theNARGFM (Pool et al. 2011). The changes in storagefor the interpretive simulations fall above and belowthis value, depending on the scenario. However, in allcases, water is pulled from storage and cumulative storagechange increases through time. This reflects an imbalancebetween water supply and demand. This imbalanceis attributed to groundwater overdrafting through wellpumping and surface water diversions. It is assumed thatbefore the development of groundwater resources in thestudy area, or before 1938, the groundwater system wasin equilibrium and there was no imbalance between watersupply and demand. It was assumed that less water waspulled from storage over time because of groundwatercapture of resources, especially of the Verde River and itstributaries.

This imbalance between supply and demand mayresult in future, unmet demands for water for both naturaland human communities, which may significantly alter theecology of the Verde River system, and may negativelyimpact human communities in the area. Communitieswill need to develop sustainable water managementstrategies to combat these issues. A more detailed analysisof these imbalances and possible solutions is includedin Reclamation’s Colorado River Basin Water Supplyand Demand study (USDI Reclamation 2012) and theArizona Department of Water Resources Water ResourcesDevelopment Commission report (ADWR 2011).

AcknowledgmentsThis study was made possible with support from the

Salt River Project. Drs. Deborah Huntzinger and PeterKroopnick, and Sharon Masek Lopez are acknowledgedfor their help and support. The Program for Climate Model

Diagnosis and Intercomparison (PCMDI) and the WCRP’sWorking Group on Coupled Modelling [sic] (WGCM)are acknowledged for their roles in making available theWCRP CMIP3 multimodel dataset. Support of this datasetis provided by the Office of Science, U.S. Department ofEnergy.

ReferencesAllen, A.R., and D.V. Chapman. 2000. A review of the impacts

of forestry on groundwater and implications for forestmanagement. In Groundwater: Past Achievements andFuture Challenges. Proceedings of the XXX IAH CongressCape Town, South Africa 12/30/99 , ed. O. Sililo et al.,863–868.

Arizona Department of Water Resources (ADWR). 2011.Water Resources Development Commission FinalReport Volume I. http://www.azwater.gov/AzDWR/WaterManagement/WRDC_HB2661/documents/WRDCFinalReportVolumeI.pdf (accessed April 12, 2013).

Baker, Jr. M.B. 1986. Effects of Ponderosa pine treatments onwater yield in Arizona. Water Resources Research 22, no.1: 67–73.

Baker, Jr. M.B. 2003. Hydrology. Friederici, P. Ecologi-cal Restoration of Southwestern Ponderosa Pine Forests .Phoenix, Arizona: Arizona Board of Regents: 161–174.

Bazan, R.A., B.P. Wilcox, C. Munster, and M. Gary. 2012.Removing woody vegetation has little effect on conduitflow recharge. Ecohydrology . DOI:10.1002/eco. 1277.

Blasch, K.W., and J.R. Bryson. 2007. Distinguishing sources ofground water recharge by using δ2H and δ18O. GroundWater 45, no. 3: 294–308.

Blasch, K.W., J.P. Hoffman, L.F. Graser, J.R. Bryson, and A.L.Flint. 2006. Hydrogeology of the Upper and Middle VerdeRiver Watersheds, Central Arizona: U.S. Geological SurveyScientific Investigations Report 2005–5198: 102, 3 plates.Reston, Virginia: USGS.

Bosch, J.M., and J.D. Hewlett. 1982. A review of catchmentexperiments to determine the effect of vegetation changeson water yield and evapotranspiration. Journal of Hydrology55: 3–23.

Coston, N. 2010. Statistical examination of water data in theCoconino plateau for use as sustainability indicators. Seniorthesis, NAU.

Covington, W.W., and M.M. Moore. 1994. Southwesternponderosa forest structure: Changes since Euro-Americansettlement. Journal of Forestry 92: 39–47.

Dixon, G.E. comp. 2002. Essential FVS: A user’s guide tothe Forest Vegetation Simulator, Internal Rep. Fort Collins,Colorado: U.S. Department of Agriculture, Forest Service,Forest Management Service Center. 226 p. (Revised 2013).

Dore, S., M. Montes-Helu, S.C. Hart, B.A. Hungate, G.W. Koch,J.B. Moon, A.J. Finkral, and T.E. Kolb. 2012. Recover ofponderosa pine ecosystem carbon and water fluxes fromthinning and stand-replacing fire. Global Change Biology18: 3171–3185. DOI:10.1111/j.1365-2486.2012.02775.x.

ESRI. 2011. ArcGIS Desktop: Release 10 . Redlands, California:Environmental Systems Research Institute.

Flerchinger, G.N., and K.R. Cooley. 2000. A ten-year waterbalance of a mountainous semi-arid watershed. Journal ofHydrology , 237, nos. 1–2: 86–99. DOI:10.1016/S0022-1694(00)00299-7.

Flint, L.E., and A.L. Flint. 2008. Regional analysis of groundwater recharge. In Groundwater Recharge in the Arid andSemiarid Southwestern United States , ed. D.A. Stonestrom,J. Constantz, T.P.A. Ferre, and S.A. Leake, 29–59. Reston,Virginia: USGS Professional Paper 1703.

Garner, B.D., D.R. Pool, F.D. Tillman, and B.T. Forbes. 2013.Human effects on the hydrologic system of the Verde

NGWA.org C.J.W. Wyatt et al. Groundwater 9

Valley, Central Arizona, 1910–2005 and 2005–2110, usinga regional groundwater flow model. U.S. Geological SurveyScientific Investigations Map 2013–5029, vi, 47 p.

Gottfried, G.J. 1991. Moderate timber harvesting increases wateryields from an Arizona mixed conifer watershed. WaterResources Bulletin 27, no. 5: 537–547.

Harbaugh, A.W. 2005. MODFLOW-2005, the U.S. GeologicalSurvey modular ground-water model – the Ground-WaterFlow Process: U.S. Geological Survey Techniques andMethods 6-A16. Reston, Virginia: USGS.

Harpold, A.A., J.A. Biederman, K. Condon, M. Merino, Y.Korgaonkar, T. Nan, L.L. Sloat, M. Ross, and P.D. Brooks.2013. Changes in snow accumulation and ablation follow-ing the Las Conchas Forest Fire, New Mexico, USA. Eco-hydrology . http://onlinelibrary.wiley.com.ezproxy.princeton.edu/doi/10.1002/eco.1363/full.

IPCC Core Writing Team. 2007. Climate change 2007: Syn-thesis report. Contribution of Working Groups I, II andIII to the Fourth Assessment Report of the Intergovern-mental Panel on Climate Change. Geneva, Switzerland:IPCC, 104 p.

Maurer, E.P., L. Brekke, T. Pruitt, and P.B. Duffy. 2007.Fine-resolution climate projections enhance regional cli-mate change impact studies. Eos, Transactions AmericanGeophysical Union 88, no. 47: 504.

Meehl, G.A., C. Covey, T. Delworth, M. Latif, B. McAvaney,J.F.B. Mitchell, R.J. Stouffer, and K.E. Taylor. 2007. TheWCRP CMIP3 multi-model dataset: A new era in climatechange research. Bulletin of the American MeteorologicalSociety 88: 1383–1394.

Neary, D.G., A.L. Medina, J.N. Rinne, (ed.). 2012. Synthesisof Upper Verde River research and monitoring 1993–2008.USDA Forest Service, Rocky Mountain Research Station,General Technical Report RMRS-GTR-291. 295 p. FortCollins, Colorado: USDA Forest Service.

Pool, D.R., K.W. Blasch, J.B. Callegary, S.A. Leake, andL.F. Graser. 2011. Regional groundwater-flow model ofthe Redwall-Muav, Coconino, and alluvial basin aquifersystems of northern and central Arizona: U.S. GeologicalSurvey Scientific Investigations Report 2010–5180. v. 1.1,101 p. Reston, Virginia: USGS.

PRISM Climate Group. 2012. Oregon State University.http://prism.oregonstate.edu (accessed Fall 2012).

Rumbaugh, J., and D. Rumbaugh, 2011. Groundwater Vistas(Version 6.17 Build 17) [software] . Reinhold, Pennsylvania:Environmental Simulations, Inc.

Scanlon, B.R., R.C. Reedy, D.A. Stonestrom, D.E. Prudic,and K.F. Dennehy. 2005. Impact of land use and landcover change on groundwater recharge and quality in thesouthwestern US. Global Change Biology 11: 1577–1593.

Stegman, S.V. 1996. Snowpack changes resulting from timberharvest: Interception, redistribution, and evaporation. WaterResources Bulletin 32, no. 6: 1353–1360.

Troendle, C.A., and R.M. King. 1987. The effect of partial andclearcutting on streamflow at Deadhorse Creek, Colorado.Journal of Hydrology 90: 145–157.

USDA Forest Service. 2012. 4FRI Draft Environmental ImpactStatement Chapter 1. 75 p. http://www.fs.usda.gov/detail/4fri/planning/?cid=stelprdb5408763 (accessed Feb-ruary 14, 2013).

USDI Reclamation. 2012. Colorado River Basin WaterSupply and Demand Study Executive Summary (Pre-Production Copy). 26 p. http://www.usbr.gov/lc/region/programs/crbstudy/finalreport/Executive%20Summary/Executive_Summary_FINAL_Dec2012.pdf (accessed Jan-uary 7, 2013).

Veatch, W., P.D. Brooks, J.R. Gustafson, and N.P. Molotch.2009. Quantifying the effects of forest canopy cover onnet snow accumulation at a continental, mid-latitude site.Ecohydrology 2, no. 2: 115–128.

Wilcox, B.P., B.D. Newman, D. Brandes, D.W. Davenport, andK. Reid. 1997. Runoff from a semiarid Ponderosa pinehillslope in New Mexico. Water Resources Research 33,no. 10: 2301–2314. DOI:10.1029/97WR01691.

Wyatt, C.J.W. 2013. Estimating aquifer response followingforest restoration and climate change along the MogollonRim, Arizona. M.S. thesis. Northern Arizona University,Flagstaff, Arizona, 100 p.

Zou, C.B., P.F. Ffolliott, and M. Wine. 2010. Streamflowresponses to vegetation manipulations along a gradient ofprecipitation in the Colorado River Basin. Forest Ecologyand Management 259: 1268–1276.

10 C.J.W. Wyatt et al. Groundwater NGWA.org