ebook img

Groundwater level response in US principal aquifers to ENSO, NAO, PDO, and AMO PDF

14 Pages·2014·3.61 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Groundwater level response in US principal aquifers to ENSO, NAO, PDO, and AMO

JournalofHydrology519(2014)1939–1952 ContentslistsavailableatScienceDirect Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol Groundwater level response in U.S. principal aquifers to ENSO, NAO, PDO, and AMO Amber Jean M. Kussa, Jason J. Gurdakb,⇑ aUniversityofCalifornia,SantaCruz,DepartmentofEnvironmentalStudies,1156HighStreet,SantaCruz,CA95064,USA bSanFranciscoStateUniversity,DepartmentofEarth&ClimateSciences,1600HollowayAve,SanFrancisco,CA94132,USA a r t i c l e i n f o s u m m a r y Articlehistory: Groundwaterwillplayanimportantroleinsociety’sadaptationtoclimatevariabilityandchange.There- Received1March2014 fore, it is particularly important to understand teleconnections in groundwater with interannual to Receivedinrevisedform26July2014 multidecadalclimatevariabilitybecauseofthetangibleandnear-termimplicationsforwater-resource Accepted25September2014 management. Here we use singular spectrum analysis (SSA), wavelet coherence analysis, and lag Availableonline5October2014 correlation to quantify the effects of the El Niño Southern Oscillation (ENSO) (2–7year cycle), North ThismanuscriptwashandledbyPeterK. Atlantic Oscillation (NAO) (3–6year cycle), Pacific Decadal Oscillation (PDO) (15–25year cycle), and Kitanidis,Editor-in-Chief,withthe assistanceofRoseannaM.Neupauer, Atlantic Multidecadal Oscillation (AMO) (50–70year cycle) on precipitation and groundwater levels AssociateEditor acrosstheregionallyextensiveCentralValley,BasinandRange,andNorthAtlanticCoastalPlainprincipal aquifers(PAs)oftheUnitedStates(U.S.).Resultsarecomparedtorecentfindingsfromasimilarclimate Keywords: variabilitystudyoftheHighPlainsaquifertoprovidethefirstnational-scaleassessmentoftheeffectsof Groundwater interannualtomultidecadalclimatevariabilityongroundwaterresourcesinU.S.PAs.Theresultsindicate Climatevariability thatgroundwaterlevelsarepartiallycontrolledbyinterannualtomultidecadalclimatevariabilityand ENSO arenotsolelyafunctionoftemporalpatternsinpumping.ENSOandPDOhaveagreatercontrolthan NAO NAOandAMOonvariabilityingroundwaterlevelsacrosstheU.S.,particularlyinthewesternandcentral PDO PAs. Findings and methods presented here expand the knowledge and usable toolbox of innovative AMO approaches that can be used by managers and scientists to improve groundwater resource planning andoperationsunderfutureclimateuncertainty. (cid:2)2014ElsevierB.V.Allrightsreserved. 1.Introduction levels and recharge are partially controlled by complex interac- tions of low frequency (interannual to multidecadal) climate Interannual to multidecadal climate variability partially variability (Dickinson et al., 2004; Hanson et al., 2004; Pool, controlsprecipitationdistributioninspaceandtime,droughtfre- 2005; Fleming and Quilty, 2006; Gurdak et al., 2007; Anderson quency and severity, snowmelt runoff, streamflow, and other andEmanuel,2008;Holmanetal.,2009,2011;Clarketal.,2011; hydrologic processes that profoundly affect surface-water Figura et al., 2011; Perez-Valdivia and Sauchyn, 2011; Tremblay resources (Ropelewski and Halpert, 1986; Cayan and Webb, et al., 2011; Venencio and Garcia, 2011; Perez-Valdivia et al., 1992;DettingerandCayan,1994;Enfieldetal.,2001;Ghil,2002; 2012). Improved understanding of the long-term fluctuations in Dettinger et al., 2000, 2002; McCabe et al., 2004; Labat, 2008, groundwater availability that is dominated by low frequency cli- 2010; Vicente-Serrano et al., 2011; Ionita et al., 2012). However, mate variability is essential for best informed management and the effects of interannual to multidecadal climate variability on policydecisions,particularlywithin thecontextofthe increasing recharge rates and mechanisms and other subsurface hydrologic use of groundwater for human consumption and irrigation processesthataffectgroundwaterquantityandqualityarelargely (Wada et al., 2010) and the uncertainty of climate change and unknowninmostaquifersoftheUnitedStates(U.S.)(Gurdaketal., related impacts on groundwater quantity and quality (Hanson 2009) and other regions of the world (Green et al., 2011; Treidel et al., 2006; Holman, 2006; Earman and Dettinger, 2011; Stoll et al., 2012). High-frequency (synoptic to seasonal) climate vari- etal.,2011;Gurdaketal.,2012). ability creates short-term hydrologic responses, but groundwater Thecomplexnatureofclimatevariabilityonalltemporalscales, including interannual to multidecadal is a major obstacle in the ⇑ reliableidentificationofglobalchangecausedbyhumanactivities Correspondingauthor.Tel.:+14153386869. (Ghil, 2002). Recent discussions about adopting nonstationary E-mailaddresses:[email protected](A.J.M.Kuss),[email protected](J.J.Gurdak). http://dx.doi.org/10.1016/j.jhydrol.2014.09.069 0022-1694/(cid:2)2014ElsevierB.V.Allrightsreserved. 1940 A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 models for water-resource risk assessment and planning is climate variability on groundwater resources in U.S. PAs, and motivated by climate change and natural low-frequency climate advances current understanding needed for effective variability(Millyetal.,2008).Thefourleadingatmospheric–ocean water-resourcemanagementandpolicydecisionsunderincreasing circulationsystemsthataffectNorthAmericaninterannualtomul- climateuncertainty. tidecadal climate variability are the El Niño Southern Oscillation (ENSO), North Atlantic Oscillation (NAO), Pacific Decadal Oscilla- 2.Background tion (PDO), and Atlantic Multidecadal Oscillation (AMO) (Ghil, 2002;McCabeetal.,2004).Thetimingandphaserelationofthese 2.1.Climatevariability quasiperiodicclimatecycleshaveteleconnectionswithhydrologic variability across much of the U.S. (McCabe et al., 2004; Hanson Climate variability is characterized in terms of anomalies, andDettinger,2005;Gurdak,2008). which are defined as the difference between the current climate In this paper we quantify the teleconnections between ENSO, conditionsandthemeanstate,whichisrepresentativeof‘‘normal’’ NAO, PDO, and AMO and the groundwater resources of principal conditions that are computed over many years (Hurrell et al., aquifers (PAs) along a west–east transect of the U.S. based on 2003).Naturalclimatevariabilityoccursonmanydifferentspatio- observed anomalies in the Pacific and Atlantic Oceans. PAs are temporalscalesandiscreatedduetointerplaybetweenmultiple regionallyextensiveaquifersand aquifersystemsofnationalsig- variables including (but not limited to) sea-level pressure (SLP) nificancebecauseoftheirhighproductivityoruseandarecritically anomalies,pressureheightvariations,fluctuationsinwindspeeds, important sources of potable water (USGS, 2003). The primary variationsintheEarth’sorbit,andvolcaniceruptions(Ghil,2002). objectivesofthispaperaretocharacterizenonstationarypatterns Climateindicesaregenerallycreatedusingavarietyofanomalies inclimatevariabilityandlagcorrelationstospatiotemporaltrends atdifferentlocations,suchaschangesinthemeandistributionof inlong-termrecordsofprecipitation(55–89years)andgroundwa- SLPattwolocations,thedeviationfrommeansea-surfacetemper- ter levels (41–93years) (Table 1) of the Central Valley aquifer atures (SSTs) at multiple locations, variations in oceanic wind (52,000km2),BasinandRangeaquifersystem(700,000km2),and strength, or a combination of multiple variables (Ghil, 2002; North Atlantic Coastal Plain aquifer system (130,000km2) Hurrelletal.,2003). (Fig. 1). Some results from these aquifers are compared to major findingsfromarecentandsimilarclimatevariabilitystudyofthe High Plains aquifer (450,000km2) (Gurdak et al., 2007) that is 2.1.1.ElNiñoSouthernOscillation(ENSO) locatedinthecentralU.S.(Fig.1).Thispaperisthefirsttoprovide TheENSOisa2–7yearquasiperiodicphenomenonthatresults national-scaletrendsontheeffectsofinterannualtomultidecadal fromlarge-scaleinteractionsbetweenthetropicalandsubtropical portionsofthePacificandtheIndianOceanbasins,whichresultsin variations in pressure, temperature, and precipitation patterns Table1 throughout the U.S. and other regions of the world (Ropelewski Hydrologictime-seriessiteinformation. and Halpert, 1986; Diaz and Markgraf, 1992; Hanson et al., StudyID AgencysiteID Periodofrecord Years N 2006). We use the Multivariate ENSO Index (MEI) (Wolter and CVP1 47292 1951–2009 58 696 Timlin,2011)tomeasurethetemporalextentandthephasedesig- CVGW1 401059122102801 1937–2004 67 128 nationofENSO(Fig.2a).Thepositive(negative)MEIisrelatedto CVP2 45385 1920–2009 89 1068 thepositive(negative)ENSOphase.TheMEIisbasedonmultiple CVGW2 391028121312501 1947–2003 56 114 variables of the Comprehensive Ocean-Atmospheric Data Set CVP3 49200 1926–2009 83 996 (COADS), and represents a weighted average of SLP, zonal and CVGW3 382458121455801 1951–2004 53 104 meridional winds, SST, air temperature, and total cloudiness CVP4 45032 1927–2009 82 984 (WolterandTimlin,2011). CVGW4 380238121091301 1962–2004 42 354 DuringthepositiveENSO(ElNiño)phase,theequatorialPacific CVP5 41244 1940–2009 69 828 experiences abnormally low SLP in the east and increased SLP in CVGW5 352228119295201 1937–2004 93 128 the west, allowing for the warm waters of the western Pacific to BRP1 266779 1950–2009 59 708 migrate eastward, thus creating a shift in the jet stream BRGW1 393737119514801 1966–2007 41 138 (Ropelewski and Halpert, 1986). A U.S. coast-to-coast continuity BRP2 264950 1931–2009 78 936 of increased precipitation (especially in the winter months of BRGW2 393310114475001 1948–2006 58 438 December–February) from North Carolina to California has been BRP3 426135 1941–2009 68 816 observed during the positive ENSO phase, with stronger correla- BRGW3 393143111523301 1935–2007 72 1951 tions in the southwest and central U.S. (Ropelewski and Halpert, BRP4 262243 1950–2009 59 708 1986;KiladisandDiaz,1989;KurtzmanandScanlon,2007).Dur- BRGW4 361843115161001 1943–2007 64 1847 ing a negative (La Niña) phase, SLPs increase and SSTs decrease BRP5 264436 1951–2009 58 696 in the eastern equatorial Pacific, creating opposing temperature BRGW5 360528115094201 1965–2007 42 1923 and precipitation patterns across the U.S (Ropelewski and NAP1 307633 1938–2009 71 852 Halpert, 1986; Kiladis and Diaz, 1989; Diaz and Markgraf, 1992). NAGW1 405308072553101 1945–2006 61 2119 Additionally, winter precipitation is generally above-normal in NAP2 307134 1931–2009 78 936 the Pacific Northwest and below-normal in the Southwestern NAGW2 405743072425701 1954–2007 53 544 U.S.duringthenegativeENSOphase,whiletheoppositeconditions NAP3 283181 1948–2009 61 732 aregenerallyobservedduringthepositiveENSOphase(Kiladisand NAGW3 402553074271701 1944–2005 61 649 Diaz,1989).Thisdipolesignatureinthewinterprecipitationofthe NAP4 72730 1948–2009 61 732 western U.S. is influenced by the phasing of the Pacific Decadal NAGW4 391949075410701 1957–2004 47 564 Oscillation(PDO)(BrownandComrie,2004). NAP5 188000 1954–2009 55 660 NAGW5 382329075263701 1967–2010 43 683 2.1.2.TheNorthAtlanticOscillation(NAO) CV, Central Valley; BR, Basin and Range; NA, North Atlantic Coastal Plain; P, The NAO is a north–south dipole of pressure anomalies, with precipitationsite;GW,groundwaterlevelsite;N,numberofdatapoints. one anomaly centered over Greenland and the other anomaly A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 1941 Fig.1. MapshowingthelocationoftheCentralValley,BasinandRange,andNorthAtlanticCoastalPlainaquifersystems; andthefiveco-locatedprecipitationand groundwaterlevelsitesineachaquifer(Table1).FindingsfromarecentclimatevariabilitystudyoftheHighPlainsaquifer(Gurdaketal.,2007)isusedheretocompletethe east–westgradientofPrincipalAquifers(PAs)acrosstheUnitedStates.ThefindingsandsiteinformationfortheHighPlainssites1–6isdetailedbyGurdaketal.(2007). Fig.2. (a)ThemonthlyMultivariateElNiñoSouthernOscillation(ENSO)index(WolterandTimlin,2011),(b)Hurrellstation-basedannualNorthAtlanticOscillation(NAO) index(Hurrell,1995),(c)monthlyPacificDecadalOscillation(PDO)index(Mantuaetal.,1997),andthe(d)KaplanSST,unsmoothedAtlanticMultidecadalOscillation(AMO) index(Enfieldetal.,2001). spanningthecentrallatitudesoftheAtlanticbetween35(cid:3)and40(cid:3) 2.1.3.ThePacificDecadalOscillation(PDO) N (Hurrell, 1995). The positive phase NAO is characterized by Decadal to interdecadal variability in atmospheric circulation, below normal geopotential heights and pressures in the North specificallythewintertimeAleutianLowpressuresystemandSSTs Atlantic and above normal heights and pressures over eastern from 20(cid:3) N poleward, is often associated with the PDO (Mantua U.S.andwesternEurope,andviceversaduringthenegativephase et al., 1997). The PDO is indexed with monthly northern Pacific NAO (Hurrell, 1995). The NAO index (Fig. 2b) is the difference SST residuals (the difference from the observed anomalies and betweennormalizedmeanwinter(December–March)SLPanoma- themonthlymeanglobalaverageSSTanomaly)(Fig.2c)(Mantua liesbetweenLisbon,PortugalandStykkisholmur,Iceland(Hurrell, andHare,2002).FluctuationsinthePDOindexoverthe20thcen- 1995). The NAO has a dominant quasiperiodic oscillation of turyweremostenergeticintwogeneralperiodicities–onefrom 3–6years with a less significant 8–10year mode (Hurrell et al., 15 to 25years and the other from 50 to 70years (Mantua and 2003). During the positive NAO, enhanced westerly flow across Hare,2002).Inthispaper,weevaluateonlythe15-to25-yearperi- the Atlantic moves warm moist air over Europe and the eastern odicityofthePDO. U.S., creating an increase in winter storms (Hurrell, 1995; The climate anomalies connected to the PDO are somewhat Ottersen et al., 2001; Hurrell et al., 2003). While the majority of similar to the ENSO with comparable shifts in the jet stream thepositiveprecipitationsignalsareseeninEurope,weakerposi- (MantuaandHare,2002).ThepositivephasesofthePDOareasso- tive correlations are observed in the eastern U.S. (Hurrell et al., ciated with warm dry periods in the Pacific Northwest and cool 2003). wet periods in the southwestern U.S. (Mantua and Hare, 2002). 1942 A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 RegimeshiftsinthePDOalsoaffectthephaseandtheoccurrence uppermost aquifer in the North Atlantic Coastal Plain aquifer oftheENSO(BrownandComrie,2004;Gutzleretal.,2002).During system (Trapp, 1992). Average annual temperature is 11(cid:3)(cid:3)C and thenegative(positive)PDO,thereisagreateroccurrenceofnega- average annual precipitation is 1044mm (Polsky et al., 2000). tive(positive)ENSOevents(Gutzleretal.,2002). The North Atlantic Coastal Plain aquifer system is ranked 13th among the PAs in terms of total withdrawals (Maupin and 2.1.4.TheAtlanticMultidecadalOscillation(AMO) Barber,2005). TheAMOisanatmospheric–oceanphenomenonwithaperiod- icityof50–70yearsthatarisesfromvariationsinSSTsintheAtlan- 3.Methods tic Ocean associated with variations in the strength of the thermohaline circulation (Kerr, 2000; Enfield et al., 2001). The The time series evaluated here include the previously men- AMO is indexed with a ten-year running mean of Atlantic SSTs tioned MEI, NAO, PDO, and AMO indices, groundwater levels from0to70(cid:3)N,withapeakvariationof0.4(cid:3)Cdeterminingaphase (1933–2009)frommonitoringwellsintheU.S.GeologicalSurvey shift (Fig. 2d) (Enfield et al., 2001). The AMO is associated with (USGS)NationalWaterInformationSystem(NWIS)(USGS,2012), variations in air temperature and precipitation regimes across and precipitation data (1920–2009) from NOAA (2012) meteoro- the U.S., with a positive (negative) phase producing decreased logicalstations(Table1).Atotalof5monitoringwellsineachPA (increased)rainfall(Enfieldetal.,2001;McCabeetal.,2004).The were selected based on the length and completeness of the AMO may also modulate the strength and occurrence of NAO water-levelrecordsandtorepresentarangeofhydrogeologiccon- (PeingsandMagnusdottir,2014)andENSOcycles,creatingaweak- ditionsintheuppermostandunconfinedaquiferunits.Inorderto ened (strengthened) El Niño during the positive (negative) AMO assessinterannualtointerdecadalclimatevariability,weselected phase (McCabe et al., 2004). Coupled influences of the PDO and wells that had at least 40years of water-level records (Table 1). theAMOmayalsoproduceextensivespatialandtemporalfluctu- Although our site selection criteria required at least annual ations,withcoincidentpositiveAMOandnegativePDOincreasing water-level values with no multi-year gaps in the records, the theoccurrenceofdroughtsintheU.S.(McCabeetal.,2004). majority of the wells have at least quarterly water-level data (Table 1). One long-term (>40years) meteorological station was 2.2.Sitedescription required to be co-located within 24km of each monitoring well. The 5 co-located meteorological stations and monitoring wells TheCentralValley,BasinandRange,andNorthAtlanticCoastal wereselectedtocharacterizethegeneralspatialpatternsineach Plainaquifersystemsarelocatedalongawest–easttransectofthe PA and not meant to be an exhaustive representation of the contiguousU.S.(Fig.1).Theaquifersaregenerallyunconfinedand heterogeneityineachPA.Theco-locatedwellsandmeteorological consistofunconsolidatedsand,gravel,andsilt,andrepresentsome stationsareidentifiedbytheaquifername(CV,CentralValley;BR, ofthemostimportantgroundwaterresourcesintheU.S.foragri- Basin and Range; and NA, North Atlantic Coastal Plain), location culture, industry, and domestic uses (Maupin and Barber, 2005). (labeled1–5fromnorthtosouth)(Fig.1),andsitetype(P,precip- McMahon et al. (2006, 2007) and Gurdak et al. (2007) provide a itation; and GW, groundwater) (Table 1). Because long-term detaileddescriptionoftheHighPlainsaquifer(Fig.1). groundwater pumping records are not publically available, we TheCentralValleyaquiferlocatedinCalifornia(Fig.1)isasin- use a simulated groundwater-pumping time series (1962–2003) gle,heterogeneousaquifersystemthatisgenerallyunconfinedin from the Central Valley aquifer that was developed by Faunt the upper hundred meters and confined at depth by overlapping etal.(2009)toquantifytheeffectsofinterannualtomultidecadal discontinuous clay beds, primarily in the southern portion climate variability on groundwater pumping. Simulated ground- (PlanertandWilliams,1995;Fauntetal.,2009).ClimateisMedi- water-pumping time series are not publically available for the terranean and Steppe, with hot summers and mild winters otherPAsinthisstudy. (Bertoldietal.,1991).Approximately85%oftheannualprecipita- tion occurs from November to April, with a steady decrease in overallprecipitationfromthenorth(average584mm)tothesouth 3.1.Pre-processing (average152mm)(Bertoldietal.,1991).Pumpingtosupportirri- gation has affected groundwater levels, with major declines in Time series analysis has been used to assess long-term varia- muchoftheaquifer(MaupinandBarber,2005). tions in hydrologic variables (Enfield et al., 2001; Hanson et al., The Basin and Range aquifer system, located primarily in Ari- 2004,2006;McCabeetal.,2004;Gurdaketal.,2007).Usingdata zona, California, Nevada, New Mexico, and Utah, includes about pre-processingmethodsoutlinedbyHansonetal.(2004),weinter- 120alluvium-filledbasinsinterspersedbetweenmountainranges polated each time series with a monthly spline to integrate any (Fig. 1) (Planert and Williams, 1995). The basins are generally irregular sampled records, and then converted these data into unconsolidatedtomoderatelyconsolidated, welltopoorlysorted cumulative departures seriesfrom the period of recordusing the bedsofgravel,sand,silt,andclaydepositedonalluvialfans,flood monthly mean. Next, the residuals of the monthly cumulative plains,andPleistocenelakes(PlanertandWilliams,1995).Thecli- departureseries wereobtained by subtracting a regression-fitted mate is arid, with average annual precipitation ranging from low-order(cubic)polynomial.Theoverallshapeofthelow-order approximately 102 to 203mm in the basins and 406–508mm in polynomial represents temporal trends (or responses) in the the mountain ranges (Sheppard et al., 2002). Hot summers and hydrologic time series to larger climatic cycles or periods of largeevapotranspirationrates,particularlyinloweraltitudes,limit anthropogenic effects (Hanson et al., 2004). Finally, the residuals rechargetoanestimated5%ofprecipitation(PlanertandWilliams, arenormalizedbythehistoricmeantofacilitatestatisticalcompar- 1995).TheBasinandRangeaquifersystemisranked4thamongthe isonsbetweenvariousdatatypesandarereferredtoasnormalized PAsintermsoftotalwithdrawals,andapproximately81%ofwith- departures (unitless). The primary goal of the data-processing drawalssupportirrigatedagriculture(MaupinandBarber,2005). stepsistoremoverednoisepriortousingsingularspectrumanal- The North Atlantic Coastal Plain aquifer system, located in ysis (SSA) to identify temporal structure in the time series that Delaware, Maryland, New Jersey, New York, North Carolina, and have a statistically significant difference from red noise. The Virginiaconsistsofsixregionalunconfinedandconfinedaquifers data-processing steps also remove much of the long-term, (TrappandHorn,1997).Thisstudyfocusedonwellsinthesurficial multidecadalanthropogenicsignalsinthegroundwater-leveltime aquiferthatconsistsofunconsolidatedsandandgravelandisthe series,suchaslong-termland-usechangeandtheimplementation A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 1943 ofimprovedirrigationtechnology,andannualanthropogenicsig- variations in both amplitude and frequencythat can be analyzed nals, such as crop rotation and other irrigated-agricultural prac- using wavelet transforms (Torrence and Compo, 1998; Grinsted tices (Hanson et al., 2004; Gurdak et al., 2007). Following the et al., 2004; Labat et al., 2000; Labat, 2005, 2008; Holman et al., pre-processingsteps,thetimeserieswereprocessedusingSSAto 2011).Theuseofwaveletcoherencecanprovideimportantinsight extracttemporalstructuresfromthenoisydataandusingwavelet inhowthestrengthofgroundwaterteleconnectionsvariesthrough analysis to characterize periodicity as a function of time, as time (Holman et al., 2011). We applied wavelet analyses on the describedindetailnext. compositeRCsfromtheclimateindices,precipitation,andground- water level time series using a MATLAB script developed by 3.2.Singularspectrumanalysis(SSA) Grinstedetal.(2004)andoutlinedinHolmanetal.(2011).Tocal- culatecoherencebetweentwoseries,thewaveletpowerspectrum UsingthemethodsoutlinedbyHansonetal.(2004)andGurdak ofeachtimeserieswasused(TorrenceandCompo,1998).Impor- et al. (2007), we applied the SSA–MTM toolkit (Dettinger et al., tantfeaturesofthewaveletpowerspectrumareidentifiedassig- 1995; Ghil et al., 2002) for time series frequency analyses using nificantatthe5%level,whichindicatesa95%confidencelevelof SSA, which is a modified form of principal component analysis coherence between the two series, identifying the relation (PCA) in the vector space of delay coordinates for a time series. between climatic and hydrologic variables (Torrence and Compo, SSA incorporates a data-adaptive method to analyze short-noisy 1998; Newman et al., 2003; Grinsted et al., 2004; Holman et al., series and extract the dominant frequencies of a time series 2011). (Vautardetal.,1992).SSAhasbeenwidelyappliedtogeophysical Weusethecontinuouswavelettransform(CWT)thatislocal- datasetsofvariouslengthsandtoidentifyquasi-periodicoscilla- izedintime(Dt)andfrequency(Dx)toidentifytheperiodicities tions, for example ENSO and other interannual to multidecadal and phases of cycles within a single time series (Grinsted et al., coupledoceanic–atmosphericphenomenon(Vautardetal.,1992). 2004;Holmanetal., 2011). Forthe best balanceoftime and fre- WeusedSSAtodecomposethedetrendedandnormalizedhydro- quency, the Morlet wavelet is used (Torrence and Compo, 1998; logic time series into temporal principal components (PCs) that Grinsted et al., 2004). The CWT establishes the power spectrum representtheprojectionoftheoriginaltimeseriesontoempirical ofeachprecipitation,groundwaterlevel,andclimateindexbefore orthogonal functions (EOFs) (Vautard et al., 1992; Ghil et al., the cross-correlation of the series is performed (Grinsted et al., 2002).Thephaseinformationofthetimeseries,oscillatorymodes, 2004). and noise are reconstructed by using linear combinations of the AftertheCWT,weusedthecrosswavelettransform(XWT)to PCs and EOFs to create the reconstructed components (RCs). No identify the cross wavelet power of two time series against the informationislostduringthereconstructionprocessbecausethe background power spectra for each of the series (Torrence and sumoftheindividualRCsequalstheoriginaltimeseries.Thevar- Compo, 1998; Grinsted et al., 2004). Because errors are present iabilityinmosthydrologictimeseriescanbeadequatelydescribed atthebeginningandendofthefinitewaveletpowerspectrum,a intermsofthefirst10RCs(withdecreasingvariancefromoneto coneofinfluence(COI)isusedtoidentifytheregionofthewavelet ten)(Hansonetal.,2004). spectrumwheretheseedgeeffectsneedtobeexcluded(Torrence WeappliedSSAtoidentifythefirst10RCsthatrepresentstatis- andCompo,1998;Grinstedetal.,2004). tically significant oscillatory modes within each of the individual While the XWT identifies areas of high common power, the climate index, precipitation, groundwater level, and simulated waveletcoherence(WTC)measuresthecross-correlationbetween pumping time series. The reader is referred to Kuss (2011) for twoseriesasafunctionoffrequencyasaquantitybetween0and1 the SSA results of all 10 RCs for each time series. We used the andidentifiesspectralcoherenceofthetwoseriesevenifthereisa Chi-Squared significance test (Allen and Smith, 1996) on each of low common power when each series is localized in time–fre- thefirst10RCstoidentifythosestatisticallysignificantoscillations quencyspace(TorrenceandCompo,1998).Althoughthespectral (Ha)againstared-noisenullhypothesis(H0).TheSSA–MTMtoolkit coherenceofthetwoseriesishighlighted,thetemporalcoherence userguideprovidesdetailsaboutthetheoryandapplicationofChi- isdiminishedascomparedtotheXWT(Labat,2005;Grinstedetal., Squaredsignificancetest(Dettingeretal.,1995;Ghiletal.,2002). 2004).InordertotestthesignificanceoftheWTC,1000randomly Foreachindividualtimeseries,wegroupandsumonlythestatis- constructedsyntheticseriesarecreatedusingMonteCarlometh- tically significant RCs according to the following period ranges: ods(Grinstedetal.,2004). 2–7year(ENSO-like),3–6years(NAO-like),15–25years(PDO-like), The three methodological steps of CWT, XWT, and WTC help and>25years(>PDO).PDOfluctuationsoverthe20thcenturywere establish the coherence in the temporal viability of the climate most energetic in two general periodicities – one from 15 to indices,precipitation,andgroundwaterlevels.Althoughthethree 25years and the other from 50 to 70years (Mantua and Hare, steps were necessarily followed, we present only the results and 2002). Here we evaluate only the 15- to 25-year periodicity of discussionontheWTC,whichprovidesthemostrobustoutcomes the PDO. Because the periodicity of the AMO is indistinguishable ofinteresttothereader(Holmanetal.,2011).Kuss(2011)presents from the lower frequencymode (50–70years) of PDO variability, detailedresultsanddiscussionfromtheCWTandXWT. we use thenaming convention proposedby Hansonet al. (2004) and we refer to all periods of climate variability greater than 3.4.Lagcorrelations 25years as greater than PDO (>PDO). By summing RCs based on similar period ranges, we create composite RCs that represents Correlation coefficients measure the strength of association the statistically significant oscillatory modes within each of the betweentwovariables(HelselandHirsch,2002)andwhenasys- raw hydrologic time series that are consistent with ENSO, NAO, temhasadelayedresponsetosomeforcingoracommondelayed PDO, or >PDO periodicity (Hanson et al., 2004; Gurdak et al., responsetoanadditionalvariable.Tocalculatethelagcorrelations 2007). The statistically significant composite RCs are used in all andtheirstatisticalsignificance,weusedtheUSGSHydrologicand subsequentwaveletandlagcorrelationsanalyses. Climatic Analysis Toolkit (HydroClimATe), which is a computer programtoassessrelationsbetweenvariableclimaticandhydro- 3.3.Waveletanalysis logictime-seriesdata(Dickinsonetal.,2014).Usingaprioriexpec- tations from regional hydroclimatology documented in the WeusewaveletanalysistocomplementtheSSAbecausemany literature, we report the strongest and statistically significant hydrologic time series are nonstationary and have temporal (95% confidence interval) lag correlation coefficients (unitless) 1944 A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 and corresponding phase lags (year) betweenthe climate indices consistentwithfindingsfromGurdaketal.(2007)thatgroundwa- andcompositeRCsfromtheprecipitationtimeseriesandbetween ter levels can be substantially influenced by PDO variability. No thecompositeRCsfromtheco-locatedprecipitationandground- SSA modes of variability with periodicities consistent with NAO water level time series. As explained by Dickinson et al. (2014), or >PDO were identified in precipitation or groundwater levels thephaselagsbetweenmanycomplexphysicalprocessesaregen- sitesintheCentralValley(Fig.3a). erallyundetermined,butthelagcorrelationanalysisisvaluablein GroundwaterlevelsintheCentralValleyaquiferareaffectedby providinginsightintothesystemwithoutusingadynamic,numer- withdrawals for irrigation (Faunt et al., 2009). We evaluated the icalmodel. responseingroundwaterpumpingtoclimatevariabilityandtofur- ther explore the hypothesis that groundwater levels are partially 4.Resultsanddiscussion controlled by climate variability and are not solely a function of temporal patterns in pumping. The results of SSA indicate that 4.1.HydroclimaticteleconnectionswithENSO,NAO,PDO,and>PDO no statistically significant RCs are present within the simulated pumpingtimeseries(Fauntetal.,2009),whichindicatesthatthe TheresultsoftheSSAindicatethatallthehydrologictimeseries simulatedpumpingisnotsubstantiallycontrolledbyclimatevar- contain variations that are partially consistent with ENSO, NAO, iability.Amorerobustanalysisonadditionalgroundwaterpump- PDO,and>PDO.Theprecipitationandgroundwaterleveltimeser- ing time series is needed to more definitively evaluate pumping ies in the Central Valley have two dominant modes of variability responsestointerannualtomultidecadalclimatevariability.How- consistent with PDO and ENSO (Fig. 3a). The amount of variance ever, it iswell documentedthat agriculturalland-use patternsin attributed to the PDO-like cycle ranged from 26.4% to 83.0% and theCentralValleyaredynamicandmayvarygraduallyorabruptly was consistently larger in the precipitation (range 38.6–83.0%, intimebecauseofthecumulativeeffectofurbanizationanddevel- average of 68.7%) than the co-located groundwater levels (range opment, free-market trends and changing crops types, resource 26.4–53.1%, average 34.5%) (Fig. 3a). The second largest amount limitations, particularly water, and climate variability (Faunt of variance in precipitation and groundwater levels (range etal., 2009). These complexand changing land-use patternspar- 7.3–20.5%, average 13.3%) was attributed to the ENSO-like cycles tiallycontrolagriculturalgroundwaterdemandwithinthecontext (Fig. 3a). These findings indicate the importance of lower- ofindividualfarmingpracticesandcropselection.TheSSAfindings frequency PDO forcings with lesser controls by ENSO and are supportthesecomplexitiesandindicatethatthe2-to7-and15-to Fig.3. Thepercentvariance(%)andperiod(years)ofthecombinedreconstructedcomponents(RC)fortheprecipitationandgroundwaterleveltimeseriesareshownforthe (a)CentralValleyaquifer,(b)BasinandRangeaquifersystem,and(c)NorthAtlanticCoastalPlainsaquifersystem.GWdenotesgroundwatersiteandPdenotesprecipitation site. A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 1945 25-year cycles within the groundwater levels are responses to vadosezonetothewatertable.ThetemporallagsbetweenENSO ENSOandPDO.Theresultsofthestatisticallysignificantlagcorre- coherenceintheprecipitationandENSOcoherenceintheground- lations, presented below, further support these teleconnections waterlevelsarefurtherquantifiedinSection4.3.Atsites1and5,a withgroundwaterlevels. 5%significantcoherenceinthe3-to7-yearperiodicityisobserved TheSSAresultsfromtheBasinandRangeaquifersystemindi- and an approximate 5-year temporal lag in the groundwater cate multiple modes of variability (Fig. 3b). The PDO-like cycle coherence behind the precipitation coherence to ENSO (Fig. 4a). was consistently observed as the leading mode of variability in Additionally, moderate to strong coherence is observed at most allthesites exceptfortwo(BRP2and BRGW2)andcontributed sites(withtheexceptionofGW3andP4),howevertheresultsdo about 20–77.2% of the variability in all Basin and Range sites notshowa5%significantcoherence.Moderatetostrongcoherence (Fig.3b).AtBRP2andBRGW2,the>PDOcontributedtothegreat- atthe9-to24-yearperiodisalsoapparentintheWTCofthepre- est amount of site variability (66.1% and 47.3%, respectively). cipitationandgroundwaterlevelsatallCentralValleysites(with Althoughthelengthofthetimeseriescanbea limitingfactorin theexceptionofGW2,GW3,andP5)(Fig.4b),whichisconsistent identifyinglow-frequencycycles,BRP2andBRGW2havemoder- withtheperiodicityofthePDO.Theextentofthe5%significancein aterecordlengths(1931–2009and1948–2006,respectively)and precipitationthatisconsistentwithPDOvariesacrosstheCentral manysiteshaveconsiderablylongerrecords.Therefore,thelength Valley, with no 5% significance found at sites P1, P5, GW2, GW3, oftherecorddoesnotappeartobeacontrollingfactorinobserving andGW4.Thenorthernlocations(sites1and2)havemoretempo- the low frequency >PDO cycle at BR P2 and GW2. The ENSO-like rally discrete 5% significance that spans the 1970s to 2000s and cycle is also observed in all of the Basin and Range sites except centers on the 1980s (Fig. 4b), which was a period of persistent forBRGW2,andcontributes5.8–27.1%ofthevarianceintheBasin positive phase PDO (Fig. 2c) associated with cool wet conditions and Range sites (Fig. 3b). Excluding the BR GW2, the ENSO-like inthesouthwesternU.S.Sites3and4locatedneartheSanFran- cyclecontributestoanaverageof13.5%inthevarianceofthepre- cisco Bay and Delta have 5% significance that spans the 1930s to cipitationandgroundwatersites(Fig.3b). present day (Fig. 4b). For sites 1–3, the 5% significance level in Thereisanapparentshiftinthedominantcyclepresentinthe groundwater levels that is consistent with PDO tends to be less hydrologictimeseriesoftheNorthAtlanticCoastalPlainascom- apparent,shorterinduration,andoftentemporallylaggedbehind pared to the Central Valley and Basin and Range aquifer systems the 5% significance level in the co-located precipitation sites (Fig.3c).The>PDOcycleisobservedatthreeoftheprecipitation (Fig.4b).TheWTCforNAOand>PDOwerenotpresentedbecause sitesintheNorthAtlanticCoastalPlain(NAP2,NAP3,andNAP4), the Central Valley precipitation and groundwater levels do not andcontributestothegreatestamountofvarianceineachrecord havestatisticallysignificantRCswithperiodicitiesconsistentwith with63%,70.7%,and65.5%,respectively(Fig.3c).SimilartoBasin NAOor>PDO. and Range sites with >PDO-like variations, the record lengths at Results of the WTC from the Basin and Range aquifer system sites NAP2, NAP3, and NAP4 are not the longest in this study. (Fig.5)havemanysimilaritiestotheWTCfromtheCentralValley The North Atlantic Coastal Plain sites have PDO-like cycles that aquifer.AllsitesintheBasinandRangehavemoderatetostrong contributeabout20–77.7%ofthevariability(Fig.3c).A2-to7-year (0.5–1) coherence with ENSO in the 3- to 7-year period range signalcontributes4.1–51.2%(NAP3andNAGW2,respectively)of (Fig.5a)andPDOinthe8-to24-yearperiodrange(Fig.5b).How- thevarianceattheNorthAtlanticCoastalPlainsites,andmaybe ever, unlike the Central Valley, site 2 in the Basin and Rangehas attributedtotheENSOortheNAO(Fig.3c).TheENSO2-to7-year coherencewith>PDOintheprecipitationandgroundwaterlevels periodicityislongerthantheNAO3-to6-yearperiodicity,butthe (Fig. 5c). The coherence between precipitation and groundwater NAO also has an observed 8- to 10-year periodicity (Kiladis and levelsandPDOhastwogeneralpatternssomewhatsimilartothose Diaz,1989;Hurrelletal.,2003).Previousstudieshavealsoshown fromtheCentralValley.Thereisstrongcoherenceatlonger(16– coherent ENSO effects in the Mid-Atlantic regions of the U.S. 25year)periodicitiesfromaboutthe1940sand1950stopresent (RopelewskiandHalpert,1986).Weusethewaveletanalysisand insitesP2andGW4,whilesitesP1,P3,GW3,andP5havestrong thelag correlationstoidentifytherelativeinfluenceoftheENSO coherence at shorter (8–14years) periodicities (Fig. 2c). Strong and (or) NAO on the 2- to 7-year cycle in the precipitation and coherenceconsistentwith>PDOwaspresentatsiteP2,particularly groundwaterlevelsintheNorthAtlanticCoastalPlain,asdescribed fromthe1940sto1960s(Fig.5c),whichisconsistentwiththelast below. fullpositivephaseofAMO(Fig.2d).However,thecoherenceofthe >PDOsignalisobserved outsidethecone ofinfluence.The phase 4.2.Coherencebetweenclimateindices,precipitation,and arrowsintheprecipitationandgroundwaterlevelWTCpointsto groundwaterlevels the left (out-of-phase), which is consistent with drought across muchoftheU.S.andespeciallythesouthwesternU.S.thatassoci- The wavelet transforms support the SSA findings and indicate atedwiththepositiveAMOphase(McCabeetal.,2004). significantcoherencebetweenENSO,NAO,PDO,and>PDOinmany ResultsoftheWTCfromtheNorthAtlanticCoastalPlainaquifer of the precipitation and groundwater levels. The WTC from the system(Fig.6)indicatemoderatetostrong(0.5–1)coherencewith Central Valley aquifer (Fig. 4), Basin and Range aquifer system ENSOinthe3-to7-yearperiodrange(Fig.6a)(withtheexception (Fig. 5), and North Atlantic Coastal Plain aquifer system (Fig. 6) ofGW1andP2),NAOinthe3-to7-yearperiodrange(Fig.6b),PDO arediscussednext. in the8- to24-year periodrange(Fig. 6c) (withthe exceptionof AllsitesintheCentralValleyhavemoderatetostrong(0.5–1) GW2), and with >PDO (Fig. 6d). The North Atlantic Coastal Plain coherencewithENSOintheWTCatthe3-to7-yearperiodrange sitesgenerallyhaveagreaterinfluenceofclimateoscillationsfrom (Fig.4a).AtCVP1,coherenceatthe5%significancelevelfrom3-to the Atlantic Ocean than from the Pacific Ocean; including coher- 7-yearsintheWTCispredominantlyfromtheearly-1970stothe ence with NAO at all 5 sites and with >PDO at 3 sites (Fig. 6b mid- to late-1980s, which is consistent with the strong 1976– and d). Each of the North Atlantic Coastal Plain sites has strong 1977 negative ENSO event and the 1982–1983 extreme positive coherenceandinsomecases(notablyatsitesP3andGW5)5%sig- ENSOevent(Fig.2a).AtCVGW1,coherenceatthe5%significance nificance level from 2 to 7years, which is consistent with ENSO levelfrom3to7yearsintheWTCisfromthelate-1970stolate- and NAO. WTCs were not computed with the NAO in the Basin 1980s, which is approximately 4- to 5-years lagged behind the and Range or the Central Valley due to the low probability of 5%significancelevelintheP1WTC(Fig.4a)duetothelagbetween influences from the NAO in the western U.S. (Hurrell et al., thearrivalofprecipitationandthetraveltimeofwaterthroughthe 2003).WhereastheCentralValleyandBasinandRangesiteshad 1946 A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 Fig.4. Waveletcoherence(WTC)betweenprecipitationandgroundwaterlevelsattheCentralValleyaquifersites1–5andthe(a)ElNiñoSouthernOscillation(ENSO)and(b) PacificDecadalOscillation(PDO).Notethatspectralpowerisdimensionless,thethickblacklinesarethe5%significancelevel,andthelessintensecolorsindicatetheconeof influence(COI).Thephaseangle(shownwithblackarrows)identifiesthephaserelationbetweentwoseries,witharight-pointingarrowindicatinganin-phaserelationanda left-pointingarrowindicatingananti-phaserelation,andarrowspointingupordownshowthatonetimeseriesisleadingtheotherby90(cid:3)(Grinstedetal.,2004;Holman etal.,2011).(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.) 5%significancecoherencewithENSOcenteredonthe1980s,most AtlanticCoastalPlainWTCgenerallyareright-pointingindicating oftheNorthAtlanticCoastalPlainshaveweaktomoderatecoher- anin-phaserelationbetweenthe>PDOcycleandtheprecipitation enceduringthe1980s(Fig.6a).MostoftheNorthAtlanticCoastal (Fig.6d),whichisconsistentwiththelessfrequentdroughtasso- Plain sites have 5% significance coherence with NAO that is cen- ciated with the positive phase AMO along the eastern coast of teredonthemid-tolate-1980s(Fig.6b),whichisconsistentwith theU.S.(McCabeetal.,2004). the shift from the 1940s to 1970s negative phase to a positive phase in the 1980s with the greatest NAO index values in 1983, 4.3.Lagcorrelationsbetweenclimateindices,precipitation,and 1989, and 1990 (Fig. 2b) (Hurrell and Van Loon, 1997). These groundwaterLevels WTC findings indicate relatively less ENSO and relatively more NAO influence on the precipitation, groundwater levels, and Results indicate that nearly all of the composite RCs from the recharge to aquifers along the North Atlantic Coast as compared precipitation time series from the selected PAs are statistically tothoseaquifersinthewesternU.S.Moderatetostrongcoherence correlated(95%confidenceinterval)withtheMEI,NAO,PDO,and (0.5–0.8) with >PDO is more apparent in the precipitation (sites AMO indices. These findings are consistent with the well estab- 2–4)than groundwater levels of theNorth AtlanticCoastal Plain, lished hydroclimatology literature presented in the background particularly from the 1940s to 1970s (Fig. 6d). Unlike the >PDO that U.S. precipitation spatiotemporal patterns have complex coherence in the Basin and Range, the phase angle in the North teleconnectionstointerannualandmultidecadalclimatevariability. A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 1947 Fig.5. Waveletcoherence(WTC)betweenprecipitationandgroundwaterlevelsattheBasinandRangeaquifersystemsites1–5andthe(a)ElNiñoSouthernOscillation (ENSO),(b)PacificDecadalOscillation(PDO),and(c)AtlanticMultidecadalOscillation(AMO).Notethatspectralpowerisdimensionless,thethickblacklinesarethe5% significancelevel,andthelessintensecolorsindicatetheconeofinfluence(COI).Thephaseangle(shownwithblackarrows)identifiesthephaserelationbetweentwoseries, witharight-pointingarrowindicatinganin-phaserelationandaleft-pointingarrowindicatingananti-phaserelation,andarrowspointingupordownshowthatonetime seriesisleadingtheotherby90(cid:3)(Grinstedetal.,2004;Holmanetal.,2011).(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtotheweb versionofthisarticle.) Ofthe21precipitationtimeseriesanalyzedfromthefourPAs,20 Fisher and Healy, 2008; Faunt et al., 2009). Results of these havecompositeRCslagcorrelatedwiththeMEI(coefficientsrange statisticalanalysesprovideinsightintothephysicalprocessesthat from0.12to0.60),5havecompositeRCslagcorrelatedwithNAO influence hydroclimatic variability in groundwater levels. Future index(coefficientsrangefrom0.13to0.27),19havecompositeRCs modeling studies are needed to explore how the lag times of lag correlated with PDO index (coefficients range from 0.14 to climate signal at co-location precipitation and groundwater level 0.83), and 8 have composite RCs lag correlated with AMO index sitesrespondtolocalvadosezonematerials(Gurdaketal.,2007), (coefficientsrangefrom(cid:2)0.27to(cid:2)0.72).OnlythecompositeRCs includingtheverticaldistributionofsandcontentorlayersofclay in precipitation that were significantly correlated to the climate andthehydrogeologicpropertiesofthesaturatedaquifer,includ- indices were used in the lag correlation analysis between the inglocaltransmissivityandstorativity(Dickinsonetal.,2004). co-locatedprecipitationandgroundwaterlevels(Fig.7). Moderatetostrongstatisticallysignificantlagcorrelationscoef- ManyCentralValleyandBasinandRangeco-locatedprecipita- ficients between co-located precipitation and groundwater levels tion and groundwater levels had moderate to strong statistically composite RCs that have periodicities consistent with PDO were significant (95% confidence interval) lag correlations (coefficients identified in the Central Valley and Basin and Range (Fig. 7a and rangefrom0.5to0.9)withcompositeRCsofsimilarperiodicities b).Interestingly,therangeofphaselags(11–27years)ofRCswith asENSOandPDOcycles(Fig.7aandb).However,onlyBasinand periodicities similar to PDO are generally larger than phase lags Range site 2 had statistically significant lag correlations between (3–14years)ofRCswithperiodicitiessimilartoENSOintheCentral precipitationandgroundwaterlevelswithperiodicitiessimilarto Valley (Fig. 7a) and Basin and Range (Fig. 7b). These statistical >PDO (Fig. 7b). In the Central Valley, the average lag correlation resultsprovideinsightsintothephysicalprocessesthatgroundwa- coefficient between precipitation and groundwater levels consis- ter levels in the western U.S. may have a relatively faster hydro- tent with ENSO periodicity was 0.5 with an average phase lag of logic response to ENSO as compared to PDO. The statistically 11years(Fig.7a).Wesuggestthattheaveragephaselagisconcep- significant phase lags between the co-located precipitation and tuallyequivalenttotheaveragetraveltimeofwaterinthevadose groundwaterlevelsintheNorthAtlanticCoastalPlainrangedfrom zone between the arrival of precipitation at land surface and the 3 to 16years (Fig. 7c) and are generally much smaller than the responseinthegroundwaterlevels,andthe11-yearaveragephase corresponding phase lags for ENSO or PDO in the Central Valley lag is reasonable given previously reported recharge rate of or Basin and Range aquifers. These statistical results point to 86–530mmyr(cid:2)1 in the Central Valley (Phillips et al., 2007; importantdifferencesinhydrologicresponsetoENSOandPDOin 1948 A.J.M.Kuss,J.J.Gurdak/JournalofHydrology519(2014)1939–1952 Fig.6. Waveletcoherence(WTC)betweenprecipitationandgroundwaterlevelsattheNorthAtlanticCoastalPlainaquifersystemsites1–5andthe(a)ElNiñoSouthern Oscillation(ENSO),(b)NorthAtlanticOscillation(NAO),(c)PacificDecadalOscillation(PDO),and(d)AtlanticMultidecadalOscillation(AMO).Notethatspectralpoweris dimensionless,thethickblacklinesarethe5%significancelevel,andthelessintensecolorsindicatetheconeofinfluence(COI).Thephaseangle(shownwithblackarrows) identifiesthephaserelationbetweentwoseries,witharight-pointingarrowindicatinganin-phaserelationandaleft-pointingarrowindicatingananti-phaserelation,and arrowspointingupordownshowthatonetimeseriesisleadingtheotherby90(cid:3)(Grinstedetal.,2004;Holmanetal.,2011).(Forinterpretationofthereferencestocolorin thisfigurelegend,thereaderisreferredtothewebversionofthisarticle.) the western groundwater levels as compared to the responses in the eastern water levels. These differences may be a function of the relatively smaller recharge fluxes in the western aquifers as compared to recharge fluxes in the eastern aquifers (McMahon et al., 2011), which may help regulate the propagation of the climate variability signals from land surface in the infiltrating watertothewatertableintherechargeflux.Additionalnumerical modeling is needed to further explore the controlling physical processes for the observed spatial and temporal differences in the phase lags of the climate variability signals in the co-located precipitationandgroundwaterlevels. 4.4.Nationaltrends Thisstudyprovidesevidenceofimportantregionalandnational patternswithrespecttotheENSO,NAO,PDO,and>PDOtelecon- nections and groundwater level variability across the U.S. In general, climate oscillations associated with the Pacific Ocean (ENSOandPDO)(Fig.8aandc)haveagreatercontrolthanAtlantic Oceanoscillations(NAOandAMO)(Fig.8bandd)onhydroclimatic variabilityinPAsacrosstheU.S. AlthoughENSOwasnotthedominantmodeofvariabilityinthe groundwaterlevels(onaverage13.3%,13.5%,and26.7%intheCen- tral Valley, the Basin and Range, and the North Atlantic Coastal Plain,respectively(Fig.3)),ENSOisastatisticallysignificantmode ofvariabilitythatwasobservedatallbut1ofthe42(precipitation and groundwater level) sites, including those in the High Plains (Fig. 8a). Interestingly, there are no statistical differences Fig.7. Summaryoflagcorrelationcoefficients(unitless)andphaselags(years)for (a=0.05,Tukey–Kramertest(Tukey,1977))inthestrengthoflag RCs that are significantly (95% confidence interval) correlated with ENSO, NAO, correlations between ENSO and hydroclimatic variability in the PDO,and>PDOfromco-locatedprecipitationandgroundwaterlevelsiteinthe(a) PAs(Fig.8a).Althoughnotstatisticallysignificant,themedianlag CentralValleyaquifer,(b)BasinandRangeaquifersystem,and(d)NorthAtlantic CoastalPlainaquifersystem.Sitesarenumber1–5andcorrespondtoTable1. correlations coefficients are slightly stronger in the PAs in the

Description:
recharge rates and mechanisms and other subsurface hydrologic processes that NAO, PDO, and AMO and the groundwater resources of principal aquifers .. orthogonal functions (EOFs) (Vautard et al., 1992; Ghil et al.,. 2002).
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.