Environ Monit AssessDOI 10.1007/s10661-011-1915-z

Assessing possible visitor-use impacts on water qualityin Yosemite National Park, California

David W. Clow · Rachael S. Peavler ·Jim Roche · Anna K. Panorska ·James M. Thomas · Steve Smith

Received: 28 March 2010 / Accepted: 27 January 2011© Springer Science+Business Media B.V. (outside the USA) 2011

Abstract There is concern that visitor-use asso-ciated activities, such as bathing, dish washing,wastewater production, and stock animal use nearlakes and streams, could cause degradation of

Electronic supplementary material The online versionof this article (doi:10.1007/s10661-011-1915-z) containssupplementary material, which is available toauthorized users.

D. W. ClowU.S. Geological Survey,Colorado Water Science Center,Denver, CO, USA

D. W. Clow (B)MS 415, Federal Center, Denver, CO 80225, USAe-mail: dwclow@usgs.gov

R. S. PeavlerGSI Water Solutions, Inc., Portland, OR, USA

J. RocheNational Park Service, Yosemite National Park,Yosemite, CA, USA

A. K. PanorskaDepartment of Mathematics, University of Nevada,Reno, NV, USA

J. M. ThomasDesert Research Institute, Reno, NV, USA

S. SmithU.S. Geological Survey, National Water QualityLaboratory, Denver, CO, USA

water quality in Yosemite National Park. A studywas conducted during 2004–2007 to assess pat-terns in nutrient and Escherichia coli (E. coli) con-centrations in the Merced and Tuolumne Riversand characterize natural background concentra-tions of nutrients in the park. Results indicatedthat nutrient and E. coli concentrations werelow, even compared to other undeveloped sitesin the United States. A multiple linear regres-sion approach was used to model natural back-ground concentrations of nutrients, with basincharacteristics as explanatory variables. Modelednitrogen concentrations increased with eleva-tion, and modeled phosphorus concentrations in-creased with basin size. Observed concentrations(±uncertainty) were compared to modeled con-centrations (±uncertainty) to identify sites thatmight be impacted by point sources of nutrients,as indicated by large model residuals. Statisticallysignificant differences in observed and modeledconcentrations were observed at only a few loca-tions, indicating that most sites were representa-tive of natural background conditions. The empiricalmodeling approach used in this study can be usedto estimate natural background conditions at anypoint along a study reach in areas minimally im-pacted by development, and may be useful for set-ting water-quality standards in many national parks.

Keywords Standards · Visitors · User-capacity ·Nutrients · E. coli · Yosemite

Environ Monit Assess

Introduction

In the mid-1980s, the U.S. Congress designatedthe Merced and Tuolumne Rivers in YosemiteNational Park, California, as Wild and ScenicRivers, reflecting the importance of these riversystems as a natural and cultural resource forthe country (Merced River, Public Law 100–149;Tuolumne River, Public Law 98–425). The Wildand Scenic River designation provides special pro-tection for the rivers’ Outstandingly RemarkableValues (ORVs) and free-flowing status while al-lowing for public use and enjoyment (16 USC1271–1287). Outstandingly Remarkable Valuesare river-related values that make a river uniqueand worthy of special protection. In Yosemite,these values include aesthetic, recreational, bio-logical, and hydrological features. Water quality isintimately linked to each of these values, and theWild and Scenic Rivers Act requires protection ofwater quality along designated reaches.

Yosemite National Park receives approxi-mately 3.4 million visitors per year, and the park’swaterfalls, streams, and lakes are a primary attrac-tion (National Park Service, Public Use StatisticsOffice, personal communication, 2008). As such,the areas around these water features receiveheavy visitor-use, especially in Yosemite Valleyand in Tuolumne Meadows, and there is con-cern that this could cause degradation of naturaland cultural resources. Extensive social trails insome high-use areas near the rivers have led toincreased bank erosion, channel width, and sedi-ment transport (Madej et al. 1994). Impacts fromcamping activities (e.g., bathing, swimming, anddish washing) and from pack animals at streamcrossings are of concern because of their poten-tial to introduce nutrients and other contaminantsinto surface waters.

Thin, poorly developed soils and an abun-dance of exposed granitic bedrock provide littlebuffering or nutrient assimilation capacity, espe-cially during snowmelt or storm events; as a re-sult, pollutants may enter streams and lakes withlittle attenuation (Clow et al. 1996; Hoffman et al.1976; Sorenson and Hoffman 1981). The prob-lem is particularly acute at high elevation, whereaquatic ecosystems are being impacted by excess

atmospheric deposition of nitrogen, a non-pointsource pollutant (Clow 2010).

Lower elevation ecosystems may be less sus-ceptible to atmospheric deposition impacts be-cause they tend to have more soil and vegetation;however, they tend to have higher visitor-use.Degradation of water quality in Yosemite Val-ley is of particular concern because of the largenumber of visitors the area receives. Althougha study of water quality in the Merced Riverduring the 1990s indicated low nutrient concen-trations in Yosemite Valley, nutrient concentra-tions increased downstream from the wastewatertreatment plant at El Portal, which processeswastewater from the valley (Brown and Short1999).

The Wild and Scenic Rivers Act requires theNational Park Service (NPS) to protect and en-hance the ORVs for which the rivers were desig-nated (including water quality), and specifies thatthis be done through managing user capacities.User capacity is defined as the type and amountof visitor-use that can be accommodated withoutadverse impacts on the ORVs of the river, or onthe quality of the recreational experience. Visitor-use management in the NPS is an adaptive processwhere limits are set based on recreational, naturaland cultural resources, and social considerations;indicators of these factors are then monitoredin order to maintain acceptable park conditions.Several similar frameworks are available for de-veloping indicators and using monitoring datato inform park management, including Limits ofAcceptable Change (Stankey et al. 1985), Visi-tor Impact Monitoring (Graefe et al. 1990), andthe Visitor Experience and Resource Protectionframework (National Park Service 1997). Theseframeworks generally have four key elements: (1)determination of desired conditions for the re-source, (2) selection of measurable indicators andstandards, (3) monitoring of the indicators andcomparison to standards, and (4) implementationof management actions to be taken if desired con-ditions are not being met. Desired conditions arequalitative descriptions linked to the rivers’ ORVsand quality of the visitor experience along theriver corridor. Indicators and standards are mea-surable quantities that allow park management

Environ Monit Assess

to determine whether the desired conditions aremet; standards reflect the minimal acceptablecondition of an indicator (National Park Service1997). The word ‘standard’ as used in the contextof user-capacity management and in this paperapplies to park-specific goals and does not referto regulatory requirements (National Park Service1997).

Selection of appropriate indicators and stan-dards that will protect and enhance the rivers’ORVs is an important step in the visitor-use man-agement process. In Yosemite, selected indicatorsinclude social and natural resource parameters(Bacon et al. 2006). User capacities usually areset using social indicators and models; naturalresource indicators, such as water quality, are usedas a check to ensure that no resource degradationoccurs. The focus of this paper is on water quality,which is a key component in establishing usercapacities because of its link to river health, andbecause water is related to all of the rivers’ ORVsand recreational uses. Water-quality constituents,such as nutrient concentrations, turbidity, bacte-ria, and organic wastewater compounds (OWCs),can be good indicators of whether desired condi-tions are being met because they are responsiveto perturbation and can be quantified with highaccuracy.

Examples of visitor-use associated nutrientsources include wastewater treatment plants,stock use near surface-water bodies, and improperwaste disposal at campsites. Stream-bank erosionmay cause an increase in turbidity, particularlyduring storms when loose soil may be washed intostreams and lakes. Escherichia coli (E. coli) is abacterium found in the lower intestine of humansand other warm-blooded animals. It can exist forshort times outside the gut and is a useful indicatorof fecal contamination. OWCs include caffeineand personal care products, which can be usefulindicators of visitor-use associated impacts; recentadvances in analytical methods have made it pos-sible to detect OWCs at very low concentrations(parts per trillion; Zaugg et al. 2006).

Development of standards is a challenging,yet critical component of visitor-use management(Hof and Lime 1997; National Park Service 1997).Ideally, a standard would be set to a “natural

background condition” that reflects conditionsprior to human influences (Smith et al. 2003).Quantification of natural background conditionsis problematic, however, because pristine refer-ence sites are scarce in the industrialized world(Smith et al. 2003). Even where point sourcesof nutrients do not exist, atmospheric depositionof nitrogen is an indirect source of nutrients towatersheds that need to be accounted for whenestablishing standards (Smith et al. 2003). As anadditional complication, nitrogen assimilation ca-pacities vary in relation to basin characteristics,such as distribution and type of vegetation, soils,and geology (Rohm et al. 2002; Smith et al. 2003).Thus, background conditions can vary spatially aswell, which should be considered when estimatingattainable conditions for stream-water quality.

Previous work on establishing reference con-ditions for nutrients has mostly been conductedat level III Ecoregion or larger scales. Clark et al.(2000) compiled nutrient data from undevelopedsites in the United States to provide baseline in-formation on natural background concentrationsacross the country. The U.S. EnvironmentalProtection Agency (USEPA) calculated referenceconditions for nutrients in level III Ecoregions ofthe United States, including the Sierra Nevada,based on the 25th percentile of a general popu-lation of streams (U.S. Environmental Protec-tion Agency 2000b). The data were aggregatedinto 14 Nutrient Ecoregions, and nutrient criteriafor total nitrogen and total phosphorus were rec-ommended for each of them (http://epa.gov/waterscience/criteria/nutrient/ecoregions/; accessedJanuary 2010). The reference conditions andnutrient criteria calculated by the USEPA repre-sent nutrient levels that are designed to protectaquatic ecosystems against adverse effects ofnutrient enrichment. Although useful at regionalscale, they are not designed to account for local-scale variability in basin characteristics that leadto spatial variability in the ability of ecosystems toassimilate nitrogen and phosphorus.

The objectives of the study described in thispaper were to (1) document current conditionsand spatial patterns in water quality in the upperMerced and upper Tuolumne River basins, (2)identify major influences on water quality, with

Environ Monit Assess

special emphasis on possible visitor-use effects,and (3) develop a method for modeling naturalbackground concentrations of nutrients that couldaccount for local variability in nutrient assimila-tion rates, and could be used to inform visitor-usemanagement in Yosemite National Park.

Study area

Yosemite National Park is located in the cen-tral Sierra Nevada of California, approximately

300 km (200 mi) east of San Francisco (Fig. 1). Thepark covers approximately 3,000 km2 (1,158 mi2),with elevations ranging from 648 to 3,997 m (2,127to 13,114 ft). Land use in most of Yosemiteis restricted to recreation; approximately 95%of the park is designated wilderness, wheremechanized travel is prohibited (http://www.nps.gov/yose/planyourvisit/wildregs.htm; accessed1/15/2011). High elevations have extensive ar-eas of neoglacial till (late Holocene till withminimal soil development), talus, and exposedbedrock with sparse alpine grasses and herbs.

Fig. 1 Map showinglocations of studysampling sites inYosemite NationalPark, California

CALIFORNIA

San Francisco

Yosemite Valley

Wawona

Foresta

Happy Isles

El Portal

1 2

58

18

19 20 21

23

1516 17

69

37 11

12

22

13

10

144

119˚20'W

119˚20'W

119˚40'W

119˚40'W

38˚10'N38˚10'N

37˚50'N37˚50'N

37˚30'N37˚30'N

Merced River

South Fork Merced River

Tuolumne River

Site Codes1. FRST

2. M140

3. POHO

4. MRYL

5. SNTB

6. HAPI

7. NEVF

8. MRML

9. MHSC

10. FLCR

11. MARC

12. MPFF

13. LMTX

14. LAK4

15. SFWC

16. SFKB

17. SFSB

18. CNCR

19. TRBC

20. 120B

21. DFGC

22. TWNB

Tuolumne Meadows

Sampling location

Roads

Rivers, lakes23. LFMC

Merced river basinSouth Fork of Merced river basin

Tuolumne river basin

Hetch Hetchy

0 10 km

Environ Monit Assess

Thick deposits of Pleistocene till occur at middleelevations. Coniferous forest dominates middleelevations, and mixed deciduous and coniferousstands occur at the lowest elevations in the park(Huber et al. 1989).

The Merced and Tuolumne Rivers drain thetwo primary river basins in Yosemite. The upperMerced River basin (upstream from the westernpark boundary) is 1,320 km2 (510 mi2), and in-cludes the main stem and the South Fork of theMerced River, both of which are west-flowingtributaries of the San Joaquin River. A U.S. Ge-ological Survey (USGS) stream gage has been inoperation on the Merced River at Happy Isles,which is at the upper end of Yosemite Valley,since 1915. Developments in the upper MercedRiver basin upstream from Happy Isles includetrails, small utility systems, wilderness campsites,and three Yosemite High Sierra Camps, whichare developed campgrounds with rustic lodg-ing surrounded by wilderness. Downstream fromHappy Isles, the main stem of the Merced Riverflows through Yosemite Valley, where infrastruc-ture includes roads, bridges, buildings, parkinglots, campgrounds, lodging units, horse camps,and electric, water, and sewer systems. BelowYosemite Valley, the Merced River flows adja-cent to the El Portal Administrative Site, whichincludes housing for approximately 640 residents,park maintenance, storage, warehouse facilities, aprivately owned hotel, and the El Portal Wastewa-ter Treatment Plant, which serves all of YosemiteValley. The South Fork of the Merced River flowsthrough Wawona, where developments includea campground, horse camp, hotel, golf course,maintenance buildings, employee housing, ap-proximately 300 privately owned residences, awastewater treatment plant, leach fields, store, gasstation, and an impoundment for the community’swater supply. Although park land upstream fromWawona on the South Fork is free of developmentor roads, adjacent Sierra National Forest landswithin the basin contain roads, informal campingareas, and evidence of logging.

The upper Tuolumne River basin (upstreamfrom the western park boundary) drains approx-imately 1,733 km2 (669 mi2) in the northernpart of the park, and has several large tribu-

taries. Development is concentrated in TuolumneMeadows, and includes a 304-site campground,store, gas station, horse camp, buildings, em-ployee housing, and a wastewater treatment fa-cility. Downstream from Tuolumne Meadows, de-velopment consists of a High Sierra Camp andHetch Hetchy reservoir, which provides drinkingwater for the City of San Francisco (San FranciscoPublic Utilities Commission; http://sfwater.org/msc_main.cfm/MC_ID/13/MSC_ID/165; accessed1/15/2011).

Methods

Water-quality sampling strategy

Water-quality surveys were conducted 6 to 12times per year on the Merced River during 2004–2007, and a similar program was performed onthe Tuolumne River during 2005–2007. Samplingsites included wilderness and non-wilderness sites,which covered a gradient of visitor-use and de-velopment. Sampling frequency was tailored tomatch the seasonal hydrograph (where access per-mitted), with the highest frequency (weekly) dur-ing the high-flow snowmelt period (May–June)and the lowest frequency (monthly to bi-monthly)during the winter low-flow period (October–April). The objectives of this sampling strategywere to characterize seasonal variability in soluteconcentrations and to obtain a suite of samplesthat would be representative of the true popula-tion of stream chemistry. Winter samples were notobtained at wilderness sites because trails wereclosed by deep snow from November throughMay. Stream-water samples also were collectedopportunistically during storms at selected sites onboth rivers; these stream samples were referredto as storm samples. The purpose of the stormsampling was to characterize how nutrient andE. coli concentrations responded to rain events;previous studies have documented substantial in-creases in nutrient concentrations during storms(Peters 1994; Tsihrintzis and Hamid 1997; Wetzel2003).

Indicator constituents included dissolved nitriteplus nitrate, hereinafter referred to as NO3; dis-solved nitrogen, DN; dissolved phosphorus, DP;

Environ Monit Assess

total (unfiltered) phosphorus, TP; and E. coli.Concentrations of NO3 are expressed in mg L−1

“as N” throughout this paper. Reconnaissancedata on the occurrence of OWCs were collectedat a subset of sites during late summer 2006(Table 1). Results of the OWC sampling indicatedthat despite the high sensitivity of the analyticalmethod, 98% of concentrations were below thelimit of quantification, and all were within therange of blanks. Thus, OWC data are not furtherdescribed in this paper, however, sampling meth-ods and results are presented in the electronicsupplement (ESM Table 3, ESM Fig. 1).

Nutrients and E. coli

Water samples were collected for nutrient analy-ses at 11 sites on the main stem of the MercedRiver, 2 tributaries to the main stem of theMerced River, 3 sites on the South Fork of theMerced River, and 6 sites in the Tuolumne River(June–October only) (Table 1). Samples for E.coli analyses were collected at a subset of sites(Table 1).

Water-quality samples were collected usinggrab-sampling methods along well-mixed reaches;adequacy of mixing was established by measur-ing specific conductivity in cross sections acrossthe stream (U.S. Geological Survey 2011). Ad-ditional details on sampling methods and fieldmeasurements are available in the USGS Na-tional Field Manual (U.S. Geological Survey2011) and 2005 User Capacity Management Pro-gram Field Guide (http://www.nps.gov/archive/yose/planning/ucmp2005fieldguide.pdf; accessed2/2/2009).

Quality-control procedures included collectionand analyses of field blanks and replicates, whichcomprised 10 percent of the total sample load.A list of analytical methods, method detectionlimits, and reproducibility of replicates is providedin Table 1 in the electronic supplement (ESMTable 1). Concentrations in blanks were belowthe Method Detection Limits (MDLs) 89% of thetime for NO3, 76% for DN, 97% for DP, and100% for TP. Median differences in replicates(n = 76) were 0.001 mg L−1 for NO3, 0.001 mg L−1

for DN, 0.0007 mg L−1 for DP, and 0.0006 mg L−1

for TP.

Statistical methods

Concentration distributions

Concentration distributions at individual siteswere characterized by calculating percentiles us-ing uncensored concentration data. By usinguncensored data to calculate concentration distri-butions, subjective reclassification of values be-low the MDL was avoided (e.g., to zero or to0.5 × MDL). It should be recognized, however,that individual data points below the MDL arenot considered quantitative. E. coli data belowthe analytical detection limit were censored, soconcentration distributions were calculated usingKaplan–Meier estimation techniques (Klein andMoeschberger 1997).

The bootstrapping technique was used to quan-tify the uncertainty of median concentrationsfor sites with at least 20 observations (Efron1981; Efron and Tibshirani 1993). In the boot-strap method, random samples of size n − 1 withreplacement are drawn from the original data,where n is the number of observations. The statis-tic of interest (median in our case) is computed foreach of those samples. Usually many such boot-strap samples are drawn and the resulting valuesof the statistic of interest are used to approximateits distribution. In particular, bootstrap distribu-tion of a statistic is often used to approximateits percentiles, mean, or variance and constructconfidence intervals for those parameters (Efron1981; Efron and Tibshirani 1993). In this study,500 bootstrap samples were drawn to obtain 95%confidence intervals around the median. For siteswith fewer than 20 observations, uncertainty wascalculated as the standard error of the median(Wilcox 2005).

Major assumptions of the bootstrap method arethat there are no trends in the data (stationarity),and that the observed data represent the true pop-ulation (Efron and Tibshirani 1993). Trends wereevaluated using Seasonal Kendall’s Tau (SKT)trend test (Helsel and Hirsch 1992; Helsel et al.2005). Results indicated no significant trends, asexpected given the short time period of the study.To obtain sample data representative of the truepopulation, samples were collected over a rangeof hydrologic conditions and throughout the year,

Environ Monit Assess

Tab

le1

Lis

tofs

ites

and

num

ber

ofsa

mpl

esan

alyz

edfo

rva

riou

sco

nsti

tuen

tsdu

ring

the

stud

y,20

04–2

007

[Ele

v.,e

leva

tion

;NO

3,n

itra

te;D

N,d

isso

lved

nitr

ogen

;DP

,di

ssol

ved

phos

phor

us;T

P,t

otal

phos

phor

us;E

.col

i,E

sche

rich

iaco

li;O

WC

,org

anic

was

tew

ater

cont

amin

ants

]

Site

USG

Ssi

teID

Site

nam

eSi

teE

lev.

Ele

v.L

atit

ude

Lon

gitu

deN

O3

DN

DP

TP

E.c

oli

OW

Ca

code

(m)

(ft)

137

4010

1194

8580

1M

erce

dR

iver

abov

eF

ores

taB

ridg

eF

RST

500

1640

37◦ 4

0′10

′′11

9◦48

′ 58′

′61

6161

6038

22

3740

1711

9473

301

Mer

ced

Riv

erab

ove

SR14

0B

ridg

eM

140

556

1825

37◦ 4

0′17

′′11

9◦47

′ 33′

′35

3535

3423

337

4259

1193

9530

1M

erce

dR

iver

abov

eP

ohon

oB

ridg

eP

OH

O11

8338

8137

◦ 42′

59′′

119◦

39′ 5

7′′

6262

6261

384

3743

5111

9363

600

Mer

ced

Riv

erbe

low

Yos

emit

eL

odge

MR

YL

1195

3921

37◦ 4

3′51

′′11

9◦36

′ 36′

′2

537

4436

1193

5200

1M

erce

dR

iver

abov

eSe

ntin

elB

ridg

eSN

TB

1204

3950

37◦ 4

4′36

′′11

9◦35

′ 20′

′35

3535

3421

611

2645

00M

erce

dR

iver

abov

eH

appy

Isle

sH

AP

I12

1940

0037

◦ 43′

54′′

119◦

33′ 3

1′′

122

109

6657

392

737

4329

1193

1540

1M

erce

dR

iver

abov

eN

evad

aF

alls

NE

VF

1826

5990

37◦ 4

3′29

′′11

9◦31

′ 54′

′20

2020

208

3744

1711

9250

401

Mer

ced

Riv

erbe

low

Mer

ced

Lak

eM

RM

L21

9572

0037

◦ 44′

17′′

119◦

25′ 0

8′′

1415

1515

937

4418

1192

4110

1M

erce

dR

iver

abov

eH

igh

Sier

raC

amp

MH

SC22

0772

4037

◦ 44′

18′′

119◦

24′ 1

1′′

77

78

1037

4423

1192

3480

0F

letc

her

Cre

ekF

LC

R22

2573

0037

◦ 44′

22′′

119◦

23′ 4

6′′

88

89

1137

4350

1192

3360

0M

erce

dR

iver

abov

era

nger

cabi

nM

AR

C22

2573

0037

◦ 43′

46′′

119◦

23′ 3

5′′

44

44

1237

4134

1192

0590

1M

erce

dP

eak

For

kat

foot

brid

geM

PF

F24

8681

5737

◦ 41′

34′′

119◦

20′ 5

9′′

66

66

1337

4224

1191

9270

0L

yell

For

kof

Mer

ced

R.a

bove

trai

lcro

ssin

gL

MT

X28

4193

2037

◦ 42′

23′′

119◦

19′ 3

0′′

99

99

1437

4215

1191

7150

0L

ake

4A1–

016

LA

K4

2908

9540

37◦ 4

2′15

′′11

9◦17

′ 18′

′5

55

515

3733

0211

9371

100

S.F

ork

Mer

ced

R.b

elow

Waw

ona

Cam

pgro

und

SFW

C11

7738

6037

◦ 33′

05′′

119◦

41′ 1

4′′

5959

5959

452

1637

3219

1193

9280

0S.

For

kM

erce

dR

.abo

veSo

uth

For

kB

ridg

eSF

KB

1204

3950

37◦ 3

2′19

′′11

9◦39

′ 28′

′34

3434

3423

1737

3219

1193

7110

0S.

For

kM

erce

dR

.abo

veSw

ingi

ngB

ridg

eSF

SB12

7441

8037

◦ 32′

19′′

119◦

37′ 1

1′′

5959

5958

461

1837

5433

1192

5090

0T

uolu

mne

Riv

erbe

low

Con

ness

Cre

ekC

NC

R23

9678

6037

◦ 54′

33′′

119◦

22′ 5

9′′

1212

1212

52

1937

5234

1192

2590

0T

uolu

mne

Riv

erab

ove

Bud

dC

reek

TR

BC

2609

8560

37◦ 5

2′34

′′11

9◦22

′ 59′

′40

4038

3813

220

3752

3411

9211

400

Tuo

lum

neR

iver

abov

eH

wy1

20B

ridg

e12

0B26

1585

8037

◦ 52′

34′′

119◦

21′ 1

4′′

1515

1515

1321

3752

4511

9180

701

Dan

aF

ork

belo

wG

aylo

rC

reek

DF

GC

2816

9240

37◦ 5

2′45

′′11

9◦18

′ 07′

′13

1313

1313

2237

5210

1191

9500

1L

yell

For

kat

Tw

inB

ridg

esT

WN

B26

4386

7037

◦ 52′

09′′

119◦

19′ 5

3′′

3030

3030

1123

3746

4011

9154

101

Lye

llF

ork

belo

wM

aclu

reC

reek

LF

MC

3109

1020

037

◦ 46′

39′′

119◦

15′ 4

4′′

1313

99

a Col

lect

edA

ugus

t–O

ctob

er20

06

Environ Monit Assess

except at wilderness sites where winter trail clo-sures precluded access.

Spatial dif ferences

Differences in concentration distributions amongsites were tested in two steps using nonparamet-ric methods. First, the Kruskall–Wallis test wasused to determine whether the medians of thedistributions of nutrient concentrations at all thesites were identical (Helsel and Hirsch 1992). Ifthe Kruskall–Wallis test result indicated that atleast one site was significantly different from theothers, an analysis of variance (ANOVA) wasperformed on ranked data with a Tukey multiple-range test to identify sites or groups of sites thatdiffered from each other in median concentration.The objective of these tests was to identify spatialpatterns in nutrient concentrations that might berelated to visitor-use. All tests for statistical sig-nificance were evaluated at a significance level of0.05 unless otherwise stated.

Characterizing natural backgroundconcentrations

To characterize spatially varying natural back-ground concentrations, statistical models were de-veloped by regressing median observed nutrientconcentrations against a suite of basin charac-teristics using stepwise multiple linear regression

(MLR). The results, which were median mod-eled nutrient concentrations for each site, reflectthe ability of the upstream basin area to assimi-late inputs of nitrogen and phosphorus. Modelednutrient concentrations were compared to ob-served concentrations to identify sites with largemodel residuals, which can be indicative of im-pacts from point sources of contamination. Thesignificance of differences between modeled andobserved concentrations was evaluated by de-termining whether the 95% confidence intervalsof modeled and observed concentrations over-lapped. If they overlapped, then differences werenot statistically significant; if they did not overlap,differences were significant.

The input variables considered in the MLRanalysis included, but were not limited to, basinarea, average elevation, slope, aspect, bedrockand surficial geology, vegetation, and atmosphericdeposition (for N only, because atmospheric de-position of phosphorus was considered negligi-ble; see Table 2 for a complete list of possibleexplanatory variables). The basin characteristicthat explained the most variance in the chemicalvariable entered the model first. The variancesexplained by the remaining explanatory variableswere recalculated, and the variable that explainedthe next greatest amount of variance enteredthe model next. This iterative process was re-peated until no additional variables showed sta-tistically significant correlations to the dependentchemical variable at p ≤ 0.1. Multicollinearity

Table 2 Basincharacteristics used inregression analysis,grouped by data layers[Atmospheric depositionof nitrogen not listedhere, but was includedin analysis]

aDatum, NAD83

Topographya Geology Vegetation

Basin area (mi2) Alluvium (%) Mixed conifer (%)Relief (ft) Diorite (%) Deciduous (%)Average elevation (ft) Granite (%) Foothill woodland (%)Max elevation (ft) Granodiorite (%) Juniper/cedar (%)Min elevation (ft) Mafic Intrusive (%) Unvegetated (%)Average slope (%) Metasediments (%) Snow (%)Slope > 30◦ (%) Metavolcanics (%) Talus (%)North (%, aspect) Neoglacial and talus (%) Water (%)Northeast (%, aspect) Pleistocene glacial till (%) Riparian (%)East (%, aspect) Other (%)Southeast (%, aspect)South (%, aspect)Southwest (%, aspect)West (%, aspect)Northwest (%, aspect)

Environ Monit Assess

among explanatory variables was evaluated us-ing the variance inflation factor [1/(1 − r2)] (Hairet al. 2005), with a threshold for exclusion of 0.2.The resulting beta coefficients (partial regressioncoefficients) for the explanatory variables repre-sent independent contributions of each explana-tory variable (Kachigan 1986). Residuals plots andnormal probability plots were used to check for vi-olation of assumptions of normality, linearity, andhomoscedasticity (Kachigan 1986). The standarderrors of the regression equations were used tocalculate 95% confidence intervals for modeledsolute concentrations.

Basin characteristics were quantified in ArcGISusing a 10-m resolution digital elevation model(DEM), 1:125,000 scale geologic and 1:24,000scale vegetation layers obtained from the NationalPark Service (NPS; http://www.nps.gov/gis/data_info/park_gisdata/ca.htm; accessed 2/5/2009), anda 1:24,000 scale hydrologic layer obtained fromthe USGS (http://nhd.usgs.gov/; accessed 2/5/2009). Geologic units with similar geochemicalcharacteristics were combined to obtain a sim-plified classification scheme, resulting in ten geol-ogy classes (Table 2); intrusive rocks covered thelargest area, but volcanic and metamorphic rockswere locally important. Important surficial unitsincluded talus and neoglacial deposits, glacialtill, and alluvium. Vegetation units were groupedinto nine classes based on major vegetation orland-cover types (Table 2); the most importantclasses by area were unvegetated, mixed conifer,juniper/cedar, and foothill woodland.

Current conditions and spatial patternsin nutrient and E. coli concentrations

Nutrient concentrations in the upper Merced,South Fork of the Merced, and Tuolumne Riverbasins were low, with median NO3, DN, DP,and TP concentrations below 0.082 (10−1.08), 0.135(10−0.87), 0.003 (10−2.52), and 0.007 (10−2.16) mgL−1, respectively, at all sites (Fig. 2a–d; ESMTable 2). For comparison, median NO3 and TPconcentrations at almost all of the study sites werebelow the medians calculated for undevelopedsites in the U.S. by Clark et al. (2000; Fig. 2a,d). Median TP concentrations at study sites also

were below reference conditions calculated bythe USEPA for the Sierra Nevada Ecoregion(0.015 mg L−1, Fig. 2d) and were below TP nutri-ent criteria established by the USEPA for Nutri-ent Ecoregion 2 (0.01 mg L−1; U.S. EnvironmentalProtection Agency 2000a). These results confirmthat nutrient concentrations in Yosemite are low,even compared to other undeveloped basins in theUnited States.

Results from the Kruskall–Wallis tests indi-cated that there were significant differences inconcentration distributions among sites for eachnutrient (NO3, DN, DP, and TP). ANOVA wasused to identify which sites or groups of sitesdiffered from each other. Results are shown inFig. 2; box plots with the same letter indicatedistributions that were not significantly differentfrom each other.

In general, NO3 concentrations tended to in-crease with elevation (Fig. 3a); NO3 concen-trations at high-elevation sites (LMTX, LAK4,LFMC) were significantly greater than at mostlow-elevation sites (Fig. 2a). There were few sig-nificant differences in NO3 concentrations amongother sites, except at the lowest elevation site inthe Merced basin, Foresta (FRST). At Foresta,although median NO3 concentrations were low,there were numerous anomalously high values(Fig. 2a), which occurred primarily during the latesummer and fall seasons. In the South Fork ofthe Merced River basin, although median con-centrations were low, concentrations were highlyvariable, particularly during late summer and fall.

Spatial patterns in DN concentrations weresimilar to those of NO3, exhibiting a general in-crease with elevation, with a few notable excep-tions (Fig. 3a–b). DN (and NO3) concentrationswere significantly higher at Foresta than at M140,3 km upstream (Fig. 2a–b). These sites are locatedapproximately 0.5 km below and 2.5 km abovethe El Portal wastewater treatment plant, respec-tively, suggesting that stream-water nitrogen con-centrations might be influenced by the treatmentplant. As with NO3, concentrations of DN wererelatively variable at Foresta and at the sites onthe South Fork of the Merced River.

The ratio of NO3 to DN declined with el-evation, and most of the decline occurred be-tween 3,200 and 2,100 m (Fig. 3c). This pattern

Environ Monit Assess

−3

−2

−1

0

log

NO 3 a

s N

(m

g L-1

)

−2

−1

0

log

TP

(m

g L-1

)lo

g D

N (

mg

L-1)

−4

−3

−2

−4

−3

−2

log

DP

(m

g L-1

)

Merced S. Fork TuolumneF

RS

T

M14

0

PO

HO

SN

TB

HA

PI

NE

VF

MR

ML

MH

SC

FLC

R

MA

RC

MP

FF

LMT

X

LAK

4

SF

WC

SF

KB

SF

SB

CN

CR

TR

BC

120B

DF

GC

TW

NB

LFM

C

52 35 53 35 113

20 14 7 8 4 6 9 5 59 35 59 12 40 15 13 31 12

AABCABABCDEFABCDABC

DEDEFGHDEFGCDDEFEFGFGHBCD GHGHH AEFGHGHDEF

GHDEFGFGH

ABABABABABABABABABABABBA ABABAB ABABBABABB

ABCDABCDABCDABCDABCDABCDABCDBCDABCDABCDABAAB DBCDABC ABCDDABCDCDCDD

ABCDEFG

BCDEFG

ABCDEFG

ABCDEFG

ABCDEFG

ABCDEFDEFGBCD

EFGBCBCDAABA CDEFG

CDEFGAB DEFGDEFGGEFGBCDEFG

upper quartile + 1.5*interquartile range75th percentilemedian25th percentilelower quartile - 1.5*interquartile range

d

c

a

b

Environ Monit Assess

� Fig. 2 Concentration distributions for a NO3, b dissolvednitrogen (DN), c dissolved phosphorus (DP), and (d) totalphosphorus (TP) at study sites in Yosemite National Park.Number of samples at each site is indicated by numberat top of plot. Box plots identif ied by the same letter atbottom of each plot indicate that the distributions were notsignificantly different. Short dashed horizontal blue lines in-dicate median NO3 and TP concentrations at undevelopedsites from Clark et al. (2000). Longer dashed horizontalred lines indicate USEPA reference conditions for TP inSierra Nevada ecoregion (U.S. Environmental ProtectionAgency 2000a, b). Dashed vertical green lines separate sitesin Merced, S. Fork, and Tuolumne river basins

is consistent with conversion of atmosphericallydeposited inorganic nitrogen to organic nitrogenin the aquatic environment. The relatively largechange in NO3:DN from high- to mid-elevationsreflects rapid conversion of inorganic nitrogen toorganic nitrogen as water moves from the alpinezone down into the montane forest.

In contrast with the nitrogen species, DPand TP tended to vary inversely with elevation(Fig. 3d–e). Although most DP and TP concen-trations were below the MDL, there were a fewsignificant differences among sites. Median DPand TP concentrations at low-elevation sites onthe Merced River (POHO, M140, and FRST) andthe South Fork of the Merced (SFWC) were sig-nificantly greater than at most high-elevation siteson the South Fork of the Merced and TuolumneRivers (Fig. 2c–d). All of the sites with elevatedDP and TP were below areas of high visitor-useor development, suggesting that disturbance orpoint sources might be influencing phosphorusconcentrations at those locations.

E. coli concentrations also were low, with me-dian concentrations below 5 MPN 100 ml−1, andthey exhibited substantial intra- and inter-sitevariability (Fig. 4; MPN is most probable number).The sites with the highest median E. coli concen-trations were the four sites on the Merced Riverin and below Yosemite Valley (excluding HAPI),and SFWC and SFKB on the South Fork of theMerced. These sites are in areas receiving highvisitor-use or are downstream from substantial

�Fig. 3 Scatter plots showing log median concentrations ofa NO3, b DN, c NO3:DN, d DP, and e TP against elevationat study sites in the upper Merced River basin

y = 0.00x - 2.04

r 2 = 0.54

-2.2

-2.0

-1.8

-1.6

-1.4

-1.2

-1.0

log

NO

3 (

mg

L-1

)

y = 0.00x - 1.23

r 2 = 0.51-1.3

-1.2

-1.1

-1.0

-0.9

-0.8

log

DN

(m

g L

-1)

b

y = -0.00x - 2.48

r 2 = 0.44-3.1

-3.0

-2.9

-2.8

-2.7

-2.6

-2.5

-2.4

log

DP

(m

g L

-1)

d

y = -0.00x - 2.09

r 2 = 0.81

-2.8

-2.7

-2.6

-2.5

-2.4

-2.3

-2.2

-2.1

0 500 1000 1500 2000 2500 3000 3500

Elevation (m)

log

TP

(m

g L

-1)

e

a

y = 0.00x + 0.10

r 2 = 0.480.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7N

O3:D

N (

mg

L-1

)c

Foresta

Environ Monit Assess

Fig. 4 Concentrationdistributions of E. coli atstudy sites during2004–2007. Number ofsamples at each siteindicated by number attop of plot. Box plotsymbology as in Fig. 2;dots indicate individualdata points. In somecases, medians overlapquartiles. Box plotsidentified by same letterindicate that thedistributions were notsignificantly different

log

E.C

oli (

MP

N p

er 1

00 m

l)0

0.5

1

1.5

2

2.5

FR

ST

M14

0

PO

HO

SN

TB

HA

PI

SF

WC

SF

KB

SF

SB

CN

CR

TR

BC

120B

DF

GC

TW

NB

S. Fork TuolumneMerced

ABC ABCD A AB ABCD ABCD ABCD BCD ABCD D ABCD CD ABCD

38 23 38 21 39 45 23 46 5 13 13 13 11

developments, including wastewater treatmentplants. E. coli concentrations typically were lowat the Tuolumne River sites. Relatively few of thedifferences among sites were statistically signifi-cant; exceptions included POHO and SNTB onthe Merced River in Yosemite Valley, where E. coliconcentrations were significantly higher than atTRBC and DFGC on the Tuolumne River.

For comparison, measured concentrations ofE. coli at sites downstream from developed ar-eas were within the range observed at undevel-oped sites, and were substantially below thoseobserved at urban, agricultural, and rural sitesin several previous USGS studies (Ebbert et al.2000; Hoos et al. 2002). No samples exceededthe USEPA single-sample standard for E. coliof 235 MPN/100 ml for moderate recreationalcontact (U.S. Environmental Protection Agency1986). Possible sources of E. coli include hu-mans, stock animals, and wildlife; it is not possibleto determine the dominant source of E. coli atYosemite sampling locations with available infor-mation. However, DNA microbial source tracking(ribotyping) might be a useful tool for identifyingthe source(s) of E. coli contamination in futurestudies (Stoeckel et al. 2004).

Seasonality in nutrient concentrationsand E. coli concentrations

Temporal variability in nutrient concentrations isdriven by a combination of seasonal hydroclimatic

and biological cycles and short-term hydrologicvariability related to storm events. Seasonal pat-terns in nutrient concentrations varied by nutrientspecies (Fig. 5). For the purpose of this study,the issue of primary concern is identifying whenexceedance of nutrient standards is most likely.

Median NO3 concentrations were highest dur-ing June–September, reflecting release of NO3

from melting snow. In contrast, median DN con-centrations were relatively low during this period,due to dilution of organic nitrogen by snowmelt.Variations in NO3 and DN concentrations weregreatest during late summer through early fall(July–October), probably because of a combina-tion of factors. Summers usually are quite dry inthe Sierra, and snowmelt is the dominant sourceof water to streams from May through Septem-ber. As snowmelt inputs and streamflow declinethrough the summer, the concentrating effect ofevapotranspiration on stream-water chemistry in-creases. By late summer and early fall, effluentfrom wastewater treatment plants can contribute5–15% of flow in reaches downstream from thetreatment plants. Wash off of dry deposition andflushing of accumulated salts from soil can providean additional source of nutrients to streams duringstorms, particularly the first storms of the fall sea-son. Opportunistic samples collected from streamsduring storms indicate that nutrient and E. coliconcentrations tend to increase substantially dur-ing storm events, particularly for the phosphorusspecies (Peavler 2008). The coefficient of varia-tion for all of the nutrient species was greatest

Environ Monit Assess

Fig. 5 Concentrationdistributions of a NO3,b DN, c DP, and d TP atstudy sampling sites bymonth, during 2004–2007.Number of samples ateach site indicated bynumber at top of eachplot. Box plot symbologyas in Fig. 2; dots indicateindividual data points

0

0.2

0.4

0.6

24 24 46 55 90 85 53 78 28 15

0

0.2

0.4

0.6

0.8

0

0.01

0.02

0

0.02

0.04

0.06

NO

3 (m

g L-1

as

N)

DN

(m

g L-1

)D

P (

mg

L-1)

TP

(m

g L-1

)

a

b

c

d

8075

J F M A M J J A S O N D

Environ Monit Assess

during October, which often coincides with thefirst rainstorms of the fall season. Median con-centrations of DP and TP generally were highestduring fall and winter (Fig. 5).

The seasonal pattern of variations in nutri-ent concentrations has implications for meetingwater-quality standards in Yosemite. Exceedanceof standards for nutrients is most likely to occurduring late summer and early fall, particularlyduring storms.

Models of natural background concentrationsof nutrients

The empirical models of nutrient concentrationssimulated observed concentrations reasonablywell, reproducing general trends with elevationand spatial patterns among the three study basins(Fig. 6). The amount of variance in observed con-centrations explained by the models was greaterfor the nitrogen species than for the phosphorusspecies (Table 3). This might be due, in part,to the very low concentrations of phosphorus inYosemite streams, which resulted in a low “signalto noise ratio.”

The selected input variables (and the sign oftheir coefficients) for NO3 included average ele-vation (positive), % Pleistocene glacial till (neg-ative), and % unvegetated terrain (positive) inthe basins (Table 3). This model explained 86%of the variance in median NO3 concentrationsand had an RMSE of 0.122. The input variablesselected by the stepwise MLR procedure are notunique in their predictive ability; other modelswith different explanatory variables can be usefulas well, providing only slightly less explanatorypower and slightly greater RMSE. For example,when average elevation and basin area were man-ually excluded from consideration in the model,the stepwise MLR substituted % neoglacial tilland talus (positive) for elevation (other explana-tory variables were the same); the adjusted r-square for this model was 0.85 and the RMSEwas 0.126 (not listed in Table 3 for brevity). Tosome extent, the interchangeability of input vari-ables reflects collinearity among them. Alternatemodels could be developed for any of the nu-trients; examination of alternate models can be

useful for gaining a better understanding of howbasin characteristics influence natural backgroundconcentrations of nutrients.

Input variables for the DN model were sim-ilar to those in the NO3 model, and included% neoglacial till and talus (positive) and %Pleistocene glacial till (negative; Table 3). Thecommonalities of input variables in the nitrogenmodels indicate that there are processes associ-ated with these basin characteristics that stronglyinfluence natural background concentrations ofnitrogen. The signs of the coefficients of the vari-ables provide information about the nature of theprocesses involved, and scaled coefficients showthe relative influence of factors in the regres-sion equations. The primary source of nitrogenin Yosemite wilderness areas is atmospheric de-position, which is a non-point source of nitrogen(Clow 2010). Nitrogen concentrations generallyincrease with elevation or as the fraction of basinarea classified as neoglacial till, talus, or unveg-etated terrain increases; this reflects the low as-similation capacities of these areas with respect toatmospherically deposited nitrogen. The positiverelation between elevation and NO3 may be duein part to a decline in growing season length withelevation, which limits the period of nitrogen up-take. Neoglacial till, talus, and unvegetated terrainhave minimal soil and vegetation, further limitingnitrogen uptake. In contrast, nitrogen concentra-tions varied inversely with % Pleistocene glacialtill, reflecting comparatively high nitrogen assim-ilation capacities in those areas, which tend tohave deep soil and well established forests. Anabundance of Pleistocene till in the DFGC basinexplains the relatively low observed and modelednitrogen concentrations in stream water at thatsite (Fig. 6).

It is noteworthy that atmospheric deposition ofnitrogen was not a significant predictor of stream-water nitrogen concentrations in the MLR mod-els. The influence of atmospherically depositednitrogen on natural background concentrationswas pointed out by Smith et al. (2003) in theirlarge-scale study of natural background con-centrations in the conterminous United States.Although the effect of nitrogen deposition onsurface-water chemistry in Yosemite is recog-nized, nitrogen deposition is difficult to quantify.

Environ Monit Assess

Fig. 6 Modeled andobserved median nutrientconcentrations at studysites in Yosemite. Errorbars indicate 95%confidence intervals inmodeled and observedmedian concentrations.Vertical dashed linesseparate basins. Sites ineach basin are orderedfrom lowest elevation onleft to highest elevationon right

-1.6

-1.4

-1.2

-1.0

-0.8

-0.6

-3.2

-3.0

-2.8

-2.6

-2.4

-2.2

-3.2

-2.8

-2.4

-2.0

-2.5

-2.0

-1.5

-1.0

-0.5

log

NO

3 (m

g L

-1)

ModeledObserved

FR

ST

M14

0P

OH

OS

NT

BH

AP

IN

EV

FM

RM

LM

HS

CF

LCR

MA

RC

MP

FF

LMT

XLA

K4

SF

WC

SF

KB

SF

SB

CN

CR

TR

BC

120B

DF

GC

TW

NB

LFM

C

log

TP

(m

g L

-1)

log

DN

(m

g L

-1)

log

DP

(m

g L

-1)

Merced SouthFork

Tuolumne

Environ Monit Assess

Table 3 Parametercoefficients for multiplelinear regression modelsof median NO3, DN, DP,and TP in surface watersin Yosemite NationalPark

Scaled coefficients are thecoefficients centered bymean, scaled by range/2

Variable Coefficient Scaled coefficient

log NO3 modelIntercept −2.50665Average elevation (m) 0.00031 0.20036Pleistocene glacial till (%) −0.01683 −0.39392Unvegetated (%) 0.05055 0.22496Adjusted r-square 0.86RMSE 0.122

log DN modelIntercept −1.01166Neoglacial till and talus (%) 0.01897 0.08443Pleistocene glacial till (%) −0.01140 −0.26679Adjusted r-square 0.85RMSE 0.061

log DP modelIntercept −2.91836Basin area (km2) 0.00040 0.20398Adjusted r-square 0.53RMSE 0.114

log TP modelIntercept −2.57483Basin area (km2) 0.00054 0.27457Neoglacial till and talus (%) −0.03421 −0.15225Pleistocene glacial till (%) −0.00785 −0.18364Adjusted r-square 0.78RMSE 0.107

This is because complex terrain causes complexspatial patterns in nitrogen deposition, which aredifficult to simulate with existing deposition models.

Basin area was the most important explana-tory variable in the MLR models for DP and TP(Table 3). The relation between basin area andDP and TP concentrations was positive, and mayreflect several inter-related basin characteristics.Mineral weathering is a common source of phos-phorus in basins lacking anthropogenic inputs(Mueller and Spahr 2006; Shacklette et al. 1971).In Yosemite, as basin size increases, average soildepth and soil temperatures increase, so mineralweathering sources of phosphorus are likely toincrease as well. Atmospheric deposition of phos-phorus might increase in concert with basin sizeif it is a substantial component of dry depositionbecause of an inverse relation between dry depo-sition and elevation; however, atmospheric depo-sition of phosphorus has not been quantified inYosemite. Additionally, land disturbance is likelyto increase with basin area, although this effectcould not be quantified with available information.

Comparison of modeled and observed nutrientconcentrations

Observed concentrations of nutrients exceededmodeled natural background concentrations bystatistically significant amounts at only a few sites(Fig. 6). Observed NO3 exceeded modeled NO3

at two adjacent sites in Yosemite Valley (SNTBand HAPI). Observed DN exceeded modeled DNat the lowest elevation site on the Merced River(FRST) and at HAPI. Observed DP exceededmodeled DP at SFWC, on the South Fork ofthe Merced River. Observed TP exceeded mod-eled TP at POHO, in Yosemite Valley, and atSFWC.

Sites where observed median concentrationsexceeded modeled median concentrations may ormay not be affected by point sources of conta-mination. It is possible that the models did notaccurately simulate natural background concen-trations at those sites; this could occur if the basincharacteristics used in the models did not reflectimportant processes or nutrient sources within

Environ Monit Assess

the basin. Secondly, although the nitrogen modelswere able to explain ≥85% of the variance innitrogen concentrations, the phosphorus modelswere not as powerful, explaining 53% and 78% ofthe variance in DP and TP concentrations, respec-tively (Table 3). Despite these caveats, it is worthnoting that sites where observed concentrationsexceeded modeled concentrations often occurredin areas downstream from possible point sourcesof nutrients. The DN exceedance at the site down-stream from the El Portal wastewater treatmentplant (FRST) is consistent with studies that havedocumented elevated nitrogen concentrations be-low wastewater treatment plants (Mueller andSpahr 2006). The TP exceedances at POHO andSFWC might reflect streambank disturbance asso-ciated with visitor-use, which is high in areas up-stream from the sites. The SFWC site also mightbe affected by leach fields at the campground,or by inputs from tributaries originating outsidethe park. It is suggested that sites where observedconcentrations exceeded modeled concentrationsmerit additional observation and study; this isparticularly true for sites where exceedances inmultiple nutrient species occurred, such as HAPIand SFWC.

Alternatives for setting nutrient standardsin Yosemite

The empirical models of natural background con-centrations of nutrients may be useful for settinggoals or standards for water quality in YosemiteNational Park and many other national parks. Themodeled nutrient concentrations reflect the inte-grated chemical signal resulting from processingof non-point sources of nutrients within the basin.The models do not account for point sources ofnutrients; thus, deviations from modeled concen-trations can be used to identify sites where pointsources of nutrients might be influencing waterquality. One benefit of the empirical modeling ap-proach is that, after model development, naturalbackground concentrations could be estimated forany stream site along the study reach based onbasin characteristics. Thus, new sites could be eas-ily incorporated into the monitoring program tofill spatial gaps or to perform model validation.

An alternative to using the empirical modelingapproach to setting standards for water quality inYosemite is to use the nutrient criteria establishedby the USEPA for Nutrient Ecoregion 2 (U.S.Environmental Protection Agency 2000b). Thereare two problems with using this approach inYosemite. First, it does not account for local-scalevariability in the capacity of watersheds to assim-ilate non-point sources of nutrients. This variabil-ity can be large, as reflected in highly variablenatural background concentrations of nutrients inthe park. Second, the USEPA nutrient criteriaare designed to reflect nutrient concentrations atrelatively undeveloped sites, rather than at siteswhere there is essentially no development. Waterquality in Yosemite is managed to a higher stan-dard; where water quality is acceptable, essentiallyno degradation is permitted (National Park Ser-vice 1997). Thus, applying the USEPA nutrientcriteria in Yosemite might not provide sufficientprotection for aquatic resources in the park.

Another alternative for setting nutrient stan-dards for use within the visitor-use managementframework is to use observed concentrations atunimpacted sites as a reference for future com-parison. Trends in nutrient concentrations couldbe evaluated periodically, or a percentile (e.g.,95th percentile) could be established as a site-specific standard. If concentrations exceed the95% percentile of concentrations at unimpactedsites, specific management actions could be im-plemented. This approach would be simpler toapply than developing empirical models of naturalbackground concentrations of nutrients becauseGIS and multiple-regression analyses would notbe required. A disadvantage of this alternative isthat identification of unimpacted sites would besubjective.

Summary and conclusions

This study documented that nutrient and E. coliconcentrations in Yosemite National Park werelow, even compared to other undeveloped sitesin the United States, and they showed consid-erable spatial and temporal variability. Nutrientconcentrations tended to be highest during latesummer or early fall, particularly during storms,

Environ Monit Assess

and exceedance of standards is most likely at thosetimes.

Spatial variations in nutrient concentrationswere strongly associated with variations in basincharacteristics. Concentrations of NO3 and DNgenerally increased with elevation, and concentra-tions of DP and TP increased with basin area. Thepositive correlation between nitrogen species andelevation is consistent with effects of atmospheri-cally deposited nitrogen and low rates of nitrogenassimilation at high elevation, as has been ob-served in previous studies in Yosemite and othermountainous areas in the Western United States.DP and TP patterns likely reflected increasingmineral weathering sources of phosphorus, or in-creasing land disturbance as basin size increased.

Empirical models of natural background con-centrations of nutrients were developed, withbasin characteristics as explanatory variables. Rel-atively few sites had significant differences in ob-served and modeled concentrations that might beattributable to point sources of pollution associ-ated with visitor-use. The empirical modeling ap-proach used in this study can be used to estimatenatural background conditions at any point alonga study reach in areas minimally impacted bydevelopment; and could be used for setting water-quality standards in Yosemite and other nationalparks.

Acknowledgements This work was supported by the U.S.Geological Survey/National Park Service Water QualityAssessment and Monitoring Program. The authors wish tothank David K. Mueller, Barbara Ruddy, and an anony-mous reviewer for helpful reviews of the manuscript.

References

Bacon, J., Roche, J., Elliott, C., & Nicholas, N. (2006).VERP: Putting principles into practice in YosemiteNational Park. The George Wright Forum, 23, 73–83.

Brown, L. R., & Short, T. M. (1999). Biological, habitat,and water quality conditions in the upper Merced Riverdrainage, Yosemite National Park, California, 1993–1996 (p. 56). U.S. Geological Survey Water-ResourcesInvestigations Report 99-4088.

Clark, G. M., Mueller, D. K., & Mast, M. A. (2000). Nu-trient concentrations and yields in undeveloped basinsof the United States. Journal of the American WaterResources Association, 36, 849–860.

Clow, D. W. (2010). Changes in the timing of snowmelt andstreamflow in Colorado: A response to recent warm-ing. Journal of Climate, 23, 2293–2306.

Clow, D. W., Mast, M. A., & Campbell, D. H. (1996). Con-trols on surface water chemistry in the upper MercedRiver basin, Yosemite National Park, California. Hy-drological Processes, 10, 727–746.

Ebbert, J. C., Embrey, S. S., Black, R. W., Tesoriero,A. J., & Haggland, A. L. (2000). Water quality in thePuget Sound basin, Washington and British Columbia,1996–98 (p. 31). U.S. Geological Survey Circular1216.

Efron, B. (1981). Nonparametric standard errors andconfidence intervals. Canadian Journal of Statistics, 9,139–172.

Efron, B., & Tibshirani, R. J. (1993). An introduction tothe bootstrap: Monographs on statistics and appliedprobability (p. 456). New York: Chapman & Hall.

Graefe, A., Kuss, F., & Vaske, J. (1990). Visitor impactmanagement: The planning framework (p. 100). Na-tional Parks Conservation Association.

Hair, J. F., Black, W. C., Babin, B., Anderson, R.,& Tatham, R. L. (2005). Multivariate data analysis(6th ed., p. 928). Upper Saddle River: Prentice Hall.

Helsel, D. R., & Hirsch, R. M. (1992). Statistical methods inwater resources (p. 522). Amsterdam: Elsevier.

Helsel, D. R., Mueller, D. K., & Slack, J. R. (2005). Com-puter program for the Kendall family of trend tests(p. 13). U.S. Geological Survey Scientific Investiga-tions Report.

Hof, M., & Lime, D. (1997). Visitor experience and re-source protection framework in the national park sys-tem: Rationale, current status, and future direction. U.S.Department of Agriculture, Forest Service, RockyMountain Research Station General Technical ReportINT-GTR-371.

Hoffman, R. J., Dong, A. E., & Keeter, G. L. (1976).Water-quality study of a reach of the Merced River inYosemite National Park and vicinity, California, April1973 through September 1974 (p. 66). U.S. GeologicalSurvey Open-File Report 76-326.

Hoos, A. B., Garrett, J. W., & Knight, R. R. (2002).Water quality of the Flint River basin, Alabama andTennessee, 1999–2000 (p. 46). U.S. Geological SurveyWater Resources Investigations Report 01-4185.

Huber, N. K., Bateman, P. C., & Wahrhaftig, C. (1989).Geologic map of Yosemite National Park and vicinity,California. U.S. Geological Survey, 1:125,000.

Kachigan, S. K. (1986). Statistical analysis (p. 589).New York: Radius.

Klein, J. P., & Moeschberger, M. L. (1997). Survivalanalysis: Techniques for censored and truncated data(p. 556). New York: Springer.

Madej, M. A., Weaver, W. E., & Hagans, D. K. (1994).Analysis of bank erosion on the Merced River,Yosemite Valley, Yosemite National Park, California,USA. Environmental Management, 18, 235–250.

Mueller, D. K., & Spahr, N. E. (2006). Nutrients in streamsand rivers across the nation—1992–2001 (p. 44). U.S.Geological Survey Scientific Investigations Report2006-5107.

Environ Monit Assess

National Park Service (1997). Visitor experience and re-source protection (VERP) framework: A handbook forplanners and managers (p. 108). National Park Service.

Peavler, R. S. (2008). Design and implementation of awater-quality monitoring program in support of es-tablishing user capacities in Yosemite National Park(p. 101). Masters thesis, University of Nevada, Reno.

Peters, N. E. (1994). Water-quality variations in a forestedPiedmont catchment, Georgia, USA. Journal of Hy-drology, 156, 73–90.

Rohm, C. M., Omernik, J. M., Woods, A. J., & Stoddard,J. L. (2002). Regional characteristics of nutrient con-centrations in streams and their application to nutrientcriteria development. Journal of the American WaterResources Association, 38, 213–239.

Shacklette, H. T., Hamilton, J. C., Boerngen, J. G.,& Bowles, J. M. (1971). Elemental composition ofsurf icial materials in the conterminous United States(p. 71). U.S. Geological Survey Professional Paper574-D.

Smith, R. A., Alexander, R. B., & Schwarz, G. E. (2003).Natural background concentrations of nutrients instreams and rivers of the conterminous United States.Environmental Science & Technology, 37, 3039–3047.

Sorenson, S. K., & Hoffman, R. J. (1981). Water-qualityassessment of the Merced River, California, in the 1977water year (p. 37). U.S. Geological Survey Water-Resources Investigations Report 80-75.

Stankey, G. H., Cole, D. N., Lucas, R. C., Petersen, M. E.,& Frissell, S. S. (1985). The limits of acceptable change(LAC) system for wilderness planning (p. 37). U.S.Department of Agriculture, Forest Service GeneralTechnical Report INT-176.

Stoeckel, D. M., Mathes, M. V., Hyer, K. E., Hagedorn,C., Kater, H., Lukasik, J., et al. (2004). Comparison

of seven protocols to identify fecal contaminationsources using Escherichia coli. Environmental Science& Technology, 38, 6109–6117. doi:10.1021/es0354519.

Tsihrintzis, V. A., & Hamid, R. (1997). Modeling and man-agement of urban stormwater runoff quality: A re-view. Water Resources Management, 11, 137–164.

U.S. Environmental Protection Agency (1986). Ambientwater quality criteria for bacteria, 1986 (p. 18). U.S.Environmental Protection Agency EPA 440/5-84-002.

U.S. Environmental Protection Agency (2000a). Ambi-ent water quality criteria recommendations: Rivers andstreams in nutrient ecoregion II (p. 39). U.S. Environ-mental Protection Agency EPA-822-B-00-015.

U.S. Environmental Protection Agency (2000b). Nutrientcriteria technical guidance manual, rivers and streams(p. 253). U.S. Environmental Protection Agency EPA-822-B00-002.

U.S. Geological Survey (2011). National f ield manual forthe collection of water-quality data. U.S. Geologi-cal Survey Techniques of Water-Resources Investiga-tions Book 9, chapters A1–A9. Available online athttp://pubs.water.usgs.gov/twri9A.

Wetzel, K. F. (2003). Runoff production processes insmall alpine catchments within the unconsolidatedPleistocene sediments of the Lainbach area (UpperBavaria). Hydrologic Processes, 17, 2463–2483.

Wilcox, R. R. (2005). Introduction to robust estimation andhypothesis testing (p. 608). Elsevier.

Zaugg, S. D., Smith, S. G., & Schroeder, M. P. (2006). De-termination of wastewater compounds in whole waterby continuous liquid–liquid extraction and capillary-column gas chromatography/mass spectrometry. U.S.Geological Survey Techniques and Methods Book 5,Chapter 4. Available online at http://pubs.usgs.gov/tm/2006/05B04.