Nitrogen Removal by Riparian Buffers along a European Climatic Gradient: Patterns and Factors of Variation

Page 1

O R I G I N A L A R T I C L E S

Nitrogen Removal by RiparianBuffers along a European ClimaticGradient: Patterns and Factors of

VariationSergi Sabater,1* Andrea Butturini,1 Jean-Christophe Clement,2 Tim Burt,3

David Dowrick,3 Mariet Hefting,4 Veronique Maıtre,5 Gilles Pinay,3

Carmen Postolache,6 Marek Rzepecki,7 and Francesc Sabater1

1Department of Ecology, Faculty of Biology, Avgda. Diagonal 645, 08028 Barcelona, Spain; 2UMR 6553 ECOBIO, University of Rennes I,Avenue du General Leclerc, F-35042 Rennes, France; 3Department of Geography, University of Durham, Durham DH1 3LE, England;

4Department of Geobotany, Utrecht University, Wentgebouw, Sorbonnelaan 16, 3584 CA Utrecht, The Netherlands; 5Laboratory ofGeology (GEOLEP), Civil Engineering Department, Swiss Federal Institute of Technology, 1015 Lausanne, Switzerland; 6Department of

Systems Ecology and Management of Natural Capital, University of Bucharest, 91–95 Splaine Independentei Avenue, 5 Bucharest,Romania; and 7Institute of Ecology PAN, Konopnickiedj 1, Dziekanow Lesny, 05-092 Lomianki, Poland

ABSTRACTWe evaluated nitrogen (N) removal efficiency by ri-parian buffers at 14 sites scattered throughout sevenEuropean countries subject to a wide range of climaticconditions. The sites also had a wide range of nitrateinputs, soil characteristics, and vegetation types. Dis-solved forms of N in groundwater and associated hy-drological parameters were measured at all sites; thesedata were used to calculate nitrate removal by theriparian buffers. Nitrate removal rates (expressed asthe difference between the input and output nitrateconcentration in relation to the width of the riparianzone) were mainly positive, ranging from 5% m�1 to30% m�1, except for a few sites where the valueswere close to zero. Average N removal rates weresimilar for herbaceous (4.43% m�1) and forested(4.21% m�1) sites. Nitrogen removal efficiency wasnot affected by climatic variation between sites, andno significant seasonal pattern was detected. When

nitrate inputs were low, a very large range of nitrateremoval efficiencies was found both in the forestedand in the nonforested sites. However, sites receivingnitrate inputs above 5 mg N L�1 showed an exponen-tial negative decay of nitrate removal efficiency (ni-trate removal efficiency � 33.6 e�0.11 NO3 input, r2 �0.33, P � 0.001). Hydraulic gradient was also nega-tively related to nitrate removal (r � �0.27, P � 0.05)at these sites. On the basis of this intersite comparison,we conclude that the removal of nitrate by biologicalmechanisms (for example, denitrification, plant up-take) in the riparian areas is related more closely tonitrate load and hydraulic gradient than to climaticparameters.

Key words: nitrate removal; riparian buffers; veg-etation; denitrification; temperature; climatic con-ditions.

INTRODUCTION

The importance of riparian zones in mitigating dif-fuse nitrate fluxes from adjacent upland areas has

been recognized for nearly 2 decades (Peterjohnand Correll 1984; Lowrance and others 1984).Since then, it has been stressed that riparian zonesare spatially and temporally heterogeneous withrespect to hydrology (Lowrance and others 1997),soil characteristics (Jacinthe and others 1998), andbiogeochemical processes (Lowrance and others

Received 15 August 2001; accepted 2 May 2002.*Corresponding author; email: [email protected]

Ecosystems (2003) 6: 20–30DOI: 10.1007/s10021-002-0183-8 ECOSYSTEMS

© 2003 Springer-Verlag

20

Page 2

1997; Hedin and others 1998; Jacinthe and others1998; Hill and others 2000). This variability affectsthe rate of nitrate removal in the riparian zonebecause the major pathway for nitrate movement isthrough subsurface flow (Hill 1996). Thus, the re-moval capacity of riparian zones is controlled bothby their hydrological characteristics, that is, waterresidence time and degree of contact between soiland groundwater (Gold and others 1998), and bytheir biological processes, that is, plant uptake anddenitrification (Groffman and others 1992; Haycockand Burt 1993; Haycock and others 1993; Groffmanand others 1996b). The relative influence of thesefactors is known to vary depending on soil charac-teristics (Groffman and others 1992, 1996a; Fliteand others 2001) and the nitrogen (N) input to theriparian zone (Hanson and others 1994a). More-over, vegetation type (that is, forest or meadow)seems to also affect the nitrate removal efficiency ofriparian zones (Groffman and others 1996a). How-ever, these conclusions remain controversial be-cause some studies have shown that meadow sitescould not remove nitrate (Osborne and Kovacic1993; Hubbard and Lowrance 1997), whereas oth-ers have demonstrated just the opposite (Haycockand Pinay 1993; Schnabel and others 1996; Addyand others 1999).

Several environmental protection agencies havedeveloped diffuse pollution control strategies, in-cluding the use of riparian zones (Haycock andothers 1997; Hession and others 2000). However,we still do not know whether it is possible to useriparian zones to control diffuse pollution across awide range of climatic conditions. The question re-mains: Is there a climatic constraint on the effi-ciency of riparian zones, or do local factors imposea stronger control on their removal capacities, re-gardless of the prevailing climatic conditions?

In this paper, we investigate variations in the Nremoval capacity of 14 riparian zones located acrossa wide range of climatic, hydrological, land-cover,and diffuse N input flux conditions typical withinthe European continent. This study was part of ajoint European research project that was designedto evaluate the natural performance of riparianzones in buffering waterborne fluxes of diffuse Npollution from agricultural to aquatic environ-ments. The study sites ranged from the Mediterra-nean to central and northern Europe. We hypoth-esized that the removal capacity of riparian zoneswould be more efficient under milder (maritime)conditions, where both processes—that is, denitri-fication and plant uptake—could occur more or lessall year round. Specifically, our aim was to deter-mine if:

1. Riparian zones could effectively remove N in-put under a wide range of climatic conditions;

2. There was a climatic gradient of nitrate removalefficiency;

3. We could measure any seasonal effect underthe different climatic conditions;

4. The type of riparian vegetation cover influ-enced the N removal capacity of the riparianzone; and

5. N removal efficiency of riparian zones corre-lated with N load.

SITE DESCRIPTION

Dissolved forms of N in groundwater and associatedhydrological parameters were studied for 1 year(1998–99) (Table 1) at 14 riparian sites throughoutEurope (Figure 1) covering a range of climatic andsoil conditions. Among the sites characterized by acontinental climate, the Polish and Romanian oneswere located, respectively, over glacial and alluvialsoils. The Swiss site, with a typical alpine climate,was also located over a glacial soil type. The Dutch,British, and French sites were characterized by ahumid maritime climate. Soils at these sites were,respectively, glacial, alluvial, and schistose, all or-ganic-rich. Finally, the Spanish site was typicallyMediterranean, its soil being schistose and organic-poor. As a whole, the sites covered from 41° to 53°(N) latitude. Mean annual temperature rangedfrom 7.8 to 12°C, and potential evapotranspiration(PET) ranged from very low (Swiss site) to ex-tremely high (Spanish site).

In all countries, a forested riparian strip was stud-ied; in some countries, a herbaceous site was stud-ied as well. Six of the study sites were covered byherbaceous or shrubby vegetation; the other eightsites were forested. Upland agricultural fields (usu-ally cereals or pasture) bordered these areas. Theriparian zones had a wide range of widths (from 10to 60 m), slopes (from 0% to 22%) and soil char-acteristics. Soils ranged from organic-rich to organ-ic-poor and from sandy to clayey. There was signif-icant variability in soil water content andsaturation, both in terms of water table heightwithin the topsoil and the extent of time that itremained saturated (Table 1).

MATERIALS AND METHODS

Hydrological parameters (hydraulic conductivity,hydrological gradient, and water flux) and chemicalvariables (chloride, nitrate, and ammonia inputs toand outputs from the riparian zone) were measured

Nitrogen Removal in European Riparian Buffers 21

Page 3

monthly in each riparian zone. Monthly measure-ments of temperature, organic matter content, wa-ter, and sand percentages in the soil were used tocharacterize the sites.

At each study site, groundwater levels were mon-itored across the riparian zone using a grid of pi-ezometers. These were placed in four rows, eachhaving from four to eight wells. The riparian plotshad different sizes (Table 2) and dipwell spacing(usually between 3 and 10 m). The piezometerswere made from PVC pipes (internal diameter 2–12cm) installed to depths ranging from 1 to 7 m; thescreened ends ranged from 20 cm to 1 m in length.The monitoring of temporal dynamics and spatialheterogeneity of water table elevation allowed us toestablish (a) the hydraulic gradient and groundwa-ter flow direction, and (b) the influence of streamwater on the riparian zone water table (Burt andothers 2002).

Saturated hydraulic conductivity (K) was mea-sured in piezometers at high water level using bail

tests at the Spanish and French sites and by pump-ing tests at the other study plots. Conductivity val-ues ranged widely across sites. Groundwater fluxacross a plane of a square meter was measuredusing Darcy’s formula:

Q � K Hgrad (1)

where Q is the specific water flux (m3 m�2 h�1), Kis the saturated hydraulic conductivity (m h�1), andHgrad is the hydraulic gradient (m m�1) calculatedas the difference in water level between piezom-eters located on the upslope and downslope bound-ary of the riparian zone.

Groundwater samples for water chemistry werecollected at least monthly between October 1998and December 1999. Prior to each sampling, thestanding water within the piezometers was re-moved. Water samples for analyses were filteredwith Whatman GF/F glass filters in the laboratoryand analyzed within 24–36 h for NO3

�-N, NH4�-N,

Table 1. Location and Physical Characteristics of the Studied Sites

Studied Site Latitude LongitudeHeight(m a.s.l.)

MeanTemp.(°C)

AnnualRainfall(mm y�1)

PET(mm y�1)

WaterSaturatedSoil(days)

Poland Forested 1 53°45�N 21°25�E 206 7.8 590 520 365

Poland Forested 2 53°45�N 21°25�E 206 7.8 590 520 365

United Kingdom Forested 54°42�N 1°23�W 110 8.5 642 537 180

United Kingdom Herb. 1 54°42�N 1°23�W 110 8.5 642 537 220United Kingdom Herb. 2 54°42�N 1°23�W 110 8.5 642 537 365

The Netherlands Forested 51°25�N 6°51�E 50 8.9 710 490 180The Netherlands Herb. 51°25�N 6°51�E 50 8.9 710 490 220

Switzerland Forested 43°36�N 6°24�E 660 9.6 1064 460 365Romania Forested 44°27�N 25°16�E 134 10.7 545 708 215

Romania Herb. 44°27�N 25°16�E 134 10.7 545 708 215

France Forested 48°3�N 2°3�E 20 11.6 850 635 228France Herb. 1 48°3�N 2°3�E 20 11.6 850 635 228

France Herb. 2 48°3�N 2°3�E 20 11.6 850 635 228

Spain Forested 41°42�N 2°34�E 550 12 550 1100 4

PET, Potential evapotranspiration; % Sand-DW, sand proportion with respect to soil dry weight; % OC-DW, organic carbon proportion with respect to soil dry weight.Sites are arranged according to the mean annual temperature. Climatological, soil, and vegetation features are also indicated.Width and slope are given as the average for every riparian zone.

22 S. Sabater and others

Page 4

and chloride. Nitrate and chloride were analyzedusing a Technicon Autoanalyzer (Technicon, NewYork, 1977) and/or the capillary electrophoresistechnique (Waters, Milford, CIA-Quanta 5000, Ro-mano and Krol 1993). Ammonia was analyzed us-ing the indophenol-blue method (Solorzano 1969).

The ability of the riparian strip to retain inorganicnitrogen (NO3

�-N and NH4�-N) entering from the

agricultural field was estimated using a simple in-put–output approach with the following formula:

�NO3 � (Ninput � Noutput) Ninput�1 �h�1 100 [m�1]

(2)

where �NO3 is the percent nitrate removed perlinear meter of riparian area, (Ninput � Noutput) isthe difference between the nitrate concentrationentering (Ninput) and leaving (Noutput) the riparianzone, and �h�1 is the linear distance between theinput and output piezometers. The input nitrate(Ninput) was obtained from the piezometers locatednear the agricultural field; the output nitrate wasobtained from piezometers located in the riparian

zones. If �NO3 is greater than zero, the nitrate re-moval is effective; if �NO3 is less than zero, nitraterelease predominates over depletion. Groundwatermonitoring allowed us to detect a dynamic zone ofdilution between groundwater and stream water atseveral sites (the French sites, the Spanish forestedsites, and the Dutch sites). To avoid any effect ofmixing between groundwater and stream water,output nitrate data were obtained from piezometerslocated in the middle of the riparian zone, beyondthe stream’s influence. �NO3 was not estimatedwhen the soil was frozen (Poland in winter) orwhen stream water was moving across the entireriparian strip (French sites in early autumn) (seeBurt and others 2002).

Chloride was used as a tracer to detect waterdilution between the input and output piezom-eters (Altman and Parizek 1995). It was assumedthat dilution from groundwater or river waterwas irrelevant when chloride concentrations atthe input and output were similar. To determinewhether or not this situation existed, chloridedata from piezometers at the output point were

Table 1. Continued

Soil Type % H2O

%Sand-

DW

%OC-DW

Width(m)

Slope(%)

Dominant PlantSpecies

sandy clay/peat 29.7 60.3 3.2 14 17–20 Urtica dioica, Alnusglutinosa

sandy clay/peat 29.7 60.3 3.2 30 17–20 Urtica dioica, Alnusglutinosa

brown soil 27.8 43.5 7.4 50 1.7–5.5 Acer pseudoplatanus,Fagus sylvatica

brown soil 32.5 31.7 5.8 40 2.2–4.4 Lolium, Poa, Trifoliumbrown soil 32.5 31.7 5.8 60 2.2–4.4 Glyceria, Myosotis,

Nasturtiumsand/peat 43.2 63.5 10.4 10–20 7.5–18 Alnus glutinosasand/peat 45.2 67.5 8.8 20 8.7–20 Glyceria maxima, Urtica

dioicareductisoils 67.4 11.25 17.1 20 0.8 Alnus glutinosa, Salix sp.chromic luvisols 21.0 43.0 14.0 20 6.3 Cornus x sanguinea,

Populus nigrachromic luvisols 20.0 32.0 10.0 10.5 3.0 Lolium, Trifolium,

Taraxacumorganic/clay 41 26.7 6.3 15 2.2 Salix sp., Quercus sp.organic/clay 47.7 23.8 4.7 15 2.2 Salix sp., Rubus

fruticosusorganic/clay 47.7 23.8 4.7 15 2.2 Holcus lanatus, Dactylis

glomeratasandy 11.4 55.2 2.8 20 22.0 Alnus glutinosa, Platanus

hybrida

Nitrogen Removal in European Riparian Buffers 23

Page 5

regressed against those at the input location.When the slope of the expression was not greatlydifferent than 1 (that is, a 1:1 relationship be-tween the input and the output), it was assumedthat dilution was not affecting the N dynamics.The difference between the theoretical and ob-served expressions was tested statistically bymeans of a two-tailed t-test (Fowler and Cohen1990). On the other hand, when changes in chlo-ride concentrations between the input and theoutput were significant, it was assumed that adilution factor must be included. We estimatedthe expected nitrate concentration for the outputpiezometer (NO3 O-exp.) when uptake was zerowith the following formula:

NO3 O-exp. � NO3 I ClO ClI�1 (3)

where NO3 I and ClI is nitrate and chloride concen-tration in the input piezometers and ClO is chlorideconcentration in the output piezometers. We thenreplaced Ninput in Eq. (2) with NO3 O-exp. Chloridedata from the Romanian sites were not available,and it was assumed that dilution was not occurringat those sites.

The Pearson product-moment correlation wasused to detect possible relationships between theanalyzed variables. Variables were transformedprior to analysis using the most suitable transfor-mation to normalize the distribution.

RESULTS

Hydrological characteristics were significantly dif-ferent among the various sites. Hydraulic conduc-tivities ranged from 0.002 to 1.8 m h�1. The highestK values (more than 1 m h�1) were recorded at theNetherlands herbaceous site and the Swiss and UKforested sites (Table 2). Hydraulic gradients rangedfrom 0.0001 (UK herbaceous 2) to 0.158 (Spainforested). As a result, specific water fluxes in theriparian areas ranged from 0.00008 m3 m�2 h�1 to0.069 m3 m�2 h�1. The lowest values were mea-sured in the Spanish forested, the Romanian her-baceous, and the UK herbaceous sites. The highestfluxes were found at the English forested site.

Dilution by river or groundwater was low tomoderate at eight of the sites (Table 3), includingthe two herbaceous and the forested French sites,the Spanish and Swiss forested sites, the two her-baceous English sites, and the herbaceous Dutchsite. Dilution was assumed to be low at the twoRomanian sites because fluxes were very small (Ta-ble 2) and because there was covariation in ground-water levels in input and output wells (Burt andothers 2002). Dilution was relevant and affected thenitrate dynamics of the remaining four sites. It oc-curred regularly in the Dutch forested site, whichwas higher in spring and summer, but it did notfollow any temporal pattern at the other three sites.

Nitrate inputs in the riparian zones ranged from0.12 mg L�1 to 35 mg L�1 N-NO3 (Table 2);whereas ammonia inputs ranged from 0.04 to 0.85mg L�1 N-NH4. There were no significant differ-ences between herbaceous and forested sites for Ninputs or hydrological characteristics, but signifi-cant differences existed with respect to the waterflowing through the riparian zone (higher in theforested sites) (ANOVA, F � 7.9, P � 0.0056, n �160).

Nitrate was the dominant form of N entering theriparian zone at most sites (Table 2). Ammoniainput was higher than nitrate only at the Englishherbaceous 2, the Polish (forested), and the Roma-nian (forested and herbaceous) sites. In absoluteterms, concentrations and fluxes of ammonia werealways low (Table 2). When it did occur, ammoniainput was in general higher in the nonforested thanin the forested sites (F � 3.44, P � 0.065, N �160).

Nitrate removal in the riparian zone was positive inmost of the studied sites (Figure 2). Annual nitrateremoval efficiency was close to 30% m�1 at theFrench forested site; whereas at four other sites (theherbaceous Romanian, Dutch, and one of the Frenchsites, and the forested Spanish site), the efficiency was

Figure 1. Location of the sites included in this study inEurope. The approximate locations of the site are identi-fied by the signs ● (forested) and * (herbaceous).

24 S. Sabater and others

Page 6

Tab

le2.

Hyd

rolo

gica

l,G

eom

orp

holo

gica

l,an

dC

hem

ical

Par

amet

ers

(In

pu

tan

dO

utp

ut

of

the

Rip

aria

nZon

e)of

the

Stu

dy

Sit

es

Hyd

r.C

on

d.

(m/h

)G

rad

ien

t(m

/m)

Sp

ecifi

cW

ater

Flu

x(m

3/m

2h

)

N-N

O3

Inp

ut

N-N

H4

Inp

ut

N-N

O3

Inp

ut

Flu

x(g

h�

1)

N-N

H4

Inp

ut

Flu

x(g

h�

1)

Wel

lS

pac

ing

(m)

pp

m�

SD

pp

m�

SD

Pola

nd

Fore

sted

10.0

918

0.0

230

0.0

021

0.0

20.0

20.8

51.0

50

.00

00

38

0.0

01

80

47

.0Pola

nd

Fore

sted

20.0

734

0.0

036

0.0

003

0.0

20.0

20.4

50.5

70

.00

00

06

0.0

00

11

83

0.0

Rom

ania

Fore

sted

0.3

400

0.0

003

0.0

001

0.1

20.1

00.3

50.4

50

.00

00

11

0.0

00

03

15

.5R

om

ania

Her

bace

ou

s0.0

030

0.0

550

0.0

002

0.1

60.3

70.2

00.1

20

.00

00

26

0.0

00

03

22

.4U

nit

edK

ingd

om

Fore

sted

1.0

100

0.0

690

0.0

697

0.5

10.7

00.5

21.3

60

.03

55

42

0.0

36

23

91

0.0

Un

ited

Kin

gdom

Her

b.1

0.1

930

0.0

480

0.0

093

0.3

50.6

80.4

01.6

40

.00

32

42

0.0

03

70

61

0.0

Sw

itze

rlan

dFore

sted

1.8

000

0.0

05

0.0

09

1.1

40.7

80.0

40.0

70

.10

30

00

0.0

04

00

01

6.0

Un

ited

Kin

gdom

Her

b.2

0.0

038

0.0

001

0.0

001

1.6

91.0

40.6

80.6

10

.00

02

00

0.0

00

10

01

0.0

Fra

nce

Fore

sted

0.0

980

0.0

085

0.0

008

3.3

54.2

00.4

00.6

10

.00

28

02

0.0

00

33

52

.9Th

eN

eth

erla

nds

Her

bace

ou

s0.0

208

0.1

700

0.0

035

8.5

48.2

00.2

10.2

11

.48

19

97

0.0

35

61

39

.5Spai

nFore

sted

0.0

005

0.1

580

0.0

0008

10.7

03.7

00.1

00.3

00

.00

08

0.0

00

00

87

.9Fra

nce

Her

b.1

0.0

200

0.0

140

0.0

003

11.6

03.9

50.3

80.4

00

.00

34

80

0.0

00

11

46

.7Fra

nce

Her

b.2

0.0

150

0.0

520

0.0

008

12.3

54.7

00.4

30.9

00

.00

96

00

0.0

00

34

46

.9Th

eN

eth

erla

nds

Fore

sted

0.0

208

0.0

980

0.0

020

35.0

010.4

00.1

60.1

20

.07

13

44

0.0

00

32

01

2.6

Site

sar

ear

ran

ged

acco

rdin

gto

the

incr

easi

ng

N-N

O3

inpu

tco

nce

ntr

atio

n.

Nitrogen Removal in European Riparian Buffers 25

Page 7

about 10% m�1. The nitrate removal efficiency de-creased to around 5% m�1 for four other sites (thetwo English herbaceous sites, one herbaceous Frenchsite, and the Swiss site). For the remaining sites, ni-trate removal efficiency was close to zero (the Dutchforested site, the two forested Polish sites, the forestedBritish, and the Romanian sites).

At the scale of the intersite comparison, vegetationtype was not a dominant factor explaining variationin nitrate or total dissolved inorganic nitrogen (DIN)removal rates (ANOVA; P � 0.1475 and P � 0.8311,respectively, n � 44). DIN removal averaged 4.43%m�1 for herbaceous and 4.21% m�1 for forested sites.However, at the French, Dutch, and Romanian sites,nitrate removal was significantly different betweenvegetation types (Table 4). However, no clear patternemerged because in some cases, forested riparianzones had higher removal efficiencies (French for-ested site); whereas in other cases, herbaceous ripar-ian zones were more retentive (the Dutch and Roma-nian herbaceous sites).

Variations in nitrate removal did not follow a sea-sonal pattern at any study site (Figure 2). This wasconfirmed by the lack of relationship with tempera-ture for most sites (Table 5). The Swiss site showed an

Table 3. Chloride Concentration (Average and Standard Deviation) at the Input and Output of theRiparian Areas

Chloride input Chloride output

r2 b P nppm �SD ppm �SD

Poland Forested 1 15.4 1.4 15.1 2.9 0.57 1.74 0.24 4Poland Forested 2 11.5 3 21.8 7.9 0.53 0.28 0.061 7Romania Forested na naRomania Herbaceous na naUnited Kingdom Forested 60.5 28.7 36.5 11.1 0.86 1.77 0.05 8United Kingdom Herb. 1 29.15 9 27.93 7.15 0.97 1.24* 0.0001 10Switzerland Forested 11.7 2.9 10.9 2.9 0.98 1.01* 0.0001 11United Kingdom Herb. 2 56.7 10.8 59.75 13.1 0.63 0.66* 0.01 9France Forested 63 20.25 60.41 15.37 0.87 0.71* 0.0006 8The Netherlands Herbaceous 22.1 4.6 31.1 1.9 0.41 0.91* 0.05 14Spain Forested 26.7 8.7 25.7 9.2 0.77 0.93* 0.0001 34France Herb. 1 46.9 3.9 47.9 5.9 0.58 1.18* 0.001 14France Herb. 2 45.2 3.9 42.9 4 0.86 0.92* 0.0001 16The Netherlands Forested 23.4 3 18.3 3.2 0.22 0.44 0.08 14

Chloride data were not available (na) for the two Romanian sites. The r2, slope (b), probability (P), and number of cases (n) of the regression line between the chloride outputand input are shown for every site. Cases when slope (b) did not differ significantly from 1 (two-tailed t-test, indicating absence of dilution) are indicated by an asterisk (*).

Figure 2. Mean seasonal nitrate removal rates and stan-dard deviation of the means for the different study sites.Acronyms are F � France, NL � The Netherlands, E �England (United Kingdom), SP � Spain, Sw � Switzer-land, R � Romania, P � Poland. Bar shadings from left toright: black is winter (January–March), gray is spring(April–June), light gray is summer (July–September),white is fall (October–December). Lower-case f and hdefine the respective forested and herbaceous study sitefor each country. In cases where more than one study siteoccurs for each category, they are defined with numbers.

Table 4. ANOVA (Tukey HSD) on theSignificant Differences between Nitrate RemovalEfficiencies for the Different Countries WhereForested and Herbaceous Riparian Zones WereStudied

Country P (Forested vs Herbaceous)

France 0.00023Netherlands 0.000588Romania 0.000352United Kingdom 0.41196

26 S. Sabater and others

Page 8

increase in nitrate removal from winter to autumn(correlation between temperature and nitrate re-moval of 0.71, P � 0.05). In some sites, seasonalchanges in nitrate removal were dramatic (for exam-ple, the Romanian herbaceous site), but variabilitywas too high to indicate any reliable seasonal pattern.When all the sites were considered, season did notaccount for significant differences in nitrate or DINremoval between sites (ANOVA; P � 0.978 and P �0.281, respectively, n � 42).

At the site scale, there were few significant relation-ships between nitrate removal rates and nitrate input(concentration or fluxes). We found a positive corre-lation between nitrate removal rates and N inputs attwo of the study sites (the Spanish and Swiss sites)(Table 5). At these sites, nitrate inputs increased fol-lowing fertilizer application in the upland catchment,and nitrate removal efficiency increased as well.

When all sites and periods were considered to-gether, nitrate removal rates were unrelated to ni-trate inputs (Figure 3a; correlation coefficient �0.04, n � 160, P 0.05), even though nitrateremoval rates were negatively correlated with ni-trate fluxes (Figure 3b, r � �0.24, n � 160, P �0.05). A more definite relationship between nitrateremoval and nitrate input concentration was ob-served when only sites receiving more than 1 mgL�1 were considered. At the forested sites, wherethe range of nitrate inputs was much higher (1–43mg L�1 NO3), nitrate removal efficiency was nega-tively correlated to input (r � 0.36, P � 0.05). Forthe herbaceous sites, with nitrate inputs rangingfrom 1 to 20 mg l�1, there was a positive relation-

ship between nitrate input and nitrate removal ef-ficiency (r � 0.51, n � 45, P � 0.05). However,forested and herbaceous sites receiving more than 5ppm showed a significant negative relationship be-tween nitrate removal and nitrate input concentra-tion (r � �0.59, P � 0.05) (Figure 4).

Correlation analysis did not show any significantrelationship between nitrate or DIN removal andthe hydrological parameters of each site, except atthe French herbaceous site, where nitrate removalefficiency was positively related to hydraulic con-ductivity and hydraulic gradient (Table 5). Hydrau-lic gradient was the only hydrological parameterthat was significantly related to nitrate removal forsites receiving more than 5 ppm N-NO3 (Table 5).

DISCUSSION

Climatic Range of Efficiency of RiparianBuffer

This comparison was designed to evaluate whetherriparian zones can effectively buffer N input from

Table 5. Correlation Coefficient (Pearson)between Nitrate Removal Efficiencies and otherDescriptors of the Riparian Zones WhenCorrelations Were Significant (P � 0.05)

R n

Nitrate removal Francex hydraulic conductivity 0.69 24x hydraulic gradient 0.69 24x water flux 0.69 24Nitrate removal Spain x nitrate

input 0.66 32Nitrate removal Switzerlandx temperature 0.71 11x DIN input 0.61 11Nitrate removal x hydraulic gradient

at herbaceous and forested siteshigher than 5 ppm �0.27 69

DIN, dissolved inorganic nitrogen

Figure 3. Relationship between nitrate removal and anitrate input and b nitrate flux for nonforested (blackcircles) versus forested (open circles).

Nitrogen Removal in European Riparian Buffers 27

Page 9

upland catchments under a wide range of climaticconditions from Mediterranean (Catalonia, Spain)to temperate (Brittany, France, and north England),and from maritime (The Netherlands and Switzer-land) to continental (Masuria in Poland and Vala-chia in Romania). Nitrate is removed in riparianzones by the microbial process of denitrification andby plant uptake (Peterjohn and Correll 1984; Hill1996), both of which ought to be closely linked toclimatic conditions. Denitrification can account for50%–90% of the total nitrate elimination undertemperate climates where soils are usually water-saturated (Haycock and Burt 1993; Nelson and oth-ers 1995). Denitrification at other sites may be con-strained by low soil temperature (continentalclimates of central Europe) or low soil moisture(Mediterranean climate). However, our resultsshowed that in most cases there was significant Nremoval (Figure 2), with rates ranging from 5%m�1 to 30% m�1, and we did not find any signifi-cant climatic pattern. The large variation in nitrateremoval efficiency between the riparian study areas(Figure 2) was not unexpected, given the high het-erogeneity within individual riparian buffers (Hilland others 2000). The removal capacity of riparianzones is controlled by several factors related to hy-drology, geomorphology, soil type, and biologicalprocesses; this results in a large range of efficiencies(Lowrance and others 1997). For instance, riparianzones with water tables close to the soil surface aremore effective at removing N compounds thanother sites where water tables are lower (Cooper

1990; Haycock and Pinay 1993; Jordan and others1993; Starr and Gillham 1993). The lack of relation-ship between climate and removal capacity under-lines the control that the geomorphological andhydrological conditions impose on the N removalcapacity of riparian zones.

Seasonality of N Removal

No significant seasonal pattern in N removal effi-ciency could be detected at the different sites (Fig-ure 2). The two main processes involved in N re-moval (denitrification and plant uptake) canoperate under different hydrological and tempera-ture conditions, and other sources of variation maymask any seasonal pattern in each of these pro-cesses (Pinay and Decamps 1988; Nelson and others1995). For instance, at the French and the Dutchsites, which have mild climatic conditions, therewas no significant difference between seasonswithin each site, despite significant differences be-tween sites. Likewise, high N removal rates weremeasured at the Spanish site, even in summerwhen the riparian soils were dry. This finding indi-cates that plant uptake was most probably the mainremoval process at that site during dry soil condi-tions. This apparent lack of seasonality reinforcesthe fact that within the range of climatic conditionstested, riparian zones can significantly remove Ninput. The two removal processes (denitrificationand plant uptake) operate either simultaneously orin isolation depending on the hydrological and tem-perature conditions.

Does Riparian Vegetation Type Matter?

Significant differences in N removal capacity weremeasured between forested and meadow riparianzones at several sites. Yet there was no consistentdifference between regions. For instance, withinthe French site, the N removal rate of the forestedzone was significantly higher than the meadowzone; whereas in the Dutch and the Romanianregions, the meadow riparian zones were signifi-cantly more efficient (Table 4). Therefore, no clearpattern could be detected when we compared theaverage removal rates of herbaceous (4.43% m�1)and forested (4.21% m�1) zones in the differentregions. These results only add to the conflictingresults already described in the literature concern-ing the respective efficiency of herbaceous and for-ested riparian zones (see for example, Haycock andPinay 1993; Osborne and Kovacic 1993; Correll1997).

It has been suggested that variation in denitrifi-cation for different herbaceous and forested ripar-

Figure 4. Relationship between nitrate removal and ni-trate input when nitrate concentrations were above 5ppm at the groundwater input in the forested and her-baceous sites. Black symbols � herbaceous sites, opensymbols � forested sites; triangles � French sites,squares � Dutch sites, rhombus � Spanish sites.

28 S. Sabater and others

Page 10

ian wetlands is fundamentally related to the organicmatter quality of the site (Groffman and others1991). However, Parry and others (1999) demon-strated that a difference in denitrification rate be-tween a pasture and a cultivated field was insteadrelated to differences in the pore space structureresulting from agricultural practices. Our resultssuggest that both types of vegetation cover canprovide similar N removal rates, and that bothcould be used to mitigate diffuse N pollution in thewide range of climatic conditions tested.

N Saturation Effect

While vegetation type and climatic factors did notpresent any consistent pattern of variation betweenthe sites, nitrate load and hydrological factors (hy-draulic gradient) were identified as the two mainfactors controlling variation between the riparianzones in nitrate removal rates. These factors werenot significant under low nitrate inputs becausenitrate removal efficiencies ranged widely for bothforested and nonforested sites (Figure 3 a). Underhigher nitrate concentration inputs (more than 5mg N/L; that is, approaching the legal limit fornitrate in fresh water within the European Union),nitrate removal efficiency was negatively correlatedwith nitrate input (r � �0.59, P � 0.05). Thisrelationship followed a pattern of negative expo-nential decay (Figure 4). Hydraulic gradient wasalso negatively correlated with nitrate removal (r ��0.27, P � 0.05) at sites receiving water withconcentrations above 5 mg/L nitrate-N. This impliesthe importance of prolonged saturation within theriparian zone for denitrification to be effective.

Lowrance (1996) stated that high rates of nitrateremoval were possible under higher loadings ofnitrate when denitrification was the primary mech-anism of nitrate removal. Indeed, denitrificationrates in wetland areas increase dramatically whenexposed to nitrate (Warwick and Hill 1988; Bengts-son and Bergwall 1995). However, the upper limitof denitrification rates has not been yet established.For instance, Hanson and others (1994b) observedclear symptoms of N saturation in a forested ripar-ian zone subjected to long-term enrichment. Thesesymptoms consisted of enrichment of total plantand microbial N pools as well as an increase in soilN process rates—namely, mineralization and nitri-fication. The negative relationship between nitrateload and riparian zone removal efficiency found inour study sites also suggests a saturation effect oflong-term nitrate loading, which exceeds the buff-ering capacity (Aber 1992) of the riparian zones. Anextreme case for this expression, the Dutch forestedsite, showed a flat response in nitrate removal effi-

ciency when the nitrate input to the riparian zoneincreased (Figure 4). The Dutch forested site is lo-cated in an area of long-lasting N enrichment (Table2), and soil processes (decomposition rates andmineralization) were significantly higher there thanat all the other sites included in this study (Pinayand Burt 2000).

CONCLUSION

Within the range of climatic conditions testedwithin Europe, riparian zones can effectively re-move nitrate inputs from uplands. Riparian zoneprotection and restoration can therefore be pro-posed as a tool to mitigate diffuse nitrate pollutionin Europe. However, because nitrate load and hy-draulic gradient are stronger predictors than cli-matic conditions or vegetation cover, close atten-tion should be given to geomorphologicalconditions within the riparian zone in order to as-sess their potential effectiveness for nitrate removal.

ACKNOWLEDGMENTS

This study arose from the European program Nitro-gen Control by Landscape Structures in AgriculturalEnvironments (NICOLAS), which was supported bya European Grant from the Environment and Cli-mate Program (ENV4-CT-97-0395) (Dr. H. Barth,EU Scientific Advisor). We appreciate the insightfulcomments of Dr. N. Bond (Monash University) onthe manuscript.

REFERENCES

Aber JD. 1992. Nitrogen cycling and nitrogen saturation in tem-perate forest ecosystems. Trends Ecol Evol 7:220–4.

Addy KL, Gold AJ, Groffman PM, Jacinthe PA. 1999. Groundwater nitrate removal in subsoil of forested and mowed ripar-ian buffer zones. J Environ Qual 28:962–70.

Altman SJ, Parizek RM. 1995. Dilution of nonpoint-source ni-trate in groundwater. J Environ Qual 24:707–18.

Bengtsson G, Bergwall C. 1995. Heterotrophic denitrificationpotential as an adaptive response in groundwater bacteria.FEMS Microbiol Ecol 16:307–17.

Burt TP, Pinay G, Matheson FE, Haycock NE, Butturini A, Clem-ent JC, Danielescu S, Dowrick DJ, Hefting MM, Hillbricht-Ilkowska A, and others. 2002. Water table fluctuations in theriparian zone: comparative results from a pan-European ex-periment. J Hydrol. 265:129–148.

Cooper AB. 1990. Nitrate depletion in the riparian zone andstream channel of a small headwater catchment. Hydrobiolo-gia 202:13–26.

Correll DL. 1997. Buffer zones and water quality protection:general principles. In: Haycock NE, Burt TP, Goulding KWT,Pinay G, editors. Buffer zones: their processes and potential inwater protection. Harpenden (UK): Quest Environmental. p7–20.

Flite OP, Shannon RD, Schnabel RR, Parizek RR. 2001. Nitrate

Nitrogen Removal in European Riparian Buffers 29

Page 11

removal in a riparian wetland of the Appalachian valley andridge physiographic province. Environ Qual 30:254–61.

Fowler J, Cohen L. 1990. Practical statistics for field biology.Chichester (England): Wiley.

Gold AJ, Jacinthe PA, Groffman PM, Wright WR, Puffer RH.1998. Patchiness in groundwater nitrate removal in a riparianforest. Environ Qual 27:146–55.

Groffman PM, Axelrod EA, Lemunyon JL, Sullivan WM. 1991.Denitrification in grass and forest vegetated filter strips. JEnviron Qual 20:671–74.

Groffman PM, Gold AJ, Simmons RC. 1992. Nitrate dynamics inriparian forests: microbial studies. J Environ Qual 21:666–71.

Groffman PM, Hanson GC, Kiviat E, Stevens G. 1996a. Variationin microbial biomass and activity in four different wetlandtypes. Soil Sci Soc Am J 60:622–9.

Groffman PM, Howard G, Gold AJ, Nelson WM. 1996b. Micro-bial nitrate processing in shallow groundwater in a riparianforest. J Environ Qual 25:1309–16.

Hanson GC, Groffman PM, Gold AJ. 1994a. Denitrification inriparian wetlands receiving high and low groundwater inputs.J Environ Qual 23:917–22.

Hanson GC, Groffman PM, Gold AJ. 1994b. Symptoms of nitro-gen saturation in a riparian wetland. Ecol Appl 4:750–6.

Haycock NE, Burt TP. 1993. Role of floodplain sediments inreducing the nitrate concentration of subsurface run-off: acase study in the Cotswolds, UK. Hydrol Proc 7:287–95.

Haycock NE, Burt TP, Goulding KWT, Pinay G. 1997. Bufferzones: their processes and potential in water protection. Harp-enden (UK): Quest Environmental.

Haycock NE, Pinay G. 1993. Groundwater nitrate dynamics ingrass and plopar vegetated riparian buffer strips during thewinter. J Environ Qual 22:273–8.

Haycock NE, Pinay G, Walker C. 1993. Nitrogen retention inriver corridors: European perspective. Ambio 22:340–6.

Hedin LO, Vonfischer JC, Ostrom NE, Kennedy BP, Brown MG,Robertson GP. 1998. Thermodynamic constraints on nitrogentransformations and other biochemical processes at soil–stream interfaces. Ecology 79:684–703.

Hession WC, Johnson TE, Charles DF, Hart DD, Horwitz RJ,Kreeger DA, Pizzuto JE, Velinsky DJ, Newbold JD, CianfraniC, and others. 2000. Ecological benefits of riparian reforesta-tion in urban watersheds: study design and preliminary re-sults. Environ Monit Assess 63:211–22.

Hill AR. 1996. Nitrate removal in stream riparian zones. J Envi-ron Qual 25:743–55.

Hill AR, Devito KJ, Campagnolo S, Sanmugadas K. 2000. Sub-surface denitrification in a forest riparian zone: interactionsbetween hydrology and supplies of nitrate and organic carbon.Biogeochemistry 51:193–223.

Hubbard RK, Lowrance R. 1997. Assessment of forest manage-ment effects on nitrate removal by riparian buffer systems.Trans ASAE 40:383–91.

Jacinthe PA, Groffman PM, Gold AJ, Mosier A. 1998. Patchinessin microbial nitrogen transformations in groundwater in ariparian forest. J Environ Qual 27:156–64.

Jordan TE, Correll DL, Weller DE. 1993. Nutrient interception bya riparian forest receiving inputs from adjacent cropland. JEnviron Qual 22:467–73.

Lowrance R. 1996. The potential role of riparian forests as bufferzones. In: Haycock NE, Goulding KWT, Pinay G, editors.Buffer zones: their processes and potential in water protection.Harpenden (UK): Quest Environmental. p 128–33.

Lowrance R, Altier LS, Newbold JD, Schnabel RR, Groffman PM,Denver JM, Correll DL, Gilliam JW, Robinson JL, BrinsfieldRB, and others. 1997. Water quality functions of riparianforest buffers in Chesapeake Bay watersheds. Environ Manage21:687–712.

Lowrance RR, Todd RL, Asmussen LE. 1984. Nutrient cycling inan agricultural watershed. II. Streamflow and artificial drain-age. J Environ Qual 13:27–32.

Lowrance R, Vellidis G, Hubbard RK. 1995. Denitrification in arestored riparian forest wetland. J Environ Qual 24:808–15.

Nelson WM, Gold AJ, Groffman PM. 1995. Spatial and temporalvariation in groundwater nitrate removal in a riparian forest.J Environ Qual 24:691–9.

Osborne LL, Kovacic DA. 1993. Riparian vegetated buffer stripsin water-quality restoration and stream management. Fresh-water Biol 29:243–58.

Parry S, Renault P, Chenu C, Lewsi R. 1999. Denitrification inpasture and cropped soil clods as affected by pore space struc-ture. Soil Biol Biochem 31:493–501.

Peterjohn WT, Correll DL. 1984. Nutrient dynamics in an agri-cultural watershed: observations on the role of a riparianforest. Ecology 65:1466–75.

Pinay G, Burt TP. 2000. Nitrogen control by landscape structures.Final report. European Commission DG XII. Brussels.

Pinay G, Decamps H. 1988. The role of riparian woods in regu-lating nitrogen fluxes between the alluvial aquifer and surfacewater: a conceptual model. Regulated Rivers 2:507–16.

Romano J, Krol J. 1993. Capillary ion electrophosesis, an envi-ronmental method for the determination of anions in waters.J Chromatogr 640:403–12.

Schnabel RR, Cornish LF, Stout WL, Shaffer JA. 1996. Denitri-fication in a grassed and a wooded, valley and ridge, riparianecotone. J Environ Qual 25:1230–5.

Solorzano L. 1969. Determination of ammonia in natural watersby phenolhypochloride method. Limnol Oceanogr 14:799–801.

Starr RC, Gillham RW. 1993. Denitrification and organic carbonavailability in two aquifers. Ground Water 31:934–47.

Technichon. 1977. Nitrate and nitrite in water and seawater.Industrial methods 158–171 WIA. Tarrytown (NY): TechniconIndustrial Systems.

Warwick J, Hill AR. 1988. Nitrate depletion in the riparian zoneof a small woodland stream. Hydrobiologia 157:231–40.

30 S. Sabater and others