Abstract An experiment was designed to fur-ther the empirical understanding of the eVects ofscale on Xuxes of water and dissolved nitrogenfrom hillslopes in semi-arid shrubland. It washypothesised that the behaviour of dissolvednitrogen is related to the scale of the contributinghillslope/catchment area and dynamics of theoverland Xow as has been demonstrated to be thecase for soil erosion (Parsons et al. 2006). Datafrom four hillslope scales (ca. 21–300 m2) and onesubcatchment (ca. 1,500 m2), collected over twomonsoon seasons, support this hypothesis anddemonstrate that the key controls of average dis-solved nitrogen yields are Xow discharge and plotscale. The slope of the best-Wt line describing therelationship between Xow discharge and total
dissolved nitrogen (TDN) yields descreases withincreasing scale, from 0.0183 at 21.01 m2, 0.0092 at56.84 m2, 0.0059 at 115.94 m2, 0.0024 at 302.19 m2
to 0.0004 at 1,500 m2. An implication of theseWndings is that care must be taken when upscalingresults describing nutrient behaviour from small,plot experiments, as this behaviour appears to bescale dependent. For example, average yields ofTDN in overland Xow increase to a maximumwith increasing plot area until an area of 50 m2 isreached, and decline with increasing plot sizethereafter. Thus, studies that rely upon Wxed plotscales may misrepresent catchment- or landscape-scale Xuxes as they do not describe the changingrelationship between overland Xow and nutrientXuxes with increasing spatial scale. Further inves-tigations into intra-event behaviour illustrate thatnitrogen losses from natural rainfall/runoV eventsare supply limited as over the course of the eventsmonitored, decreasing concentrations illustrate apattern of nutrient exhaustion. When events arecompared at the same sites through the monsoonseason, however, the anticipated seasonal exhaus-tion eVect is not present. This work provides anempirical basis to upscale the understanding ofdissolved nitrogen behaviour from small hillslopeplots to catchment scales in degraded semi-aridenvironments.
R. E. Brazier (&)Department of Geography, University of Exeter, Rennes Drive, Exeter, EX4 4RJ, UKe-mail: [email protected]
A. J. Parsons · J. WainwrightSheYeld Centre for International Drylands Research (SCIDR), Department of Geography, University of SheYeld, Western Bank, SheYeld, S10 2TN, UK
D. M. PowellDepartment of Geography, University of Leciester, University Road, Leciester, LE1 7RH, UK
W. H. SchlesingerInstitute of Ecosystem Studies, 65 Sharon Turnpike Millbrook, New York 12545, USA
1 3
266 Biogeochemistry (2007) 82:265–278
Introduction
Understanding the movement of water, sedimentand nutrients within a landscape is critical inorder to address the eVects of land degradation,which include erosion (Frederickson et al. 1998;Wainwright et al. 2000), nutrient depletion(Schlesinger et al. 1999) and the consequent pol-lution of waterways (Brazier 2004). In semi-aridareas, which are often subject to high levels ofland degradation, rates of sediment and nutrientmovement are highest where shrub species domi-nate (Parsons et al. 1996; Schlesinger et al. 1999,2000; Wainwright and Thornes 2003; Wainwright2004; Boardman et al. 2003). In US south-west,over the past 150 years, a shift in dominance ofshrub species over grass species has occurred,resulting in severe erosion and potentially high-nutrient losses (BuVington and Herbel 1965; Con-ley et al. 1992; Frederickson et al. 1998; Parsonset al. 2003; Schlesinger et al. 1990). Understand-ing the implications of this shift is therefore offundamental importance if the currently unsus-tainable levels of land degradation and nutrientdepletion in such environments are to be arrested(Frederickson et al. 1998; Peters 2002). Further-more, as semi-arid areas occupy some 17% of theglobal land area and are home to 1/6th of the glo-bal population (UNEP 1992) the extrapolation ofthis understanding will provide a much neededinsight into degradation of drylands throughoutthe world.
Although some eVort has been made to under-stand the distribution of nutrients within semi-arid soils (see Schlesinger et al. 1996; E. N. Müller2004, unpublished data; Müller et al. in press) andthe redistribution of nutrients in overland Xowfrom diVerent vegetation types (for example, Par-dini et al. 2003; Parsons et al. 2003, submitted;Schlesinger et al. 1999, 2000), very little work hasaddressed the dynamics of nutrient Xuxes at bothdiVerent spatial and temporal scales under natu-ral rainfall events within areas dominated byshrub species. Although Schlesinger et al. (2000)report Wndings from natural rainfall events, dataare limited to one spatial scale and are presentedas event totals and annual yields. Consequently,the issue of scale in understanding nutrientbehaviour has been poorly addressed, particularly
so for spatial scale (though see Meixner and Fenn2004 for a catchment scale budgetting approach).Furthermore, as the understanding of nutrientbehaviour is currently-based upon small-scaleexperiments (see aforementioned papers andWainwright et al. 2000 for examples), but isrequired at larger scales to meet the needs ofrecent legislation (Dressing et al. 2003), there is afundamental requirement to improve understand-ing at a wider range of spatial and temporalscales. This paper builds upon recent workreported by Parsons et al. (2004, 2006) who estab-lish the link between hillslope scale and erosionrates in semi-arid environments, and it postulatesthat a similar relationship may be found betweenhillslope scale and nutrient dynamics. Herein wedescribe the behaviour of dissolved nutrients, pri-marily nitrogen, as dissolved phosphorus concen-trations in runoV have been shown to benegligible in such semi-arid environments (Schle-singer et al. 1999).
It has been shown that as hillslope lengthincreases and runoV coeYcients decrease sedi-ment yield from interrill areas initially increases(up until »7 m downslope) and then decreases(Parsons et al. 2004; Parsons and Wainwright2005) as the travel distance of particles anddecrease in runoV coeYcients controls rates oferosion. Consequently, spatial scaling of erosionis partly due to spatial scaling of runoV (Parsonset al. 2006; Wainwright and Parsons 2002; Yairand Kossovsky 2002; Yair and Raz-Yassif 2004).Thus, it is an objective of this paper to testwhether such spatial scaling behaviour is also trueof nitrogen that is transported by the same Xow asit is hypothesised that the increase in scale anddecreases in runoV coeYcients (promoting reten-tion of water, sediment and nutrients within thelandscape) that control erosion should also con-trol nitrogen yields. This objective is tested viainterpretation of Weld data collected during 14natural rainfall events over two monsoon seasonsin the semi-arid south west of the USA.
Additionally, the behaviour of nitrogenentrained and transported by overland Xow isdescribed at a range of temporal scales in order toimprove understanding of the intra- and inter-event dynamics of water and nitrogen yields fromsemi-arid hillslopes. It is hypothesised that a
1 3
Biogeochemistry (2007) 82:265–278 267
decline in nutrient concentrations will be evidentboth during events (Schlesinger et al. 1999) asexhaustion of nitrogen supply occurs and throughmonsoon seasons (Schlesinger et al. 2000) as veg-etation cover increases eVecting a reduction in themobilisation of nitrogen from the soil.
Methods
The study was undertaken at a Weld site within theLucky Hills catchment, which is part of the Wal-nut Gulch Experimental Watershed, southernArizona, USA (CanWeld and Goodrich 2003;Renard et al. 1993). It is a highly eroded land-scape that is strongly connected to the neighbour-ing San Pedro River system and riparian zone(Kepner et al. 2004) with consequently seriousimplications for the oV-site impacts of sedimentand nutrients delivered from the hillslopes withinthe catchment area (Ritchie et al. 2005). Vegeta-tion cover is predominantly woody shrubs domi-nated by Larrea tridentata and Acacia constricta(Weltz et al. 1994) that have invaded the grass-land community over the past 100–150 years(Renard et al. 1993). Soil types are Lucky Hills-McNeal sandy loams (Ustochreptic Calciorthids),which typically contain large proportions of rockfragments and have developed a distinct stonepavement across much of the catchment (Parsonsand Abrahams 1992; Wainwright et al. 1999)resulting in increased runoV rates and high levelsof erosion throughout the catchment (Ritchieet al. 2005). Rainfall events are monsoonal; highintensity and short duration, the majority ofwhich fall between July and September, withmean annual rainfall of 356 mm (Nichols et al.2002).
Within the Weld site, pairs of plots were con-structed to monitor soil erosion at four locations(Parsons et al. 2006). In this study, we are con-cerned with data obtained from the larger plot ofeach pair (named Laurel, Abbott, Dud andWise). These plots had dimensions of 21.01,115.94, 56.84 and 302.19 m2, with lengths of 4.12,14.48, 18.95 and 27.78 m, respectively. They wereinstrumented to monitor intra-event behaviour ofrainfall, Xow and nitrogen Xuxes on a 1-min time-step (see Fig. 1 for an example). In addition to
these four plots, a 1,500 m2 Wrst-order subcatch-ment (Cleese) was instrumented to monitor thesame variables at a larger spatial scale to includethe eVects of concentrated (rill) Xow as well asunconcentrated (interrill) Xow. Although inclu-sion of catchment-scale alongside hillslope-scaleresults may add further complexity to the inter-pretation of results (for example, due to the pres-ence of concentrated Xow lines within the Cleesesubcatchment, pavement cover is, on average,appreciably lower than on the hillslopes), it isconsidered a goal of this paper to upscale under-standing across as wide a range of spatial scales aspossible.
Schlesinger et al. (1996) suggest that the pat-tern of vegetation within a study area exerts astrong control on the spatial distribution andredistribution of the nutrients by overland Xowwithin that area. Consequently, surface coveraVorded by vegetation (as well as stone pavementand Wne particles) was kept as constant as waspractical when plot location was determined(Table 1). Other variables such as soil type, andplot gradient were also kept as constant as possi-ble, though despite these eVorts, some variationbetween plots is inevitable (Table 1). As this vari-ability is by no means consistent between plots(Parsons et al. 2006), however, it is argued that itdoes not unduly bias the analysis of the eVects ofplot characteristics on nitrogen behaviour.
Plot boundaries and gutters were fabricatedfrom sheet steel and aluminium Xashing, buriedwithin the hillslope to ensure that no surface Xowleft the plot other than through a Xume sited atthe downslope end. The Xume concentrated Xowto allow measurement of Xow depth and permitthe extraction of runoV samples for nitrogen anal-ysis. Flow depth was measured using the bubblermodule of an ISCO 6700 pump sampler in con-junction with a small stilling well located 100 mmupstream of the Xume outlet. Flow depth was con-verted to discharge using rating equations-basedupon a series of in situ calibration experimentswhich involved circulating Xow through the Xumeat various discharges (Parsons et al. 2006). RunoVsamples were taken via an intake mounted 50 mmdownstream of the stilling well and between 10and 20 mm from the Xume bed depending on plotsize as smaller plots tended to produce lower Xow
1 3
268 Biogeochemistry (2007) 82:265–278
depths. Sample collection when the intake wascompletely submerged, was triggered by the bub-bler module. Samples were removed from theWeld site for Wltering as soon as was practical anddepending upon event timing always within 4 h ofthe end of each event.
At the Cleese subcatchment site, a Santa Rita-type Xume (Smith et al. 1982) with a stilling wellwas installed at the catchment outlet. A bubbler-module stage recorder from an ISCO 6700 pumpsampler was mounted within the stilling well. Theintake for the pump sampler was mounted in theXoor of the Xume and set to trigger at 25 mm Xowdepth in a similar fashion to the hillslope set up.Flow depth and rainfall data were recorded at a 1-min timestep, with Xow being converted to dis-charge using a rating curve based on the Xumedimensions and gradient (Smith et al. 1982).
On return to the laboratory, 30 ml subsamplesof the 1-l runoV samples were taken and Wlteredthrough pre-rinsed 0.45 �m Millipore HA Wltersinto polypropylene sample bottles. These sampleswere then analysed for ammonium (NH4-N) andnitrate (NO3–N) using standard methods on aTraacs 800 Autoanalyser. Total inorganic N wastaken as the sum of NH4–N + NO3–N. Each sam-
ple was then subjected to a persulphate digestion(D’Elia et al. 1977) and re-analysed. The diVer-ence between the digested and undigested con-centrations was assumed to represent dissolvedorganic forms of N (DON). See Schlesinger et al.(2000) for further details and note compatibilityof analytical techniques with previous work bythese authors.
While 35 plot-events occurred during 2001 and2002, due to mechanical breakdowns, only 14events had complete records of chemistry, rainfalland discharge to include in the numerical analysis.Accepting this limitation, the dataset presentedhere is thought to provide a good representationof the dynamics of dissolved nitrogen behaviourduring natural rainfall/runoV events in a semi-aridenvironment as events monitored range in runoVmagnitude from 0.018 to 33.3 l m¡2 with returnperiods of between 1 and 5 years (Abrahamset al. 2006).
Results
As an example of the event data collected duringthe 2-year monitoring period, a dataset for the
Fig. 1 The Laurel hills-lope plot (21.02 m2) Lucky Hills watershed, Walnut Gulch, AZ, USA. Flume and collecting res-ervoirs are shown bottom centre, pump sampler and housing centre right
Table 1 Variablesdescribing plot character-istics, modiWed from Par-sons et al. (2005)
event recorded on the Abbott hillslope plot on 04/08/02 is shown in Fig 2a, b. One-minute rainfallintensities of up to 198 mm h¡1 were observedover a period of 38 min, generating a total Xowof 3,133 l over 35 min with a peak Xow of270 l min¡1. The resultant hydrograph is charac-teristic of hillslope hydrographs in semi-arid areaswith a steep rising limb, short lag-time betweenpeak rainfall and peak Xow (in this examplewithin 1 min) and a prolonged (compared to therising limb) recession. In general suspended sedi-ment dynamics tend to precede changes to Xow ina clockwise hysteretic fashion. In this example,highest loads of 925 g min¡1, occur prior to peakXow concentrations by 3 min and total yields»7,596 g over the event. Nitrogen dynamics(shown throughout as gross yields, i.e. uncor-rected for atmospheric deposition) are also at
their highest—total dissolved nitrogen(TDN) = 1.8 mg m¡2 min¡1 1 min prior to peakXow—and tend to follow the form of the hydro-graph closely. This behaviour is true of all nitro-gen fractions observed here, though proportionsof each fraction contributing to TDN vary widelyas is illustrated later.
Spatial scaling of dissolved nitrogen behaviour
To investigate the relationships between nitrogenbehaviour, hillslope characteristics and dischargean analysis of all 14 events from the Wve spatialscales was performed. In the Wrst instance thiswork focused on a regression analysis of nitrogenyields (mg m¡2 min¡1) against plot areas. Figure 3illustrates the relationship between instanta-neous TDN yield from all events and plot area.
Fig. 2 (a) Typical event data describing rainfall intensity (mm h¡1), dis-charge (l min¡1) shown here as the solid line and sediment loads (g min¡1) shown as crosses, from the Abbott hillslope plot, 4 August 2002. (b) Typical event data describing rainfall intensity (mm h¡1), discharge (l min¡1) and nutrient concentrations (mg l¡1) from the Abbott hillslope plot, 4 August 2002
0
50
100
150
200
250
300
84:
21
05:
21
25:
21
45:
21
65:
21
85:
21
00 :
31
20:
31
40:
31
60:
31
80 :
31
01:
31
21 :
3 1
41 :
31
61:
31
81:
31
02:
31
22:
31
42:
3 1
Time
01002003004005006007008009001000
g)(d
aoltne
mide
S
0
100
200
Rainfall intensity (mm hr-1)
0
0.2
0.4
0.6
0.8
1
1.2
1.4
84:
21
05:
21
25:
21
45:
21
65:
21
85:
21
00:
31
20:
31
40:
31
60:
31
80:
31
01 :
31
21:
31
41:
31
61 :
31
81:
31
02:
31
22:
31
42:
31
Time
Nut
rient
con
cns
(mgl
-1)
Q (
lmin
-1)
0
50
100
150
200
250
nim/l(
Q)
NH4-N (mg/l)
NO3-N (mg/l)
TN (mg/l)
DON (mg/l)
0
b
a
100
200
300
Rainfall intensity (mm hr-1)
1 3
270 Biogeochemistry (2007) 82:265–278
For all nitrogen fractions observed a wide rangeof yields is evident from all plot sizes with largestandard deviations around the mean and weak,positive relationships with plot area (see Tables 2,3 for details). When described by the mean yieldswith 95% conWdence limits (shown as box andwhisker plots), Fig. 3 shows a pattern of decreas-ing TDN yields with plot area. Furthermore, one-way analysis of variance, to compare meansbetween scales shows a statistically signiWcantdiVerence at the 95% level between the popula-tions—F(13.22) > Fcrit(2.42).
To consider the inXuence of overland Xowupon nitrogen behaviour instantaneous nitrogenyields were plotted against instantaneous dis-charge for all events across all spatial scales(Fig. 4). Although there is some evidence for avery weak positive relationship between instanta-neous yields and discharges across all spatialscales (see Table 1), there is a high degree of vari-ability within the dataset, particularly for low dis-charges: those less than 300 l min¡1 produce TDNyields of between 0.03 and 1.87 mg m¡2 min¡1.Such behaviour is similar for all fractions of N(Table 2) apart from NO3–N, which shows a rea-sonably strong, signiWcant relationship betweendischarge and nitrogen yields. However, if theinXuence of discharge on nitrogen yields at eachseparate spatial scale is considered, much strongerrelationships are found (Fig. 4, Table 4), suggest-ing that the relationship between nitrogen yieldsand discharges identiWed by previous authors
(Schlesinger et al. 2000), may in fact be scaledependent, as the slope of the best Wt linedecreases with an increase in scale. Thus, the datacollected here demonstrate a general increase inTDN yield with increasing discharge, though therate at which this increase occurs is signiWcantlyreduced as plot size increases.
Temporal scaling of dissolved nitrogen behaviour: intra-event dynamics
The characteristics of a typical event are exempli-Wed in Fig. 5a, which describes rainfall, Xow, andnitrogen concentrations from the Cleese catch-ment on 26/07/02. Schlesinger et al. (1999)observed declines in nitrogen concentrationsthroughout a series of simulated rainfall events onsemi-arid shrubland plots. Similar behaviourunder natural rainfall events is observed here,with initially high concentrations of TDN (peak-ing at 2.03 mg l¡1) declining, with increasing
Fig. 3 Instantaneous TDN yields (mg m¡2 min¡1) and mean values as a function of plot area (m¡2) for all events. Box and whiskers describe 95% conWdence limits around means
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
10 100 1000 10000
Area (m2)
TD
N y
ield
s (m
g m
2 min
-1)
Cleese
Laurel
Abbott
Dud
Wise
Means with 95%confidence limits
Table 2 R2-values describing relationships betweeninstantaneous Xow (l min¡1), plot areas (m2) and instanta-neous nutrient yields (mg m¡2 min¡1)
Fig. 4 Instantaneous TDN yields (mg m¡2 min¡1) as a func-tion of instantaneous dis-charge (l min¡1) for all events. Best Wt lines de-scribe changing linear relation between TDN yields and discharge with scale
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0 500 1000 1500 2000 2500
Discharge (l min-1
)
TD
N y
ield
s (m
g m
-2 m
in-1)
Cleese
Laurel
Abbott
Dud
Wise
Cleese
WiseAbbott
Dud
Laurel
Table 4 R2-values and equations describing best linear relationships between instantaneous discharges (l min¡1) and nutri-ent yields (mg m¡2 min¡1) from all plots
Bold type denotes relationships that are signiWcant at P < 0.01
R2-values Discharge versus TDN yield
Equation describinglinear best-Wt
Discharge versusDON yield
Discharge versusNH4–N yield
Discharge versusNO3–N yield
Laurel 0.82 y = 0.0183x + 0.1697 0.51 0.95 0.01Dud 0.89 y = 0.0092x ¡ 0.0104 0.76 0.92 0.23Abbott 0.89 y = 0.0059x + 0.0401 0.87 0.89 0.43Wise 0.97 y = 0.0024x + 0.0064 0.91 0.97 0.09Cleese 0.54 y = 0.0004x + 0.0782 0.31 0.42 0.60
1 3
272 Biogeochemistry (2007) 82:265–278
tions average 0.72 mg l¡1. Furthermore, thisdecline is also evident within the fractions ofNH4–N declining on average from 0.49 to0.31 mg l¡1 and DON declining on average from0.38 to 0.19 mg l¡1. However, the concentration ofthe NO3–N fraction tends to stay relatively stableduring events with a slight increase on average of0.1–0.16 mg l¡1. For all events, at each separatescale, it has been shown that discharge exerts astrong control upon nitrogen yields (Fig. 4),therefore, it is important to study the intra-eventdynamics of TDN yields relative to the observedinstantaneous discharges to further understand-ing of the temporal scaling of dissolved nitrogenbehaviour within events.
Rainfall, Xow and nitrogen yields for the Cle-ese 26/07/02 event are illustrated in Fig. 5a, b. In
contrast to the decrease in nitrogen concentrationwithin-events described above, there is a strongsimilarity between nitrogen yields and the form ofthe hydrograph illustrating how nitrogen yieldsare positively related to discharge during events(as was also seen on the hillslopes, see Fig. 2b foran example). Peak yields of all fractions coincidewell with peak Xows, which supports the Wndingsof Schlesinger et al. (2000), who conclude thatnitrogen yields are well correlated with dischargefrom small plots under natural rainfall in the Chi-huahuan desert. This behaviour is consistentbetween events as was seen in Fig. 4 for TDN, butit is also true for DON and NH4–N, with theexception of yields from the Cleese catchment,where coeYcients of determination are low andinsigniWcant. NO3–N on the other hand, does not
Fig. 5 (a) Rainfall inten-sity (mm h¡1), discharge (l min¡1) and nutrient concentrations (mg l¡1) from the Cleese subcatch-ment, 26 July 2002. (b) Rainfall intensity (mm h¡1), discharge (l min¡1) and nutrient yields (mg m¡2 min¡1) from the Cleese subcatch-ment, 26 July 2002
0
0.5
1
1.5
2
2.5
04:
02
34:
02
64:
02
94:
02
25:
02
55:
02
85:
02
10:
12
40:
12
70:
12
01:
12
31:
12
61 :
12
91:
12
22:
12
52 :
12
82:
12
13:
12
43:
12
73:
12
04:
12
34:
12
64:
12
94:
12
25:
12
55:
12
85:
12
10:
22
Time
Nut
rient
con
cns
(mgl
-1)
Nut
rient
yie
lds
(mg
m-2
min
-1)
0
500
1000
1500
2000
2500
Q (
l min
-1)
Q (
l min
-1)
NO3-N (mg/l)
TN (mg/l)
DON (mg/l)
NH4-N (mg/l)
0
50
100
Rainfall intensity (mm hr-1
)
0
0.2
0.4
0.6
0.8
1
1.2
04:02
34:02
64:02
94:02
25:02
55:02
85:02
10:12
40:12
70:12
01:12
31:12
6 1:12
91:12
22:1 2
52:12
82:12
13:12
43:12
73:12
04:12
34:12
64:12
94:12
25:12
55:1 2
85:12
10: 22
Time
0
500
1000
1500
2000
2500
NO3-N (mg/m2/min)
TN (mg/m2/min)
DON (mg/m2/min)
NH4-N (mg/m2/min)
0
50
100
Rainfall intensity (mm h-1)
a
b
1 3
Biogeochemistry (2007) 82:265–278 273
seem to be positively related to discharge as thenear constant yields observed in Fig. 5b are repli-cated for all events at all scales with the exceptionof those at Cleese.
Temporal scaling of dissolved nitrogen behaviour: inter-event dynamics
To investigate the contrast in behaviour betweenevents, successive events on the same hillslopeplot (Wise) from the 26/07/02 and the 04/08/02 areshown in (Fig. 6). The events are characterised bysimilar rainfall totals (29.5 and 31 mm), thoughvery diVerent peak 1-minute (76 and 213 mm h¡1)and average event (28 and 53 mm h¡1) intensitieswhich generate distinct hydrograph responses.The hydrograph of the 26/07/02 event produces atotal discharge of 1,385 l over a period of 28 min,with peak Xows of 140 l min¡1, whereas the hyd-rograph of the 04/08/02 event yields a total dis-charge of 3,730 l over a period of 25 min with apeak Xow of 738 l min¡1. Such Xow characteristicslead to diVerences in nitrogen responses betweenevents. Total N yield is 1,027 mg from the 26/07/02 event and 2,675 mg from the 04/08/02 event,thus, given an increase in total Xow by a factor of2.7, there is a very similar (2.6 factor) increase inTDN yield between events. Due to the limitedtemporal coverage of the dataset, it is not clearwhether these results are true of all successiveevents, as only three sets of successive eventswere captured. However, the data shown for theWise plot in Figs. 4 and 6a, b conWrm that laterevents do not produce lower values of TDN thanmight be expected from the relationship betweendischarge and TDN, which is linear, and suggeststhat there is no notable decline in nitrogen yieldson an inter-event basis.
To understand the dynamics of each nitrogenfraction that contributes to the behaviour ofTDN, it is illustrative to look at the proportions ofTDN from each of the three fractions analysed(NH4–N, NO3–N and DON), for the same eventsas detailed above, over all plots where observa-tions were made (Dud, Abbott, Wise and Cleesefor the 26/07/02 event and Laurel, Dud, Abbottand Wise for the 04/08/02 event). As is shown inFig. 7a, b for the smaller event, NH4–N dominates(43–57% of TDN), with the remaining contribu-
tion from DON (18–38% of TDN) and NO3–N(20–25% of TDN). If we consider the proportionsof TDN from each fraction from the larger event,however, consistently diVerent behaviour is evi-dent from all plots, with a marked increase inNH4–N (52–78% of TDN) and a decrease inNO3–N (1–7% of TDN), though a similar propor-tion of TDN from DON (21–41% of TDN). Ananalysis of the factors which might be expected tocontrol the proportions within TDN from twoevents separated by only 8 days (total dischargeand plot area), however, shows weak, insigniWcantrelationships between all TDN fractions and bothtotal discharge and plot area and the combinedeVect of both. Consequently, though it has beenshown that plot areas and Xow discharges controlnitrogen yields from a hydrological perspective, itis not clear from the results presented here, whatthe controls upon the proportions of TDN comingfrom each fraction are. Although it might berelevant to consider the role of atmosphericdeposition and soil mineralisation/nitriWcation ascontrolling factors on nitrogen Xuxes, no suchdata were collected here.
Discussion
Schlesinger et al. (2000) monitored nitrogenyields from small (4 m2) plots under natural rain-fall events and used these results in combinationwith atmospheric deposition data, to calculate netgains/losses of N to/from the landscape in units ofkg ha¡1 year¡1. Results showed that losses of dis-solved nitrogen in runoV were potentially signiW-cant (0.43 kg ha¡1 year¡1) if extrapolated to thelandscape scale. However, data presented heresuggest that the relationship between the size ofplot where observations are made and the magni-tude of those observations may be more complexthan has previously been assumed. In fact, therelationship between discharge and nitrogenlosses varies signiWcantly with the scale of obser-vation across the range of plot sizes from 21 to1,500 m2. This result brings into question theapproach of upscaling results from such smallplots to the landscape scale, which is an area ofwork that requires much care (Wainwright et al.2000; Müller et al., in press). A key Wnding of this
1 3
274 Biogeochemistry (2007) 82:265–278
study is that extrapolation of results from smallspatial-scale experiments using either simulatedor natural rainfall events may be misleading ifonly one spatial scale of observation is used. Par-sons et al. (2004) have demonstrated that for soilerosion it is the explicit consideration of hillslopelength (and therefore of particle travel distance),that allows scaleable predictions of sedimentyields to be made. Herein it is suggested that asimilar approach is taken to upscaling under-standing of nitrogen yields, with direct consider-ation of plot/catchment areas and thereforevariability in runoV behaviour being made inorder to provide larger scale estimates of dis-solved nitrogen losses in overland Xow that do notunderestimate potential losses by ignoring theeVect of spatial scale.
It is likely that the most direct control ondecreasing nitrogen yields with increasing spatialscale, is the reduction in runoV coeYcients as plot
size increases, or the spatial scaling of runoV thatwas observed by Parsons et al. (2006), Wain-wright and Parsons (2002), Yair and Kossovsky(2002) and Yair and Raz-Yassif (2004) in relationto soil erosion. As spatial scale increases,observed increases in discharge will occurthrough more concentrated Xowpaths, and deeperoverland Xow. Dissolution of nitrogen held in thesoil, however, may only occur in the lower layersof runoV that are in contact with the soil leadingto a lower concentration of nutrients in larger vol-umes of water. Furthermore, in dry environmentswhere NO3–N in particular has accumulated insoils over the dry season from nitriWcation andmineralisation processes, the upper layers of thesoil may be the main source of TDN (Rigganet al. 1985; Avila et al. 1992; Holloway and Dahl-gren 2001). Consequently, as catchment areasincrease, the eVect of this Xushing of NO3–N fromthe soil is reduced, with the reduced runoV coeY-
Fig. 6 (a) Rainfall inten-sity (mm h¡1), discharge (l min¡1) and nutrient concentrations (mg l¡1) from the Wise hillslope, 26 July 2002. (b) Rainfall intensity (mm h¡1), dis-charge (l min¡1) and nutrient concentrations (mg l¡1) from the Wise hillslope, 4 August 2002
0
0.2
0.4
0.6
0.8
1
1.2
1.4
00:
12
30:
12
60:
12
90:
12
21:
12
51:
12
81:
12
12:
12
42:
12
72:
12
03:
12
33:
12
63:
12
93:
12
24:
12
54:
12
84:
12
15:
12
Time
Nut
rient
con
cns
(mg/
l)N
utrie
nt c
oncn
s (m
g/l)
020406080100120140160
nim/l(
Q)
NH4-N (mg/l)
NO3-N (mg/l)
TN (mg/l)
DON (mg/l)
0
50
100
Rainfall intensity (mm hr-1)
0
0.2
0.4
0.6
0.8
1
1.21
5:2
1
35:
21
55:
21
75:
21
95:
21
10:
31
30:
31
50:
31
70:
31
90:
31
11:
31
31:
31
51:
31
71:
31
91:
31
12:
31
32:
31
52:
31
Time
0
100
200
300
400
500
600
700
800
nim/l(
Q)
NH4-N (mg/l)
NO3-N (mg/l)
TN (mg/l)
DON (mg/l)
0b
a
100
200
300
Rainfall intensity (mm hr-1)
1 3
Biogeochemistry (2007) 82:265–278 275
cients. Similar observations were made by Fennand Poth (1999), Meixner et al. (2001) andMeixner and Fenn (2004) to explain spatialpatterns at larger spatial scales. This is a tentativeexplanation for the reduction in the slope of thebest Wt line describing the relationship betweenXow discharge and TDN yields (Fig. 4) and it isalso a mechanism which would explain the con-trol exerted by the form of the event hydrographsobserved upon the accompanying nitrogen yields.Clearly, more detailed datasets describing N min-eralisation/nitriWcation and the deposition (bothwet and dry) of N, and the interaction of theseprocesses with surface runoV are required toenhance understanding in this area.
Across all spatial scales, observations of nitro-gen dynamics during single, natural rainfall eventssuggest that initial Xuxes, associated with the ris-ing limb of event hydrographs produce the high-est concentrations of all nitrogen fractions. TheseWndings support the conclusions of Schlesingeret al. (1999) who found concentrations of bothTDN and DON in runoV to be at their highestwithin the Wrst 5 min of simulated rainfall eventsand to decline thereafter. Supply limited behav-iour such as this has commonly been observed inrelation to suspended sediments in dryland
channels (Alexandrov et al. 2003) and erosionrates on semi-arid hillslopes over both simulatedevents (Parsons et al. 1994) and during naturalrainfall events (Parsons et al. 2006). Similarobservations in relation to nutrient behaviourwere made by Schlesinger et al. (2000) from smallplots under natural rainfall in semi-arid NewMexico. In addition these authors found that suchbehaviour led to higher nutrient yields fromshrubland areas, due to increases in discharge,rather than concentration. Data collected heresupport these Wndings at a range of spatial scales(from 21 to 1,500 m2), demonstrating that theexhaustion of nitrogen during events occurs inde-pendent of spatial scale, and that the yields of dis-solved nitrogen from natural events at eachspatial scale increase with discharge, though at adecreasing rate as plot size increases.
Little support is evident, however, for anyinter-event or seasonal decline in nitrogen losses,as was observed by Fisher and Grimm (1985) whorecorded signiWcant decreases in DON over thecourse of three events within the Sonoran desert.Results described here also contradict the Wnd-ings of Li et al. (2006), who published data from achaparral watershed demonstrating seasonalexhaustion of nitrogen where successive storms
Fig. 7 (a) Proportions of mean TDN from NH4N, NO3N and DON from Dud, Abbott, Wise and Cleese during the 26 July 2002 event. (b) Propor-tions of mean TDN from NH4N, NO3N and DON from Laurel, Abbott, Dud and Wise during the 4 August 2002 event
0%
20%
40%
60%
80%
100%
Plot
b
a
DONNO3NNH4N
0%
20%
40%
60%
80%
100%
Laurel
Plot
Pro
port
ion
Pro
port
ion
DONNO3NNH4N
Abbott Dud Wise
CleeseAbbottDud Wise
1 3
276 Biogeochemistry (2007) 82:265–278
occur closely together (within 3–4 months). How-ever, these Wndings are supported by the work ofParsons and Wainwright (2005) who found a sea-sonal decline in NH4–N concentrations in runoVunder grassland, but found no signiWcant seasonalvariation in concentrations of NH4–N under tar-bush and creosote bush shrub species (similarvegetation types to this study). Here, DON con-tributes largely the same proportion of TDNbetween the consecutive events observed, thoughunfortunately none of the plot datasets covers anentire season to elucidate whether or not a sea-sonal exhaustion eVect is prevalent. From the lim-ited data available on successive events, it seemsunlikely that this is the case, however, as nitrogenyields appear to vary consistently with hydrologi-cal behaviour regardless of event timing or posi-tion within the season.
Conclusions
This paper describes the behaviour of dissolvednitrogen associated with overland Xow at a rangeof temporal and spatial scales. In so doing itaddresses the problem of up-scaling results fromsmall-plot experiments to provide larger scaleestimates of nitrogen yields by demonstrating thatobserved nitrogen behaviour is particularly sensi-tive to the spatial scale at which it is monitored.Dissolved nitrogen yields increase with increasingdischarge but at a declining rate related toincreasing plot size. Therefore, if results fromsmall-scale plot experiments are extrapolated tothe landscape scale, it is likely that misleadingestimates of nitrogen yields (overestimations) willbe made. Small spatial-scale experiments do notconsider the retention of water-borne nutrientswithin the landscape, nor changing abilities ofdeeper Xows to pick up proportionally morenutrients.
Observations describing the temporal scalingof nitrogen behaviour show that event-basednitrogen concentrations are initially high anddecline through events. This Wnding indicates thatnitrogen exhaustion occurs, in a system that canbe described as supply limited on an event basis.Nitrogen yields are strongly controlled by event
hydrology, so that even when concentrations areat their highest (typically on the rising limb of thehydrograph), dissolved yields are low, only reach-ing their peak when discharge increases. As des-ert soils tend to be low in nutrients and, in thepresent case have suVered from accelerated ero-sion over the last 100–150 years, such behaviour isunderstandable and agrees with the existing data-sets that describe event dynamics at single spatialscales from rainfall simulation experiments.
On an inter-event basis, there appears to be lit-tle support for the hypothesis that nitrogen yieldswill decline either between events or across themonsoon season. This Wnding is in line with an,albeit limited, literature and suggests that thoughnitrogen supply may be exhausted in the short-term (i.e. during an event), seasonal exhaustiondoes not occur. Therefore, large events, wheneverthey fall during a season, will yield signiWcantnitrogen losses. This may be due to the role ofnitrifying bacteria in replenishing the supply ofNO3–N in the soil, once the rainy season hasstarted (Hartley and Schlesinger 2000) and theaccumulation of extremely high amounts of soilNH4–N during the dry season due to the minerali-sation of organic N in the soil (Fisher et al. 1987).
A key outcome of this work is an improvedempirical understanding of the eVect of spatialscale and hydrological behaviour upon thedynamics of dissolved N in overland Xow. Ongo-ing work will incorporate this understandingwithin a modelling framework to permit the pre-diction of nitrogen losses at larger spatial andlonger temporal scales. This work will utilise theempirical understanding presented here in addi-tion to further data (currently being collected),which describe the role of N mineralisation/nitriW-cation and wet/dry deposition of N to addresslandscape scale questions.
Acknowledgements This research has been funded bythe Natural Environment Research Council (Grant GR3/12754). We thank Susan Moran for permission to use thefacilities of the USDA-ARS Walnut Gulch Field Station,Burt Devere for permission to conduct our experiments onhis ranch, John Smith, Howard Larsen, Art Dolphin andJim Smith for their invaluable support and assistancethroughout, Beth Thomas and Heather Hemric for chemi-cal analysis of the runoV samples and the many Weld assis-tants who helped with the data collection.
1 3
Biogeochemistry (2007) 82:265–278 277
References
Abrahams AD, Neave M, Schlesinger WH, Wainwright J,Howes DA, Parsons AJ (2006) Biogeochemical Flux-es. In: Havstad K, Huenneke L, Schlesinger WH (eds)Structure and function of a chihuahuan desert ecosys-tem. The jornada basin long-term ecological researchsite. Oxford University Press, Oxford
Alexandrov Y, Laronne JB, Reid I (2003) Suspended sedi-ment concentration and its variation with water dis-charge in a dryland ephemeral channel, northernNegev, Israel. J Arid Environ 53:73–84
Avila A, Pinol J, Roda F, Neal C (1992) Storm solutebehaviour in a montane Mediterranean forestedcatchment. J Hydrol 140:143–161
Boardman J, Parsons AJ, Holland R, Holmes PJ, Washing-ton R (2003) Development of badlands and gullies inthe Sneeuberg, Great Karoo, South Africa. Catena50:165–184
Brazier RE (2004) Quantifying soil erosion by water in theUK: a review of monitoring and modelling approach-es. Prog Phys Geogr 28(3):1–26
BuVington LC, Herbel CH (1965) Vegetational changes ona semidesert grassland range from 1858 to 1963. EcolMonogr 35:139–164
CanWeld HE, Goodrich DC (2003) Studies for scale andprocesses in hydrologic modelling on Lucky Hills Wa-tershed. In: Renard KC, McElroy SA, Gburek WJ,CanWeld HE, Scott RL (eds) First interagency confer-ence on research in the Watersheds. USDA Agricul-tural Research Service Special Publication,Washington, DC, pp 444–450
Conley W, Conley MR, Karl TR (1992) A computationalstudy of episodic events and historical context inlong-term ecological processes: climate and grazingin the northern Chihuahuan desert. Coenoses 7:55–60
D’Elia CF, Steudler PA, Corwin N (1977) Determinationof total nitrogen in aqueous samples using persulfatedisgestion. Limnol Oceanogr 22:760–764
Dressing SA et al (20 authors) (2003) Nationalmanagement measures to control nonpoint source pol-lution from agriculture. United States EnvironmentProtection Agency (EPA) 841-B-03-004. http://www.epa.gov/owow/nps/agmm/index.html
Fenn ME, Poth MA (1999) Temporal and spatial trends instreamwater nitrate concentrations in the san bernadi-no mountains, southern california. J Environ Qual28:822–836
Fisher SG, Grimm NB (1985) Hydrologic and materialbudgets for a small Sonoran desert watershed duringthree consecutive cloudburst events. J Arid Environ9:105–118
Fisher FM, Parker LW, Anderson JP, Whitford WG (1987)Nitrogen mineralization in a desert soil: InterfacingeVects of soil moisture and nitrogen fertilizer. Soil SciSoc Am J 51:1033–1041
Frederickson E, Havstad KM, Estell R, Hyder P (1998)Perspectives on desertiWcation: south-western UnitedStates. J Arid Environ 39:191–207
Hartley AE, Schlesinger WH (2000) Environmental con-trols on nitric oxide emission from north Chihuahuandesert soils. Biogeochemistry 50:279–300
Holloway JM, Dahlgren RA (2001) Seasonal and event-scale variations in solute chemistry for four Sierra Ne-vada catchments. J Hydrol 250:106–121
Kepner WG, Semmens DJ, Bassett S, Mouat DA, Good-rich DC (2004) Scenario analysis for the San Pedro riv-er, analyzing hydrological consequences of a futureenvironment. J Environ Monit Assess 94:115–127
Li X, Meixner T, Sickman JO, Miller AE, Schimel JP, Me-lack JM (2006) Decadal-scale dynamics of water, car-bon and nitrogen in a California chaparral ecosystem:DAYCENT modeling results. Biogeochemistry77:217–245
Meixner T, Fenn ME, Poth MA (2001) Nitrate in pollutedmountainous catchments with Mediterranean cli-mates. The ScientiWc World 1: DOI 10.1100/tsw.2001.324
Meixner T, Fenn ME (2004) Biogeochemical budgets in amediterranean catchment with high rates of atmo-spheric N deposition—importance of scale and tempo-ral asynchrony. Biogeochemistry 70:331–356
Müller EN, Wainwright J, Parsons AJ (in press) The im-pact of connectivity on the modelling of water Xuxes insemi-arid shrubland environments. Water ResourcesResearch
Nichols MH, Renard KG, Osborn HB (2002) Precipitationchanges from 1956 to 1996 on the Walnut Gulch exper-imental watershed. J Am Water Resour Assoc 38:161–172
Pardini G, Gispert M, Dunjó G (2003) RunoV erosion andnutrient depletion in Wve Mediterranean soils of NESpain under diVerent land use. Sci Total Environ309:213–224
Parsons AJ, Abrahams A (1992) Field investigations ofsediment removal in interrill overland Xow. In: Par-sons AJ, Abrahams AD (eds) Overland Xow: hydrau-lics and erosion mechanics. UCL Press, London, pp307–334
Parsons AJ, Abrahams AD, Wainwright J (1994) Rain-splash and erosion rates in an interill area on semi-aridgrassland, Southern Arizona. Catena 22(3):215–226
Parsons AJ, Abrahams AD, Wainwright J (1996) Respons-es of interrill runoV and erosion rates to vegetationchange in southern Arizona. Geomorphology 14:311–317
Parsons AJ, Wainwright J, Schlesinger WH, Abrahams AD(2003) The role of overland Xow in sediment and nitro-gen budgets of mesquite duneWelds, southern NewMexico. J Arid Environ 53:61–71
Parsons AJ, Wainwright J, Powell DM, Kaduk J, BrazierRE (2004) A conceptual model for determining soilerosion by water. Earth Surf Process Landforms29:1293–1302
Parsons AJ, Wainwright J (2005) The impact of water andnutrient Xuxes on the stability of vegetation bound-aries in a semi-arid ecosystem, the Jornada Basin, NewMexico. In: Proceedings of the geophysical researchabstracts 7, EGU General Assembly, Vienna
1 3
278 Biogeochemistry (2007) 82:265–278
Parsons AJ, Brazier RE, Wainwright J, Powell DM (2006)Scale relationships in hillslope runoV and erosion.Earth Surf Process Landforms 31(11):1384–1393
Peters DC (2002) Plant species dominance at a grassland-shrubland ecotone: an individual-based gap dynamicsmodel of herbaceous and woody species. Ecol Model152:5–32
Renard KG, Lane LJ, Simanton JR, Emmerich WE, StoneJJ, Weltz MA, Goodrich DC, Yakowitz DS (1993)Agricultural impacts in an arid environment. HydrolSci Technol 9:145–190
Riggan PJ, Lockwood RN, Lopez EN (1985) Depositionand processing of ariborne nitrogen pollutants in Med-iterranean-type ecosystems of southern California.Environ Sci Technol 19:781–789
Ritchie JC, Nearing M, Nichols M, Ritchie CA (2005) Pat-terns of soil movement on Lucky Hills Watershed,Walnut Gulch Arizona. Catena 61(2–3):122–130
Schlesinger WH, Reynolds JF, Cunningham GL, Hu-enneke LF, Jarrell WM, Virginia RA, Whitford WG(1990) Biological feedbacks in global desertiWcation.Science 247:1043–1048
Schlesinger WH, Raikes JA, Hartley AE, Cross AF (1996)On the spatial pattern of soil nutrients in desert eco-systems. Ecology 77:364–374
Schlesinger WH, Abrahams AD, Parsons AJ, WainwrightJ (1999) Nutrient losses in runoV from grassland andshrubland habitats in Southern New Mexico: I. Rain-fall simulation experiments. Biogeochemistry 45:21–34
Schlesinger WH, Ward TJ, Anderson J (2000) Nutrientlosses in runoV from grassland and shrubland habitatsin Southern New Mexico: II. Field plots Biogeochem-istry 49:69–86
Smith RE, Chery DL Jr, Renard KG, Gwinn WR (1982)Supercritical Xow Xumes for measuring sediment-lad-en Xow. US Department of Agriculture, TechnicalBulletin No. 1655. 72pp. Washington, DC, USDA
UNEP (1992) World atlas of desertiWcation. Edward Ar-nold, Sevenoaks, UK
Wainwright J, Parsons AJ, Abrahams A (2000) Plot-scalestudies of vegetation, overland Xow and erosion inter-actions: case studies from Arizona and New Mexico.Hydrol Process 14:2921–2943
Wainwright J, Parsons AJ (2002) The eVect of temporalvariations in rainfall on scale dependency in runoVcoeYcients. Water Resour Res 38(12):10 1029/2000WR000188
Wainwright J (2004) History and evolution of Mediterra-nean desertiWcation. Adv Environ Monit Model1(4):1–87
Wainwright J, Parsons AJ, Abrahams A (1999) Field andcomputer simulation experiments on the formation ofdesert pavement. Earth Surf Process Landforms24:1024–1037
Weltz MA, Ritchie JC, Fox HD (1994) Improved modelsfor estimating soil erosion rates from caesium-137measurements. J Environ Qual 28:611–622
Yair A, Kossovsky A (2002) Climate and surface proper-ties: hydrological response of small arid and semi-aridwatersheds. Geomorphology 42:43–57
Yair A, Raz-Yassif N (2004) Hydrological processes in asmall arid catchment: scale eVects of rainfall and slopelength. Geomorphology 61:155–169
