Upload
al-james
View
212
Download
0
Embed Size (px)
Citation preview
Journal of Hydrology 377 (2009) 351–366
Contents lists available at ScienceDirect
Journal of Hydrology
journal homepage: www.elsevier .com/ locate / jhydrol
Antecedent moisture conditions and catchment morphology as controlson spatial patterns of runoff generation in small forest catchments
A.L. James a,b,*, N.T. Roulet a,b,c,1
a Department of Geography, McGill University, 805 Sherbrooke St. W., Montreal, QC, Canada H3A 2K6b Global Environment and Climate Change Centre (GEC[3]), McGill University, 805 Sherbrooke St. W., Montreal, QC, Canada H3A 2K6c McGill School of Environment, McGill University, 805 Sherbrooke St. W., Montreal, QC, Canada H3A 2K6
a r t i c l e i n f o s u m m a r y
Article history:Received 5 May 2009Received in revised form 23 August 2009Accepted 30 August 2009
This manuscript was handled byK. Georgakakos, Editor-in-Chief, with theassistance of Enrique R. Vivoni, AssociateEditor
Keywords:Runoff generationScaleIsotope hydrograph separationAntecedent moisture conditions
0022-1694/$ - see front matter � 2009 Elsevier B.V. Adoi:10.1016/j.jhydrol.2009.08.039
* Corresponding author. Present address: Departmmental Resources, North Carolina State University, Ca27695 USA. Tel.: +1 919 513 2511; fax: +1 919 515 6
E-mail addresses: [email protected] (A.L. Ja(N.T. Roulet).
1 Tel.: +1 514 398 4945; fax: +1 514 398 7437.
Although existing empirical studies of runoff generation in headwater catchments have provided evi-dence of runoff mechanisms, contributing sources and active flowpaths from forest catchments aroundthe world, our understanding of how hydrologic and biological processes vary and aggregate in spacewithin headwater systems remains poor. In this study, we examine the spatial patterns of storm runoffgeneration from eight small nested forest catchments ranging in size from 7 to 147 ha, as a function ofantecedent moisture conditions and catchment morphology. The catchments, located in the formerly gla-ciated terrain of Mont Saint-Hilaire, Quebec, Canada, are complex in both their topographic and subsur-face characterization. Hydrologic response from the eight catchments shows a strong nonlinear changewith antecedent moisture conditions consistent with the hypothesis of different ‘states-of-wetness’. Withthe transition from wet to dry conditions, local groundwater stores become depleted in some ephemeralcatchments, variable source areas shrink and new-water delivered by shallow-subsurface stormflow andthe transient development of perched water in valley-bottoms can account for much larger percentagesof total runoff (up to 76% of total runoff). For the five storm events, no consistent pattern in percent new-water delivery was observed. However, for the storms observed under dry conditions, larger magnitudesof new water were generated from the three largest catchments attributable to basin morphology, whilestorms observed under wet conditions exhibited no consistent pattern, with larger variability among thesmaller catchments. The results presented here illustrate the complexity of influences of antecedentmoisture conditions and catchment morphology on spatial patterns of runoff generation in headwatersystems.
� 2009 Elsevier B.V. All rights reserved.
Introduction
Empirical studies of runoff generation have investigated runoffmechanisms, contributing sources and active flowpaths withinheadwater catchments but our understanding of how hydrologicprocesses vary and/or aggregate in space within headwater sys-tems remains poor. A principal weakness in our abilities to addressthe issues of scale, spatial variability and spatial aggregation ofhydrologic response is the absence of spatially distributed datasetsneeded to test existing hillslope and watershed models (Bonell,1998), identifying a clear need for field based studies. The multi-scale studies of runoff generation that do exist have begun to ad-
ll rights reserved.
ent of Forestry and Environ-mpus Box 8008, Raleigh, NC193.mes), [email protected]
dress the spatially limited observational record (e.g. Mulhollandet al., 1990; Sidle et al., 1995; Brown et al., 1999; Shanley et al.,2002; McGlynn et al., 2004; Inamdar and Mitchell, 2007) but com-parisons among studies and generalizations from these studies re-mains challenging due to variation in conditions of the observationrecords such antecedent moisture conditions (AMCs), storm sizeand catchment morphology. In this paper we report on the resultsof a nested catchment study in the cool temperate region of NorthAmerica where we examined AMCs, catchment morphology andstorm response.
Process-based studies at both hillslope and catchment scaleshave long identified antecedent moisture conditions and stormsize as influential controls on runoff generation resulting in nonlin-ear streamflow response (e.g. Wipkey and Kirkby, 1978). Recentstudies have highlighted a threshold response of hillslopes withstorm size (e.g. Tromp-van Meerveld and McDonnell, 2005; Uchidaet al., 2005) and antecedent moisture conditions (e.g. Buttle et al.,2001; Buttle and Turcotte, 1999; Buttle and Peters, 1997).Similarly, catchments scale studies have shown the evidence of
352 A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366
threshold moisture storage (Sidle et al., 1995) before whichephemeral catchments will contribute runoff. Observations ofthreshold response with AMCs have led to hypotheses of varyingstates of catchment wetness (Grayson et al., 1997) (wet versusdry) attributed to changing hydrologic connectivity, defined byStieglitz et al., 2003 as the active connection of moving water,and may be a catchment-scale example of nonlinearity due toswitching of controlling processes (Dooge, 2004).
Direct observation and inferential evidence including new/oldisotope hydrograph separation (IHS) suggest changing dominantdelivery mechanisms across the AMC threshold. Under wet condi-tions, such as those observed during periods of snowmelt (Wadd-ington et al., 1993; Shanley et al., 2002) or within continuouslywet rainforest catchments (Pearce et al., 1986; McDonnell, 1990)old water contributions dominate the hydrograph, suggesting deepflowpaths are involved (Brown et al., 1999). Shanley et al. (2002)hypothesized that under dry conditions high storage deficits re-duce catchment-scale transmissivity and may promote greaterinfiltration of rainfall, resulting in less new-water delivery to thestream channel, also suggesting that deep flowpaths may continueto dominate. But field studies have shown that shallow saturatedsubsurface storm flow (SSSF) and saturated overland flow (SOLF)occurs under dry conditions within humid regions, possibly dueto hydrophobicity of organic soils (Biron et al., 1999; Buttle andTurcotte, 1999; Buttle et al., 2005), reduced permeability at depthor storm intensity. Both Sklash and Farvolden (1979) and Buttleand Turcotte (1999) observe overland flow on forested soils in hu-mid regions given very intense storms on dry basins. Sklash andFarvolden (1979) found both the hydrograph and the overland flowcomponent were composed predominantly of event water. Simi-larly, Brown et al. (1999) observed quick activation of surfaceand SSSF under dry conditions, with high instantaneous peaks ofnew water.
In reviewing the varied findings in scaling patterns of new/oldwater delivery, both Buttle (2005) and McGlynn et al. (2004) reportno consistent trend with catchment size. Shanley et al. (2002)hypothesize that new-water scaling patterns will vary as a func-tion of the combined effects of topography, AMCs (including sub-surface storage deficits), differences in soil compaction orproperties, and the presence of groundfrost during snowmeltevents. Assuming a trend in basin morphology in which largercatchments collect larger areas of lower slope (larger valley bot-toms), promoting greater saturated overland flow (SOLF), they pro-posed a topographic scale effect (based on the topographic wetnessindex ln (a/tan b) of Beven and Kirkby, 1979) in which percent newwater would be expected to increase with catchment size. .Theyproposed that dryer AMCs (with high storage deficits) would pro-mote more infiltration of new water with less traveling to thestream channel (for a given storm event), and that greater infiltra-tion would occur in catchments with more transmissive underly-ing till, resulting in greater old water displacement during astorm event. As a result, spatial patterns of percent new-waterdelivery would coincide with spatial patterns in transmissivity ofcatchment tills (higher new-water delivery from catchments withthe least transmissive till). Under wet conditions, they suggestedany differences in till transmissivity may matter less since flowthrough the underlying till is less important in controlling stream-flow, with high water tables promoting flow in the more transmis-sive and shallower part of the soil profile (Kendall et al., 1999), andas a results more uniform new-water delivery across scale.
In this paper, we quantitatively examine of the spatial patternsof storm runoff generation as a function of AMCs and catchmentmorphology within the forested headwater system of MontSaint-Hilaire (MSH), Quebec. Our objectives are to: (1) quantifyand compare hydrometric response from the eight catchmentsacross the nonlinear change in AMCs, (2) integrate hydrometric,
hydrochemical and isotopic response to provide event-based inter-pretation of runoff generation for five individual storm events, and(3) contribute an additional case study to the current discussion onspatial variability and scale dependence of runoff generation andnew/old water delivery.
Site description
This study was conducted within the Westcreek watershed ofthe Mont Saint-Hilaire (MSH) UNESCO Biosphere reserve, a maturebeech-maple forest, located in southern Quebec (Lat: 45�3204900N,Long: 73�1000700W). Eight sub-catchments are defined, ranging insize from 0.07 to 1.47 km2. Hillslope soils are well drained andrange in depth from �0 to 1.5 m. In the valley bottoms soils canbe >2 m (i.e. depth unknown), and a low-permeability layer offragipan (Dingman, 1994) within 30–50 cm of the surface can belocally present. Precipitation is relatively uniform throughout theyear (�80 mm/month), but the regional climate of eastern Quebechas a large seasonality with daily mean temperature for Januaryand July of �10.3 �C and 20.8 �C, respectively. The region receivesan average annual precipitation of 940 mm, 22% of which comesin the form of snow (Environment Canada, 2009). As a result, ante-cedent moisture conditions are wettest in the spring due to melt.During wet AMCs, valley-bottom areas show extensive areas of sat-uration, a function of high water tables and perched water tablespromoted by the localized shallow fragipan. As the growing seasonprogresses, storage is depleted due to stream flow and evapotran-spirative demand, and a strong relative change in AMCs can result,especially during summers with little rain. Under these increas-ingly dry conditions, saturated areas shrink, some disappearingaltogether. During storm events, transient saturation can occur inthe valley-bottoms on the fragipan, potentially a function of bothshallow-subsurface stormflow and impeded infiltration at thedepth of the fragipan, connecting and delivering water to thestream channel. End-member-mixing-analysis of stream waterchemistry (James and Roulet, 2006) identified perched water,groundwater and throughfall as important contributing end-mem-bers to streamflow where perched water is defined as the shallowsubsurface and/or surface water that develops in the valley-bottompositions on both seasonally and storm-based time-scales. Furtherdescription of MSH can be found in James and Roulet (2006, 2007).
Methods
Catchment delineation and characterization
Eight nested catchments located within the Westcreek wa-tershed at MSH (Fig. 1) were delineated using standard GIS soft-ware and a high-resolution (1 m � 1 m) digital elevation model(DEM). The DEM was created from a LIDAR dataset of MSH flownwith an Optech ALTM 2050, 50 kHz (50,000 pulses/s) system inearly spring 2003. Table 1 shows catchment characteristics gener-ated from DEM analysis for each gauged catchment including totaland mean sub-catchment area, fractions of hillslope and valley-bottom area, mean slope, minimum and maximum elevation andmean topographic wetness index (Beven and Kirkby, 1979). Basedon the approach of McGlynn et al. (2004) mean sub-catchment sizewas calculated as the mean of the distribution of upslope contrib-uting area for each pixel of the stream channel in a given catch-ment (Fig. 2). Accumulated area for each pixel upon which thestream network is based was calculated using a single-directionflow algorithm based on the algorithm of Jenson and Domingue(1988). Appropriate threshold area for the initiation of the streamnetwork was determined by comparing the DEM-generated streamnetwork with field information collected from the Aw and Vc
Fig. 1. Eight nested catchments at Mont Saint-Hilaire (MSH), Quebec. A perennial spring is located immediately above the Sc gauging station, and downstream of the YVgauging station. Transect 1 located within the Aw catchment includes riparian water table wells (B–E) used as surrogates of AMCs (altered from James and Roulet, 2006).
Table 1Catchment characteristics. Topographic-based metrics derived from the 1 m � 1 m resolution LIDAR-derived DEM. See explanation in text for definition of mean sub-catchmentarea, hillslope and valley-bottom areas.
Catchment Totalarea (ha)
Meansub-catchmentarea (ha)
Averageslope (%)
Elevationrange (m)
Mean topographicwetness Index
Valley-bottomarea (ha)
Dynamicstorage Vd
b
(mm)
Max baseflowQo
a (mm/d)Time constantab (1/d)
Lk 147 22.8 15.9 237 3.71 38.7 41 1.47 0.04Ef 91 24.0 16.2 231 3.69 21.5 54 1.91 0.04Pw 48 15.9 15.7 235 3.72 15.3 50 1.93 0.04Sc 38 12.5 18.2 224 3.58 7.3 24 2.27 0.10Yv 30 14.8 19.5 220 3.54 5.2 9 1.45 0.16Vc 11 4.1 14.7 202 3.79 2.4 19 1.10 0.06Aw 11 6.9 23.1 158 3.39 1.2 11 1.96 0.17Sb 7 4.2 13.5 117 3.73 1.8 1 0.37 0.43
a Maximum baseflow estimated from streamflow records.b Dynamic storage capacity Vd, and time constant a, estimated from recession analysis.
A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366 353
catchments. Total hillslope and valley-bottom area were estimatedusing a simple slope-based criterion consistent with detailed fieldobservations. We define valley bottom by the break in slope of 8�.
Streamflow
During the 2001 and 2002 field seasons streamflow was moni-tored from the eight nested catchments using a series of V-notchweirs and existing culverts (Table 2). Stage readings, recorded on15 min intervals, were converted to discharge using rating curvesdeveloped for each station. A wide and shallow streambed pre-vented weir installation or electronic monitoring for the Ef catch-ment (91 ha), located immediately above the confluence of theeast and west branches of the Westcreek stream system (Fig. 1).As a result, discharge for this station was calculated by subtractingdischarge from the west branch, measured immediately upstreamat the Pw station (53 ha) from the Lk station (147 ha) located
approximately 20 m downstream. Runoff ratios (total event dis-charge/throughfall) were calculated for individual storm hydro-graphs using total event discharge (no graphical hydrographseparation) where the start and end of hydrograph rising andrecession limbs, respectively were identified by changes in the firstderivative (dQ/dt) or slope of the hydrograph. In the case of therecession limb, end time is defined with a dQ/dt of zero or a valueestimated as indicating minimal change. Timing of response for theeight catchments including lag times of initial and peak streamflowand d18O tracer response were calculated based on time since startof throughfall (time zero). Resolution of lag times is based on a 15-min sampling interval.
Streamflow records were also used to estimate catchment-scaledynamic storage using recession analysis (Vitvar et al., 2002). Inthis approach, baseflow recession periods observed between stormevents are modelled using an exponential relationship,
Qt ¼ Q o � e�at ð1Þ
(a) (b)
Fig. 2. Comparison of cumulative frequency distributions of sub-catchment area for individual catchments: (a) sub-catchment area in units of hectares (ha) and (b)normalized with respect to total catchment area. Median (50th percentile) sub-catchment area for individual catchments is indicated at the intersection of frequencydistributions and the 0.5 vertical line.
Table 2Catchment gauging station ID, catchment size, and station discharge monitoringequipment.
Catchment (size) Equipment
Lk (147 ha) 1.0 m diam. culvert; Keller 173-L pressure transducer(0–2.5 psi range)
Ef (91 ha) n/aa
Pw (53 ha) 90 V-notch weir; stilling well with potentiometerSc (38 ha) 0.8 m cm diameter culvert; Keller 173-L pressure
transducer (0–2.5 psi range)Yv (30 ha) 90 V-notch weir; stilling well with potentiometerVc (11 ha) 0.5 m culvert; potentiometerAw (11 ha) 53.8 V-notch weir; stilling well with potentiometerSb (7 ha) 53.8 V-notch weir; stilling well with potentiometer
a Discharge at the Ef catchment is estimated by the difference between Pw and Lkcatchment discharge.
354 A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366
where Qt is discharge (m3 s�1) at time t, Qo is the maximum base-flow, and a is the baseflow recession coefficient (day�1). Time to fullrecession is estimated using an approximation of minimum base-flow and solving for t in Eq. (1). Catchment dynamic storage capac-ity Vd is then estimate as
Vd ¼Q o
a: ð2Þ
For each catchment, maximum baseflow was estimated duringwet AMCs in late spring. Full baseflow recession was approximatedusing the lowest discharge observed under driest AMCs, late in thegrowing season. For the three ephemeral catchments (Sb, Vc, Yv) aminimum baseflow of 0.01 l/s was used to approximate fullrecession.
Antecedent moisture conditions (AMCs)
AMCs were quantified by three different methods. An anteced-ent precipitation index (API7), generated by recorded throughfallfor the 7-day period prior to storm events was calculated fromon-site tipping bucket gauges (15-min interval) located in theAw and Lk catchments. Tipping bucket gauges were calibratedagainst manual throughfall measurements. In two of the eightnested catchments (Aw and Vc catchments) local water table ele-vations were monitored for seasonal and storm-based changes in
catchment storage. Here, riparian water table elevations and meanshallow soil moisture collected from spatial surveys within the Aw(11 ha) catchment (Fig. 1) are used as additional surrogates ofchanging AMCs throughout the eight nested catchments. Addi-tional information on water table elevation data and soil moisturesurveys can be found in James and Roulet (2007).
Geochemical and isotopic tracers
Streamwater from the eight catchments (including baseflowand stormflow) and potential contributing end-members weresampled for d18O, DOC, electrical conductivity (EC) and a suite ofgeochemical tracers (see James and Roulet, 2006, for discussionof end-member identification). Throughfall was collected from 15manual collectors distributed throughout the 147 ha watershed.Samples were collected in 250 ml HDPE bottles and 20 ml HDPEbottles with cone shaped plastic liner for isotopic analysis. Eachbottle was rinsed with sample water three times prior to samplecollection. The 250 ml samples were filtered using 0.45 lm cellu-lose acetate syringe filters and separated into vials for cation andanion analysis. Cation samples were acidified to 2–3 pH and refrig-erated at 4 �C until ion exchange chromatography analysis wasperformed; anion samples were frozen until analysis. The remain-ing unfiltered sample was set aside for electrical conductivity mea-surements. EC measurements were made at 25 �C, allowingsamples to equilibrate to controlled laboratory room temperature.
Samples were analyzed for a suite of anions and cations, dis-solved organic carbon (DOC), electrical conductivity (EC) andd18O. DOC was measured using a Shimadzu TOC 5050 organic car-bon analyzer. Ion exchange chromatography (Dionex DX-500) wasperformed at University of Toronto, Mississauga GeographyDepartment. Isotopic analysis of d18O was performed by the Uni-versity of Waterloo’s Environmental Isotope Lab. Blind samplesprovided a mean repeatability of d18O compositions of ±0.09‰.
Isotopic hydrograph separation (IHS)
Using d18O compositions, IHS (Pinder and Jones, 1969; Sklashand Farvolden, 1979) was performed for each storm hydrographto estimate contributions of event or new (fn) and pre-event orold (fo) water contributions. The composition of new-water deliv-ered during each storm was calculated from volume weighted
A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366 355
average of throughfall compositions collected at the 15 manualcollectors. Standard deviation in new water composition due tospatial variability was within 0.16‰ with the exception of oneevent where standard deviation was 0.42‰. Old water compositionwas defined by an average of baseflow collected from each gaugingstation and the natural spring prior to each storm event. Spatialvariability in old water composition due to spatial variability waswithin 0.25‰.
For each storm we estimate uncertainty in new water contribu-tions due to uncertainty in instantaneous runoff (WQ) and isotopiccompositions of new (WCn) and old water (WCo). A confidenceinterval for instantaneous runoff is estimated from ±1 standarddeviations of the best-fit rating curve coefficients. At high dis-charge, uncertainty in runoff (WQ) is largest. For each sampling ofstream water, uncertainty in the fraction of old or new water(uncertainties are the same for both old and new water contribu-tions) due to the isotopic composition of old, new and streamwater was estimated using the root-mean-square-error formula-tion following Genereux (1998), and expressed here as uncertaintyin new water (Wfn):
Wfn¼f0
ðCo�CnÞWCo
� �2
þ fn
ðCo�CnÞWCn
� �2
þ 1ðCo�CnÞ
WCs
� �2( )1=2
ð3Þ
where C refers to the isotopic signature and subscripts o, n, and s re-fer to old, new and stream water, respectively. The two sources ofuncertainty are then propagated using the same formulation for aproduct of two independent variables to give an estimate of uncer-tainty in new water contributions (Lyons, 1991). Instantaneous newwater runoff (Xn) is the product of instantaneous total runoff (Q)and the fraction of new water (fn):
Xn ¼ Qfn ð4Þ
where uncertainty in new water runoff (WXn) is estimated by:
WXn ¼ XnWQ
Q
� �2
þ Wfn
fn
� �2" #1=2
ð5Þ
Table 3Storm event characterization.
Storm No. 5 85-June-02 2
Magnitude (mm) 5.5 1Duration (h) 16.6Avg intensity (mm per 15 min interval) 0.08Max intensity (mm per 15 min interval) 0.8API7
a (mm) 65 1A-priori baseflow Aw (11 ha) catchment (mm/day) 1.75Mean shallow soil moisture (vol/vol) 0.24Aw (11 ha) riparian water table elevation (m) well D �0.33 �
a Antecedent precipitation index defined here by throughfall records from the previou
Table 4d18O isotopic compositions (per mil) of new/event and old/pre-event waters.
5-June-02 23-June-02(5) (8)
New watera �4.26 �6.90(Spatial std. deviation) (0.13) (0.14)Old waterb �11.87 �11.90(Spatial std. deviation) (0.17) (0.18)New–old difference 7.61 5.00
a Volume weighted average of spatial samples with standard deviations of compositiob Average of all gaugings stations and natural spring prior to storm with spatial stand
Finally, uncertainty in total new-water delivered during the stormis estimated by summing the errors for each instantaneous mea-surement during the storm hydrograph.
Three-component-mixing-model
We apply a three-component mixing model to estimate the rel-ative contributions of throughfall, a perched water or shallow sub-surface flow component and groundwater for each individualstorm event. In addition to d18O, DOC is used here as a tracer dueto its association with shallow subsurface flowpaths, particularlyunder dry AMC (Brown et al., 1999). Independent field samplingof end-members was performed to determine characteristic end-member DOC concentrations and d18O signatures.
Results
Storm selection
From the full hydrometric record of nine recorded storms, fivewere selected for detailed analysis on the basis of storm size,AMCs, completeness of observational records and d18O source sep-aration between event and pre-event waters (Table 3). Storms var-ied in origin (frontal versus convective), ranged in size from 5 to38 mm and delivered maximum intensities of 0.8–12.2 mm/15 min. API7, values ranged from 65 mm to 0 mm. Surrogate ripar-ian water table elevations ranged from depths of �0.33–0.88 m(well D) below ground surface and mean shallow soil moisture var-ied from 0.24 to 0.11 vol/vol. For each of these five storms, a largeisotopic difference between event and pre-event water signatureswas observed, ranging from 5.00‰ to 7.61‰ (Table 4).
Runoff ratios
Runoff ratios of the full hydrometric record (nine storms),including the five listed in Table 3, provide evidence of a thresh-old-like change in hydrologic response over a small change inAMCs (Fig. 3), shown here by plotting runoff ratios by a priori mean
10 11 13-June-02 17-July-02 3-September-02 16-June-01
4.1 38.11 7.03 25.161.2 2.4 2.9 10.32.0 3.8 0.6 0.66.6 6.3 2.6 12.25 5 0 12.57 0.28 0.35 0.070.24 0.20 0.11 0.230.29 �0.81 �0.88 �0.76
s 7-day period.
17-July-02 3-September-02 16-June-01(10) (11) (1)
�5.73 �5.47 �5.36(0.16) (0.42) (0.12)�11.75 �11.99 �12.96
(0.25) (0.17) (0.17)6.02 6.52 7.60
ns included in brackets.ard deviation in brackets.
(a) (b)
(c)
Fig. 3. Change in runoff ratio as a function of AMCs (a and b) and both AMC and storm size (c). Storms 1, 5, 8, 10 and 11 are identified. AMCs are quantified by (a) meanshallow soil moisture content from snap-shot spatial surveys (James and Roulet, 2007), and (b) various riparian water table elevations. In plot (c) riparian well D (see also b) isused to represent AMCs.
356 A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366
soil moisture (3a) and a priori riparian groundwater elevations(3b). Fig. 3c illustrates a krigged surface highlighting the changesin runoff ratios as a function of both storm size and AMCs. Wetconditions produced significantly larger runoff ratios than dry con-ditions. Storm 5 (5.5 mm) (Table 5a) produced the largest runoffratios of 20–59% and storm 8 (14 mm) (Table 5b) generated runoffratios of 3–15%. On dry conditions, the two largest storms (10 and1) (Table 5c and d) produced much smaller runoff ratios, rangingfrom 0.2% to �2.8%, and from 0% to 2.1%, respectively. Storm 11(7 mm) (Table 5e) produced the smallest runoff ratios (0–1.4%).Note that three of the smaller catchments (Vc, Sb and Yv) areephemeral, generating no baseflow during the dryest AMCs.
Using the uncertainty analysis to inform the absence or pres-ence of significant scaling trends (i.e. reporting correlations onlyif uncertainties indicate different values across multiple catch-ments as shown in Fig. 7), correlations between runoff ratio and se-lect catchment characteristics (Table 1) can be observed for thetwo large storms on dry AMCs. Storm 10 (dry) shows distinctlyhigher runoff ratios from the three largest catchments (Lk, Ef,and Pw), with increased runoff ratio with total area (r2 = 0.56,p = 0.03), sub-catchment area (r2 = 0.73, p = 0.007) and dynamicstorage (r2 = 0.84, p = 0.001). For storm 1, there is a strong correla-tion with total area (r2 = 0.90, p = 0.014), mean sub-catchment area
(r2 = 0.85 p = 0.026), and dynamic storage capacity (r2 = 0.97p = 0.002), although it must be noted that only five of the eightcatchments had complete data for this storm (Table 5d).
New/event water
For each of the five storms on which IHS was performed, event/new water exhibited a heavier (less negative) d18O signature thanbaseflow resulting in strong dilution of stream water d18O compo-sitions, illustrated here for storms 8 (14 mm on wet AMCs) and 10(38 mm on dry AMCs) in Figs. 4 and 5, respectively. Instantaneousrunoff separated into new and old water contributions is illus-trated in Fig. 6 for storm 10 at the Lk (147 ha) catchment. Panel6a shows the rating curve for the Lk catchment with a 68% (±1r)confidence interval. The corresponding uncertainty in the instanta-neous runoff is illustrated in panel 6b. Uncertainty in new watercomposition (6b) results from propagation of uncertainty fromboth runoff and isotopic compositions. Integrated through time,we estimate the uncertainty in total new water contributions foreach catchment for each storm, allowing a comparison of newwater contributions across all catchments. Table 5 summarizesuncertainties in total runoff and new water contributions (mmand %) for each storm and catchment.
Table 5Storm runoff response, new water contributions and three-component-mixing-model relative contributions.
Catchment Area (ha) Total runoff (mm)a Runoff ratio (%) New water (mm)b New Water (%) Ground water (%) Perched water (%) Thfall (%)
(a) Small storm (storm 5)c on wet antecedent moisture conditions (AMCs)Lk 147 3.24 ± 0.41 58.9 ± 7.5 0.09 ± 0.043 3 ± 1 99 0 1Ef 91 – – – – – – –Pw 48 – – – – – – –Sc 38 2.02 ± 0.61 36.7 ± 11.2 0.05 ± 0.045 3 ± 2 98 1 0Yv 30 1.81 ± 0.03 32.9 ± 4.7 0.07 ± 0.033 4 ± 2 97 2 1Vcd 11 1.63 ± 0.22 29.6 ± 4.0 0.10 ± 0.035 6 ± 2 97 4 0Aw 11 3.40 ± 0.33 61.8 ± 6.0 0.05 ± 0.046 2 ± 2 98 1 1Sb 7 1.11 ± 0.15 20.2 ± 2.8 0.09 ± 0.025 8 ± 2 99 1 0
(b) Medium storm (storm 8) on wet antecedent moisture conditionsLk 147 2.27 ± 0.28 15.4 ± 2.1 0.22 ± 0.08 10 ± 4 79 14 6Ef 91 2.08 ± 0.51 14.8 ± 3.7 0.29 ± 0.09 9 ± 5 79 15 6Pw 48 1.85 ± 0.40 13.1 ± 2.9 0.28 ± 0.09 15 ± 6 73 17 9SC 38 0.34 ± 0.10 2.4 ± 0.8 0.07 ± 0.02 21 ± 9 83 4 13Yv 30 1.38 ± 0.02 9.8 ± 0.5 0.19 ± 0.05 14 ± 4 83 5 12Vc 11 0.39 ± 0.05 2.8 ± 0.4 0.11 ± 0.02 27 ± 6 36 57 7Aw 11 1.64 ± 0.16 11.6 ± 1.3 0.11 ± 0.06 7 ± 4 82 14 4Sb 7 0.52 ± 0.07 3.7 ± 0.5 0.12 ± 0.02 23 ± 6 81 7 12
(c) Large storm (storm 10) on intermediate antecedent moisture conditionsLk 147 0.92 ± 0.12 2.4 ± 0.3 0.35 ± 0.05 39 ± 8 56 24 20Ef 91 0.98 ± 0.23 2.6 ± 0.6 0.31 ± 0.09 31 ± 12 67 19 14Pw 48 1.07 ± 0.23 2.8 ± 0.6 0.49 ± 0.11 46 ± 15 47 28 26Sc 38 0.23 ± 0.07 0.60 ± 0.18 0.09 ± 0.03 38 ± 16 64 3 33Yv 30 0.44 ± 0.006 1.20 ± 0.02 0.18 ± 0.01 41 ± 3 60 5 34Vc 11 0.24 ± 0.03 0.60 ± 0.08 0.10 ± 0.01 44 ± 9 53 30 16Aw 11 0.30 ± 0.03 0.80 ± 0.08 0.13 ± 0.02 44 ± 7 45 31 25Sb 7 0.06 ± 0.008 0.20 ± 0.02 0.04 ± 0.007 76 ± 16 22 10 68
(d) Large storm (storm 1)e on dry antecedent moisture conditionsLk 147 0.52 ± 0.066 2.1 ± 0.26 0.13 ±.019 24 ± 5 70 22 8Ef 91 – – – – – –Pw 48 – – – – – –Sc 38 0.28 ± 0.085 1.1 ± 0.34 0.07 ± 0.022 25 ± 11 68 13 20Yv 30 0.13 ± 0.002 0.50 ± 0.01 0.04 ± 0.002 33 ± 2 71 4 25Vc 11 – – – – – – –Aw 11 0.06 ± 0.006 0.20 ± 0.02 0.02 ± 0.003 34 ± 6 60 15 25Sb 7 0 0 0 0 0 0 0
(e) Small storm (storm 11) on dry antecedent moisture conditionsLk 147 0.091 ± 0.011 1.3 ± 0.16 0.008 ± 0.005 9 ± 6 83 10 6Ef 91 0.10 ± 0.019 1.4 ± 0.28 0.012 ± 0.007 12 ± 7 91 3 6Pw 48 0.035 ± 0.008 0.5 ± 0.11 0.004 ± 0.001 11 ± 4 84 8 8Sc 38 0.048 ± 0.015 0.7 ± 0.21 0.007 ± 0.003 15 ± 8 79 7 14Yv 30 0 0 0 0 0 0 0Vc 11 0 0 0 0 0 0 0Aw 11 0.061 ± 0.006 0.9 ± 0.08 0.009 ± 0.004 15 ± 6 75 17 8Sb 7 0 0 0 0 0 0 0
a Error associated with uncertainty in power model coefficients.b Error propagated from runoff and isotopic compositions. Error associated with isotopic compositions alone (Genereux, 1998) ranges between 1% and 6%.c Ef, Pw have incomplete data.d Vc has a second peak at 24.75 h.e Ef, Pw and Vc have incomplete data.
A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366 357
Analysis of the five storms shows a wide range of percent newwater contributions with AMCs and storm events but given uncer-tainties, there are only a few examples of quantifiable differencesacross catchments (Table 1, Fig. 7). Storm 5 (Table 5a), the smalllow intensity storm on wet AMCs, generated similar percent con-tributions of new water across the catchments with only the small-est catchment (Sb, 7 ha) delivering a distinctly higher percentcontribution (8%) from all but the Yv and Vc catchments. Storm 8(Table 5b), a 14 mm moderate intensity storm on wet AMCs pro-duced percent new water contributions ranging from 7% to 27%.Two of the smaller catchments, Sb and Vc, delivered distinctly lar-ger percent new water than some of the largest catchments (e.g. Lk,Ef, Yv). Storm 10 (Table 5c), the largest and on average, most in-tense storm, occurring on dry AMCs (see metrics, Table 3), pro-duced the highest percent new water observed for the fivestorms (31–76%). Prior to this storm ephemeral catchments Yv,Vc and Sb were dry, a further indication of depleted storage condi-
tions. New water percent from the smallest and ephemeral Sb(7 ha) catchment is distinctly higher than for any other catchment(76%). Storm 1 (Table 5d), a large, low intensity storm occurringwith dry AMCs during the previous summer (16-June-01) gener-ated new water contributions ranging from 0% to 34% with dis-tinctly larger fraction from the smaller Aw catchment (34%) thanfrom the largest Lk (24%) catchment. For storm 11 (Table 5e), asmall storm on dry AMCs occurring in early fall of 2002, percentnew water is indistinguishable for the 5 flowing catchments (9–15%) given uncertainties and the three ephemeral catchments(Yv, Vc, Sb) delivered no runoff.
There is significant correlation between percent new water con-tributions and storm size (r2 = 0.472, p < 0.001) indicating a trendtowards higher percent new water with larger storms, regardlessof AMCs. Although there are no identifiable trends in percentnew water contributions given uncertainties, there are a numberof correlations between new water (mm) and catchment charac-
Fig. 4. Storm 8 (14 mm on wet AMCs) hydrographs and d18O stream watercompositions. Throughfall (mm) is shown in the top panel. Total runoff (mm) isillustrated on the left hand axis (open symbols) and d18O stream water composition(‰) on the right hand axis (filled symbols).
Fig. 5. Storm 10 (38 mm on dry conditions) hydrographs and d18O stream watercompositions. Throughfall (mm) is shown in the top panel. Total runoff (mm) isillustrated on the left hand axis (open symbols) and d18O stream water composition(‰) on the right hand axis (filled symbols).
358 A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366
teristics for the two large storms on dry AMCs. For storm 10, dis-tinctly more total new water (mm) is delivered from the largest3 catchments generating correlations with catchment area(r2 = 0.41, p = 0.09), mean sub-catchment area (r2 = 0.54, p = 0.04),and dynamic storage capacity (r2 = 0.72, p = 0.007). Storm 1 showscorrelation with total area (r2 = 0.09, p = 0.014), sub-catchmentarea (r2 = 0.89 and p = 0.017) and dynamic storage capacity(r2 = 0.96, p = 0.003). For storm 1 it should be noted that three ofthe catchments are excluded from analysis due to incomplete data(Table 5d).
Timing of catchment response
Fig. 8 illustrates lag time to initial and peak streamflow re-sponse (left hand panels) and a comparison of peak streamflowand d18O response (right hand panels). For the five storms, thereare no consistent trends in lag times of initial or peak streamflowwith catchment characteristics, as timing of response shows con-
siderably variability not explained by basic topographic metrics.For instance, during storms 5, 8 and 10, the Yv (30 ha) catchmentis quickest to reach peak streamflow, while the smaller Aw(11 ha) catchment responds with similar lag times of the largestcatchments. As an example of additional complexity in response,the Yv catchment exhibits a strong double peak (shown in Fig. 5,third panel from top) not observed elsewhere. This second peakis likely a result of its downstream location from the Aw catch-ment, where after dropping over a small waterfall just below theAw gauging station, the stream travels a long elongated valley be-tween the two gauging stations (see Fig. 1). However, under dryAMCs, lag times can show some overall correlations with catch-ment metrics. Storms 10 and 1 (Fig. 8, left hand panels) show in-crease in lag time of peak streamflow with catchment total area(r2 = 0.47 p = 0.059; r2 = 0.61, p = 0.067, respectively). Under thesedry conditions long initial lag times and even no response areexhibited by ephemeral catchments; the previously dry ephemeralSb catchment initiates flow only after 1 h 30 min during storm 10,and produces no flow during storms 1 and 11.
Stage (mm)
Dis
char
ge (
l/s)
Day of Year
Inst
anta
neou
s D
isch
arge
(m
m/h
r)
0
20
40
60
80
100
120
0 50 100 150 200 250
Data
Mean
upper confidence interval
lower confidence interval
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
198.7 198.9 199.1 199.3
0
5
10
15
20
25
30
35
40
Total Runoff
New Water
ThFall
(a) (b)
Fig. 6. Quantifying runoff and new water compositions for storm 10 (38 mm on dry AMCs) at the Lk (147 ha) catchment. Panel (a) shows the rating curve for the Lk catchmentwith a 68% (±1r) confidence interval. The corresponding uncertainty in the instantaneous runoff, propagated from both instantaneous runoff measurements and isotopiccompositions, is illustrated in panel 5b.
A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366 359
Timing of new-water delivery is significantly correlated withcatchment metrics (total area and mean sub-catchment area,respectively) for one storm 8 on wet AMCs (r2 = 0.62 p = 0.02;r2 = 0.69, p = 0.011) and storm 1 on dry AMCs (r2 = 0.61 p = 0.07;r2 = 0.63, p = 0.06). But again, there is considerably inter-catchmentvariability. For example, for storm 8, the Sc (38 ha) catchmentshows quicker new water peaks than the Yv (30 ha) catchment.For this storm, Yv shows the largest lag between catchment spe-cific Q and new water (�1 h after peak Q), potentially due to traveltime of water from the upstream Aw catchment (see also storm10). During storm 5, lag time to peak in d18O tracer is much longerfor the smaller Vc and Sb catchments (�30 h compared to �15 hfor the largest catchments), suggesting transport limited deliveryof new water due to longer travel times to the gauging stationsfrom these relatively lower-relief catchments (Table 1). In contrast,for the high-energy Aw catchment, d18O peaks the fastest andapproximately 4 h before Q. This trend for the Aw catchment is alsoreproduced for storm 8, where d18O peaks before Q, in this instanceby �30 min.
Dissolved organic carbon (DOC) and three-component-mixing-analysis
The magnitude and time-evolution of DOC concentrations in thestream channel during a storm provides supporting evidence of ac-tive flowpaths, contributing source waters and their variation withstorm event, AMCs and catchment. Water moving through the or-ganic rich, O-horizon can deliver high concentrations of DOC to thestream channel due to rising water tables (Hornberger et al., 1994)or under dry conditions, rapid lateral movement of water throughmacropores or high permeability O-horizon layer in the shallowsubsurface (Brown et al., 1999). Solute–solute plots of DOC andd18O (Fig. 9) illustrate the time-evolution of stream water boundedby end-members of groundwater, throughfall and perched waterfor each of the five storms.
In the case of storm 5, stream water shows only small changesfrom groundwater, with changes in DOC < 200 leq/l (Fig. 9) withno correlation between DOC and d18O (n = 79). Three-component-mixing-analysis estimates throughfall contributions to be verysmall (0–1%) with larger contributions of perched water (up to�4%) and dominant contributions of groundwater (97–99%) (Table5a). Storm 8 exhibits correlation of DOC and d18O (0.59; p-va-lue < 0.001; n = 81) suggesting a stronger contribution of new-water delivery via a shallow subsurface flowpath from which highDOC and d18O compositions can originate (e.g. Brown et al., 1999).
The temporal evolution of stream water concentrations (Vc catch-ment shown as an example in Fig. 9) are bounded by end-mem-bers, initially most similar to groundwater, progressing towards amixture dominated by groundwater and perched water with somethroughfall, and finally returning towards groundwater. Catch-ments deliver 73–83% groundwater, with the exception of the Vcwhich delivers only 36% groundwater and 57% perched water.Throughfall is a relatively small component, ranging between 4%and 13%.
Storm 10 has the highest correlation of DOC and d18O (0.69;p-value < 0.001; n = 67), Evolution of stream water concentra-tions are shown for Lk and Vc catchments. Estimates of perchedwater contributions range between 3% and 31%. Throughfall isestimated to contribute between 14% and 68%, with the largest68% derived from the ephemeral Sb catchment. Groundwatercontributions range from 22% (Sb catchment) to 67% (Ef catch-ment). Storm 1 shows smaller but still significant correlation be-tween DOC and d18O (0.38; p-value < 0.001; n = 53). There is asimilarly strong departure of stream water concentrations fromgroundwater towards perched water and throughfall end-mem-bers but a greater variability in initial concentrations betweencatchments than observed during storm 10. Groundwater contri-butions range from 60% to 71%. The three smaller catchmentsdeliver distinctly more throughfall (20–25%) than perched watercompared to the largest Lk catchment (8%). Perched water con-tributions range from 4% to 22% with the largest catchmentdelivering the most. For storm 11, there is a low but significantcorrelation between DOC and d18O (0.33; p-value < 0.001; n = 39).Stream water concentrations depart from groundwater to aweaker degree than storms 10 and 1. Groundwater, perchedwater, and throughfall contributions range from 75% to 91%, 3to 10% and 6 to 14%, respectively for those catchments exhibit-ing flow.
Discussion
Hydrometric response
The observational record collected at MSH captures a nonlinearchange in hydrologic response with AMCs, similar to that observedin very different climate and catchment settings. The dramatic in-crease in runoff ratios with wetter AMCs indicates much greaterefficiency in the ability of water to be transported to the streamchannel, i.e. greater hydrologic connectivity for AMCs character-ized by shallower water table depths (e.g. local riparian water table
0.00
0.01
0.02
0.03
0.04
0.05
Lk Ef Pw Sc Yv Vc Aw Sb
0.00
0.01
0.02
0.03
0.04
0.05
Lk Ef Pw Sc Yv Vc Aw Sb
0.0
0.2
0.4
0.6
0.8
1.0
Lk Ef Pw Sc Yv Vc Aw Sb
0.0
0.2
0.4
0.6
0.8
1.0
Lk Ef Pw Sc Yv Vc Aw Sb0.0
0.2
0.4
0.6
0.8
1.0
Lk Ef Pw Sc Yv Vc Aw Sb
0.0
0.2
0.4
0.6
0.8
Lk Ef Pw Sc Yv Vc Aw Sb
Storm 5 (5.5 mm on wet)
New
wat
er (
mm
)
Run
off
ratio
(fr
actio
n)
Storm 8 (14 mm on wet)
Storm 1 (25 mm on dry)
Storm 11 (7 mm on dry)
Storm 10 (38 mm on intermediate)
0.0
0.2
0.4
0.6
0.8
1.0
Lk Ef Pw Sc Yv Vc Aw Sb
New
wat
er (
frac
tion)
New
wat
er (
mm
)
0.0
0.2
0.4
0.6
0.8
Lk Ef Pw Sc Yv Vc Aw Sb0.0
0.2
0.4
0.6
0.8
1.0
Lk Ef Pw Sc Yv Vc Aw Sb
0.0
0.2
0.4
0.6
0.8
1.0
Lk Ef Pw Sc Yv Vc Aw Sb
New
wat
er (
mm
)
0.0
0.2
0.4
0.6
0.8
Lk Ef Pw Sc Yv Vc Aw Sb
New
wat
er (
mm
)
0.00
0.01
0.02
0.03
0.04
0.05
Lk Ef Pw Sc Yv Vc Aw Sb
New
wat
er (
mm
)
WE
T A
MC
DR
Y A
MC
New
wat
er (
frac
tion)
New
wat
er (
frac
tion)
New
wat
er (
frac
tion)
New
wat
er (
frac
tion)
0.0
0.2
0.4
0.6
0.8
Lk Ef Pw Sc Yv Vc Aw Sb
Run
off
ratio
(fr
actio
n)
0.0
0.2
0.4
0.6
0.8
Lk Ef Pw Sc Yv Vc Aw Sb
0.00
0.01
0.02
0.03
0.04
0.05
Lk Ef Pw Sc Yv Vc Aw Sb
Run
off
ratio
(fr
actio
n)R
unof
f ra
tio (
frac
tion)
Run
off
ratio
(fr
actio
n)
Fig. 7. Runoff ratios and new-water delivered (mm and %) by catchment and storm.
360 A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366
was within 0.33 m of the ground surface, higher mean shallow soilmoisture >0.24 vol/vol and higher API7).
The MSH dataset reports on storms ranging in size from 5 to38 mm, a similar range as those reported by Inamdar and Mitchell(2007) (4–24 mm) and Brown et al. (1999) (9–34 mm), both stud-ies carried out in the humid North Eastern United States. However,
climatic normals (1971–2000) for extreme daily rainfall for May toNovember (Montreal PET International Airport) range from 46 to94 mm (Environment Canada, 2009), suggesting that the MSHstudy is missing observations of the largest storms and in particu-lar, observations of large storms on wet AMCs. We would expectthat large storms occurring on wet conditions similar to storm 5
Fig. 8. Lag times (h) for initial and peak response in discharge Q (left hand panels) and peak discharge and d18O signature change (right hand panels). Initial response lag timeis defined as the time between start of storm and change in slope of stream discharge. Zero indicates no flow from ephemeral catchments. N/a indicates incomplete data. Notefor storm 5 the vertical scale (time) is different.
A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366 361
within 1–2 months of snowmelt would produce the highest runoffratios, as observed at the hillslope scale by Buttle et al. (2001), lift-ing the back right corner of the krigged surface interpolated inFig. 3c.
Although quantitative metrics of AMCs are often cited in pro-cess-based studies (see for instance McGlynn et al., 2004; Inamdarand Mitchell, 2007, citing of API7 and antecedent groundwaterindices), it often remains unclear in what ‘state’ (using the termi-
362 A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366
nology of Grayson et al., 1997) the hydrologic system exists or ex-actly how the AMC ‘state’ has been determined – i.e. a quantitative3-D visual similar to Fig. 3c (e.g. Buttle et al., 2001) is not provided,making comparative analysis challenging. For instance, althoughInamdar and Mitchell (2007) observe the influence of AMCs onstorm response, only one of their 10 storms (September 22,
Fig. 9. Stream water solute–solute plots of DOC (leq/l) versus d18O (per mil) for eachtemporal variation for years 2001 or 2002 and spatial uncertainties represented by errvariation observed during the 2001 spring-fall field season).
2003) appears to represent a ‘dry’ state as defined by a thresholdchange in runoff ratio, suggesting that this dry state is underrepre-sented in their analysis.
The hydrometric response observed from the headwater catch-ments at MSH is complex and highly variable. Storms observed onwet AMCs shows strong exceptions to any scaling trends among
storm. Independently measured geographical sources of water are included withor bars (e.g. Spring 01 indicates spring water where error bars represent temporal
A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366 363
the smaller catchments (Table 5, Fig. 7). For instance, one of thehighest runoff ratios (e.g. Aw) is generated from the steepest yetone of the smallest catchments during storm 5 on wet AMCs. How-ever, the results indicate runoff ratios can show trends with catch-ment topographic-based metrics (total area or sub-catchmentarea) for larger storms on dry AMCs. As groundwater stores are de-pleted (dry AMCs), the largest three catchments (Pw, Ef, Lk) canproduce distinctly more runoff (mm or %) than the smaller catch-ments for a given storm (e.g. storm 10, Table 5c). These three larg-est catchments exhibit distinct differences in valley-bottom areaand distributions of sub-catchment areas indicating how upslopecontributing area is accumulated (Fig. 2; Table 1). They also exhibitdistinctly larger dynamic storage volumes as estimated by reces-sion analysis (Table 1), and it was with this catchment characteris-tic for these two storms that runoff ratios was most stronglycorrelated. For the MSH catchments, there is moderate correlationbetween mean dynamic storage capacity and both catchment areaand mean sub-catchment area (r2 = 0.64 p = 0.017, r2 = 0.67,p = 0.013, respectively) that likely contributes to the spatial trendsin runoff generation with topographic metrics observed under dryAMCs. It should be noted that the higher correlation observed forstorm 1 could be affected by the smaller number of catchments in-cluded in the analysis (three catchments excluded due to incom-plete data.
Unlike observations of McGlynn et al. (2004) at the Maimaicatchments, timing of response (either initial or peak) does notconsistently relate to catchment area (or mean sub-catchmentarea) for the varied storm events (Fig. 9). Lag times to peak Q forstorms observed under dry AMCs can exhibit some trend withcatchment total area (storms 10 and 1) but the variability of re-sponse across the MSH catchments makes generalization difficult.Under these conditions (storms 1, 10 and 11), three of the smallercatchments (<50 ha) have minimal to no flow prior to storm re-sponse (depleted groundwater storage) and can exhibit strong lagsto initial Q response (e.g. Sb catchment, storm 10).
Fig. 10. Pre-event contributions from a range of studies with a
Scaling of new/old -water delivery
The percent new/old water delivery observed at the MSH catch-ments spans the entire range of literature values reported by Buttle(2005) (Fig. 10) and reports one of the largest % new-water deliv-ered (76% from the ephemeral Sb catchment) and largest variation(45%) among catchments, recorded here during a summer storm(storm 10) on dry AMCs. Given uncertainties, no trends in percentnew-water delivered were observed with catchment area undereither wet or dry AMCs, similar to results McGlynn et al. (2004)from the Maimai, NZ catchments. However, total amounts of newwater (mm) observed under dry AMCs show some topographic-scaling effect. For three of the storms (8, 10 and 1), distinctly largermagnitudes of new water (mm) (Table 5) are delivered from one ormore of the largest catchments, consistent with a topographic-scaling effect in which increasing valley-bottom area with catch-ment size results in more SOLF and great magnitudes of new-waterdelivered to the stream channel. Unlike the original topographic-scaling effect suggested by Shanley et al. (2002), for each of thesethree storms, it is one or several of the smaller catchments forwhich new water is the larger proportion of runoff, giving the lar-ger percent new water contributions (e.g. storm 10%, 76% from theSb (7 ha) catchment compared to 31% for the Ef (91 ha) catch-ment). In this sense, the coarsest of trends in new-water deliveryat MSH are more similar to those observed by Brown et al.(1999) in which peak percent new water contributions were high-est for the small catchments (11–62%) where groundwater storeswere likely smaller. At MSH, maximum instantaneous new watercontributions were as high as 99% (storm 1, Aw catchment).
Lower catchment transmissivity due to depleted storage insmall catchments may be an explanation for the higher percent-ages of new water. With depleting storage in small catchments,subsurface transmissivity also decreases and the shallow subsur-face flow path activated under high storm intensities (e.g. storm10) overwhelms the effect of the lower transmissivity emphasizing
dditional MSH storm response (altered from Buttle, 2005).
Tabl
e6
Synt
hesi
sof
stor
mev
ent
resp
onse
.
Stor
m5
810
111
Stor
msi
ze/a
vgIn
ten
sity
Smal
l/lo
win
ten
sity
Med
ium
/mod
erat
ein
ten
sity
Larg
e/h
igh
inte
nsi
tyLa
rge/
low
inte
nsi
tySm
all/
low
inte
nsi
ty
AM
CW
etW
etD
ryD
ryD
ry
Res
pon
se(r
un
off
rati
os)
Larg
est
(20–
59%
)M
oder
ate
(3–1
5%)
Smal
l(0
.2–2
.8%
)Sm
all
(0–2
.1%
);1
eph
emer
alca
tch
men
t�
0fl
owSm
alle
st(0
–1.4
%);
3ep
hem
eral
catc
hm
ents�
0fl
ow
Tim
ing
Lon
gla
gti
mes
tobo
thpe
akQ
(>7
h);
peak
new
wat
erla
ggin
gQ
Shor
tla
gti
mes
tope
akQ
(<2
h);
sim
ilar
new
wat
erti
min
g(w
ith
exce
ptio
ns)
Var
iati
onin
init
ial
Qti
min
g;si
mil
arpe
akQ
and
new
wat
erti
min
g
Un
ifor
mti
min
gof
init
ialQ
;sh
ort
lag
tope
akQ
(<2
h);
sim
ilar
new
wat
erti
min
g
Var
iati
onin
init
ial
Qti
min
g;pe
akQ
>2
h;
new
wat
erpe
aks
fast
erth
anQ
New
/old
New
wat
er�
2–8%
;u
nif
orm
insp
ace
(�0.
1m
m)
New
wat
er(7
–27%
);n
osc
alin
gpa
tter
nN
eww
ater
(31–
76%
);sc
alin
gof
new
wat
erN
eww
ater
(24–
33%
);to
po-
scal
ing
ofn
eww
ater
New
wat
er(0
–15%
);n
osc
alin
gpa
tter
n
Sou
rces
GW
(97–
99%
);pe
rch
ed(0
–4%
);Th
fall
(0–1
%)
GW
(36–
83%
);si
zabl
epe
rch
ed(5
–57%
);th
fall
(4–1
3%)
GW
(22–
67%
);pe
rch
ed(3
–31%
);th
fall
(14–
68%
)G
W(6
0–71
%)
for
flow
ing
catc
hm
ents
;si
mil
arpe
rch
ed(4
–22
%)
and
thfa
ll(8
–25%
)
GW
(75–
91%
)fo
rfl
owin
gca
tch
men
ts;
sim
ilar
perc
hed
(3–1
7%)
and
thfa
ll(6
–14%
)
Expl
anat
ion
Hig
hru
nof
f;de
eper
subs
urf
ace
flow
dom
inat
esdu
eto
hig
hex
isti
ng
stor
age
and
smal
lst
orm
Stor
age
depl
etin
g,re
duci
ng
con
nec
tivi
tyin
smal
lca
tch
men
ts;
shal
low
subs
urf
ace
flow
path
and/
orSO
LFac
tiva
ted
Dep
lete
dst
orag
e,re
duci
ng
con
nec
tivi
tyin
smal
lca
tch
men
ts;
shal
low
subs
urf
ace
flow
path
and/
orSO
LFac
tiva
ted
Dep
lete
dst
orag
e,re
duce
dco
nn
ecti
vity
;gr
eate
rin
filt
rati
on,
gen
tler
trig
geri
ngof
shal
low
subs
urf
ace
flow
path
and/
orSO
LF
Dep
lete
dst
orag
e,re
duce
dco
nn
ecti
vity
(eph
emer
alca
tch
men
ts);
grou
ndw
ater
dom
inat
esdu
eto
smal
llo
win
ten
sity
stor
m
364 A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366
the storm-based temporal connectivity created in the riparian areaabove the fragipan.
Trends in timing of new-water delivery across the catchmentsare also variable. Unlike the findings of McGlynn et al. (2004) inwhich the delay in new water peak delivery increased with catch-ment size, lag times to new water peak delivery at MSH show lit-tle relation to catchment size. Lag times of new-water deliverycan be quite different for individual catchments (e.g. storm 5,Vc and Sb catchment). The strong correlation McGlynn et al.(2004) observed between new water lag times and catchmentarea, reflecting changing connectivity across scale, is generallynot seen at MSH. This may result from very different catchmentorganization of MSH as compared to the well-organized Maimaicatchment.
The number of storms for which detailed runoff generationanalysis is available across scale at MSH, although comparableto those reported in other studies (e.g. five storms in Brownet al., 1999; 3 in Shanley et al., 2002; 10 in Inamdar and Mitch-ell, 2007), provides a limited study of the range of spatial patternof response that may exist at MSH under varying AMC condi-tions, storm intensities and sizes. For instance, further observa-tions would be valuable to address if a stronger topographic-scaling effect can be observed for a larger storm on wet AMCsat MSH. This study presents a limited case study of spatial pat-terns in runoff generation and new-water delivery from smallforest catchments.
Explanation of storm events
Explanation of the observed response during these five stormevents under changing AMCs are synthesized in Table 6. Storm 5illustrates response during a small storm on wet ‘state’ AMCs, gen-erating the largest runoff ratios. Groundwater delivered by deepersubsurface flow dominates streamflow generation (97–99%) due tohigh existing storage in the catchments early in the growing seasonwithin several months of snowmelt. In the case of storm 8, a larger,more intense storm delivered on wet conditions, perched/O-hori-zon soil water generated by the activation of a shallow subsurfaceflowpath and/or SOLF can provide significant contributions (5–57%).
As the catchment system shifts into a ‘dry’ state with de-pleted storage across the catchments, we observe the potentialfor an increased dominance of stormflow by shallow processesgenerating perched/O-horizon water and throughfall, particularlyduring large events of variable intensity like storms 10 and 1.Clearly these shallow pathways can be activated during wet con-ditions (e.g. storm 8) but with depleted storage and decreasedhydrologic connectivity in the saturated subsurface, they cancontribute an increasing percentage of total stream flow duringdry conditions, particularly for ephemeral catchments. In thecase of storm 10, groundwater contributions drop to 22–67%.Streamflow generated from the smallest catchment (Sb), ephem-eral with no flow prior to the storm, is 68% throughfall and 10%perched water, consistent with �76% new-water delivery. Stormsof lower volume and intensity suggest greater infiltration with agentler triggering of shallow subsurface/SOLF flowpaths (e.g.storms 1 and 11).
Conclusion
With five of the eight catchments <40 ha in size, the MSH studyclearly shows the highly variable nature of small headwater catch-ment storm response in this physiographic setting. Although thenumber of storms for which detailed runoff generation analysishas been provided is comparable or larger to other reported stud-
A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366 365
ies, generalized patterns are difficult to identify with this limitednumber of events. For the complex setting of MSH study, a largerpopulation of storms, ranging across AMCs, storm size and inten-sity would be helpful to further describe the shifting patterns ofnew-water delivery for the MSH catchment system.
The MSH study provides an examination of spatial patterns inrunoff generation from eight nested catchments observed acrossthe threshold change in AMCs. Analysis of the five storms pre-sented suggest that AMCs (along with storm size and intensity)play an important role in determining contributing runoff mecha-nisms, relative source contributions and observable scaling pat-terns of runoff generation. The MSH study provides the followingfindings:
� For the MSH catchments, relationships between morphology(topographic-based and storage based) and hydrologic response(e.g. runoff ratios) were observed but only for larger storms ondry AMCs. Independent measures of dynamic storage capacityderived from recession analysis provided added support forwhy scaling trends were observed for these catchments. Underthese dry AMCs, the timing of peak flow can also exhibit somerelationship with catchment total area but the variability ofresponse across the MSH catchments makes generalization diffi-cult. It must be noted that large storms on wet AMCs are absentfrom the analysis.
� For sizable storms on dry AMCs, the perched/O-horizon watergenerated by shallow-subsurface stormflow and/or SOLF fromvalley-bottom areas can contribute a significant component ofrunoff as smaller catchments deplete groundwater stores, someto the point of ephemeral stream flow.
� The percent new water generated during the five storms fromthe MSH catchments spans the entire range of literature values.Taking into consideration uncertainties, no scaling in percentnew-water delivery was observed with catchment area undereither wet or dry AMCs. Although no scaling was observed, theresults show that a large storm on dry AMCs, a higher percentnew water is generated from some of the smallest (possiblyephemeral) catchments due to depleted groundwater stores. Inthis sense, results suggest that, smaller catchments can producelarger percent new water than larger catchments, in contrast toresults of Shanley et al. (2002) and more similar to Brown et al.(1999). This results from the combined effects of depleted stor-age and the activating of shallow subsurface flow and/or SOLFdelivering new water to the stream channel.
� The MSH results are not inconsistent with an AMCs-topographicframework as suggested by Shanley et al. (2002). The eightcatchments exhibit morphological differences that contributeto the spatial patterns of runoff generation observed under dryAMCs. Under these conditions, given a sizable storm, we observea distinct topographic-scaling effect in total amount of newwater (mm) delivered consistent with site-specific catchmentmorphology.
Acknowledgements
This work was funded in part by the National Sciences and Engi-neering Research Council of Canada (NSERC), a McGill-McConnellFellowship, a McGill Graduate Studies Fellowship, and the McGillGlobal Environment and Climate Change Centre (GEC3). We wouldlike to thank the staff of the Mont Saint-Hilaire nature research forfield support, Dr Martin Lechowicz and the ECONET project forfunding of the lidar dataset and Benoit Hamel for data processingand DEM creation. We would also like to thank an invaluable groupof field assistants Catie Burlando, Nathan Deustch, Raissa Marksand Sheena Pappa who contributed to storm-event field samplingand processing.
References
Beven, K.J., Kirkby, M.J., 1979. A physically based, variable contributing area modelof basin hydrology. Hydrological Sciences Bulletin 24 (1), 43–68.
Biron, P.M., Roy, A.G., Courschesne, F., Hendershot, W.H., Cote, B., Fyles, J., 1999. Theeffects of antecedent moisture conditions on the relationship of hydrology andhydrochemistry in a small forested watershed. Hydrological Processes 13,1541–1555.
Bonell, M., 1998. Selected challenges in runoff generation research in forests fromthe hillslope to headwater drainage basin scale. Journal of the American WaterResources Association 34 (4), 765–785.
Brown, V.A., McDonnell, J.J., Burns, D.A., Kendall, C., 1999. The role of event water, arapid shallow flow component and catchment size in summer stormflow.Journal of Hydrology 217, 171–190.
Buttle, J.M., 2005. Chapter 116: Isotope Hydrograph Separation of Runoff Sources.In: Anderson, M.G. (Ed.), Encyclopedia of Hydrological Sciences. John Wiley &Sons, Ltd.. 2 p.
Buttle, J.M., Peters, D.L., 1997. Inferring hydrological processes in a temperate basinusing isotopic and geochemical hydrograph separation: a re-evaluation.Hydrological Processes 11, 557–573.
Buttle, J.M., Turcotte, D.S., 1999. Runoff processes on a forested slope on theCanadian Shield. Nordic Hydrology 30, 1–20.
Buttle, J.M., Lister, S.W., Hill, A.R., 2001. Controls on runoff components on aforested slope and implications for N transport. Hydrological Processes 15,1065–1070. doi:10.1002/hyp.450.
Buttle, J.M., Creed, I.F., Moore, R.D., 2005. Advances in Canadian forest hydrology,1999–2003. Hydrological Processes 19, 169–200.
Dingman, S.L., 1994. Physical Hydrology. Prentice-Hall, Upper Saddle River NJ. 575 pp.Dooge, J.C.I., 2004. Bringing it all together. Hydrology and Earth System Sciences
Discussions 1, 41–73 (SRef-ID: 1812-2116/hessd/2004-1-41).Environment Canada, 2009. Canadian Climate Normals 1971–2000. National
Climate Data and Information Archive [Internet] [modified 2009-04-30; cited2009-08-03] <www.climate.weatheroffice.ec.gc.ca>.
Genereux, D., 1998. Quantifying uncertainty in tracer-based hydrographseparations. Water Resources Research 34 (3), 915–919.
Grayson, R.B., Western, A.W., Chiew, F.H.S., Bloëschl, G., 1997. Preferred states inspatial soil moisture patterns: local and non-local controls. Water ResourcesResearch 33 (12), 2897–2908.
Hornberger, G.M., Bencala, K.E., McKnight, D.M., 1994. Hydrological controls ondissolved organic carbon during snowmelt in the Snake River near Montezuma,Colorado. Biogeochemistry 25, 147–165.
Inamdar, S.P., Mitchell, M.J., 2007. Contributions of riparian and hillslope waters tostorm runoff across multiple catchments and storm events in a glaciatedforested watershed. Journal of Hydrology 341, 116–130.
James, A.L., Roulet, N.T., 2006. Investigating the applicability of end-member mixinganalysis (EMMA) across scale: a study of eight small, nested catchments in atemperate forested watershed. Water Resources Research 42, W08434.doi:10.1029/2005WR004419.
James, A.L., Roulet, N.T., 2007. Investigating hydrologic connectivity and itsassociation with threshold change in runoff response in a temperate forestedwatershed. Hydrological Processes 21, 3391–3408. doi:10.1002/hyp. 6554.
Jenson, S.K., Domingue, J.O., 1988. Extracting topographic structure from digitalelevation data for geographic information system analysis. PhotogrammetricEngineering and Remote Sensing 54, 1593–1600.
Kendall, C., Shanley, J.B., McDonnell, J.J., 1999. A hydrometric and geochemicalapproach to test the transmissivity feedback hypothesis during snowmelt.Journal of Hydrology 219, 188–205.
Lyons, L., 1991. A practical guide to data analysis for physical science students.Cambridge University Press. 95p.
McDonnell, J.J., 1990. A rational for old water discharge through macropores in asteep, humid catchment. Water Resources Research 26 (11), 2821–2832.
McGlynn, B.L., McDonnell, J.J., Seibert, J., Kendall, C., 2004. Scale effects onheadwater catchment runoff timing, flow sources, and groundwater-streamflow relations. Water Resources Research 40, W07504. doi:10.1029/2003WR002494.
Mulholland, P.J., Wilson, G.V., Jardine, P.M., 1990. Hydrogeochemical response of aforested watershed to storms: effects of preferential flow along shallow anddeep pathways. Water Resources Research 26 (12), 3021–3036.
Pearce, A.J., Stewart, M.K., Sklash, M.G., 1986. Storm runoff generation in humidheadwater catchments 1. Where does the water come from? Water ResourcesResearch 22 (8), 1263–1272.
Pinder, G.F., Jones, J.F., 1969. Determination of the groundwater component of peakdischarge from chemistry of total runoff water. Water Resources Research 5,438–445.
Shanley, J.B., Kendall, C., Smith, T.E., Wolock, D.W., McDonnell, J.J., 2002. Controls onold and new water contributions to stream flow at some nested catchments inVermont, USA. Hydrological Processes 16, 589–609.
Sidle, R.C., Tsuboyama, Y., Nogushi, S., Hosoda, I., Fujieda, M., Shimizu, T., 1995.Seasonal hydrologic response at various spatial scales in a small forestedcatchment, Hitachi Ohta, Japan. Journal of Hydrology 168, 227–250.
Sklash, M.G., Farvolden, R.N., 1979. The role of groundwater in storm runoff. Journalof Hydrology 43, 45–65.
Stieglitz, M., Shaman, J., McNamara, J., Engel, V., Shanley, J., Kling, G.W., 2003. Anapproach to understanding hydrologic connectivity on the hillslope and theimplications for nutrient transport. Global Biogeochemical Cycles 17 (4), 1105.
366 A.L. James, N.T. Roulet / Journal of Hydrology 377 (2009) 351–366
Tromp-van Meerveld, I., McDonnell, J.J., 2005. Comment to Spatial correlation of soilmoisture in small catchments and its relationship to dominant spatialhydrological processes. Journal of Hydrology 303, 307–312.
Uchida, T., Tromp-van Meerveld, H.J., McDonnell, J.J., 2005. The role of lateral pipeflow in hillslope runoff response: an intercomparison of non-linear hillsloperesponse. Journal of Hydrology 311 (1–4), 117–133.
Vitvar, T., Burns, D.A., Lawrence, G.B., McDonnell, J.J., Wolock, D.M., 2002.Estimation of baseflow residence times in watersheds from the runoff
hydrograph recession: method and application in the Neversinkwatershed, Catskill Mountains, New York. Hydrological Processes 16,1871–1877.
Waddington, J.M., Roulet, N.T., Hill, A.R., 1993. Runoff mechanisms in aforested groundwater discharge wetland. Journal of Hydrology 147, 37–60.
Wipkey, R.Z., Kirkby, M.J., 1978. Flow within the soil. Hillslope Hydrology, JohnWiley & Sons, Ltd., pp. 121–144 (Chapter 4).