Journal of Hydrology 380 (2010) 473–480

Contents lists available at ScienceDirect

Journal of Hydrology

journal homepage: www.elsevier .com/ locate / jhydrol

Influence of land-use change on near-surface hydrological processes:Undisturbed forest to pasture

Sonja Germer a,*, Christopher Neill b, Alex V. Krusche c, Helmut Elsenbeer d

a Berlin-Brandenburg Academy of Sciences and Humanities, Jägerstr. 22/23, 10117 Berlin, Germanyb The Ecosystems Center, Marine Biological Laboratory, 7 MBL St., Woods Hole, MA 02543, USAc Centro de Energia Nuclear na Agricultura, University of São Paulo, PO Box 96, 13.400-970 Piracicaba, SP, Brazild Institute of Geoecology, University of Potsdam, Karl-Liebknecht-Str. 24-25, 14476 Golm, Germany

a r t i c l e i n f o

Article history:Received 26 June 2009Received in revised form 30 September2009Accepted 14 November 2009

This manuscript was handled byK. Georgakakos, Editor-in-Chief, with theassistance of V. Lakshmi, Associate Editor

Keywords:Overland flowPerched water tableVariable source areaFlow pathsDeforestationAmazonia

0022-1694/$ - see front matter � 2009 Elsevier B.V. Adoi:10.1016/j.jhydrol.2009.11.022

* Corresponding author. Tel.: +49 30 20370 313; faE-mail address: germer@bbaw.de (S. Germer).

s u m m a r y

Soil compaction that follows the clearing of tropical forest for cattle pasture is associated with lower soilhydraulic conductivity and increased frequency and volume of overland flow. We investigated the fre-quency of perched water tables, overland flow and stormflow in an Amazon forest and in an adjacent25-year-old pasture cleared from the same forest. We compared the results with the frequencies of thesephenomena estimated from comparisons of rainfall intensity and soil hydraulic conductivity. The fre-quency of perched water tables based on rainfall intensity and soil hydraulic conductivity was expectedto double in pasture compared with forest. This corresponded closely with an approximate doubling ofthe frequency of stormflow and overland flow in pasture. In contrast, the stormflow volume in pastureincreased 17-fold. This disproportional increase of stormflow resulted from overland flow generationover large areas of pasture, while overland flow generation in the forest was spatially limited and wasobserved only very near the stream channel. In both catchments, stormflow was generated by saturationexcess because of perched water tables and near-surface groundwater levels. Stormflow was occasionallygenerated in the forest by rapid return flow from macropores, while slow return flow from a continuousperched water table was more common in the pasture. These results suggest that deforestation for pas-ture alters fundamental mechanisms of stormflow generation and may increase runoff volumes overwide regions of Amazonia.

� 2009 Elsevier B.V. All rights reserved.

Introduction

Quantifying the dynamics of perched water tables can improveour understanding of runoff processes. The variable source area(VSA) concept, first proposed by Hewlett and Hibbert (1967), de-scribes the spatially and temporally variable runoff response toprecipitation and how relatively small areas of saturation excesscan substantially increase overland flow. The frequency of theoccurrence of perched water tables, in addition to their durationand extent, influences overland flow and stormflow generation(Cox et al., 1996; Gburek et al., 2006). Several recent studies sug-gest how these characteristics of perched water tables are influ-enced by soil properties. For example, soils with deep impedinglayers tend to develop seasonally perched water tables, while soilswith near-surface impeding layers develop short-lived, event-based perched water tables (McDaniel et al., 2001, 2008).

Land-use change can also alter the presence and extent ofperched water tables and alter the generation of overland flow.

ll rights reserved.

x: +49 30 20370 444.

In southern Australia, deforestation for pasture increased the ex-tent of near-surface perched water tables over the past 125 years(Cox et al., 1996). In northern Idaho (USA) deforestation increasedthe height and duration of seasonal perched water (Rockefelleret al., 2004). In tropical forest regions, land-use change from undis-turbed forest to agricultural land has dramatically modified exten-sive areas around the world (Achard et al., 2002). Soil compactioninduced by deforestation, pasture installation and cattle tramplingincreases bulk density and penetration resistance and reducesmacroporosity, infiltration rates and hydraulic conductivities(Alegre and Cassel, 1996; Chauvel et al., 1991; Martínez and Zinck,2004; McDowell et al., 2003; Neill et al., 1997; Reiners and Olson,1984).

In the Amazon, conversion of forest to pasture is known todecrease hydraulic conductivity and consequently increase thefrequency and volume of overland flow (Germer et al., 2009;Moraes et al., 2006; Ziegler et al., 2004; Zimmermann et al.,2006). Zimmermann et al. (2006) examined how land use influ-ences the infiltrability and hydraulic conductivity of soils. By com-paring soil hydraulic conductivity with rainfall intensities, theyinferred an expected frequency of perched water table generation

474 S. Germer et al. / Journal of Hydrology 380 (2010) 473–480

at different soil depths and concluded that saturation overlandflow must be generated on pastures, at least occasionally, an infer-ence supported by Chaves et al. (2008) entirely on hydrochemicalgrounds. In this paper, we monitored the frequency and duration ofperched water tables in paired forest and pasture catchments inthe western Brazilian Amazon state of Rondônia. We comparedthese measurements with predictions generated from rainfallintensity and soil hydraulic conductivity. We also directly mea-sured the frequency and volume of overland flow and used the re-sults to suggest how the mechanism of stormflow generation andthe volume of stormflow are altered on lands deforested forpasture.

Study area and methods

Study area

The study site of Rancho Grande is located approximately 50 kmsouth of Ariquemes (10�180S, 62�520W, 143 m a.s.l.) in the Brazilianstate of Rondônia, in the southwestern Amazon Basin of Brazil. Thearea is part of a morphostructural unit known as the SouthernAmazon Dissected Highlands (Planalto Dissecado Sul da Amazônia,Peixoto de Melo et al., 1978), which is characterized by pro-nounced topography with an altitudinal differential of up to150 m. Remnant ridges of Precambrian basement rock composedof gneisses and granites of the Xingu (Leal et al., 1978) or JamariComplex (Isotta et al., 1978), are separated by flat valley floors ofvarying width. The climate is tropical wet and dry (Köppen’s Aw,Köppen, 1936). The mean annual temperature between 1984 and

Fig. 1. Map of the study site (a) and detailed maps for the forest (b) and pasture (c) ccontour lines (m). Wells for recording groundwater levels and piezometers for percheinstrument nest.

2003 was 27 �C. The mean annual precipitation during the sameperiod was 2300 mm a�1 with a marked dry period from July toSeptember (Germer et al., 2006).

We selected adjacent forest (1.37 ha) and pasture (0.73 ha)catchments approximately 400 m apart. Both catchments aredrained by 0-order streams and underlain by Kandiudults (Zim-mermann et al., 2006). While the stream in the forest is ephemeral,that in the pasture is intermittent. The forest vegetation at RanchoGrande is predominantly terra firme undisturbed open tropicalrainforest (Floresta Ombrófila Aberta) with a large number ofpalms. Open tropical rainforest is the predominant vegetation typewithin the transition zone from dense rainforest to cerrado vegeta-tion (savanna) in the southern Amazon (IBGE, 2004), and amountsto 55% of the total vegetated area in Rondônia (Fernandes and Gui-marães, 2001). The pasture was cleared and burned twice, in 1980and 1981. Since 1984 the pasture has been grazed by one head ofcattle per hectare, tilled once, in 1993, and planted with Brachiariahumidicola (Zimmermann et al., 2006).

Field measurements

A similar hydrological sampling design (Fig. 1) was set up in Au-gust 2004 in the forest and pasture catchments. The instruments inthe pasture were protected against disturbance by cattle withfences. A tipping bucket rain gauge (Hydrological Services P/L, Liv-erpool Australia) with a resolution of 0.254 mm and a Campbelllogger recorded 5-min rainfall intensity values adjacent to thepasture catchment. Both watersheds were equipped with 1.0 ftH-flume-type weirs for stream-flow monitoring (Blaisdell, 1976;

atchments, showing instrumentation sites, observed overland flow frequency, andd water table monitoring at depths of 12.5, 20 and 50 cm were situated in each

040

80

Sep Jan May

Cum

ulat

ive

sum

s [%

]

Forest

040

80

Sep Jan May Sep Jan May

Pasture

Sep Jan MaySep Jan May

010

0020

00R

ainf

all [

mm

]

Sep Jan May

05

1020

For

est s

torm

flow

[mm

]

Sep Jan May

010

030

0P

astu

re s

torm

flow

[mm

]

Fig. 2. Cumulative sums of rainfall (black line) and stormflow (grey line) fromAugust 2004 to July 2005.

S. Germer et al. / Journal of Hydrology 380 (2010) 473–480 475

Gwinn and Parsons, 1976) and water levels were recorded withTruTrack loggers every 5 min from September through November2004 and from January to mid-April, and every 15 min in Decem-ber 2004 and from mid-April to July 2005. Because of limited accu-racy of low flow in the H-flumes, only events with a maximaldischarge stage greater than 1 cm were classified as stream flowdischarge events.

Thirty piezometers were installed in each catchment in nests ofthree to monitor the occurrence of perched water tables at depthsof 12.5, 20 and 50 cm. Hydraulic conductivity at these depths wassurveyed previously (Zimmermann et al., 2006). The piezometerswere composed of a PVC tube with 24 mm inner diameter, whichwas hammered into the soil to avoid gaps and vertical flow betweentube and soil. The piezometer hammer consisted of a metal rod of23 mm diameter and slide hammer. Four piezometer triplets percatchment were equipped with TruTrack WT-HR water height dataloggers recording water levels every 5 min. Only four instrumentnests per catchment could be equipped with loggers at the sametime, so logger positions were rotated over the study period to ob-tain water levels for a subset of events for each piezometer triplet.

Overland flow frequency was detected with non-recordingoverland flow detectors that were made of PVC tube (24 mm innerdiameter) and consisted of a detector section that was perforatedon one side and connected at right angle to a reservoir section(Kirkby et al., 1976). The perforated tube was in contact with thesoil such that overland flow could enter the tube and be divertedto the reservoir. One overland flow detector was placed in eachinstrument nest in the pasture, while in the forest overland flowdetectors were placed at 15 locations along the channel. Previousfield observations (February 2004) during intense rainfall eventsindicated that overland flow did not occur in the forest at a dis-tance of more than 2 m from the channel. This observation wasverified during the study period by observing the forest site duringall intense daytime events. Groundwater levels were recorded withTruTrack loggers in the forest and Orphimedes loggers (OTT Mess-technik) in the pasture at the 10 instrument nests in each catch-ment. Wells were drilled to a depth of 2–4 m with a hand augerand equipped with a 5 cm diameter PVC casing.

Rainfall intensity and stream flow were monitored from August2004 to July 2005. For the same hydrological year, we loggedperched water levels from the end of September to the end ofMarch and groundwater levels from January to July. Overland flowwas monitored from mid-October to mid-March, excludingDecember and the first 10 days of January.

Results

Runoff response

The total incident rainfall at Rancho Grande from August 2004to July 2005 was 2286 mm, similar to the mean annual rainfall of2300 mm between 1984 and 2003. Total annual discharge was

Table 1Number and frequency (%) of stormflow events and perched water table (PWT) response duevents. The column ‘‘both depths” indicates the response frequency of P2 piezometerfrequencies and total volume of stormflow and relative frequencies of PWT in relation tomedian, lower confidence level of the median (LCL) and upper confidence level of the medievent.

Observed stormflow eventfrequency and volume

Observed PWT frequency(response of P2 piezometers)

Expecte

12.5 cm 20 cm Both depths Median

Forest 32(23%, 22 mm)

14(10%)

6(4%)

15(11%) 0

Pasture 60(43%, 378 mm)

34(24%)

34(24%)

41(29%) 27

much smaller in the forest (24 mm) than in the pasture(416 mm) because of lower runoff frequency and lower volumeper event (Table 1). The cumulative curves of rainfall and storm-flow illustrated that total discharge in the forest was dominatedby two consecutive events that delivered 59% of total annual dis-charge, while in the pasture the same events generated just 17%of the total annual discharge (Fig. 2). Smaller events were muchmore important for total annual discharge in the pasture than inthe forest. Baseflow added 4% and 2% to total stormflow in forestand pasture, respectively. Continuous baseflow was detected fromMarch 6 to March 9 in the forest and from January 20 to March 16in the pasture. For events that generated stormflow in both catch-ments, stormflow response times were the same in pasture andforest for 25 out of 37 events. Stormflow response times to rainfall,however, were shorter in the pasture when observed groundwaterlevels in the central well of the downslope transect remained be-tween 50 cm soil depth and the soil surface (February 21–March20), compared with all other days (p < 0.01, Wilcoxon rank sumtest). No temporal differences were recorded in the forest.

Piezometric response

Piezometric responses were observed in both catchments.While the 50 cm piezometers only responded to rising groundwa-ter levels (see Section ‘Groundwater response’), the piezometers atthe 12.5 and 20 cm soil depths reflected the development ofperched water tables, independent of groundwater levels.Although the change in water level in the piezometers varied inmagnitude for different rainfall events, the response patterns forindividual sites were consistent over time. The differences betweenforest and pasture piezometric responses were exemplified in a64 mm rainfall event on November 14, 2004 (Figs. 3 and 4), duringwhich piezometer water levels were logged in the central parts ofboth catchments (all nests of the middle transect and middle nestof the upslope transect). In the forest, the response of three out offour 12.5 cm piezometers overlapped with the duration of storm-flow during the November 14 event. Only one out of four 20 cmpiezometers in the forest responded during the same event, butthe perched water table persisted after stormflow ended (Fig. 3).

ring the period from September 25, 2004 to March 29, 2005 with a total of 139 rainfalls per site, independent of depth. Observed values in parentheses indicate relative

the 139 rainfall events. Expected PWT frequencies were derived by comparing thean (UCL) hydraulic conductivity of soils with maximum 30-min rainfall intensities per

d PWT frequency

12.5 cm (%) LCL–UCL 12.5 cm (%) Median 20 cm (%) LCL–UCL 20 cm (%)

0–6 27 15–55

20–34 52 44–61

108

64

20

Rai

n [m

m 5

−min

−1]

0.0

0.1

0.2

0.3

0.4

0.5

05:00 09:00 13:00

Dis

char

ge [L

s−1

] (a)

3025

2015

105

0

05:00 09:00 13:00

(b)

Soil

dept

h [c

m]

3025

2015

105

0

3025

2015

105

0

05:00 09:00 13:00

(c)

Soil

dept

h [c

m]

3025

2015

105

0

3025

2015

105

0

05:00 09:00 13:00

(d)

Soil

dept

h [c

m]

3025

2015

105

0

3025

2015

105

0

05:00 09:00 13:00

(e)

Soil

dept

h [c

m]

3025

2015

105

0

Fig. 3. Forest stormflow hydrograph (plot a, line) response to rainfall (plot a, bars) for a 64 mm event on November 14, 2004 and the corresponding perched water tableresponses at the eastern (plot b), middle (plot c) and western (plot d) instrument nest on the midslope transect and the middle nest on upslope transect (e). Perched waterlevels in the piezometers relative to the soil surface (grey line) are given for the soil depth of 12.5 cm (solid black line) and 20 cm (dashed black line).

476 S. Germer et al. / Journal of Hydrology 380 (2010) 473–480

In the forest, all sites exhibited downward hydraulic gradientthroughout this event. The double-peak rainfall event was mir-rored by the stormflow hydrograph and the piezometric responseat the 12.5 cm depth on the west side of the forest catchment,but not by the other three responding piezometers located withinthe channel. In the pasture, seven out of eight piezometers re-sponded during the same event, and the perched water table per-sisted after stormflow ceased (Fig. 4). While the piezometers atmost pasture sites (Fig. 4b, d and e) indicated downward waterflow throughout the event, the middle triplet of the midslope tran-sect indicated a gradient change from downward to upward flowdirection after 8:00 AM. At both depths, piezometric response overtime periods with several rainfall events, indicated that perchedwater tables generally persisted longer in the pasture comparedto the forest (Table 2).

Overland flow response

Similar differences were observed for overland flow in the forestand pasture as for piezometeric response. Overland flow detectors(OFD) responded about twice as often in the pasture as in the forest(Fig. 1). Overland flow in the forest was only observed near thechannel. In the pasture, overland flow was observed over almostthe whole catchment area for some events. For the first events afterthe end of the dry season, overland flow was detected in fewerplaces in the pasture than for later events with similar intensities(open circles vs. solid dots in Fig. 5). This was not observed inthe forest. In general, the greater the rainfall intensities, the largerwas the number of OFD responses. Most of the events with morethan 50% OFD responses in the pasture had maximum 15-min rain-

fall intensities above 40 mm h�1 (Fig. 5). In the forest, for someevents stormflow was recorded but no overland flow was observedwithin the monitored channel section. In the pasture, each storm-flow-generating event also resulted in overland flow. Overlandflow could, however, develop without a perched water table at12.5 cm depth (two thirds of all monitored OFD events) or at bothdepths (one third of all monitored OFD events).

Groundwater response

Groundwater level rose faster in the forest than in the pasture.At the beginning of groundwater monitoring (January 22, 2005)water levels were observed at 2 and 1 m depths in the middle wellsof the downslope instrument nest transects in the forest and pas-ture, respectively (Fig. 6). Groundwater levels peaked in bothcatchments after the two big events on March 5 and 6 with 58and 72 mm rainfall, respectively. Thereafter rainfall events exceed-ing 20 mm were rare and did not occur before mid-April, whengroundwater levels had fallen below 2 m in most of the wells.Therefore, groundwater level response to rainfall was minor afterthe beginning of March. Groundwater levels rose up to the soil sur-face only in the middle well of the downslope instrument nesttransects in both catchments (Fig. 6). Groundwater levels remainedat the soil surface for 1 week in the forest, but only for 2 days in thepasture. In both catchments, however, these wells recorded near-surface water levels for periods of similar length (e.g. above0.5 m during 1 month: forest: February 22–March 27, pasture: Feb-ruary 19–March 20). For all wells, after the peak of the wet seasongroundwater levels decreased faster in the forest compared withthe pasture. Despite similar groundwater levels at a distance of

108

64

20

Rai

n [m

m 5

−min

−1]

05

1015

20

05:00 09:00 13:00

Dis

char

ge [L

s−1

] (a)

3025

2015

105

0

05:00 09:00 13:00

(b)

Soil

dept

h [c

m]

3025

2015

105

0

3025

2015

105

0

05:00 09:00 13:00

(c)

Soil

dept

h [c

m]

3025

2015

105

0

3025

2015

105

0

05:00 09:00 13:00

(d)

Soil

dept

h [c

m]

3025

2015

105

0

3025

2015

105

0

05:00 09:00 13:00

(e)

Soil

dept

h [c

m]

3025

2015

105

0

Fig. 4. Same as Fig. 3 for pasture. The middle triplet of the midslope transect indicate a gradient change from downward to upward flow direction after 8:00 AM.

Table 2Comparison of perched water table durations (min) in the forest and pasture monitored at 12.5 and 20 cm soil depths in the catchments at the midslope and upslope transects.Perched water table durations are listed for single events and as total duration for two different periods (4th–20th November 2004 and 4th December 2004–11th January 2005).For these two periods all events with perched water tables are listed.

S. Germer et al. / Journal of Hydrology 380 (2010) 473–480 477

35 m from the H-flumes during the wet season, long after theperched water table ceased to exist baseflow was observed fre-quently in the pasture but only once in the forest. Baseflow during

perched water table conditions is expected to be a combination ofsubsurface flow and groundwater flow; otherwise it is sustainedonly by groundwater flow.

Fig. 5. Number of overland flow detector (OFD) responses per event in the pastureplotted against the maximum 15 min rainfall intensity. The first three events ofoverland flow occurrence after the dry season are highlighted (open circles).

100

6020

0R

ainf

all [

mm

5−m

in−1

]

2.5

1.5

0.5

Wat

er le

vel [

m]

Forest

Jan Mar May

2.5

1.5

0.5

Wat

er le

vel [

m]

Pasture

Fig. 6. Groundwater levels in the forest (a) and pasture (b) of the middle well of thedownslope transect. Water levels are given in depth relative to the soil surface. Barsfrom the top of each figure indicate the total rainfall per event.

478 S. Germer et al. / Journal of Hydrology 380 (2010) 473–480

Discussion

Land-use change from undisturbed tropical forest to pasture isknown to change physical soil characteristics, including bulk den-sity, penetration resistance, porosity (Martínez and Zinck, 2004;Moraes et al., 1996; Neill et al., 1997; Reiners and Olson, 1984)and near-surface hydraulic conductivity (Lal, 1996; Moraes et al.,

Table 3Median and median absolute deviation (MAD) of the infiltrability and saturated hydraZimmermann et al., 2006).

Infiltrability (mm h�1) Ksat at 12.5

Median MAD Median

Forest 1690.2 465.9 130.9Pasture 112.7 50.5 22.1

2006; Zimmermann et al., 2006). At our site the sharper decreaseof hydraulic conductivity in pasture compared to forest provokeda more frequent development, and longer persistence, of perchedwater tables once they developed. We also observed an increaseof overland flow frequency and volume as a consequence. Thisobservation is in line with an increased frequency of subsurfacestormflow, overland flow and perched water table occurrencefound in a pasture compared with forest under a Haplustox in East-ern Amazonia (Moraes et al., 2006). Rockefeller et al. (2004) dem-onstrated that deforestation for pasture land increased theduration and height of seasonal perched water tables in Fragipansoils in Idaho/USA. The authors attributed these changes to greaterinterception by and subsequent evaporation from the forest can-opy, which they observed but didn’t measure. In their study soilcompaction was less relevant and soil saturation was induced bya fragipan at 0.8–1 m depth under both land uses and perchedwater tables never got near to the soil surface. In our study, how-ever, a change in physical soil properties was the main reason forincreases in perched water table conditions. At our site, rainfallminus the sum of throughfall (89%, Germer et al., 2006) and stem-flow (8%, based on wet season events only, Werther, 2007) resultsin 3% interception. As this value represents average interceptionand interception is much smaller for greater events with elevatedrainfall intensities, we assume that interception had a negligibleeffect on differences in perched water conditions at our site.

Observation of saturation excess

Stormflow and overland flow were observed twice as often inthe pasture compared with forest (Table 1, Fig. 1). This relationshipbetween both catchments was expected based on the difference ofmedian hydraulic conductivity at 20 cm depth. The comparison ofexpected ponding at 20 cm depth in both catchments based on the95% confident limits of median hydraulic conductivity, however,suggests no significant difference (Table 1). The actual frequencyof perched water table generation, however, was significantly low-er than expected at the 20-cm depth in forest and pasture (Table 1).This might be due to seasonal differences in antecedent moistureand, hence, in overland flow generation, as events with similarrainfall intensities generated less overland flow during the transi-tion from dry to wet season than in the wet season (Fig. 5).

During November 2004 all of the events that generated perchedwater tables had maximum 30 min rainfall intensities above themedian Ksat (22 mm h�1, Table 3), measured at 20 cm and12.5 cm soil depth in the forest and pasture, respectively. Giventhe small differences in depth at which the same Ksat was mea-sured (7.5 cm) and hence the small differences in water depthneeded to saturate the soil (e.g. 50% porosity: 10 mm maximumin forest and 6.25 in pasture) combined with the high rainfall to-tals, one could expect perched water table development in all mon-itored piezometers in both catchments during November 2004.This was, however, only the case for the pasture (Table 2). Variableperched water tables over short (10 m) distances have been re-ported previously for shallow soils in southwestern Australia,where the thickness and hydraulic conductivity of the soil abovethe impeding layer were the most important factors affecting the

ulic conductivity at 12.5 and 20 cm depth under forest and pasture (data source:

cm (mm h�1) Ksat at 20 cm (mm h�1)

MAD Median MAD

148.3 21.6 26.111.3 6.2 5.5

3025

2015

105

0R

ainf

all [

mm

5−m

in −1

]

18:0

0

23:0

0

04:0

0

09:0

0

14:0

0

19:0

00.01

110

0D

isch

arge

[L s

−1]

Forest

Pasture

Fig. 7. Comparison of forest (grey solid line) and pasture (black dashed line)stormflow hydrographs for a rainfall event on March 5–6, 2005. Discharge is plottedon a logarithmic scale to show small flow differences more clearly. Bars in the top ofthe figure indicate 5-min rainfall sums.

S. Germer et al. / Journal of Hydrology 380 (2010) 473–480 479

occurrence of perched water tables (Cox and McFarlane, 1995). In-creased median Ksat in combination with high Ksat median absolutedeviation (MAD) in the forest compared to the pasture is expectedto lead to increased variability in the occurrence of perched watertables, and thus a decrease in the predictive power of the Ksat-pre-cipitation relationship (Table 3). A possible explanation for thisvariability might be local fast bypass flow through the unsaturatedsoil, as reported by Radulovich et al. (1992) for microaggregatedtropical soils at La Selva Biological Station, Costa Rica. Macroporesdue to bioturbation were abundant up to a soil depth of 15 cm inour forest (Sobieraj et al., 2004) and return flow from soil pipeswas observed during our study in the forest stream channel. Thiscould explain: (1) why in the forest perched water tables were ob-served more frequently at 12.5 than at 20 cm soil depth and (2)why observed frequency of perched water at 12.5 cm was signifi-cantly higher and at 20 cm was significantly lower than expected(Table 1), as an impeding layer is situated between these twodepths. In the pasture the same perched water frequency was ob-served in both depths, due to the small difference in hydraulic con-ductivity between these depths and hence the absence of an abruptchange of permeability.

The homogenous pattern of perched water table occurrence andthe longer persistence of perched water tables in the pasture ex-plain the observation of saturation-excess overland flow over a lar-ger area in the pasture catchment compared to its limitedoccurrence next to the channel in the forest catchment, whichagrees with the VSA concept (Hewlett and Hibbert, 1967). At ourpasture site, overland flow was generated by rainfall falling on sat-urated soil above an impeding layer. The greater number of OFD re-sponses observed at higher rainfall intensities (Fig. 5) indicatesvariable contributing areas of saturation excess and overland flowduring rainfall events. Such a VSA process, which is driven byimpeding layers, was recently reported for soils with fragipansand/or strong contrasting clay layers (Gburek et al., 2006; Needel-man et al., 2004).

Land-use effect on stormflow generation

Our study illustrates fundamental differences in the mecha-nisms and amount of runoff generation in Amazon forest and pas-ture. First, stormflow and overland flow occurred twice as often inpasture than in the forest. Second, the volume of stormflow was17-fold greater in pasture (Table 1). The relative frequency increaseof perched water table and overland flow occurrence can be esti-mated by comparing rainfall intensities with hydraulic conductiv-ities under both land uses, but no direct conclusion can be drawnfor the change in total stormflow volumes. The much greater in-crease in the volume of flow than in the frequency of overland flowoccurrence was probably due to a much greater contributing areain the pasture.

Return flow was another mechanism of overland flow genera-tion that could be observed in the forest and pasture, but theunderlying processes differed between catchments. Duringperched water conditions return flow is difficult to separate fromsaturation overland flow. After saturation overland flow ceased re-turn flow might substantially contribute to baseflow. The March 6,2005 event was the only one during which forest baseflow(0.64 mm) exceeded pasture baseflow (0.25 mm) (Fig. 7). Fast re-turn flow through macropores was not observed for the firststormflow event on March 5, 2005 in the forest, but only at theend of the second stormflow event on March 6, 2005, when forestbaseflow exceeded that in pasture (personal observation). Returnflow in the forest is, hence, attributed to fast return flow throughmacropores. In contrast, frequent upward water flow in the middlenest of the midslope transect in the pasture indicated slow lateralsubsurface stormflow and return flow. For this event, the contribu-

tion of return flow to total stormflow was smaller for pasture thanfor forest due to low near-surface conductivities in the pasture.Baseflow observed long after perched water conditions was, how-ever, attributed to groundwater flow and was more common in thepasture than in the forest. Stormflow tended to start simulta-neously in both catchments, but it rose more rapidly in the pastureduring times of near-surface groundwater levels. This indicatesthat saturation excess caused by rainfall on an area of groundwatertable emergence (return flow) (Dunne and Black, 1970a,b) ex-panded faster in the pasture than in the forest during rainfallevents.

In addition to insights into land-use change impacts on storm-flow generation, this study demonstrates that two different VSAprocesses can occur at the same site. The process behind the VSAconcept, as described by Hewlett and Hibbert (1967), is linked tonear steam areas with high soil water content or shallow ground-water levels but it can coexist with a VSA process of runoff gener-ation driven by saturation excess over impeding layers. In bothcases the source areas expand with increasing rainfall volumeand intensity.

Even though land-use change had a great influence on overlandflow generation at our site, where frequency and total volume ofoverland flow in undisturbed forest were low, if not negligible, thisreference condition of little overland flow cannot be generalized. Ifanything, the published evidence suggests the prominent role ofoverland flow in undisturbed tropical forests, from Queensland(Bonell and Gilmour, 1978; Elsenbeer et al., 1995) to Central Amer-ica (Godsey et al., 2004) and to western Amazonia (Elsenbeer andVertessy, 2000; Johnson et al., 2006). To illustrate this point, a max-imum 15-min rainfall intensity of 40 mm h�1 was required at ourpasture site to trigger a response in more than 50% of overland flowdetectors. The same response was achieved by only 29 mm h�1 and6 mm h�1 at two undisturbed forest sites in Panama (Godsey et al.,2004)! Where perched water tables and overland flow are the rulerather than the exception under undisturbed conditions, forestconversion can only amplify a preexisting condition, but it cannotchange the mode of runoff generation.

Conclusions

The frequency of stormflow doubled in pasture compared withforest while the volume of stormflow increased 17-fold. This dis-proportional increase of stormflow volume was caused by overlandflow generation over larger areas in the pasture than in the forest.Processes of overland flow generation differed between the forest

480 S. Germer et al. / Journal of Hydrology 380 (2010) 473–480

and pasture catchments. Saturation-excess overland flow causedby both perched water table and groundwater occurrence existedin both catchments. While the latter process reflects the originalversion of the variable source area concept, the former processcan also be linked to variable source areas. Both variable sourcearea processes can coexist. Fast return flow through soil pipeswas only observed in the forest, while slow return flow was gener-ated from a continuous perched water table in the pasture.

The transformation of undisturbed forest into pasture increasedthe frequency of occurrence of perched water tables and overlandflow by soil compaction and an increase in hydraulic conductivity.These results show that conversion of undisturbed forest to pas-ture not only increases the frequency and volume of stormflow,but also the contributing area and the manner in which the watertravels through the soil towards the stream channel. These resultssuggest greater impact of land-use change on hydrology thanwould be expected by comparing average hydraulic conductivities.Further research could provide insight into the relationship ofstormflow volume increase due to land-use change on differentsoils and at different scales.

Acknowledgements

This study was partially supported by the US National ScienceFoundation Grant DEB-0315656, and NASA LBA Grant NCC5-285to the MBL and by Grants from Brazilian agencies FAPESP # 03/13172-2 and CNPq # 420199/2005-5 to CENA. We thank the Sch-mitz family for logistical support and the opportunity to work ontheir land. For dedicated help during field work we would like tothank Sérgio Goveia Neto, Lisa Werther, Tobias Vetter and SonyaRemington and several other short-term helpers.

References

Achard, F., Eva, H.D., Stibig, H.-J., Mayaux, P., Gallego, J., Richards, T., Malingreau, J.-P., 2002. Determination of deforestation rates of the world’s humid tropicalforests. Science 297, 999–1002.

Alegre, J.C., Cassel, D.K., 1996. Dynamics of soil physical properties under alternativesystems to slash-and-burn. Agriculture, Ecosystems and Environment 58, 39–48.

Blaisdell, W., 1976. Discussion: discharge equation for HS, H, and HL flumes (byGwinn, Wendell R. and Parsons, Donald A.). Journal of the Hydraulics Division102 (HY11), 1704–1706.

Bonell, M., Gilmour, D.A., 1978. The development of overland flow in a tropicalrainforest catchment. Journal of Hydrology 39, 365–382.

Chauvel, A., Grimaldi, M., Tessier, D., 1991. Changes in soil-pore space distributionfollowing deforestation and revegetation: an example from the central Amazon,Brazil. Forest Ecology and Management 38, 259–271.

Chaves, J., Neill, C., Germer, S., Gouveia Neto, S., Krusche, A., Elsenbeer, H., 2008.Land management impacts on runoff sources in small Amazon watersheds.Hydrological Processes 22, 1766–1775.

Cox, J.W., Fritsch, E., Fitzpatrick, R.W., 1996. Interpretation of soil features producedby ancient and modern processes in degraded landscapes. 7. Water duration.Australian Journal of Soil Research 34 (6), 803–824.

Cox, J.W., McFarlane, D.J., 1995. The causes of waterlogging in shallow soils andtheir drainage in southwestern Australia. Journal of Hydrology 167 (1–4), 175–194.

Dunne, T., Black, R.D., 1970a. An experimental investigation of runoff production inpermeable soils. Water Resources Research 6 (2), 478–490.

Dunne, T., Black, R.D., 1970b. Partial area contributions to storm runoff in a smallNew-England watershed. Water Resources Research 6 (5), 1296–1311.

Elsenbeer, H., Lorieri, D., Bonell, M., 1995. Mixing model approaches to estimatestorm flow sources in an overland flow-dominated tropical rain forestcatchment. Water Resources Research 31 (9), 2267–2278.

Elsenbeer, H., Vertessy, R.A., 2000. Stormflow generation and flowpathcharacteristics in an Amazonian rainforest catchment. Hydrological Processes14, 2367–2381.

Fernandes, L.C., Guimarães, S.C., 2001. Atlas Geoambiental de Rondônia. SEDAM,Porto Velho.

Gburek, W.J., Needelman, B.A., Srinivasan, M.S., 2006. Fragipan controls on runoffgeneration: hydropedological implications at landscape and watershed scales.Geoderma 131 (3-4), 330–344.

Germer, S., Elsenbeer, H., Moraes, J.M., 2006. Throughfall and temporal trends ofrainfall redistribution in an open tropical rainforest, south-western Amazonia(Rondônia, Brazil). Hydrology and Earth System Sciences 10, 383–393.

Germer, S., Neill, C., Vetter, T., Chaves, J., Krusche, A.V., Elsenbeer, H., 2009.Implications of long-term land-use change for the hydrology and solute budgetsof small catchments in Amazonia. Journal of Hydrology 364, 349–363.

Godsey, S., Elsenbeer, H., Stallard, R., 2004. Overland flow generation in twolithologically distinct rainforest catchments. Journal of Hydrology 295, 276–290.

Gwinn, W.R., Parsons, D.A., 1976. Discharge equations for HS, H, an HL Flumes.Journal of the Hydraulics Division 102 (HY1), 73–88.

Hewlett, J.D., Hibbert, A.R., 1967. Factors affecting the response of small watershedsto precipitation in humid areas. In: Sopper, W.E., Lull, H.W. (Eds.), InternationalSymposium on Forest Hydrology. Pergamon Press, New York, pp. 275–290.

IBGE, 2004. Mapa de Vegetação do Brasil. Instituto Brasiliero de Geografia eEstatística, Rio de Janeiro.

Isotta, C.A.L., Carneiro, J.M., Kato, H.T., Barros, R.J.L., 1978. Projeto ProvínciaEstanífera de Rondônia. DNPM/CPRM, Porto Velho.

Johnson, M.S., Lehmann, J., Conto, E.G., Filho, J.P.N., Riha, S.J., 2006. DOC and DIC inflowpaths of Amazonian headwater catchments with hydrologically contrastingsoils. Biogeochemistry 81, 45–57.

Kirkby, M., Callan, J., Weyman, D. and Wood, J., 1976. Measurement and modeling ofdynamic contributing areas in very small catchments, Working Paper No. 167.University of Leeds, School of Geography, Leeds.

Köppen, W.P., 1936. Das geographische System der Klimate. Nendeln/Lichtenstein,Berlin. 252 pp.

Lal, R., 1996. Deforestation and land-use effects on soil degradation andrehabilitation in western Nigeria. 1. Soil physical and hydrological properties.Land Degradation & Development 7 (1), 19–45.

Leal, J.W.L., Silva, G.H., dos Santos, D.B., Texeira, W., Lima, M.I.C., Fernandes, C.A.C.,Pinto, A.d.C., 1978. Geologia – Folha SC. 20. Projeto Radambrasil, Porto Velho.pp. 219–355.

Martínez, L.J., Zinck, J.A., 2004. Temporal variation of soil compaction anddeterioration of soil quality in pasture areas of Colombian Amazonia. Soil andTillage Research 75, 3–17.

McDaniel, P.A., Gabehart, R.W., Falen, A.L., Hammel, J.E., Reuter, R.J., 2001. Perchedwater tables on Argixeroll and Fragixeralf hillslopes. Soil Science Society ofAmerica Journal 65 (3), 805–810.

McDaniel, P.A., Regan, M.P., Brooks, E., Boll, J., Bamdt, S., Falen, A., Young, S.K.,Hammel, J.E., 2008. Linking fragipans, perched water tables, and catchment-scale hydrological processes. Catena 73 (2), 166–173.

McDowell, R.W., Drewry, J.J., Paton, R.J., Carey, P.L., Monaghan, R.M., Condron, L.M.,2003. Influence of soil treading on sediment and phosphorus losses in overlandflow. Australian Journal of Soil Research 41 (5), 949–961.

Moraes, J.F.L., Volkoff, B., Cerri, C.C., Bernoux, M., 1996. Soil properties underAmazon forest and changes due to pasture installation in Rondônia, Brazil.Geoderma 70, 63–81.

Moraes, J.M., Schuler, A.E., Dunne, T., Figueiredo, R.O., Victoria, R.L., 2006. Waterstorage and runoff processes in plinthic soils under forest and pasture in easternAmazonia. Hydrological Processes 20 (12), 2509–2526.

Needelman, B.A., Gburek, W.J., Petersen, G.W., Sharpley, A.N., Kleinman, P.J.A., 2004.Surface runoff along two agricultural hillslopes with contrasting soils. SoilScience Society of America Journal 68 (3), 914–923.

Neill, C., Melillo, J.M., Steudler, P.A., Cerri, C.C., deMoraes, J.F.L., Piccolo, M.C., Brito,M., 1997. Soil carbon and nitrogen stocks following forest clearing for pasture inthe southwestern Brazilian Amazon. Ecological Applications 7 (4), 1216–1225.

Peixoto de Melo, D., da Costa, R.C., Natali Filho, T., 1978. Geomorfologia – FolhaSC.20. Projeto Radambrasil, Porto Velho, pp. 187–250.

Radulovich, R., Sollins, P., Baveye, P., Solorzano, E., 1992. Bypass water-flow throughunsaturated microaggregated tropical soils. Soil Science Society of AmericaJournal 56 (3), 721–726.

Reiners, W.A., Olson, R.K., 1984. Effects of canopy components on throughfallchemistry – an experimental-analysis. Oecologia 63 (3), 320–330.

Rockefeller, S.L., McDaniel, P.A., Falen, A.L., 2004. Perched water table responses toforest clearing in northern Idaho. Soil Science Society of America Journal 68 (1),168–174.

Sobieraj, J.A., Elsenbeer, H., Cameron, G., 2004. Scale dependency in spatial patternsof saturated hydraulic conductivity. Catena 55, 49–77.

Werther, L., 2007. Stemflow in an open tropical rainforest in Rondônia (Brazil).M.Sc. Thesis (unpublished), University of Potsdam, Potsdam, 81 pp.

Ziegler, A.D., Giambelluca, T.W., Tran, L.T., Vana, T.T., Nullet, M.A., Fox, J., Vien, T.D.,Pinthong, J., Maxwell, J.F., Evett, S., 2004. Hydrological consequences oflandscape fragmentation in mountainous northern Vietnam: evidence ofaccelerated overland flow generation. Journal of Hydrology 287, 124–146.

Zimmermann, B., Elsenbeer, H., Moraes, J.M., 2006. The influence of land-usechanges on soil hydraulic properties: implications for runoff generation. ForestEcology and Management 222, 29–38.