earth surface processes and landforms hydrological ... · pdf filecontrasting ﬂow pathways...

Hydrological pathways within rain forest slopes 735

Copyright © 2005 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 30, 735–753 (2005)

Earth Surface Processes and LandformsEarth Surf. Process. Landforms 30, 735–753 (2005)Published online 7 April 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/esp.1173

Contrasting flow pathways within tropical forestslopes of Ultisol soilsNick A. Chappell* and Mark D. SherlockLancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, UK

AbstractThere are very few experimental studies identifying hydrological pathways within rainforest slopes. Such knowledge is, however, necessary to understand why forest disturbanceaffects rainfall–riverflow response and nutrient migration. This study examines flow path-ways within lowland rain forest slopes comprising Udults of the Ultisol soil order. Experi-mentation was conducted on four SE Asian hillslope units (each 5 ××××× 5 m in plan) in the BukitTimah catchment (Singapore Island), and in the W8S5 catchment (Sabah, Borneo Island).The flow pathways were identified by artificial tracer experiments. We evaluated how wellhydrometric calculations based on tensiometry and permeametry measurements predictedthe tracer patterns.

The tracer work indicated much faster subsurface flows at Bukit Timah than W8S5 forthe storms studied. Some explanation of the greater subsurface waterflows at Bukit Timah incomparison to W8S5 is afforded by the less steep moisture release curves which maintainhydraulic conductivity as the soil dries. Vertical flow of the tracer through the upper 1 m ofsoil predominated (>>>>>90 per cent of percolation) in the Bukit Timah slopes. In some contrast,a major component (approximately 60 per cent) of the tracer percolation was directed later-ally within the W8S5 slopes. The flow vectors calculated using the hydrometric methods did,however, grossly under-estimate the degree of lateral deflection of waterflow generated atW8S5 and to a lesser extent over-estimated it at Bukit Timah. In part, these errors mayrelate to the inability of traditional hydrometric techniques to fully characterize the effect ofthe large and small ‘natural soil pipes’ present within both catchments. In conclusion, thestudy indicates that marked variations in flow vectors exist within the Udult great group ofSE Asian soils and hydrometric calculations may be poor predictors of these dominantpathways. Copyright © 2005 John Wiley & Sons, Ltd.

Key words: humid tropics; macropores; permeability; soil pipes; tracers; Udult; Ultisol

*Correspondence to:N. A. Chappell, LancasterEnvironment Centre,Lancaster University,Lancaster, LA1 4YQ, UK. E-mail:n.chappell@ lancaster.ac.uk

Received 1 April 2004;Revised 4 August 2004;Accepted 29 September 2004

Introduction

Knowledge of hillslope flow pathways within natural rain forest catchments in the humid tropics remains scarce(Bonell, 2004). Several key studies on hillslope hydrological flow pathways have helped shape our current theories ofstreamflow generation in these landscapes, e.g., the shallow lateral flows interpreted by Bonell and Gilmour (1978) inNortheast Queensland (Australia), and the predominantly vertical flows interpreted by Nortcliff and Thornes (1981) inCentral Amazonia. Knowledge of hydrological processes operating from plot to catchment scales is essential in thesenatural environments, where pressures for land-use conversion are extremely high (FAO, 2001). Ideally, land manage-ment and conversion strategies should be able to incorporate, and be conditioned by, the knowledge gained fromsuch hydrological investigations (Bruijnzeel, 1992; Bonell, 1998; Chappell et al., 2004b; Thang and Chappell, 2004).Further detailed process studies are needed to identify hillslope flow pathways and their physical controls in rainforests. This paper seeks to address this dearth of knowledge by examining the results of replicated tracer andhydrometric experiments conducted on hillslopes within two areas of SE Asia. Both areas have USDA Ultisol soilorders, specifically the USDA Udult great group (Sherlock, 1997; Chappell et al., 1999b; Soil Survey Staff, 1999).

736 N. A. Chappell and M. D. Sherlock


Study Sites

The study focused on two rain forest catchments in southeast Asia – the Bukit Timah catchment in central Singapore,and the W8S5 catchment in the interior of Sabah, Malaysian Borneo (Figure 1). The dominant characteristics of bothenvironments are summarized in Table I which shows that the two research sites have similar climates, topographyand forest type. The geology from which the soils are derived is, however, different.

Bukit Timah catchment, Singapore IslandThe study catchment (1°21′ N and 103°46′ E) on Singapore Island is small (0·05 km2), is located on the northwesternslopes of Bukit Timah hill, and ranges in altitude from 90 to 164 m. It is called the ‘Jungle Falls catchment’ bySherlock et al. (1995) and Sherlock (1997), but is referred to simply as the ‘Bukit Timah catchment’ in this paper.

Table I. Dominant characteristics of the Bukit Timah and W8S5 catchments

Site characteristic Bukit Timah catchment W8S5 catchment

Location 104° E 1° N 118° E 4° NClimate Humid tropical Humid tropicalGross annual precipitation 2369 mm 2778 mmAnnual evapotranspiration 1200–1500 mm 1200–1500 mmCatchment area 0·05 km2 1·7 km2

Topography Gentle crest zones. Moderate midslope zones Steep crest zones. Moderate midslope zonesBoth gentle and steep riparian zones Both gentle and steep riparian zones

Geology Granite Sandstone-dominatedSoil Udult UdultVegetation Lowland dipterocarp forest Lowland dipterocarp forest

Figure 1. SE Asian study site locations in (a) the Bukit Timah catchment, Singapore Island, and (b) the W8S5 catchment, Sabah,Malaysian Borneo.



Slopes are typically gentle (<10°) in the hillslope crest zones, becoming very steep (>20°) close to the channel-heads.In Singapore, rainfall is typically delivered during localized convective storms. Hourly rainfall intensities generallyrange between 20 and 50 mm h−1, while short-term intensities can exceed 100 mm h−1 (Sherlock, 1997). The catch-ment lies within the Bukit Timah Nature Reserve (BTNR) that covers an area of 0·81 km2, and comprises a remnant ofSingapore’s original climax vegetation, lowland dipterocarp forest. The Udult of Bukit Timah is derived from anunderlying granite geology of the Bukit Timah Granite Formation (Rahman, 1992). The profile characteristics of thissoil as derived by Sherlock (1997) are summarized in Table II. The Bukit Timah soil has a base saturation of 7 percent and cation exchange capacity of 4·9 cmol(+) kg−1 in the Bt horizon, and 28 per cent clay in the A horizon and43 per cent clay within the Bt horizon (Rahman, 1992). Some 80 per cent of the clay minerals in the Bt horizon arekaolinite (Rahman, 1992).

W8S5 catchment, Borneo IslandThe study catchment (5°01′ N and 117°48·75′ E) on Borneo Island is 1·7 km2 in area and ranges in altitude from 150to 300 m. The catchment lies 1 km west of the Danum Valley Field Centre (DVFC) within the East Malaysian State ofSabah. Slopes are often gentle (10°) adjacent to the stream channel, and become very steep (>25°) towards thehillslope crest zones. Down-cutting of the stream channel has, however, resulted in the formation of almost verticalsoil walls along localized stretches of the upper channel margins. Locally around the experimental site, slopes range

Table II. Generalized soil profile description of the Udult of (a) the Bukit Timah catchment and (b) the W8S5 catchment

(a)Location: Bukit Timah catchment, Central Singapore (104° E 1° N)Drainage: moderately well drainedVegetation: lowland rain forest (Dipterocarpacae sp.)Parent material: granite of the Bukit Timah Granite FormationLanduse: conservation and recreation

Horizon Depth (cm) Description

L 0–6 Forest litter layer (Dipterocarpacae sp.).A 6 –10 Dark brown (10YR 3/3), stoneless sandy clay loam; moist; weakly developed coarse granular structure;

some small voids; abundant medium and fine roots; sharp transition to B1 horizon.B1 10–30 Yellowish brown (10YR 5/6) stoneless sandy clay; a few faint brownish yellow mottles (10YR 6/6); strong

to medium blocky structure; moist; few voids; few very fine roots; gradual smooth transition to B2 horizon.B2–B4 30–90 Brownish yellow (7·5YR 6/8) stoneless sandy clay; a few very faint yellowish brown (10YR 5/8) mottles;

strong blocky structure; moist; few voids; few very fine roots.

(b)Location: W8S5 catchment, Ulu Segama region, Sabah, Malaysia (118° E 4° N)Drainage: Moderately well drainedVegetation: Lowland rain forest (Dipterocarpacae sp.)Parent material: Sandstone of the Kuamut FormationLanduse: Conservation and research

Horizon Depth (cm) Description

L 0–4 Forest litter layer (Dipterocarpacae sp.).A 4–8/10 Dark brown (10YR 3/3); stoneless sandy loam; weak granular structure; moist; abundant medium to fine

roots <2 cm diameter. Sharp transition to B1 horizon.B1 8/10–40 Bright yellowish brown (10YR 6/8) stoneless clay loam; medium blocky structure; moist; abundant roots

<1 cm diameter ; gradual smooth transition to B2 horizon.B2 40–60 Bright brown (7·5YR 5/8) clay with some bright reddish brown (2·5YR 5/8) gravel fragments; strong

blocky structure; moist; few fine roots; gradual transition to B3 horizon.B3–B4 60–100 Bright reddish brown (2·5YR 5/8) clay loam with many bright reddish brown (2·5YR 5/8), dull yellow

orange (10YR 6/4) and greyish yellow brown (10YR 4/2) gravel fragments; strong blocky structure; moist;few roots.



from gentle (<10°) to steep (>20°). Rainfall received by the W8S5 catchment is typically delivered during localizedconvective storms, lasting for less than 15 min and having intensities of less than 10 mm h−1 equivalent, when sampledat 5 min intervals (Chappell et al., 2001; Bidin, 2001). Large storms do, however, produce intensities in excess of100 mm h−1 equivalent, again sampled at 5 min intervals (Bidin, 2001). The W8S5 catchment lies towards the easternmargin of the 348 km2 area of protected forest called the Danum Valley Conservation Area. The catchment is alsocovered by lowland dipterocarp forest. The catchment’s Udult is derived from a geology locally dominated bysandstones but also including mudstone, tuff and chert that constitute the Miocene Kuamut Formation (Leong, 1974;Clennell, 1996; Chappell et al., 1999b). The characteristics of this soil (Sherlock, 1997) are also summarised inTable 2. The W8S5 soil has a base saturation of 13·7 per cent and cation exchange capacity of 58 cmol(+) kg−1 clayin the Bt horizon, and 15·5 per cent clay in the A horizon and 20·0 per cent clay within the Bt horizon (Chappell et al.,1999b). Approximately one third of the crystalline silicate clays are as kaolinite-smectite, one third as 2:1 illite, andone third as 2:1 vermiuculite (Chappell et al., 1999b). Unstable or expansive clays have been found within the area(Chappell et al., 1999b).

Field Methodology

Some recent studies have highlighted the uncertainties associated with calculating waterflow within hillslopes usingdata from tensiometry and permeametry (e.g. Sherlock et al., 1995, 2000). Other research has questioned the accuracyof the Darcy–Richards approach, particularly where natural soil pipes are present, when flow estimates derived fromsoil-water calculations have not matched measurements of throughflow or streamflow generation (e.g. Sloan et al.,1983; Tanaka et al., 1998; Koide and Wheater, 1991). Despite this, few alternative theories for the prediction ofsubsurface flow from internal hillslope characteristics are available. Notable exceptions are those approaches usingsoil moisture content dynamics (e.g. Mdaghri Alaoui et al., 1997), but these have not been evaluated widely. Hillslopestudies utilizing tensiometry and permeametry continue to be used in the tropics (Noguchi et al., 1997; Dykes andThornes, 2000), and indeed elsewhere (McGlynn et al., 2002; Uchida et al., 2002), and many physics-based catchmentmodels continue to use the results of hillslope studies to evaluate their internal dynamics (e.g. Anderton et al., 2002).This study aims to use tensiometry and permeametry in the estimation of subsurface flow routes and magnitudes.These estimates are, however, evaluated against direct measurements of the relative magnitude and direction of soil-water flow monitored with the use of artificial water tracers. Very few tropical studies have used artificial soil-watertracers (notable exceptions include Bonell et al., 1984; Cabellero et al., 2002; Reichenberger et al., 2002), and evenfewer have compared hydrometric predictions against observed tracer paths (Sherlock et al., 1995; Sherlock, 1997).

In this study, hydrometric and artificial tracer approaches were used to attempt to identify dominant flow pathwayswithin several study plots in the Bukit Timah and W8S5 micro-catchments. Each study plot was divided into two sub-plots: one sub-plot comprised arrays of vacuum samplers for water tracing experiments and the other sub-plot wasinstrumented with arrays of mercury manometer tensiometers (Figure 2). Two plots were instrumented in the BukitTimah catchment and four plots in the W8S5 catchment (Sherlock, 1997). Within this paper, the results from the 15°and 27° sloping plots in the Bukit Timah catchment and the 21° and 23·5° sloping plots in the W8S5 catchment arepresented in Figure 1. The storms used to advance the tracer into the soil had rainfall totals of 14·6 mm, 25·0 mm,18 mm and 0·8 mm for the 15°, 27°, 21° and 23·5° plots, respectively.

Hydrometric measurement and predictionsThe study plots were instrumented with five nests of mercury manometer tensiometers. Each nest comprised of fivetensiomenters installed to depths of 10, 30, 50, 70 and 90 cm, and were fitted with 0·5 bar, high-flow ceramic tips (SoilMoisture Equipment Corporation, Santa Barbara, USA) to increase the response times. The 27° and 15° Bukit Timahplots were monitored for one and eight months, respectively, and the 23·5° and 21° W8S5 plots were monitored forone and four months, respectively. Over selected rain events, the tensiometers were read at up to 5 min intervals, inorder to capture rapid changes in capillary potential during intense rainfall. Once monitoring of each study plot hadceased, a series of saturated hydraulic conductivity or permeability (Ksat) measurements was undertaken within bothmicro-catchments. In both catchments, Ksat was measured at various depths using a ring permeameter (Chappell andTernan, 1997). This technique was evaluated in Sherlock et al. (2000) by comparing the Ksat data with that obtainedfrom a Guelph permeameter (Reynolds et al., 1983). To calculate the flow vectors, hydraulic conductivity functions(K) were derived by matching the Ksat data to depth-specific relative hydraulic conductivity curves (Kr) derived fromMillington and Quirk (1960) analysis of moisture release curves (Sherlock et al., 2000). A total of five cores werecollected and moisture release curves derived for the Bukit Timah soil, and a further 22 for the W8S5 soils (Sherlock,



1997). Hydraulic gradients were calculated from tensiometer data, and combined with the hydraulic conductivity datato estimate specific flux magnitude in the downslope and vertical planes. Using the approach of Harr (1977), flowvectors were calculated for each time-step of tensiometer measurement.

Artificial water tracer testsA water tracing experiment was conducted within half of each study plot (Figure 2). A pulse of chemically conserva-tive sodium chloride (NaCl) solution was sprayed over a 1 × 5 m area of the upslope sampling array (Figure 2), usinga line source injection apparatus (Sherlock, 1997). Flow regulators were set such that 160 litres of NaCl solutionwas delivered in 15 min. This delivery rate over a 5 m2 area is equivalent to a depth of 32 mm rainfall in 15 min(128 mm h−1 rainfall intensity equivalent). This intensity of tracer addition is not atypical for the intensity of largerain-storms at Bukit Timah (Sherlock, 1997) or W8S5 (Bidin, 2001). Tracer injection at both study plots commencedapproximately 5 min following the onset of rainfall events selected to minimize the artificial circumstances of typicaltracer tests. Indeed, this may be the first study conducted in the humid tropics where natural storms are used toadvance tracer through the soil profile. The high tracer input concentration used (36 507 mg l−1 NaCl) was deemednecessary given the potential for dilution by event rainfall or pre-event soil water.

Tracer migration through the soil matrix was monitored using nests of vacuum samplers at all sites (Sherlock,1997). The vacuum samplers were fitted with 0·5 bar, high-flow ceramics (Soil Moisture Equipment Corporation,Santa Barbara, USA) as with the tensiometers. Within both catchments, these were installed at depths of 10, 30, 50and 70 cm, corresponding to the A, B1, B2 and B3 soil horizons (Table II), in arrays spaced downslope at 1 m internals(Figure 2). Data from the upslope two or three instrument arrays are presented in this paper, though all vacuumsamplers in each study plot were routinely sampled. The vacuum sampler solutions were extracted daily, except whensoils were too dry for sample collection.

Flow Pathways Interpreted

The evidence used to interpret water pathways within the two soils comes from: (i) direct observations of artificialtracer migration (the actual pathway of the water); (ii) the magnitude of the saturated and unsaturated hydraulicconductivity (shown by modelling to have a significant impact on waterflow magnitude and direction; e.g. Zaslavskiand Sinai, 1981); and (iii) flow-nets estimated from hydrometric data comprising tensiometer and hydraulic conductiv-ity measurements.

Figure 2. Instrument layout in the study plots of the Bukit Timah and W8S5 catchments.



Flow pathways based upon artificial water tracer evidenceThe hydrologic behaviour of the NaCl tracer was very different between the study plots at Bukit Timah (Figure 3a,b)and those at W8S5 (Figure 4a,b). Within the slopes at Bukit Timah, tracer breakthrough in the B1 horizon (30 cmdepth) vertically below the line source injection was rapid (Figure 3a,b) at 2·0 mm h−1 in the 15° plot and 14 mm h−1

in the 27° plot (Table III). Within the 27° plot, the initial breakthrough of tracer at 50 cm depth (B2 horizon) occurredafter only 2 h 13 min (Figure 3bA) giving an initial breakthrough velocity (VINITIAL) of 225·5 mm h−1. The time ofarrival of the tracer peak (VPEAK) at this point equates with a velocity of 228 mm h−1 (Table III). Such a rapid responsemay have resulted from water migrating to depth within ‘natural soil pipes’. Significant lateral tracer movement wasnot observed 1 m downslope of the NaCl source by the vacuum samplers, with no breakthroughs in the 15° plot(Figure 3aB) and only very small responses observed 1 m downslope in the A, B2 and B3 horizons of the 27° plot(Figure 3bB; Table III). Thus, even on the very steep slopes (27°) in this catchment, tracer inputs at the soil surfacefollowed a predominantly vertical flow pathway to at least 0·7 m depth (Figure 3a,b; Sherlock et al., 1995; Sherlock,1997).

Within the upslope array of the samplers in the Bukit Timah plots (Figure 3aA, 3bA), breakthrough curves peakedduring the monitoring period in all horizons except one. Within the 15° plot, the most damped tracer response was inthe B2 horizon (50 cm depth), while within the 27° plot it was the B3 horizon (70 cm). This observation might beexplained by these two samplers abstracting water from soil in poor contact with the active pipes responsible for the

Figure 3. NaCl tracer breakthrough curves derived using vacuum samplers in: (a) the Bukit Timah 15° plot from 10 April 1993(A) vertically below and (B) 1 m downslope of the line source injection; (b) the Bukit Timah 27° plot from 29 May 1993 (A)vertically below and (B) 1 m downslope of the line source injection.



rapid tracer migration to the other samplers, with tracer reaching these two samplers by dispersion from soil pipesthrough several centimetres of soil matrix. No overland flow was observed in the plots during the studied storms.

At W8S5, high concentrations of NaCl were observed throughout the subsoil (B1 to B3 horizons) at the 0, 1 and 2 mdownslope sampling arrays after the first to third days of sampling the 23·5° plot (Figure 4a,b). This marked 2 mlateral migration of tracer within the upper 1 m of soil in W8S5 (Figure 4a,b; Table IV) is in marked contrast with thelargely vertical flow seen at Bukit Timah (Figure 3a,b; Table III). The rate of vertical migration of tracer beneath theline source at W8S5 was, however, slower than that seen at Bukit Timah, though there is considerable variabilitybetween the response of the replicate plots at each site (Tables III and IV). As with Bukit Timah, no overland flow wasobserved in the W8S5 plots during the studied storms. This observation is supported by direct measurements ofoverland flow on slopes within 2 km of W8S5 which have less than 5 per cent of the rainfall travelling over slopesaway from channels (Sinun et al., 1992; Chappell et al., 1999a, 2004a).

To identify the physical factors controlling the different tracer behaviour between the two sites, hydraulic conduc-tivity data and then the predicted flow-nets are compared.

Hydraulic conductivity comparison with tracer evidencePrevious analyses have shown that the Ksat data collected at Bukit Timah and W8S5 are log-normally distributed(Sherlock, 1997; Chappell et al., 1998), and so best summarized with a geometric mean (Figure 5). In both catch-ments, Ksat generally decreased with depth. The magnitude of Ksat in the upper part of the Bukit Timah soil (<0·4 m

Figure 3. Continued



Figure 4. NaCl tracer breakthrough curves derived using vacuum samplers in (a) the W8S5 23·5° plot from 9 June 1994 (A)vertically below and (B) 1 m downslope of the line source injection, (b) the W8S5 21° plot from 9 September 1993 (A) verticallybelow and (B) 1 m downslope, and (C) 2 m downslope of the line source injection.

depth) was approximately four-fold greater than that at W8S5 (Figure 6). In the Bukit Timah catchment, the combinedpermeability of the litter layer and organic A horizon was very high though within the range of published studies forother tropical regions (cf. Bonell et al., 1983; Chappell and Ternan, 1992). These findings are consistent with theobserved greater rate of vertical migration of tracer to depth (70 cm) below the injectors in the Bukit Timah soil(Table III) compared to the W8S5 soil (Table IV).

The Ksat at each depth can be crudely compared with rainfall intensities to assess the likelihood of infiltration-excessoverland flow or perched saturation. As the litter layer and topsoil of the Bukit Timah catchment has a Ksat of2510 mm h−1 (Figure 5), then peak rainfall intensities of 100 mm h−1 (over 10 min) should infiltrate readily. Suchrainfall intensities would percolate through the B1 and B2 soil horizons where the minimum Ksat is 143 mm h−1.Similarly, infiltration and subsequent percolation into the W8S5 soil should not be impeded by its 539 mm h−1 topsoilpermeability or by that in the B1 and B2 soil horizons (min. Ksat is 103 mm h−1). Within the W8S5 soil, percolation intothe B3 horizon could be impeded as the permeability has reduced to 28 mm h−1 and reduces further to 0·37 mm h−1 bya depth of 1·0 m (Figure 5).

The hydraulic conductivity functions for the different horizons (derived from the Millington and Quirk analysis ofmoisture release curves: Sherlock et al., 2000) of the Bukit Timah and W8S5 soils are presented in Figure 6a,b.During wet conditions (<200 cm water suction or negative capillary potential), the rate of decline of the hydraulic



Figure 4. Continued



Table III. Peak tracer concentration velocity (VPEAK) in and peak tracer concentration in the Bukit Timah 15° and 27° plotsderived from vacuum sampler breakthroughs

Depth (m) Horizon VPEAK (mm h−−−−−1) Peak tracer conc. (mg l−−−−−1)

15° plotUpslope array (0 m) 10 A 0·51 420·1

30 B1 2·0 882·150 B2 <0·43 >451·870 B3 2·9 1 113·5

Array 1 m downslope 10 A nb nb30 B1 nb nb50 B2 nb nb70 B3 nb nb

27° plotUpslope array (0 m) 10 A 45 8 874

30 B1 14 1 8 94350 B2 23 1 7 07870 B3 <2·2 >5 367

Array 1 m downslope 10 A 48 19430 B1 nb nb50 B2 4·5 8670 B3 6·1 260

Array 2 m downslope 10 A nb nb30 B1 nb nb50 B2 nb nb70 B3 nb nb

nb = No breakthrough of tracer observed

Table IV. Peak tracer concentration velocity (VPEAK) and peak tracer concentration in the W8S5 23·5° (‘W13-2’) and 21°(‘W13-1’) plots derived from vacuum sampler breakthroughs

Depth (m) Horizon VPEAK (mm h−1) Peak tracer conc. (mg l−1)

23·5° plotUpslope array (0 m) 10 A 0·19 4908

30 B1 <0·066 >249650 B2 <0·109 >174370 B3 <0·15 >3223

Array 1 m downslope 10 A 2·04 159130 B1 <0·25 >90850 B2 <0·28 >239270 B3 <0·31 >2284

21° plotUpslope array (0 m) 10 A 0·72 5108·6

30 B1 0·84 4113·750 B2 0·91 3501·770 B3 <0·32 >1274·8

Array 1 m downslope 10 A 9·5 2392·330 B1 17·0 6522·750 B2 1·0 1010·270 B3 2·0 2126·4

Array 2 m downslope 10 A 94·3 1165·430 B1 5·9 1295·850 B2 1·4 318·970 B3 1·7 122·3



Figure 5. Saturated hydraulic conductivity profiles for the Bukit Timah and W8S5 soils. All saturated hydraulic conductivity valuesare the geometric means per soil horizon (see Table II) in centimetres per hour.

conductivity function in the Bukit Timah soils is much less than that for the W8S5 soils. This means that at 100 cmsuction the Bukit Timah soil profile is between 0·5 and 2 orders of magnitude (depending on horizon) more conductivethan the W8S5 soils (Figure 6a,b). Thus the differences between Bukit Timah and W8S5 seen within the Ksat data areeven larger within the unsaturated hydraulic conductivity values. This is again consistent with the observed greaterrate of vertical migration of tracer to depth (70 cm) below the injectors in the Bukit Timah soil (Table III) comparedto the W8S5 soil (Table IV). However, these hydraulic conductivity differences (Figure 6a,b) are much larger than thedifferences seen in the timing of the tracer peaks at depth below the injectors (Tables III and IV).

Flow-net comparison with tracer evidenceThe flow-nets calculated from the hydraulic conductivity functions and capillary potential data are given in Figures 7and 8 for the example Bukit Timah storms and Figures 9 and 10 for the example W8S5 storms. Other storm-basedflow-nets are given in Sherlock (1997) and Sherlock et al. (1995). It is clear from these flow-nets that the magnitudeof the calculated waterflow (shown by arrow size) is much greater during the example storms in Bukit Timah than inW8S5. This is strongly related to the differences in the slopes of the hydraulic conductivity functions and is consistentwith the data on the rate of vertical travel of the tracer directly below the injectors which show that Bukit Timah soilshave greater water velocities. Again, the relative difference in the water flow velocities predicted by the flow-net ismuch larger than those seen in the tracer data (i.e. VPEAK: Tables III and IV).

The critical discrepancy between the subsurface flows estimated by the hydrometric methods and those of tracermigration relates to the dominant direction of the water movement. Within the Bukit Timah soils, the downslopecomponent of flow (i.e. flow parallel to the ground surface) was predicted by the flow-net analyses to be as large asthe vertical component of flow at many locations (Figures 7 and 8), yet no significant tracer concentrations were seenat locations downslope of the tracer injections (Figure 3a,b). At W8S5, where significant downslope tracer migrationwas observed (Figure 4a,b), most of the flow vectors in the flow-net show predominantly vertical flow (Figures 9 and10). The same results were seen within the other replicated plots (Sherlock, 1997). Thus the calculated directions ofwater movement, using flow-nets based on hydrometric data, are largely inconsistent with the observed tracer migra-tion in these two contrasting catchment sites. Thus the interpretations of the water pathways from our hydrometricdata (i.e. combined tensiometer and hydraulic conductivity readings) are subject to considerable error. Uncertainty inthe patterns and magnitude of shallow subsurface flow may, therefore, be larger than errors associated with measure-ments of soil water potential and permeability seen in other studies (e.g. Sherlock et al., 2000).



Figure 6. Hydraulic conductivity curves for each soil horizon (see Figure 5 and Table II) for (a) the Bukit Timah soils, and (b) theW8S5 soils.

Discussion

The artificial tracer experiments show that the Udult soils examined have very different water pathways, both inmagnitude and direction, between the two catchments in tropical SE Asia. Predominantly vertical flows were observedduring storms in the Bukit Timah Udult on Singapore Island, while significant downslope flow components wereseen in the W8S5 Udult on Borneo Island. Only one other artificial tracer study undertaken on tropical Udult slopes isavailable for comparison with our studies. The Bonell et al. (1984) study mapped the migration of a tritium tracerapplied to a 14° plot of Udult in the South Creek Catchment, Queensland. This Australian study showed a 50:50split of vertical to lateral tracer flow (Bonell et al., 1984). The findings from our study, together with those of Bonellet al. (1984), therefore, add useful information on water pathways within a global region lacking much hillslopehydrological data.

Relation of soil type and direction of subsurface flowAdditionally, these findings add to the debate about whether there is a simple relationship between soil type (here aUSDA great group) and the dominant direction of subsurface flow (cf. Chappell and Ternan, 1992; McDonnell, 2003;Bonell, 2004). Here, both Bukit Timah and W8S5 soils are classified under the same USDA Udult group, whichequates to the Acrisol-Alisol soil group under the FAO–UNESCO system (Landon, 1991). The suggestion by Elsenbeer(2001) that Acrisol soils (or Acrisol-Alisol soils using the post-1990 differentiation) have a 40:60 split of vertical to



lateral flow (see figure 4 in Elsenbeer, 2001) is inconsistent with our findings from Bukit Timah, which show a 90:10split of vertical to lateral tracer flow (Table III; Sherlock, 1997, p. 263). Furthermore, examining the Ksat data for thesoils in the W8S5 catchment and surroundings (data taken from Chappell et al., 1998), Elsenbeer (2001) suggests thatthese soils are actually intergrades between the Acrisol and Ferralsol groups, and should have a 70:30 split of verticalto lateral flow. Our tracer measurements in the W8S5 catchment do, however, show a 15:85 split of vertical to lateralflow in the A, B1 and B3 soil horizons (table 9.6 in Sherlock, 1997). Additionally, the W8S5 soils are specificallyHaplic Alisols (Chappell et al., 1999b). According to Driessen and Dudal (1991) such Alisols, rather than being‘intergrades’, are likely to be hydraulically more dissimilar to Ferralsols than soils of the (post-1990) Acrisol class.

Nature and role of soil piping in the A to B3 horizonsThe relative difference in the magnitude of the tracer migration between Bukit Timah and W8S5 (Tables III and IV)was poorly predicted by the hydraulic conductivity data (Figures 5 and 6). The dominant direction of tracer migration,particularly in the W8S5 soils, was similarly poorly estimated by the flow-net calculations based on the hydraulicconductivity and capillary potential data (Figures 7–10). If the errors in the point measurements using the ringpermeameter and high-flow tensiometers are correctly quantified at −11 per cent and 5 per cent, respectively (Chappelland Ternan, 1997; Sherlock et al., 2000), then these errors only partially explain the discrepancies seen between theactual tracer migration and estimates based on hydrometric measurements. Perhaps point measurements themselvespoorly characterize the effective behaviour of larger volumes of soil. This can be the case where ‘natural soil pipes’are present within soils. These features are continuous tunnels cut by water or fauna that allow percolation to short-circuit the hydraulic characteristics of most of the soil matrix. These pipes have been observed in the Bukit Timah

Figure 7. Resultant flow vectors in the Bukit Timah slopes in the 15° study plot during a 14·6 mm rain event on 10 April 1993.



Figure 8. Resultant flow vectors in the Bukit Timah slopes in the 27° study plot during a 25·0 mm rain event on 29 May 1993.

soils (Sherlock et al., 1995; Sherlock, 1997) and W8S5 soils (e.g. Sinun, 1991; Chappell and Binley, 1992; Chappellet al., 1998, 1999a,b).

Soil pipes approximately 10 cm in diameter were observed in two out of the five soil pits excavated within theBukit Timah catchment (Sherlock, 1997, p. 60). More extensive observations of soil piping have been undertaken bythe authors in and around the W8S5 catchment in Sabah, Borneo Island. At depth within the soils, typically 0·50 to0·80 m (B2 to B3 horizon) below the ground surface just above the contact with the C-soil horizon, large lateral soilpipes are present. These pipes typically range in size from 0·1 to 0·6 m in diameter and are typically associated withchannel head or zero-order basin areas (Bidin, 1995; Sherlock, 1997; Chappell et al., 1999a). Baillie (1975) similarlyfound pipes in channel head locations in Ultisols of neighbouring Brunei. At Danum large lateral pipes are also foundon planar slopes away from channels. Unlike some pipe systems in temperate UK which have an associated surfacesubsidence depression along the line of the pipe (see Gilman and Newson, 1980), there is very little surface expressionof these large lateral pipes. In a few hillslope hollow areas which also show the build-up of dark leaf matter, a lineardepression marks the course of each pipe. More commonly, small collapse features (10–30 cm in diameter) form atirregular intervals along the course of the lateral pipes. Outlets from these large lateral pipes are present in some



Figure 9. Resultant flow vectors in the W8S5 slopes in the 23° study plot during an 18·0 mm rain event on 9 June 1994.

channel banks and heads, but also appear on some planar slopes; these are identifiable from the presence of lag gravelsat the outlets and visual observation of emerging water during storms. Some of these lateral pipes have been gaugedfor water-flow and turbidity estimation (Chappell et al., 1999a; Sayer et al., 2004). The selectively logged Barucatchment lies within 2 km of the W8S5 catchment. Within the Baru, where road cutting close to the steep divides hasgenerated landslides across active channels, large lateral soil pipes have developed within the landslide materials. Onthe undisturbed slopes, the upslope extent of the lateral pipes (beyond some 5 to 10 m) is poorly defined, as are thecritical pathways of the water to these lateral pipes. With careful observation of the surface of the soils in the W8S5catchment and surrounding areas, many vertical holes, 2 to 3 cm in diameter, can be seen. Many of these holes are theburrows formed by cicada nymphs (cf. O’Green and Busacca, 2001). Sinun (1991) counted the number of ‘soilchimneys’ above the newly cut cicada burrows on six 20 m2 plots just outside the W8S5 catchment on three occasionsin 1990. His work showed that the burrowing activity was very variable between the plots with an area of some 2 km2

and at different times of the year (Table V). The chimneys eventually erode away completely (Sinun, 1991), but leavethe 2–3 cm diameter vertical shafts in the soil surface (for an unknown period of time). The average of 15 soilchimneys per square metre of slope (Table V) means that there is at least this density of vertical shafts on the soilsurface. As yet, the precise nature of the connectivity of these vertical biopores and the larger lateral pipes is notknown. Some are certainly connected to the lateral pipes, as we have been able to pump water at rates of several litresper second into these vertical pipes/burrows without completely filling them. During storms, some of these verticalshafts serve as outlets for extended pipe systems and maintain 20 cm high ‘fountains’ of water.

Given that large lateral pipes are extensive with the W8S5 soils and possibly within the Bukit Timah soils also, andmust have a good connection with the surface inputs of water (via cicada burrows?) to generate the fast response timesobserved (Bidin, 1995; Sayer et al., 2004), then it is not unreasonable to assume that they impact on the direction and



Figure 10. Resultant flow vectors in the W8S5 slopes in the 21° study plot during an 0·8 mm rain event on 9 September 1994.

Table V. Presence of ‘soil chimneys’ produced by new cicada burrowing activity on six20 m2 plots on planar slopes close to the Danum Valley Field Centre, Sabah, Malaysia (afterSinun, 1991)

Chimneys recorded on measurement date

Plot 6 April 1990 24 June 1990 18 August 1990 Mean (m−−−−−2)

P1 18 20 41 1·3P2 691 577 375 27P3A 40 77 134 4·2P3B 60 123 230 21P4A 104 175 306 9·8P4B 326 556 734 27

Mean 15



magnitude of subsurface flow. The 7200 cm3 undisturbed cores used for ring permeametry, even if they contain a 2–3 cm vertical pipe, cannot characterize the effective behaviour of the vertical pipes (let alone a large lateral pipe) thatextends over tens of metres or even tens of centimetres. This has already been shown by Chappell et al. (1998)working in the same area. They demonstrated that saturated hydraulic conductivity estimated from tests on wholehillslopes containing vertical and lateral pipes can be two orders of magnitude larger than that derived from integrat-ing the results from 7200 cm3 undisturbed cores. The response of a tensiometer, which samples a soil volume ofperhaps 1000 cm3, is likely to depend on the proximity of that sampled soil volume to the pipe network. Given theway that the characteristics of natural soil pipe systems are poorly described by small-scale permeametry andtensiometery, then it is not surprising that flow-nets derived from these measurements within piped soils poorly matchobserved tracer plumes. It may be that the observed difference in tracer migration between the W8S5 and Bukit Timahsoils is more related to differences in the vertical and lateral distribution of soil pipes, than to differences in thehydraulic conductivity distributions. New methods of quantifying the effect of pipe networks on the hillslope charac-teristics such as effective permeability or the response of tensiometers and piezometers are needed to allow observedtracer flumes to be modelled using the intrinsic soil characteristics.

Conclusions

Following the seminal work of Bonell et al. (1984), the value of using artificial tracers to identify water pathways intropical soils has been illustrated further, as has the role of tracers in the evaluation of the hydrometric or flow-netapproach to subsurface flow calculation (Sherlock et al., 1995; Reichenberger et al., 2002). Our results would suggestthat there is currently insufficient information in the humid tropics to allow simple generalizations about the magni-tude and direction of water movement within specific soil groups. Such generalizations (Chappell and Ternan, 1992;Sherlock, 1997; Elsenbeer, 2001) are critical for rapid site evaluation as part of environmental impact assessments(EIAs), but need to be based on a much greater set of hillslope experiments than is currently available. The lack ofcorrespondence between observed tracer plumes and hydrometric predictions means that these new hillslope experi-ments must contain synchronous tracer studies. New ways of properly capturing the effects of small and large soilpipes within simple site characteristics such as ‘effective block permeability’ (Chappell et al., 1998; Lancaster, 1999)and capillary potential are required urgently.

AcknowledgementsThe authors would like to thank Dr Tony Greer and Professor Teo Siew Eng (National University of Singapore), Dr Kawi Bidin(Universiti Malaysia Sabah) and Dr Waidi Sinun and others of the Research and Development Division of Yayasan Sabah. Dr TanWee Kiat, the Director of the National Parks Board of Singapore, is thanked for granting permission to conduct research within theBukit Timah Nature Reserve, Singapore. The Danum Valley Management Committee, the Economic Planning Unit of the PrimeMinister’s Department of Malaysia (Research psdd 0095), the Sabah State Secretary and the Sabah Chief Minister’s Department arethanked for permission to conduct research in the Danum Valleu area of Dabah, Malaysia. Comments by Professor Keith Beven andRoy Sidle are much appreciated. This work has been undertaken as part of the Danum Valley Rainforest Research and TrainingProgramme under project 110 with the close collaboration of Dr Waidi Sinun and Dr Kawi Bidin. The study is also a core elementof the hydrology programme of the southeast Asia rainforest research programme of the Royal Society of London (publicationreference A/363). Financial support for this work has been provided by the UK Natural Environment Research Council studentshipgrant GT/AAPS/28 and Lancaster Universty.

References

Anderton S, Latron M, Gallart F. 2002. Sensitivity analysis and multi-response, multi-criteria evaluation of a physically based distributedmodel. Hydrological Processes 16: 333–353.

Baillie IC. 1975. Piping as an erosion process in the uplands of Sarawak. Journal of Tropical Geography 41: 9–15.Bidin K. 1995. Suburface Flow Controls of Runoff in a Bornean Natural Rainforest. MSc thesis, University of Manchester, UK.Bidin K. 2001. Spatio-temporal variability in rainfall and wet-canopy evaporation within a small catchment recovering from selective

tropical forestry. PhD thesis, University of Lancaster, UK.Bonell M. 1998. Selected challenges in runoff generation research in forests from the hillslope to headwater drainage basin scale. Journal of

the American Water Resources Association 34: 765–785.Bonell M. 2004. Runoff generation processes in tropical forests. In Forests, Water and People in the Humid Tropics, Bonell M, Bruijnzeel

LA (eds). Cambridge University Press: Cambridge; 314–406.Bonell M, Gilmour DA. 1978. The development of overland flow in a tropical rainforest catchment. Journal of Hydrology 39: 365–382.



Bonell M, Gilmour DA, Casells DS. 1983. A preliminary survey of the hydraulic properties of rainforest soils in tropical north-eastQueensland and the implications for the runoff processes. In Rainfall Simulation, Runoff, and Soil Erosion, De Ploey J (ed.). CatenaSuppl. 4: 3–24.

Bonell M, Casells DS, Gilmour DA. 1984. Tritiated water movement in clay soils of a small catchment under tropical rainforest in NorthEast Queensland. In ISSS Symposium on Water and Solute Movement in Heavy Clay Soils, Bouma J, Raats PAC (eds). ILRL PublicationNo. 37: 197–201.

Bruijnzeel LA. 1992. Managing tropical forest watersheds for production: where contradictory theory and practice co-exist. In WiseManangement of Tropical Forests, Proceedings of the Oxford Conference on Tropical Forests 1992, Miller FR, Adam KL (eds). OxfordForestry Institute: Oxford; 37–75.

Caballero Y, Jomelli V, Chevallier P, Ribstein P. 2002. Hydrological characteristics of slip deposits in high tropical mountains (CordilleraReal, Bolivia). Catena 47: 101–116.

Chappell NA, Binley A. 1992. Impact of rain forest disturbance upon near-surface groundwater flow: modelling of hillslope flow experi-ments. Annales Geophysicae 10(II): c330.

Chappell NA, Ternan JL. 1992. Flow-path dimensionality and hydrologic modelling. Hydrological Processes 6: 327–345.Chappell NA, Ternan L. 1997. Ring permeametry: design, operation and error analysis. Earth Surface Processes and Landforms 22: 1197–

1205.Chappell NA, Franks SW, Larenus J. 1998. Multi-scale permeability estimation for a tropical catchment. Hydrological Processes 12: 1507–

1523.Chappell NA, McKenna P, Bidin K, Douglas I, Walsh RPD. 1999a. Parsimonious modelling of water and suspended-sediment flux from

nested-catchments affected by selective tropical forestry. Philosophical Transactions of the Royal Society of London Series B 354: 1831–1846.

Chappell NA, Ternan JL, Bidin K. 1999b. Correlation of physicochemical properties and sub-erosional landforms with aggregate stabilityvariations in a tropical Ultisol disturbed by forestry operations. Soil and Tillage Research 50: 55–71.

Chappell NA, Bidin K, Tych W. 2001. Modelling rainfall and canopy controls on net-precipitation beneath selectively-logged tropical forest.Plant Ecology 153: 215–229.

Chappell NA, Douglas I, Hanapi JM, Tych W. 2004a. Source of suspended-sediment within a tropical catchment recovering from selectivelogging. Hydrological Processes 18: 685–701.

Chappell NA, Nik AR, Yusop A, Tych W, Kasran B. 2004b. Spatially-significant effects of selective tropical forestry on water, nutrient andsediment flows: a modelling-supported review. In Forests, Water and People in the Humid Tropics, Bonell M, Bruijnzeel LA (eds).Cambridge University Press: Cambridge; 513–532.

Clennell B. 1996. Far-field and gravity techtonics in Miocene basins in Sabah, Malaysia. In Tectonic Evolution of Southeast Asia, Hall R,Blundell D (eds). Special Publication No. 106. Geological Society: London; 307–320.

Driessen PM, Dudal R. 1991. The Major Soils of the World. Agricultural University Wageningen: Wageningen.Dykes AP, Thornes JB. 2000. Hillslope hydrology in tropical rainforest steeplands in Brunei. Hydrological Processes 14: 215–235.Elsenbeer H. 2001. Runoff generation in tropical rainforests – a review. Hydrological Processes 15: 1751–1759.FAO. 2001. State of the World’s Forests 2001. FAO: Rome.Gilman K, Newson MD. 1980. Soil Pipes and Pipeflow; A Hydrological Study in Upland Wales. British Geomorphological Research Group

Research Monograph No. 1. Geo Books: Norwich.Harr RD. 1977. Water flux in soil and subsoil on a steep forested slope. Journal of Hydrology 33: 37–58.Koide S, Wheater HS. 1991. Subsurface flow simulation of a small plot at Loch-Chon, Scotland. Hydrological Processes 6: 299–326.Lancaster J. 1999. Multi-scale estimation of effective permeability within the Greenholes Beck catchment. PhD thesis, Lancaster University,

UK.Landon JR. 1991. Booker Tropical Soil Manual. A Handbook for Soil Survey and Agricultural Land Evaluation in the Tropics and

Subtropics. Longman Technical: New York.Leong KM. 1974. The Geology and Mineral Resources of the Upper Segama Valley and Darvey Bay Area, Sabah, Malaysia. Geological

Survey of Malaysia, Memoir 4 (revised). Government Printer: Kuching.McDonnell JJ. 2003. Where does water go when it rains? Moving beyond the variable source area concept of rainfall-runoff response.

Hydrological Processes 17: 1869–1875.McGlynn BL, McDonnell JJ, Brammer DD. 2002. A review of the evolving perceptual model of hillslope flowpaths at the Maimai

catchments, New Zealand. Journal of Hydrology 257: 1–26.Mdaghri Alaoui A, Germann P, Lichner L, Novak V. 1997. Preferential transport of water and 131-iodide in a clay loam assessed with TDR-

techniques and boundary-layer flow theory. Hydrology and Earth System Sciences 4: 813–822.Millington RJ, Quirk JP. 1960. Permeability of porous solids. Transactions of the Faraday Society 57: 1200–1207.Noguchi S, Abdul Rahim N, Baharuddin K, Tani M, Sammori T, Morisada K. 1997. Soil physical properties and preferential flow pathways

in a tropical rain forest, Bukit Tarek, Peninsular Malaysia. Journal of Forest Research 2: 115–120.Nortcliff S, Thornes JB. 1981. Seasonal variations in the hydrology of a small forested catchment near Manaus, Amazonas, and the

Implications for its management. In Tropical Agricultural Hydrology, Lal R, Russell EW (eds). John Wiley: Chichester.O’Green AT, Busacca AJ. 2001. Faunal burrows as indicators of paleo-vegetation in eastern Washington, USA. Palaeography Palaeoclimatology

Palaeoecology 169: 23–37.Rahman A. 1992. Soils of Singapore. In Physical Adjustments in a Changing Landscape: The Singapore Story, Gupta A, Pitts J (eds).

Singapore University Press: Singapore; 144–189.



Reichenberger S, Amelung W, Laabs V, Pinto A, Totsche KU, Zech W. 2002. Pesticide displacement along preferential flow pathways in aBrazilian Oxisol. Geoderma 110: 63–86.

Reynolds WD, Elrick DE, Topp GC. 1983. A re-examination of the constant head well permeameter method for measuring saturatedhydraulic conductivity above the water table. Soil Science 136: 250–268.

Sayer AM, Walsh RPD, Bidin K. 2004. Pipeflow suspended sediment dynamics and their contribution to stream sediment budgets in smallcatchments in Sabah, Malaysia. In Forests and Water in Warm, Humid Asia, Sidle RC, Tani M, Nik AR, Taddese TA (eds). Proceedings ofa IUFRO Forest Hydrology Workshop. Disaster Prevention Research Institute: Uji; 170–173.

Sherlock MD. 1997. Plot-scale hydrometric and tracer characterisation of soil water flow in two tropical rain forest catchments in SoutheastAsia. PhD thesis, Lancaster University, UK.

Sherlock MD, Chappell NA, Greer AG. 1995. Tracer and Darcy-based identification of subsurface flow, Bukit Timah Forest, Singapore.Singapore Journal of Tropical Geography 16: 197–215.

Sherlock MD, Chappell NA, McDonnell JJ. 2000. Effects of experimental uncertainty on the calculation of hillslope flow paths. Hydrologi-cal Processes 14: 2457–2471.

Sinun W. 1991. Hillslope hydrology, hydrogeomorphology and hydrochemistry of an equatorial lowland rainforest, Danum Valley, Sabah,Malaysia. MSc thesis, University of Manchester, UK.

Sinun W, Wong WM, Douglas I, Spencer T. 1992. Throughfall, stemflow, overland flow and throughflow in the Ulu Segama rain forest,Sabah, Malaysia. Philosophical Transactions of the Royal Society of London Series B 335: 389–395.

Sloan PG, Moore ID, Colthorpe GB, Eigel JD. 1983. Modelling surface and subsurface stormflow on steeply-sloping forested watersheds.Research Report 142. Water Resources Research Institute, University of Kentucky: Lexington.

Soil Survey Staff. 1999. Soil Taxonomy: A basic system of soil classification for making and interpreting soil surveys (second edition).USDA, US Government Printing Office: Washington DC.

Tanaka T, Yasuhara M, Sakai H, Marui A. 1998. The Hachioji experimental basin – storm runoff processes and the mechanisms of itsgeneration. Journal of Hydrology 102: 139–164.

Thang HC, Chappell NA. 2004. Minimising the hydrological impact of forest harvesting in Malaysia’s rain forests. In Forests, Water andPeople in the Humid Tropics, Bonell M, Bruijnzeel LA (eds). Cambridge University Press: Cambridge; 852–865.

Uchida T, Kosugi K, Mizuyama T. 2002. Effects of pipe flow and bedrock groundwater on runoff generation in a steep headwater catchmentin Ashiu, central Japan. Water Resources Research 38: 1119.

Zaslavski D, Sinai G. 1981. Surface hydrology: IV. Flow in sloping, layered soil. American Society of Civil Engineers 107(HY1): 53–64.

earth surface processes and landforms hydrological ... · pdf filecontrasting ﬂow pathways...

Documents