Gully erosion in the Siwalik Hills, Nepal: estimation of sediment production from active ephemeral gullies

Gully erosion in Nepal 155

Copyright © 2006 John Wiley & Sons, Ltd. Earth Surf. Process. Landforms 31, 155–165 (2006)

Earth Surface Processes and LandformsEarth Surf. Process. Landforms 31, 155–165 (2006)Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/esp.1320

Gully erosion in the Siwalik Hills, Nepal: estimationof sediment production from active ephemeralgulliesSohan K. Ghimire,1* Daisuke Higaki2 and Tara P. Bhattarai31 Ministry of Water Resources, Department of Irrigation, Kathmandu, Nepal2 Department of Regional Environment Science, Hirosaki University, Japan3 Department of Geology, Tribhuwan University, Nepal

AbstractA simple field-based monitoring programme was established in a small catchment (area4·6 km2) to find the rates of gully erosion in the Siwalik Hills, Nepal. The rates are used toestimate the amount of sediment produced by gully erosion in the catchment. Three largeand active gullies were selected with areas ranging from 0·44 to 0·78 ha. Aerial photographstaken in 1964, 1978 and 1992 were ortho-rectified and used to study the dynamics of gullyheads. The same gullies were also monitored manually using an orthogonal reference systemfixed by erosion pins around the gully heads. Results from the aerial photos indicated thatthe gullies expanded remarkably over the period from 1964 to 1992, by 34 to 58 per cent.Head-retreat rates during that period were 0·48, 0·55 and 0·73 m a−−−−−1 and average annualsediment evacuation was estimated as 2534 ±±±±± 171, 959 ±±±±± 60 and 2783 ±±±±± 118 m3 a−−−−−1 for thethree gullies respectively. From the field measurement, estimated volumes were found tovary from 731 ±±±±± 57 to 2793 ±±±±± 201 m3 a−−−−−1 over the monitoring period of two years. It was alsofound that the gullies produce sediment which accounts for up to 59 per cent of the sedimentproduced from surface erosion in the headwater catchment. The findings are useful forplanning and executing appropriate control measures and constructing a sediment hazardmap at the catchment scale. Copyright © 2006 John Wiley & Sons, Ltd.

Keywords: gully erosion; gully-head retreat; sediment production; Siwalik Hills; Nepal.

Introduction

Gullies are relatively deep, unstable, eroding channels that form at the head, side or floor of valleys where no well-defined channel previously existed (Schumm et al., 1984). Gully expansion occurs mainly by gully-head erosion. As aresult of this process, gullies move upslope, releasing sediment to the channels and exposing new channel walls toerosion. Hence, depending upon the rate of retreat, gully erosion may represent an important sediment source in arange of environments (Poesen et al., 2003) and may also act as a sensitive indicator of environmental change(Oostwoud Wijdenes and Bryan, 2001). Therefore, the rate of gully growth is one of the most important indicatorsused for assessing the area destroyed by gully erosion and evaluating the many kinds of damage it causes.

In the recent past, many studies have been carried out to determine rates of gully-head expansion and producesediment budgets either by repeated measurements over time (e.g. Oostwoud Wijdenes and Bryan, 2001; Derose et al.,1998; Thomas and Welch, 1988) or by using sequential series of aerial photographs (Vandekerckhove et al., 2003;Nachtergaele and Poesen, 1999; Vandaele et al., 1996). Nachtergaele and Poesen (1999) outlined the potential of aerialphotos in the study of gully development. The main advantage of this technique is that it is time-saving and cost-effective. Aerial photographs also permit the ephemeral gully-erosion survey to be extended in time. The maindrawback is that small gullies are invisible. In contrast to the aerial photography method, high accuracy is possible byfield survey. However, field measurement is costly and the area or number of gullies to be covered is limited.

Throughout the Nepalese Himalaya, gully erosion is a widespread form of land degradation which is highlysensitive to climate and land-use changes (LRMP, 1986). Yet research on gully erosion is very limited. A recent study

*Correspondence to: S. K.Ghimire, Ministry of WaterResources, Department ofIrrigation, Kathmandu, Napal.E-mail: [email protected]

Received 10 February 2004;Revised 28 January 2005;Accepted 11 August 2005

156 S. K. Ghimire, D. Higaki and T. P. Bhattarai


was carried out by Higaki et al. (1998) in the Mid-Hills where they studied the rates and processes of gully-headexpansion in the laterite soil, and potential countermeasures to check gully erosion.

A catchment called Khajuri (area 4·6 km2) was selected in the Siwalik Hill region in order to investigate land-degradation problems, where gully erosion is predominant. The Siwalik Hills, which are the southernmost hill range ofthe Himalaya, exhibit a very immature topography with highly rugged terrain dissected by numerous gullies (LRMP,1986). With a weak geological formation consisting of Tertiary to early Quaternary unconsolidated sandstones, mudstonesand conglomerates and heavy monsoon rains, the gullies produce and deliver a significant amount of sediment to thetorrents. Flash floods in the torrents are highly sediment-laden with potential downstream effects such as avulsion andinundation leading to environmental degradation (Ghimire et al., 2003a). Moreover, gullies sometimes bring disasterby destroying areas of settlement and leaving many people homeless (DPTC, 1993). In order to assess the hazardlevels of different types of sediment sources in a catchment and to mitigate sediment hazards, it is essential to estimatetheir erosion potential in specific climates, topographies and land uses. Thus, the main objective of this paper is topresent the findings of a field-based study on the rates of gully erosion in the Siwalik Hills. The rates are then used toestimate the amount of sediment produced by gully erosion in a small catchment. The data pertain to the rate ofsediment production, rather than the sediment yields, because we have no means of reliably assessing the sediment-delivery ratios in the catchment.

Study Area and Gully Characteristics

The research site is located in Trijuga river basin in Udayapur district in eastern Nepal (Figure 1). Elevation frommean sea level varies from 165 m on the floodplain to 370 m at the hilltop. Mean maximum and minimum monthlytemperatures are 32 and 14 °C respectively in the months of July and January. Annual average rainfall is 1500 mm(ICIMOD, 1996). The area was selected for study because it is one of the areas facing the problems of bank cuttingand inundation by flash floods generated from a large network of gullies and steep channels.

A wide range of gullies exist in terms of size and geometry, and their erosional activity varies widely. However, inview of the importance of the large active gullies in generating significant sediment in a catchment, the study wasfocused on such gullies. The gullies are characterized by actively eroding tall headwalls with minimal or no vegetationcover and large amounts of debris deposited in the fan area. Three such gullies – Khajuri-1 (KG1), Khajuri-2 (KG2)and Musahar-1 (MG1) – were selected within a small area so that topography, land use and lithology could be

Figure 1. Location map of the study area.



Figure 2. Gully head of the KG1 site.

assumed to be similar. Study of the distribution of the sediment sources has indicated that such active gullies arewidely distributed in the headwater reach of many catchments in the study area (Ghimire, 2005). Hence, the studygullies can be considered representative of many active gullies in the area.

The gully-head catchment areas of the three gullies KG1, KG2 and MG1 are 0·60, 0·36 and 0·88 ha respectively.Gully heads are typically high (maximum height up to 50 m) and are formed in different geological beds, mainlyalternate layers of sandstone and mudstone with thick intermediate layers of gravel and boulders (DPTC, 1998). Theyconsist of nearly vertical faces which are dissected by numerous rills (Figure 2). When the rills coalesce downstream,a series of drainage channels is formed on the foot of the headwall. The gully head therefore consists of four distinctcomponents: head-rim, free-face, drainage channel and pediment slopes on the foot of channels. Drainage channels,which are usually V-shaped, are composed mainly of unconsolidated gravel and boulder layers. Side slopes of thechannels are almost bare which indicates they are in an active stage of erosion (Figure 3). Main gully channels areshort and narrow, and sideslopes are covered with good vegetation. It should be noted here that sediment productionfrom the main gully channels has not been addressed in this study because a large proportion of the sediment isproduced from the gully-head area.

Methodology

Analysis of gully enlargement derived from aerial photographsBlack and white aerial photographs taken in 1964 (scale 1:12 500) were obtained from the Department of Forestry,and those of 1978 and 1992 (scale 1:50 000), were obtained from the Department of Surveying, Katmandu. Thetopographic map (scale 1:25 000) compiled from the ground-verified 1992 aerial photograph, which was prepared bythe Land Resource Mapping Project (LRMP), Nepal, was taken as a reference map. First, the aerial photographs andtopographic map were scanned with a resolution of 800 dots per inch (315 dots per centimetre) and were saved inTIFF format. Using ERDAS Imagine version 8.5, the topographic map was first georeferenced to the UniversalTransverse Mercator (UTM) system by selecting ground control points such as road intersections and mature trees.Polynomial rectification of the 1992 aerial photo was performed using the same software by registering many groundcontrol points (GCPs) on the topographic map as well as on the photograph. This process of rectification involves thestretching or compression of the image in as uniform a manner as possible in order to match the base map locations ofground control points. The 1992-rectified photo was subsequently used as reference image for the rectification of 1978and 1964 aerial photos in the same way. All rectified images were then imported to Adobe Illustrator software version10. Gullies were traced and parameters were measured by overlaying all the rectified image layers and topographicbase map. Two variables were investigated: area and the retreat rate of the gully head. Growth rates were calculated bymeasuring the difference in gully-head area or head retreat over the time period between successive aerial photographs.



Field measurement of gully erosionGully-head retreat was measured as the change in distance between the edge of the gully head and benchmark pinsaround the gully head. Erosion from the channels was measured by inserting iron pins (1 cm diameter, 30–50 cmlength) perpendicular to the side-slope surface and repeatedly measuring the exposed segment (Ghimire et al., 2003b).

Based on form and process, erosion sources were divided into two types: retreat of vertical faces (headwallerosion), and channel-side-slope erosion. In order to look for the retreat pattern during the rainy season (from June toSeptember), field measurements were carried out before and after the rainy season as well as during the dry season. Itshould be noted here that head retreat of only two gullies – KG1 and MG1 – could be measured because the gullyhead of KG2 was inaccessible. For the measurement of channel erosion, one representative channel from each gullywas selected.

Estimation of sediment productionSediment volume derived from head retreat has been calculated in two ways: from aerial photography and fieldmeasurement. From aerial photography, difference in gully-head area is calculated from 1964 and 1992 aerial photo-graphs and total sediment volume is calculated by multiplying this difference by the average height of the gully head.The height of the gully head is taken as the distance between the gully-head rim and a level below which slopingchannels are formed, and it is assumed to be constant. Average height was calculated by averaging the heightsmeasured from different points on the head-rim. Channel erosion is excluded from this estimation because change inchannel forms and dimension was difficult to detect from the photographs.

From field measurements, eroded sediment volume was calculated for the two monitoring years by also consideringside-slope erosion from the channels. Figure 4 shows the measured parameters used to estimate the sediment produc-tion. Sediment displaced by headwall erosion has been estimated by multiplying length of head rim (Lh), average

Figure 3. Illustration of an active drainage channel of the KG1 gully system. Remarkable changes are evident on the side slopes(marked by white circles) such as undercutting and mass failure.



height of gully head (Hh) and average headwall erosion (amount of retreat) (Eh). Similarly, sediment volume from theerosion of channel side slopes has been estimated by multiplying the length of channel (Lc), height of side slopes (Hc)and the average depth of side-slope erosion (Ec). Repeated survey indicated that there were no significant changes inthe channel-bed level, hence sediment production from the bed has not been considered. Because channel forms andmaterial properties do not differ much between and within the gullies, average channel dimensions could give reason-able estimates of sediment volume.

Results

Gully-area enlargement from aerial photographyGully heads traced from the rectified aerial photos were analysed for the head-area enlargement rate. Area anddistance were measured by using a digital planimeter. Figure 5 shows the tracing sketches of head-rims from the threesequential photographs, and Table I summarizes changes in the gully-head catchment area and retreat rates for theanalysis period.

Figure 4. Sketch depicting the measured parameters used to estimate the sediment production from a gully head.

Table I. Increase in gully-head area and head retreat from 1964to 1992

Year/period KG1 KG2 MG1

Head area (m2) 1964 3842 2436 65751978 4672 2896 79421992 6059 3555 8801

Head area (%) 1964–1978 21·6 18·9 20·81978–1992 29·7 22·7 10·81964–1992 57·7 45·9 33·8

Head retreat (m) 1964–1978 2·9 5·9 12·71978–1992 10·5 9·7 7·71964–1992 13·3 15·6 20·4

Av. rate (28 yrs) (m a−1) 0·48 0·55 0·73



Figure 5. Enlargement of the gully-head catchment area of the study gullies derived from aerial photograph interpretation.

Figure 5 indicates that the enlargement rate of the head catchment area varied significantly in space and time.Maximum head retreat occurred in the headward direction; however, sideward expansion was also noticed. Table Ireveals that the catchment area of the gully heads increased by about 19 to 22 per cent from 1964 to 1978, and theareas further increased by 11 to 30 per cent from 1978 to 1992. Thus, in the whole period of 28 years, KG1 increasedby 58 per cent, KG2 by 46 per cent and MG1 by 34 per cent. Total head retreats of 13·3 m, 15·6 m and 20·4 moccurred in the whole period. This resulted in the average rates of head retreat, which were 0·48, 0·55 and 0·73 m a−1

for the three gullies respectively.

Field measurement of gully erosionErosion measurements undertaken during the study periods are summarized in Table II, which indicates that gullyheadwalls did not retreat uniformly along the perimeter. There was also a considerable temporal variation in erosionamounts. From June to September 2002, maximum headwall erosion was 89 cm and 30 cm in KG1 and MG1;however, it was 26 and 9 cm in the same period of 2003. Similarly, there was considerable temporal and spatialvariation in channel-side-slope erosion. It also indicates that there was virtually no erosion during winter or dryseasons. Differences in amounts of erosion in the two years can therefore be explained by the difference in rainfallpattern. Figure 6 shows that monthly rainfall in July 2002 was 680 mm whereas it was 390 mm in the same month in2003. Also, more intense and bigger rainstorms were observed in July 2002. Although the contribution of a singlerainstorm to gully development is not clear, it seems that higher erosion rates in 2002 could be as a result of higherand more intense rainfall events.

Estimation of sediment productionComputations of the displaced volume of sediment using the two methods are shown in Tables III and IV.

Table III shows that the amounts of sediment generated from the gullies KG1, KG2 and MG1 were 2534 ± 171,959 ± 60 and 2783 ± 118 m3 a−1 respectively. Two types of errors may occur in the volume computations: random andsystematic errors. Random errors were a result of the photogrammetric rectification. Residual georeferencing errorwas estimated by fixing as many ground control points as possible leaving others ‘free’ and then comparing actual andmapped location of the free points, the method followed by Micheli and Kirchner (2002). This procedure was repeatedfor various ground control points to generate spatially variable uncertainty estimates. Residual spatial error (differencebetween actual and mapped location of a feature) estimated for the 1992 image was ±7 m. This value is contrasted toerrors of up to 35 m estimated for non-rectified images of the same scale. The spatial errors for 1978 and 1964 images



Figure 6. Monthly rainfalls in the monitoring years 2002 and 2003.

Table III. Estimation of sediment volume from gully head by aerial photography

Area in Area in Diff. Av. height* ∆∆∆∆∆ volumeSediment production

Gully 1964 (m2) 1992 (m2) (m2) (m) (m3 ××××× 103) (m3 a−−−−−1) (t a−−−−−1)†

KG1 3842 6059 2217 32 ± 2·1 70·9 ± 4·8 2534 ± 171 3547 ± 239KG2 2436 3555 1119 24 ± 1·5 26·8 ± 1·7 959 ± 60 1343 ± 84MG1 6575 8801 2226 35 ± 1·4 77·9 ± 3·3 2783 ± 118 3896 ± 165

* The error limits are 95 per cent confidence limits.† Bulk density considered as 1·4 t m−3.

Table II. Monitoring data of headwall retreat and channel-side-slope erosion (all values are in cm)

2 Jun. 02– 1 Oct. 02– 16 Apr. 03– 7 Jun. 03–Gully 30 Sep. 02 15 Apr. 03 6 Jun. 03 25 Sep. 03

KG1 Headwall, n = 7 Min. 3·0 0·0 0·0 0·0Max. 89·0 0·0 6·0 20·0Mean 28·0 ± 22·0 0·0 ± 0·0 1·4 ± 1·4 6·0 ± 5·7

Channel, n = 8 Min. 4·0 0·0 0·0 0·0Max. 40·0 0·0 0·0 16·0Mean 22·0 ± 9·5 0·0 ± 0·0 0·0 ± 0·0 8·2 ± 4·4

KG2 Headwall – – – –Channel, n = 13 Min. 3·5 0·0 0·0 0·0

Max. 28·0 0·0 0·0 11·0Mean 8·8 ± 4·4 0·0 ± 0·0 0·0 ± 0·0 2·9 ± 2·2

MG1 Headwall, n = 6 Min. 2·0 0·0 0·0 0·0Max. 30·0 3·0 0·0 9·0Mean 16·0 ± 10·0 1·0 ± 1·0 0·0 ± 0·0 4·0 ± 3·0

Channel, n = 10 Min. 20·0 0·0 0·0 6·0Max. 35·0 0·0 5·0 20·0Mean 27·0 ± 2·5 0·0 ± 0·0 1·0 ± 1·0 10·3 ± 3·4

n, Number of erosion pins; Min. and Max., minimum and maximum erosion; Mean, mean of all erosion pins. Error limits are 95 per cent confidence limits.



Table IV. Estimation of sediment volume by field measurement

2002 2003

KG1 KG2 MG1 KG1 KG2 MG1

Headwall erosionLength of head, Lh (m) 273 208 359 273 208 359Height of head, Hh (m) 32 24 35 32 24 35Av. retreat, Eh (m) 0·28 – 0·16 0·07 – 0·04Volume (m3) 2446 ± 126 – 2010 ± 50 612 ± 40 – 503 ± 15

Channel sideslope erosionNo. of channel, N 9 12 11 9 12 11Av. length of channel, Lc (m) 25·0 18·0 21·0 25·0 18·0 21·0Av. height of sideslope, Hc (m) 3·5 3·0 4·5 3·5 3·0 4·5Av. erosion Ec (m) 0·22 0·09 0·27 0·08 0·03 0·11Volume* (m3) 347 ± 75 117 ± 26 561 ± 21 126 ± 32 39 ± 13 229 ± 42

Total eroded volume (m3) 2793 ± 201 – 2572 ± 71 738 ± 72 – 731 ± 57Total eroded weight (t)† 3910 ± 281 – 3600 ± 99 1033 ± 101 – 1024 ± 80Gully head area (ha) 0·60 0·36 0·88 0·60 0·36 0·88Erosion rate (t ha−1) 6517 ± 468 – 4091 ± 113 1722 ± 168 – 1164 ± 91

* Volume = 2 × N × Lc × Hc × Ec (2 for two sideslopes).† Bulk density of sediment as 1·4 t m−3.The error limits are 95 per cent confidence limits.

were ±8 and ±5·5 m respectively. In addition, there might be some errors in tracing the head-rim, but working in adigital format allowed us to zoom in on the features, which could limit the potential errors significantly. For all aerialphotographs, tracing was done by a single person who is very familiar with the actual field environment so as to keepthe potential errors in tracing as minimal as possible. Since these random errors tend to cancel out each other whenmany control points are taken around the gullies, they were not considered in computing errors in the estimation ofsediment volume.

Systematic errors result from deficiencies in measurement or processing, and will affect the degree to which thefinal data agree with reality. These errors resulted from the inaccuracies in height measurement of the gully heads. Thestandard error was estimated as the 95 per cent confidence limits in the height measurement which was ±2·1, ±1·5 and±1·4 m for gullies KG1, KG2 and KG3 respectively. The errors were then propagated into the final estimate ofsediment volume.

Table IV indicates significant differences in total displaced volume in the two monitoring years. A total erosion of2793 ± 201 m3 in KG1 and 2572 ± 71 m3 in MG1 took place in 2002; however, it decreased to 738 ± 72 m3 and731 ± 57 m3 respectively in 2003. It also illustrates that the contribution of channel erosion to total erosion is less thanthat of gully-head erosion. The error limits in volume were estimated by 95 per cent confidence limits in the measure-ment of height of gully head and measurement of erosion (head retreat and side-slope erosion).

Contribution of gully erosion at catchment scaleThe headwater zone of the study catchment (area 4·6 km2) where gullies are formed consists of degraded forest andshrubland. In the catchment, monitoring results using erosion pins indicated average annual soil losses of 1·0 and1·2 mm from forest and shrubland respectively (Ghimire, 2005). This corroborates the findings of Laban (1978) whoestimated the denudation rate to be 1 mm a−1 from degraded forest in the far western Siwalik region. It should be notedthat even though stream channels could make a significant contribution to the total sediment production from thecatchment, they have not been considered for comparative purposes since the monitoring data are lacking.

Estimates of sediment production by surface erosion in the catchment are presented in Table V. This shows thataverage sediment volume produced from the headwater catchment was 4750 m3 a−1. If the sediment volume producedby individual gullies (Table III) is compared to the total volume, KG1, KG2 and MG1 produced about 53, 20 and59 per cent of total surface erosion respectively. When all gullies are considered in the catchment scale, their totalcontribution would be about 1·3 times more than the surface erosion. This clearly indicates that the gullies serve asimportant sources of sediment in the catchment.



Table V. Estimation of surface erosion from catchment

Area Av. erosion Volume Weight RateLand use (ha) (mm a−−−−−1) (m3 a−−−−−1) (t a−−−−−1) (t ha−−−−−1 a−−−−−1)

Forest 391 1·0 3910 5474 14·0Shrub land 70 1·2 840 1176 16·8Total 461 4750 6650 14·4

Discussion

Comparison of erosion rates with other studiesThere is virtually no information available on the rate and process of gully erosion in the Siwalik Hill region, so wecould not compare the erosion rates of the study gullies to other similar gullies in the same geophysical environment.However, it is noteworthy that erosion studies by Higaki et al. (1998) in a degraded river terrace of the Mid-Hillregion measured gully-head retreat using the method of erosion pins. The gullies were relatively smaller in size withgully-head heights ranging from 10 to 15 m. The material was composed of red soil mixed with gravel and boulders.The average rate of gully retreat was found to vary from 3 to 82 cm a−1. Despite the difference in size, topography andland use, these rates are in a similar range to our measured rates. One of the reasons for this similarity could be therainfall pattern, which is similar for both regions.

Difference in sediment estimation from the two methodsThe estimated sediment amount from the two methods differed significantly. In aerial photography, only headwallretreat was considered as a source of sediment. Channel-side-slope erosion was also considered in the field measure-ment method even though its contribution to the total sediment production seemed less. If only sediment volumesfrom headwall estimated by the two methods are compared, they would differ by 8 per cent for KG1 and 80 per centfor MG1 in 2002. As the measured erosion was consistently less, the difference would be even more for 2003. Thereare a number of possible factors to explain this discrepancy. Average retreat rates derived from long-term intervalsmask seasonal effects or other periods of stagnation or activity, but so do persistent trends (Oostwoud Wijdenes andBryan, 2001). In the short term, retreat rates show much variation since factors other than runoff, such as thedevelopment of tension cracks and deposition of sediment below the headcut, become increasingly significant. Manystudies, such as that carried out by Vandekerckhove et al. (2001), indicated less variability in long-term retreatmeasurements than in short-term ones. In our case, variation in erosion amounts in the two years was significant,possibly due to the difference in rainfall pattern as explained before. However, it would be less in the case of a longertime period because rainfall characteristics may average out over time.

Retreat of the headwall is often referred to as the shifting of the head-rim upstream. However, erosion may occur atthe foot of the headwall which may not result in collapse of the complete wall, so no retreat is measured. This type ofpartial block collapse was evident in many locations of gullyheads. It may be possible that the propagation of the massfailure on the headwall up to the head-rim may take a longer time or require several storm events, especially in thecase of tall gully heads.

Effect of rainfall on erosionIn most of the studies related to sediment production by headcut advance, rainfall is used as a predicting parameter;however, explicit relationships between the rainfall amount and head retreat rates are often not clear. OostwoudWijdenes and Bryan (2001), for example, found no apparent relationship between rainfall amount and gully-headretreat rates from 32 storm events. They attributed this poor correlation to the differential erosion of the head, whichcould be particularly true for the high gully heads. In calculating sediment production it was assumed that retreatoccurred over the whole contour of the gully head. But measured retreat rates differed widely along the head-rim. Thissuggests that only two years of data would not be sufficient to explain the long-term sediment production rates. Themonitoring data thus simply represent a short-term erosion process, in which controlling parameters such as rainfallhave a strong dominance producing short-term fluctuations in erosion rate.



Process of erosionThe erosion process largely depends on the geological formation of the headwall. The gully headwall consists ofalternate beds of unconsolidated silt, sand and boulders derived from siltstone, sandstone and conglomerate. Atsaturation, selective washout of sand and silt induces freefall of boulders, which ultimately leads to mass failure. Thecollapsed materials fill up the plunge pool and bed of drainage channels temporarily, which could contribute somewhatto checking the undercutting of drainage channels. Since the channels are very steep, this effect lasts for a short timeuntil a few rainstorms occur. This process may result in episodic mass movements of the headwall and drainagechannels. The most conspicuous feature of the failure process is the formation of cracks on the head-rim. A similarprocess of gully-head retreat has been reported by Higaki et al. (1998). They observed that cracks were formed on thehead-rims during the dry season followed by mass failure in the rainy season.

Conclusions

This paper describes the results of a study aimed at estimating the sediment production from large active gullies in asmall catchment of Siwalik Hills. Historical aerial photographs and recent field monitoring data were used in thestudy. Long-term change analysis by sequential aerial photo comparison indicates that the gullies expanded remark-ably over the period between 1964 and 1992 by 34 to 58 per cent. Maximum retreat rates during the period for thethree gullies KG1, KG2 and MG1 were 0·48, 0·55 and 0·73 m a−1 respectively. Estimated eroded volumes were2534 ± 171, 959 ± 60 and 2783 ± 118 m3 a−1 from the gullies. From the field measurement, the eroded volumes esti-mated for KG1 and MG1 were 2793 ± 201 and 2572 ± 71 m3 respectively in 2002, and 738 ± 72 and 731 ± 57 m3 in2003. Significant variation of erosion rates in the two monitoring years indicated that gully erosion largely dependsupon the controlling factors such as rainfall pattern in the short-term period. Differences in estimated volume from theaerial photography and field measurement could be due to the fact that average retreat rates derived from long-termintervals mask seasonal effects. Also, there could be differential erosion along the head-rim as a result of rainfall in ashort-term period; however, the effects of rainfall characteristics on headwall erosion may average out in the longterm.

This paper has outlined the complexities in headwall and drainage channel erosion pattern, especially in multi-bedgeological formations. Mass failure is found to be the main process of erosion. Sediment production from these gulliesseemed episodic in nature depending on the rainfall patterns.

The study has also identified such large gullies as being significant sediment sources, contributing 20 to 59 per centof total surface erosion in these headwater catchments. The gullies therefore can be considered as a dominant sourceof sediment in the catchment. These results highlight the regional importance of gully erosion and draw attention tothe need for conservation measures, especially in the degraded catchments with dense gully networks. Further inves-tigation is required to identify and quantify other types of erosion processes such as rill erosion and stream channelerosion, which could be important for constructing a sediment-hazard map at the catchment scale.

AcknowledgementsThe research was supported by funding from Sabo Technical Centre, Japan, which is gratefully acknowledged. We also acknowledgethe help and cooperation from the Department of Water Induced Disaster Prevention (DWIDP), Katmandu during the field works.

References

Derose RC, Gomez B, Marden M, Trustrum NA. 1998. Gully erosion in Mangatu forest, New Zealand, estimated from digital elevationmodels. Earth Surface Processes and Landforms 23: 1045–1053.

DPTC. 1993. Photo album disaster of July 1993 in Nepal. Water Induced Disaster Technical Prevention Centre: Nepal.DPTC. 1998. Landslide and erosion study in Khajuri Khola and Musahar Khola watersheds, Udayapur district. Report. Water Induced

Disaster Technical Prevention Centre: Nepal.Ghimire SK. 2005. Hillslope and fluvial erosion processes in the micro-scale catchments of flash-flood streams in the Siwalik Hills of Nepal

Himalaya. PhD Thesis, United Graduate School, Iwate University, Japan.Ghimire SK, Higaki D, Bhattarai TP. 2003a. Flash floods in the Siwaliks: Consequences and countermeasures, a case study in Nepal.

Proceedings of 1st International Conference on Hydrology and Water Resources in Asia Pacific Region (APHW), 13–15 Mar, KyotoJapan: 833–838.

Ghimire SK, Higaki D, Sato G. 2003b. Measurement of surface, rill, gully and streambank erosion in a Siwalik watershed of Nepal. TohokuJournal of Natural Disaster Science 39: 171–176.



Higaki D, Karki KK, Gautam CS. 1998. Soil erosion problems in the midland of Nepal with reference to the Trisuli model site, Nuwakotdistrict, central Nepal. News Bulletin of Nepal Geological Society 15: 33–38.

ICIMOD. 1996. Climatic and Hydrological Atlas of Nepal. International Centre for Integrated Mountain Development: Katmandu, Nepal.Laban P. 1978. Field measurements of erosion and sedimentation in Nepal. Paper no. 5. Department of Soil Conservation and Watershed

Management: Katmandu.LRMP. 1986. Land Systems, Land Utilization and Agriculture-Forestry Reports. Land Resource Mapping Project. Kenting Earth Sciences

Ltd: Ottawa, Canada.Micheli ER, Kirchner JW. 2002. Effects of wet meadow riparian vegetation on streambank erosion. 1. Remote sensing measurements of

streambank migration and erodibility. Earth Surface Processes and Landforms 27: 627–639.Nachtergaele J, Poesen J. 1999. Assessment of soil losses by ephemeral gully erosion using high-altitude (stereo) aerial photographs. Earth

Surface Processes and Landforms 24: 693–706.Oostwoud Wijdenes DJ, Bryan R. 2001. Gully-head erosion processes on a semi-arid valley floor in Kenya: a case study into temporal

variation and sediment budgeting. Earth Surface Processes and Landforms 26: 911–933.Poesen J, Nachtergaele J, Verstraeten G, Valentin C. 2003. Gully erosion and environmental change: importance and research needs. Catena

50: 91–133.Schumm SA, Harvey MD, Watson CC. 1984. Incised Channels. Water Resources Publications: Colorado; 200–208.Thomas AW, Welch R. 1988. Measurement of ephemeral gully erosion. Transactions of the ASAE 31(6): 1723–1728.Vandaele K, Poesen J, Govers G, Van Wesemael B. 1996. Geomorphic threshold conditions of ephemeral gully incision. Geomorphology 16:

161–173.Vandekerckhove L, Poesen J, Oostwoud Wijdenes DJ, Gyssels G. 2001. Short-term bank gully retreat rates in Mediterrarean environments.

Catena 44(2): 133–161.Vandekerckhove L, Poesen J, Govers G. 2003. Medium-term gully headcut retreat rates in Southeast Spain determined from aerial photo-

graphs and ground measurements. Catena 50: 329–352.

Documents

Gully erosion in the Siwalik Hills, Nepal: estimation of sediment production from active ephemeral gullies