Vous êtes sur la page 1sur 12

Hydrological Sciences-Journal-des Sciences Hydrologiques, 42(6) December 1997


Sediment yield estimation using GIS

Department of Civil Engineering, University ofRoorkee, Roorkee 247 667, UP, India

National Institute of Hydrology, Roorkee 247667, India
Abstract A method has been developed in the present study for the determination of
the sediment yield from a catchment using a GIS. The method involves spatial
disaggregation of the catchment into cells having uniform soil erosion characteristics. The surface erosion from each of the discretized cells is routed to the
catchment outlet using the concept of sediment delivery ratio, which is defined as a
function of the area of a cell covered by forest. The sediment yield of the catchment
is defined as the sum of the sediments delivered by each of the cells. The spatial
discretization of the catchment and the derivation of the physical parameters related
to erosion in the cells are performed through a GIS technique using the Integrated
Land and Water Information Systems (ILWIS) package.

Estimation de l'exportation de sdiments par utilisation d'un SIG

Rsum Dans cette tude une mthode d'estimation de l'exportation de sdiments
d'un bassin utilisant un SIG a t dveloppe. La mthode repose sur une
dicrtisation du bassin en mailles dont les caractristiques d'rosion du sol sont
uniformes. L'rosion de surface de chacune de ces mailles est route l'exutoire du
bassin en utilisant le concept de fraction sdimentaire exporte, dfinie comme
fonction de la surface de maille recouverte par la fort. L'exportation de sdiments
du bassin est dfinie comme la somme des sdiments fournis par chacune des
mailles. La discrtisation spatiale du bassin et la dtermination des paramtres
physiques relatifs l'rosion dans les mailles sont raliss grce une technologie
fonde sur les SIG utilisant le logiciel ILWIS (Integrated Land and Water

Estimates of sediment yield are needed for studies of reservoir sedimentation, river
morphology, soil and water conservation planning, water quality modelling and
design of efficient erosion control structures. The process of soil erosion by rainfall
and runoff mainly consists of the detachment and transport by raindrops and runoff.
Models available in the literature for sediment yield estimation can be grouped into
two categories: (i) physically-based models; and (ii) lumped models. In the
physically-based models the ground surface is generally separated into inter-rill and
rill erosion areas. Detachment over inter-rill areas is considered to be by the impact
of raindrops because flow depths are shallow, while runoff is considered to be the
dominant factor in rill detachment and sediment transport over both rill and inter-rill
areas. The physically-based models include AGNPS (Young et al., 1987),
ANSWERS (Beasley et al., 1980), WEPP (Hearing et al., 1989) and SHE (Abbott et
al., 1986; Wicks & Bathurst, 1996). Physically-based models are expected to

Open for discussion until 1 June 1998


Umesh C. Kothyari & Sanjay K. Jain

provide reliable estimates for the sediment yield. However, these models require the
coordinated use of various sub-models related to meteorology, hydrology, hydraulics
and soil. As a result the number of input parameters for some of the models may be
as high as 50, e.g. in the case of the WEPP model (Nearing et al., 1989). Therefore,
the practical application of these models is still limited because of uncertainty in
specifying model parameter values and also due to the difference between the scales
of application i.e. a catchment vs a field (Hadley et al., 1985; Wu et al., 1993).
Alternatively, lumped models such as the universal soil loss equation (USLE)
(Wischmeier & Smith, 1978), modified universal soil loss equation (MUSLE)
(Williams, 1978) or revised universal soil loss equation (RUSLE) (Renard et al.,
1991), combine the erosion from all processes over a catchment into one equation.
Rainfall characteristics, soil properties and ground surface conditions are represented
by empirical constants in these methods. The lumped methods of sediment yield
estimation are in frequent use in many parts of the world (Bogardi et al., 1986;
Julien & Tanago, 1991; Kothyari et al, 1994, 1996). There is ample evidence that
the USLE yields a good estimate of the amount of detached soil (surface erosion) at
the plot scale (Wischmeier & Smith, 1978). However, in the case of a catchment,
part of the eroded soil is deposited within the catchment before its outlet.
Nevertheless, the catchment can be sub-divided into sub-areas for representing
spatial heterogeneity. Surface erosion as computed using the USLE in the sub-areas
can be routed to the catchment outlet using any appropriate procedure.
Some studies have been carried out wherein a Geographic Information System
(GIS) was used for the determination of the potential for soil erosion in different plot
size areas (Bocco & Valenzuela, 1988; Omakupt, 1989; Jurgens & Fander, 1993;
Jain & Saraf, 1995; Dutta et al., 1995). A GIS technique, however, is best suited for
quantification of the heterogeneity in the topographic and drainage features of a
catchment (Schumann, 1993; Schultz, 1994; Beven, 1989). Therefore in the present
study a GIS technique has been utilized for the spatial discretization of a catchment
into hydrologically homogeneous cells. The GIS technique is also utilized for the
determination of those physical parameters of individual cells that are related to soil
erosion. Surface erosion is then computed within the individual cells using the
USLE. Next, the eroded sediment is routed to the catchment outlet by rationally
accounting for the process of sediment delivery. Thus a GIS technique is used in the
present study in conjunction with a method for prediction of sediment yield from a
catchment. The results are reported below.

The data on sediment yield resulting from storm events used herein are from the
Karso catchment in Bihar, India. The catchment area is 27.9 km2. The catchment lies
in a sub-humid tropical climate. On average it receives some 1240 mm of rainfall
annually. The temperature in the catchment varies from 42.9C (maximum) to 2.4C
(minimum). Its soil mainly consists of light sandy loam, while about 40% of its area
is forested. Agriculture is practised on about 50% of the total area. The data for this

Sediment yield estimation using GIS


catchment include the variation of rainfall, runoff, sediment yield and various
parameters of the USLE. In addition, the land contour map and the land-use map
were also available. The catchment runoff was gauged by Damodar Valley
Corporation (SWCD, 1991, vol. II). Rainfall was measured using recording
raingauges. An automatic water level recorder was used to measure the stream stage
and the runoff was derived using the rating curve. The sediment yield was measured
using a Coshocton wheel silt sampler. Data for five storm events were compiled. Out
of these, two storm events were used for verification of the present method after data
for the remaining storm events had been used for the determination of coefficients
(i.e. calibration). Dates of the storms were 3-4 August 1991; 4-5 August 1991; 1719 August 1991; 27-28 August 1991; and 28-29 August 1991.

Apart from rainfall and runoff, the erosion rates from an area are also strongly
dependent upon its soil, vegetation and topographic characteristics. These characteristics vary greatly within the various segments of a catchment. Therefore, a
catchment should be discretized into smaller homogeneous units before making the
computations for soil loss. A grid- or cell-based discretization is the most commonly
used procedure both in physically-based models and in lumped models (Beven et al.,
1984). For the present study, a cell-based discretization procedure was adopted. The
cell size to be used for discretization should be small enough such that the cell
encompasses a hydrologically homogeneous area. Keeping this in mind a cell size of
500 m x 500 m was used. There is ample evidence in the literature that the USLE
produces acceptable estimates of surface erosion over small areas (Wischmeier &
Smith, 1978). Therefore, soil erosion within each cell was estimated with the USLE.
The USLE method was expressed by Williams & Berndt (1972) as:
S' = R-KL-S-CP


in which 5' = computed soil loss per unit area, R = rainfall factor, K = soil
erodibility factor, L = slope length, S = gradient factor, C = cropping management
factor and P = erosion-control practice factor. Values of the USLE parameters were
derived following Kothyari et al. (1994). Only the parameter R is found to vary with
storm event. In each storm event, the soil loss was computed using equation (1) for
each rain amount of duration 15 min. The algebraic sum of these losses over the
storm duration became the computed soil loss for the storm event within that
particular cell. The eroded sediment was routed from each cell to the catchment
outlet using the concept of sediment delivery ratio as described below.

Sediment delivery ratio

In the case of a catchment, part of the eroded soil is deposited within the catchment
before its outlet. The ratio of sediment yield to total surface erosion is termed the

Umesh C. Kothyari & Sanjay K. Jain


sediment delivery ratio, DR. Empirical formulae have been developed for DR for a
catchment, in terms of the catchment area and length of relief, by various
investigators (Maner, 1958; Roehl, 1962; Williams & Berndt, 1972; Walling, 1983,
1988; Richards, 1993). In a cell-based approach for the determination of sediment
yield the DR values for individual cells are first defined. King (Hadley et al., 1985)
considered the DR for two adjacent cells to be the ratio of the average land slope of a
given (draining) cell to that of the adjacent (receiving) cell. In an approach for
sediment yield determination based upon the time-area segmentation of a catchment,
Kothyari et al. (1994, 1996) considered the DR of two adjacent time-area segments to
be a combined function of the slope ratio, the area ratio, and the ratio of percent
forest cover of the corresponding segments. In the present method, the effects of area
and slope on sediment delivery are considered by choosing uniform size cells and
deciding on the drainage direction within the cells on the basis of steepest descent.
The DR of the z'th cell is therefore defined as below:


where Cx is a coefficient and FA is the fraction of the area of the z'th cell covered by
forest. Now let S' be the amount of erosion produced within the z'th cell estimated
using equation (1). Then sediment yield S for the catchment during the storm event is
then obtained via:

s = I>;,s/


where N is the total number of cells in the catchment, and the term D'R is the
fraction of S't that ultimately reaches the catchment outlet. The value of D'R for the
z'th cell is derived from:



The GIS package used in the present study was ILWIS (Integrated Land and Water
Information System, ITC, 1992), which was developed at the International Institute
of Aerospace Survey and Earth Sciences (ITC), Ensehede, The Netherlands
(Biesheuvel & Hemker, 1993). The parameters needed for the estimation of surface
erosion, sediment delivery ratio and sediment yield are generated and stored in
ILWIS. The base map depicting drainage pattern and elevation contours of the study
area were prepared using the Survey of India topographical map at a scale of
1:15 000. These maps were then converted into digital form with the help of a
digitizer. The catchment was also discretized into a grid network, using ILWIS, as
shown in Fig. 1 by adopting square cells having equal sides of 500 m. The elevation
contour map was rasterized and, using interpolations from isolines, converted into a

Sediment yield estimation using GIS



\1 \





-^ \


v^ /


X jl

k /

X *- Fig. 1 Flow directions of the study area.

Fig. 2 Elevation (DEM) map of the study area.


Umesh C. Kothyari & Sanjay K. Jain

digital elevation model, DEM. This is depicted in Fig. 2 in the form of elevations for
the cells.
In ILWIS, water from any grid cell is permitted to flow to one of its eight nearest
neighbouring cells. Using the grid of terrain elevations, a grid of flow directions was
created for the study area with one direction for each cell which represents the
direction of steepest descent amongst the eight permitted choices. The grid map
which depicts the drainage directions finally obtained is shown in Fig. 1. As can be
seen a unique drainage direction is obtained for each cell of the network.
A land-use map of the study area was prepared and converted into digital form in
ILWIS (Fig. 3). For computations regarding the extent of land use in each grid cell,
Fig. 3 was reclassified into two categories viz. forest (vegetation) and others. Next,
for each cell, the percentage of forest cover was computed after overlaying thus
yielding a land-use map over the grid network. The DR value for each cell was then
computed via equation (2). The D'R values were also determined for each cell from
equation (4) and are depicted in Fig. 4.
The parameter values of the USLE, viz. K, L, S, C and P were finally
determined for each cell by overlaying the pertinent maps on the grid network. The
combined term KLSCP does not vary with the storm events in a given cell of the
catchment. Thus it represents the erosion potential of that cell. The term KLSCP was
determined and is depicted in Fig. 5 for the grid network. This Figure indicates the
potential for erosion in different segments of the catchment.

Fig. 3 Land-use map of the study area.

Sediment yield estimation using GIS

Fig. 4 Sediment delivery ratio values of the study area.

Fig. 5 KLSCP values of the study area.

Umesh C. Kothyari & Sanjay K. Jain


First, the data selected for calibration were used to determine the optimum value of
the coefficient Cx in equation (2). The method of grid search was used for this
purpose. In this method, different values were given to C, and the sediment yields
for these storm events obtained using equations (l)-(4) for each value. The C, value
giving minimum error as defined by equation (5) was considered as the optimum
one. If Sj and SS, are the corresponding observed and estimated values of the
sediment yield then the error of the estimate is defined as:



where m is the number of storm events. The C, value in this grid search was varied
between zero and one at intervals of 0.001. The optimum value of C, thus obtained was
0.94. A close scrutiny of equation (2) reveals that a value of Cx equal to 0.94, would
indicate almost negligible delivery of eroded sediment from the forested cell areas.
This result is realistic as the overland flow velocity is low in such cells because of the
high frictional resistance offered to the flow by vegetation present in these areas
(Beasley et al., 1980; Cooper et al, 1987). Likewise equation (2) indicates that, other
conditions remaining the same, almost all the soil eroded in the non-forested cells is
transported away by the flow. This is attributed to the larger overland flow velocity
resulting from less frictional resistance (Henderson, 1966). Using this value of C, the
sediment yield was computed for the storm events selected for calibration as well as for
verification. The computed sediment yields were compared with the corresponding
observed values as shown in Table 1. A study of Table 1 indicates that sediment yields
were overestimated for events of smaller duration, i.e. storm events 1, 2 and 4, while
underestimation occurred for storm events having larger duration viz. storm events 3
and 5. However, the R values for the latter storm events are smaller indicating an
observed value for the delivery ratio to be more than 100 in these events. This can be
attributed to remobilization of the sediment stored in the catchment during preceding
storm events (Piest et al., 1975; Walling, 1983).

Table 1 Computed and observed values of sediment yield.


Rainfall event date


Sediment yield (t):



3-4 August 1991 (C)

4-5 August 1991 (C)
17-19 August 1991 (C)
27-28 August 1991 (V)
28-29 August 1991 (V)




C = calibration; V = verification.


Sediment yield estimation using GIS


Table 1 also indicates that the present method produced sediment yields with a
reasonable accuracy in two storm events selected for calibration and in one storm
event selected for verification. The larger error for one of the storm events was
attributed to the uncertainty of observations. Nevertheless, the prediction accuracy of
the proposed method was reasonable particularly considering that predictions from
some of the process-based models show still larger scatter in a plot between
measured and computed sediment yields (Wu et al., 1993). The accuracy of the
proposed method is considered to be satisfactory considering the given estimation,
measurement and cartographic errors that are endemic in such analysis. It should be
noted that the proposed method depends on calibration against a record of past and
present conditions and hence it cannot be directly used to predict the impacts of
future changes in catchment land use or climate.
Figures 4 and 5 were overlaid in ILWIS to identify the source areas for sediment
reaching the outlet from within the catchment. Through such an overlaying the areas
producing large sediment amounts in the catchment were identified and are indicated in
Fig. 6. It may be emphasized that these regions would need special priority during the
implementation of erosion control measures. A comparison of Fig. 6 with Figs 2 and 3
reveals that, as expected, the sources of sediment in the catchment coincide with the
less vegetated and steep areas.

Fig. 6 Sediment source areas in the study area.


Umesh C. Kothyari & Sanjay K. Jain

A grid- or cell-based approach has been used along with a GIS for the determination
of the sediment yield from a catchment. The GIS technique was used for
discretization of the catchment into a network of cells which possess unique drainage
directions. Also, the physiographic parameters of the grids were determined using
the GIS. Surface erosion in the individual cells was determined using the USLE. The
eroded sediment was routed to the catchment outlet using the concept of sediment
delivery ratio as defined by equation (2). Reasonable results were obtained when the
proposed method was used for the determination of sediment yield for several storm
events in one catchment in India. The method depends on calibration against a record
of existing conditions and hence it can be used for the estimation of sediment yield in
other such ungauged catchments which have similar hydrometeorological and land
use conditions.

Acknowledgement The authors wish to thank the anonymous reviewers whose

comments greatly improved the quality of the paper.

Abbott, M. B., Bathurst, J. C , Cunge, J. A., O'Connell, P. E. & Rammussen, J. (1986) An introduction to the
European Hydrological System-System Hydrologique Europen, SHE. /. Hydrol. 87, 45-59.
Beasley, D. B., Huggins, L. F. & Monke, E. J. (1980) ANSWERS: A model for watershed planning. Trans. ASCE 23,
Beven, K. J. (1989) Changing ideas in hydrologythe use of physically based models. /. Hydrol. 105, 157-172.
Beven, K. J., Kirkby, M. J., Schofield, N. & Tagg, A. F. (1984) Testing a physically based flood forecasting model
(TOPMODEL) for three UK catchments. J. Hydrol. 49, 119-143.
Biesheuvel, A. & Hemker, C. J. (1993) Groundwater modelling and GIS: integrating MICRO-FEM and ILWIS.
Application of GIS in Hydrology and Water Resources Management, ed. by K. Kovar & P. Nachtnebel, (Proc.
Vienna Conf., April 1993), 289-296. IAHS Publ. no. 211.
Bocco G., & Valenzuela, C. R. (1988) Integration of GIS and image processing in soil erosion studies using ILWIS. TTC
Journal. 4, 309-319.
Bogardi, I., Bardossy, A., Fogel, M. & Duckstein, L. (1986) Sediment yield from agricultural watersheds. J. Hydraul.
Engng. Proc. ASCE 112(HY 1), 64-70.
Cooper, J. R., Gilliam, J. W., Damsels, R. B. & Robarge, W. P. (1987) Riparian areas as filters for agricultural
sediment. Soil Sci. Soc. Am. J. 51, 416-420.
Dutta, D., Das, S. N. & Sharma, J. R. (1995) Potential use of remote sensing and GIS for mapping of soil erosion
through USLE. Proc. Nat. Symp. on Remote Sensing of Environment with Special Emphasis on Green Revolution,
(22-24 November, Ludhiana, India), 267-278.
Hadley, R. F., Lai, R., Onstad, C. A., Walling, D. E., & Yair, A. (1985) Recent developments in erosion and sediment
yield studies. Report UNESCO (IHP), Paris, France.
Henderson, F. M. (1966) Open Channel Flow (Chapters 8 and 9). Macmillan, New York, 285-404.
ITC (1992) The Integrated Land and Water Information System (ILWIS) (3rd edn). Int. Inst, for Aerospace Survey and
Earth Sciences, Enschede, The Netherlands.
Jain, S. K. & Saraf, A. K. (1995) GIS for the estimation of soil erosion potential. /. GIS, India 4(1), 3-6.
Julien, P. Y. & Tanago, M. G. D. (1991) Spatially varied soil erosion under different climates. Hydrol. Sci. J. 36(6)
Jurgens, C. & Fander, M. (1993) Soil erosion assessment and simulation by means of SGEOS and ancillary digital data.
Int. J. Remote Sensing 14(15), 2847-2855.
Kothyari, U. C , Tiwari, A. K. & Singh, R. (1994) Prediction of sediment yield, J. Irrg. & Drain. Engng Div. ASCE
120(6), 1123-1131.
Kothyari, U. C , Tiwari, A. K. & Singh, R. (1996) Temporal variation of sediment yield. J. Hydrol. Engng Div. ASCE
1(4), 169-176.

Sediment yield estimation using GIS


Maner, S. B. (1958) Factors affecting sediment delivery ratio in the Red Hill physiographic area. Trans. AGI] 39(4),
Nearing, M. A., Foster, G. R., Lane, L. J. & Finkener, S. C. (1989) A process based soil erosion model for USDA
water erosion prediction project technology. Trans. ASCE 32(5), 1587-1593.
Omakupt, M. (1989) Soil erosion mapping using remote sensing data and GIS. Proc. 10th ACRS, 23-29 November,
Kuatalampur, Malaysia.
Piest, R. F., Kramer, L. A. & Heinemann, H. G. (1975) Sediment movement from loessial watersheds. In: Present and
Prospective Technology for Predicting Sediment Yields and Sources, 130-141. US Dept. of Agric. Publ. ARS-S40,
Renard, K. G., Foster, G. R., Weesies, G. A. & Porter, J. P. (1991) RUSLE, Revised Universal Soil Loss Equation. /.
Soil and Water Conserv. January-February, 30-33.
Richards, K. (1993) Sediment delivery and the drainage network. In: Channel Network Hydrology, ed. by K. Beven &
M. J. Kirkby, 222-254. Wiley, Chichester, UK.
Roehl, i. W. (1962) Sediment source areas, delivery ratios and influencing morphological factors. In: Symposium ofBari
(1-8 October 1962), 202-213. IAHS Publ. no. 59.
Schultz, G. A. (1994) Meso-scale modelling of runoff and water balances using remote sensing and other GIS data.
Hydrol. Sci. J. 39(2), 121-142.
Schumann, A. H. (1993) Development of conceptual semi-distributed hydrological models and estimation of their
parameters with the aid of GIS. Hydrol. Sci. J. 38(6), 519-528.
SWCD (Soil and Water Conservation Division) (1991) Evaluation of the Hydrological Data (vol. I and II), 269.
Ministry of Agriculture, Govt, of India, New Delhi, India.
Walling, D. E. (1983) The sediment delivery problem. J. Hydrol. 65, 209-237.
Walling, D. E. (1988) Erosion and sediment yieldsome recent perspectives. J. Hydrol. 100, 113-141.
Wicks, J. M. & Bathurst, J. C. (1996) SHESED: A physically based, distributed erosion and sediment yield component
for the SHE hydrological modelling system, /. Hydrol. 175, 213-238.
Williams, R. & Berndt, H. D. (1972) Sediment yield computed with universal equation. J. Hydraul. Engng Div. ASCE
98(HY12) 2087-2098.
Williams, J. R. (1978) A sediment graph model based on instantaneous unit sediment graph. Wat. Resour. Res. 14(4),
Wischmeier, W. H. & Smith, D. D. (1978) Predicting Rainfall Erosion Losses. Agriculture Handbook no. 537, USDA,
Science and Education Administration, 58.
Wu, T. H., Hall, J. A. & Bonta, J. V. (1993) Evaluation of runoff and erosion models. J. Irrig. Drain. Engng Div.
ASCE 119(4), 364-382.
Young, R. A., Onstad, C. A., Bosch, D. D. & Anderson, W. P. (1987) AGNPS: An agriculture nonpoint source
pollution model. Conservation Research Report 35, USDA/ARS, Washington, DC, USA.
Received 26 July 1996; accepted 24 March 1997