African Journal of Agricultural Research Vol. 6(18), pp. 4185-4198, 12 September, 2011 Available online at http://www.academicjournals.org/AJAR DOI: 10.5897/AJAR11.418 ISSN 1991-637X ©2011 Academic Journals

Full Length Research Paper

Evaluation of integrated KW-GIUH and MUSLE models to predict sediment yield using geographic information

system (GIS) (Case study: Kengir watershed, Iran)

Saleh Arekhi1*, Afshin Shabani2 and Sayed Kazem Alavipanah3

1Department of Watershed Management, Agriculture College, University of Ilam, Ilam, Iran.

2Department of Cartography, Geography College, University of Tehran, Tehran, Iran.

3Department of Cartography, Faculty of Geography, University of Tehran, Tehran, Iran.

Accepted 12 August, 2011

In this study, the Kinematic Wave-Geomorphologic Instantaneous Unit Hydrograph-Modified Universal Soil Loss Equation (KW-GIUH-MUSLE) model was tested for its sediment yield estimation potential on the Kengir watershed in Iyvan city of Ilam Province, Iran. The runoff factor was calculated for six storms during 2000. The spatial distribution of soil erodibility factor was taken from the attribute data pertaining to the soils of the study area. The topographic factor (LS) was calculated by multiplying the length (L) and slope (S) factor from the created maps. On the other hand, cropping management and erosion control practice factors, for different land use, were taken from the satellite based land use/land cover attribute data and geographic information system (GIS), respectively. Subsequently, all the parameters were substituted in the KW-GIUH-MUSLE in order to derive event-wise sediment yields. Sediment yield at the outlet of the study watershed was simulated for 6 storm events spread over the year 2000 and validated with the measured values. The percent deviations between the sediment yield measurements and observations vary in the range of -22.50 to 5.83%. The high coefficient of determination value (0.99) indicates that model sediment yield predictions are satisfactory for practical purposes. Key words: Sediment yield modelling, geographic information system (GIS), kinematic wave-geomorphologic instantaneous unit hydrograph-modified universal soil loss equation (KW-GIUH-MUSLE), storm event, Kengir watershed.

INTRODUCTION Soil erosion is a widespread land degradation problem in many parts of the world. The adverse influences of widespread soil erosion on soil degradation, agricultural production, water quality, hydrological systems, and environments, have long been recognized as severe problems for human sustainability (Lal, 1998). Accurate estimation of sediment-transport rates, in general, depends on an accurate a-priori estimation of overland flows. Thus, any error in the estimation of overland flows would be magnified through grossly inaccurate erosion estimations (Clarke, 1994). Globally more than 50% of pasture lands and about 80% of agricultural lands suffer

*Corresponding author. E-mail: saleh148@yahoo.com. Tel: 0098-918-843-0318. Fax: 0098-841-2227015.

from soil erosion (Pimentel et al., 1976). It is reported (Dudal, 1981) that worldwide, about six million ha of fertile land is being lost every year, due to just soil erosion and related factors. At this rate, it is estimated that currently about 1,964.4 Mha of total land area has already been degraded (UNEP, 1997). Of this, about 1,903 and 548.3 Mha are affected with water and wind erosion problems, respectively. Also, land degradation by soil erosion is a serious problem in Iran with an estimated soil loss of 2500×10

6 t/y

and about 94% of arable lands

and permanent rangelands are in the process of degradation (FAO, 1994; Program and Budget Organization, 1996; Masoudi et al., 2006).

In terms of erosion, Iranian soils are under a serious risk due to hilly topography, soil conditions facilitating water erosion (that is low organic matter, poor plant co-verage due to arid and semiarid climate), and inappropriate

4186 Afr. J. Agric. Res. agricultural practices (that is excessive soil tillage and cultivation of steep lands). This widespread problem threatens the sustainability of watershed, which is the main surface source of drinking water for Iyvan City of Ilam Province, Iran. Water and soil losses are the main reasons for sediment entering the reservoirs and these processes potentially reduce water quality. Soil erosion in this area strongly influences the ecological health of the city. Thus, accurate estimation of soil loses and sediment yield from watershed areas is extremely important for determining suitable land use and designing appropriate resource management or soil/water conservation measures.

Due to paucity of actual long-term flow records (total runoff, peak runoff and peak runoff time), especially for the under-developed regions or remote areas, their estimation is usually difficult for diverse watershed areas. In order to resolve this problem, modeling approaches, using minimum parameters, has emerged as a promising strategy for modeling location-specific rainfall-runoff. Regional instantaneous unit hydrograph (IUH) technique is one such approach commonly used for estimating design flood hydrographs for un-gauged watersheds. However, even this approach requires some level of gauging on a watershed somewhat similar to the one, where it actually needs to be applied. To overcome this difficulty, a geomorphologic instantaneous unit hydrograph technique was developed (Rodriquez-Iturbe and Valdes, 1979) for describing rainfall-runoff pattern of a watershed. Geomorphological instantaneous unit hydrograph (GIUH) based approach, which uses the geomorphologic parameters, is used to simulate watershed runoff. Rainfall occurring in the overland regions is converted into runoff, which is transmitted through streams and after joining it reaches outlet. The quantity and its time distribution of runoff are depending on topography of falling surface (overland region) as well as the transmission surface (channels). GIUH, which uses the geomorphological parameters, is used to simulate sub watershed runoff. The travel time of excess rainfall, in a given order channel or overland area, is assumed to follow either an exponential (Gupta et al., 1980) or gamma (Jain, 1992) probability distribution function (PDF).

Various methods have been used to determine the time scale to be used with these PDFs. Rodriguez-Iturbe and Valdes (1979) determined these time parameters through regression equations developed from actual discharge records, while Agnese et al. (1988) derived a time scale formula through actual experimental data. In contrast to this, Cheng (1982) used a conveyance factor for estimating mean flow velocity, and hence travel times. Review of literature revealed successful application of this approach on particularly the large watersheds of USA (Gupta et al., 1980; Beven and Wood, 1983; Hebson and Wood, 1982), Saudi Arabia (Sorman, 1995), India (Bhaskar et al., 1997; Kurothe et al., 1997; Goel et al.,

2000) (Jira, Sei and Narmada catchments), Palestine (Shadeed et al., 2007), Japan (Chiang et al., 2007), India (Kumar and Kumar, 2008), Russia (Lee et al., 2009) and China (Cao et al., 2010) under various climatic-topographic conditions. However, it was observed that even this approach is empirical, site specific and not applicable to an un-gauged watershed, as it makes use of either experimentally or empirically determined travel times. This problem was overcomed by a geomor-phologic data based kinematic wave (KW) model proposed by Lee and Yen (1997) for estimating both overland and channel flow travel times. The introduction of geomorphologic data based kinematic wave model not only led to easy determination of travel times for un-gauged watersheds, but also relaxed the linearity restrict-tion of the unit hydrograph theory, thereby increasing this method’s scope for application on even smaller watersheds.

Well-validated watershed scale hydrologic models are excellent predictive tools for obtaining accurate estimates of sediment yield from diverse watersheds. Physical and empirical models are the two widely used approaches for soil erosion assessment. Although physical-soil erosion models are more detailed, yet sometimes even these models have been found to be relying on the same empirically derived geology and vegetation factors (Foster, 1982) as in many empirical Universal Soil Loss Equation (USLE) (Wischmeier and Smith, 1978), MUSLE (Williams, 1975) and RUSLE (Renard et al., 1997) models. Further, application of physically based models to large watersheds, for which insufficient sediment yield and runoff data are available, is not of practical interest and the subdivision of a large watershed into rills and interrill areas is practically impossible. Besides, physical models contain equations whose constants and exponents need to be determined through calibration exercises on each test watershed. Thus, if data on sediment yield and runoff from a watershed do not exist, calibretion and application of even these models are impossible.

Simple empirical methods such as the USLE (Musgrave, 1947; Wischmeier and Smith, 1965), the Modified USLE (MUSLE) (Williams, 1975), or the Revised USLE (RUSLE) (Renard et al., 1991) are frequently used for the estimation of surface erosion and sediment yield from watershed areas (Ferro and Minacapilli, 1995; Ferro, 1997; Kothyari and Jain, 1997) because simple structure and ease of application. Although USLE/RUSLE may not replicate, the real picture of erosion process as they are based on coefficients computed or calibrated on the basis of observations, it has been extensively applied all over the world, mainly due to the simplicity in the model formulation and easily available data-set (Bartsch et al., 2002; Jain and Kothyari, 2001; Jain et al., 2001). The USLE (Wischmeier and Smith, 1978) was developed for estimation of the annual soil loss from small plots of an average length of 22 m, its application for individual

storm events and large areas leads to large errors (Hann et al., 1996; Kinnel, 2005), but its accuracy increases if it is coupled with a hydrological rainfall-excess model (Novotny and Olem, 1994). In the USLE model, there is no direct consideration of runoff, although erosion depends on sediment being discharged with flow and varies with runoff and sediment concentration (Kinnell, 2005). It has been observed that delivery ratios to determine sediment yield from soil loss equation can be predicted accurately but that vary considerably. The reason for this may be due to the variation in rainfall distribution over time from year to year. As a result of uncertainty in the delivery ratio, Williams and Berndt (1972) proposed MUSLE with the replacement of the rainfall factor with a runoff factor.

Particularly, this model is intended to estimate the sediment yield on a single storm basis for the outlet of the watershed based on runoff characteristics, as the best indicator for sediment yield prediction (ASCE, 1970; William, 1975; Hrissanthou, 2005). MUSLE increases sediment yield prediction accuracy and also, it eliminates the need for delivery ratios. The MUSLE has been used previously by many researchers (Banasik and Walling, 1996; Kinnel and Risse, 1998; Tripathi et al., 2001; Sadeghi and Mizuyama, 2007), and in some cases subjected to different modifications. The sediment yield model like MUSLE is easier to apply because the output data for this model can be determined at the watershed outlet (Pandey et al., 2009). Cambazoglu and Gogus (2004) estimated sediment yield using MUSLE and USLE in the Western Black Sea region of Turkey. Tripathi et al. (2001) estimated sediment yield from a small watershed of India using MUSLE and GIS; and the estimated values were very close to the observed values of sediment yield.

However, scanning of literature showed no integration of KW-GIUH theory based run-off simulating model with a soil loss estimating such as MUSLE model for storm based total sediment loss estimations from especially un-gauged/inadequately gauged watersheds. Thus, with this in the background, the present investigation is mainly aimed at testing the application potentiality of the KW-GIUH and MUSLE based event scale hybrid model for storm-wise sediment yield estimation on Kengir watershed in Iyvan city of Ilam province. MATERIALS AND METHODS

Description of study area

The study area lies between 46° 17' 11" to 46° 27' 35" E longitude and 33° 41' 14" to 33° 50' 57"N latitude with elevation ranging from 995 to 2549 m above mean sea level (Figures 1 and 2). The geographical area of the watershed is approximately 38804 ha. The mean annual precipitation of the watershed, based on the data collected for the period 1976 to 2005 at Iyvan climatology station located in the central part of the watershed, is about 674 mm of which 90% falls between late October and early April. The watershed is influenced by the dry and cold Mediterranean climate, mainly covered by agricultural areas in mid-and down-stream and forest in the up-stream areas and also rangeland in downstream

Arekhi et al. 4187 areas. The minimum and maximum temperatures vary from -3.1 to 32.4°C. The mean relative humidity varies from a minimum of 19% in September to a maximum of 55% in the month of January.

Erosion problem is prevalent in the study area due to rolling topography and improper agricultural management practices. The watershed is equipped with a hydrometric station at the main outlet, where the water level is recorded continuously, whereas sediment sampling is only occasional. In order to apply the KW-GIUH-MUSLE model in the study watershed, six storm events are considered for which the flow discharge and sediment flux data are collected through the development of hydrographs and sediment graphs, respectively. GIS based digital delineation of test watershed and drainage networks

This comprised digital delineation of the test watershed from the digital elevation model (DEM) (Djokic and Ye, 1999) and extraction of geomorphic information on the drainage areas, average slopes, average stream lengths, numbers, slopes and widths of streams in the study watershed as inputs for the model leading to the runoff simulation. Figure 3 gives a pictorial depiction of the delineated test watershed and streams. Strahler’s method (1952) was adopted for ordering channel/stream network to identify the order of the stream flowing through it. In the present study, geomorphological data is obtained through the application of a standard avenue script (named Hydrologic) of Arc-View Spatial Analyst GIS software (ESRI, 1999) on the DEM. Channel/overland roughness coefficients are derived through standard look-up table values for specific channel/overland conditions prevailing in test watershed area. Channel conditions are assessed rather subjectively through actual ground surveys (Cowan and Woody, 1956) while overland conditions are determined through actual land use maps for the test watershed (Engman, 1986). Sediment yield prediction using modified universal soil loss equation (MUSLE)

MUSLE is the most widely used sediment yield prediction model proposed by Williams (1975). It can be used to obtain accurate sediment yield estimates (Y, in metric tons) on single storm basis and is generally expressed as:

Y=11.8(Q .qp)0.56

(K).(L).(S).(C).(P) (1)

Where, Q is total runoff volume (in m3), qp is peak runoff rate in

(m3/s), K is soil erodibility factor (t ha h/ ha Mj mm), LS is slope

length and gradient factor; C is cropping management factor and P is erosion control practice factor. These input parameters, required for predicting sediment yield (Y) through Equation (1), are estimated as per the following procedures.

Total runoff volumes (m3) and peak runoff rates (m

3/s)

prediction through kinematic wave theory based geomorphic instantaneous unit hydrograph (KW-GIUH) model Based on the Strahler stream-ordering scheme, a watershed of

order Ω (where Ω is the highest order stream in the watershed) can be divided into 2

Ω-1 flow paths. Every raindrop falling on an overland

area moves from lower order to higher order channels, in succession, to finally reach the watershed outlet. According to Rodriguez–Iturbe and Valdes (1979), if ‘w’ represents a particular

flow path: Xoi →Xi →Xj → XΩ, then the probability of a raindrop adopting this flow path can be expressed as:

P (w) = POAi *PXoiXi*PXiXj…………....*PXkXΩ (2)

4188 Afr. J. Agric. Res.

Figure 1. Location of the Kengir watershed, Iran.

Where, Xoi represents ith-order overland region and Xi, Xj , Xk or XΩ

denotes ith, j

th, k

th or highest-order channels, respectively. POAi (that

is initial state probability) is the probability of a raindrop to (initially) fall on an i

th order overland region and is equal to the ratio of the

total area of ith order overland region to the total watershed area.

PXoiXi (that is transitional state probability) is the probability of a raindrop to move from an i

th order overland area to i

th order channel.

By definition, this is always equal to one, while, PXiXj is the transitional state probability of a raindrop to move from an i

th (that is

lower) order channel to a jth (that is higher) order channel and is

generally expressed as:

i

ji

xxN

NP

ji

,= (3)

Where ji

N , is number of ith order channels flowing into j

th order

channels and iN is number of i

th order channels.

If Tw represents total time taken by a raindrop to reach the watershed outlet, after traversing through path ‘w’ and TXk are raindrop-travel times in states Xk then Tw can be expressed as:

Tw = TXoi + TXi + TXj + …….. + TXΩ (4)

According to this expression, the raindrop travel times for different states in the watershed are assumed to be statistically independent and represented as PDFs of type fXk(t), with TXk as mean travel times value for each state Xk Hence, Equation (4) reduced to:

Tw = fxoi (t) + fxi (t) + fxj (t) +….…+ fXΩ (t) (5) Assuming these PDFs as exponential type (Gupta et al., 1980), Equation (5) can be re-written as:

Tw = aoiexp (-t/TXoi) + biexp (-t/Txi) + bjexp (-t/Txj) …+bΩexp (-t/TxΩ) (6)

Where, aoi, bi, bj, ……, bΩ are the coefficients determined through Laplacian transformation.

Combination of Equations (2) and (6) in the following manner yields a geomorphologic unit hydrograph-uw (t), for flow path (w) of a watershed as:

uw(t) = aoiexp (-t/TXoi) + biexp (-t/Txi) + bjexp (-t/Txj) …+bΩexp(-

t/TxΩ)*P(w) (7) The so generated individual geomorphic instantaneous unit hydrographs (IUH), uw(t), for each flow path (w) in the total path

Arekhi et al. 4189

Figure 2. Digital elevation model (DEM) of the Kengir watershed, Iran.

space (W) are then summed up to generate total geomorphic IUH, U(t), at the watershed outlet as: U (t) = u1 (t) + u2 (t) + u3 (t) + …….uw (t) (8) Where, w = (1, 2, 3, ……., 2

Ω-1) is flow path in a watershed.

Estimating overland/Channel flow travel times Estimation of watershed-geomorphology based excess rainfall-travel times for overland/channel areas in un-gauged/inadequately gauged watersheds is the most challenging task in geomorphic run-off simulation. Lee and Yen (1997) applied the concepts of kinematic wave (KW) theory to estimate these travel times. Thus, based on KW-approximations (Wooding, 1965), time taken by excess rainfall to travel through an i

th order sub-watershed (Txoi) is

obtained as:

Txoi=

m

m

eoi

oio

iiS

Ln

/1

12/1

− (9)

Where, no is overland roughness coefficient, Soi is mean i

th-order

overland slope (in fractions), m is an exponent (= 5/3 from Manning’s equation); ie is excess rainfall intensity (m/min) and Loi is

mean overland flow length (m). The mean overland flow length (Loi) in Equation (9) is expressed as: Loi = (A *PoAi)/(2* Ni * Lci) (10) Where, Ni is number of i

th order streams, Lci is mean i

th order

channel length (m) and A is total area of watershed (km2). However,

time taken by excess rainfall to travel through an ith-order channel is

expressed as:

Tx i=

−

+ coi

m

ici

cioicem

coioie

i hB

LLni

Li

B

Sh

/1

2/1

2

2

(11)

where, nc is channel roughness coefficient, Sci is mean i

th order

channel slope; Bi is width of ith order channel and hcoi is in-flow

depth of ith order channel due to water transported from upstream

reaches. As no channel flow is transported from upstream reaches for a (i =) 1

st order channel, therefore:

hcoi = 0 for i =1 (12a)

hcoi =

m

ii

oAiice

ciBN

APANni

Si

/1

2/1

)(

−

for (1< i < Ω) (12b)

4190 Afr. J. Agric. Res.

Figure 3. Stream order network for the Kengir watershed, Iran.

Estimating excess rainfall The soil conservation curve number (SCS-CN) method (SCS, 1956), an event-based, lumped rainfall-runoff model, is based on the water balance equations as:

QFIP a ++= (13)

Under two hypotheses:

S

F

IP

Q

a

=−

(14)

and

SI a λ= , (15)

Where P is the total precipitation, Ia is the initial abstraction, F is the cumulative infiltration after time to ponding, Q is the direct runoff, S

is the potential maximum retention, and λ is the initial abstraction

coefficient. 2.0=λ (a standard value). Equation (14) combined

with Equation (13) leads to:

SIP

IPQ

a

a

+−

−=

2)( (16)

Which is valid for P ≥ Ia and Q = 0, otherwise. Coupling of Equation (16) with (15) enables a determination of S/P as (known as CN-independent method) (Mishra and Singh, 1999a, b):

2

2

2

]4)1([)]1(2[

λ

λλλλ +−−−+=

CCC

P

S (17)

Where, PQC /= = runoff factor, λ= initial abstraction ratio.

It has been observed through many studies on Iranian

watersheds that a value of λ= 0.2 gives reasonable estimates of initial abstractions (Malekian et al., 2005). In the present study, average value of runoff factor (Q/P) for the test watershed is obtained from their annual rainfall-runoff records for 1996, 1997 and 1998 periods (Regional water resources organization, 1999). From Equations (17) and (15), it is possible to calculate S (maximum potential retention) and Ia (initial abstraction) and these two

Arekhi et al. 4191

Table 1. Land use/land cover statistics of the Kengir watershed, Iran.

Land use Area (ha) Area (%)

Densed forest 21536.31 55.50

Densed rangeland 1613.367 4.15

Dry farmland 1092.629 2.82

Semi-densed forest 7435.753 19.16

Semi-densed rangeland 2780.524 7.17

low-densed rangeland 2411.322 6.21

Barren land 1934.571 4.98

parameters are input data into the KW-GIUH model for calculation of runoff hydrograph (total runoff volume, peak runoff volume).

Soil erodibility factor (K)

The K factor is related to the integrated effects of rainfall, runoff, and infiltration on soil loss, accounting for the influences of soil properties on soil loss during storm events on upland areas (Renard et al., 1997). It represents average soil loss from a specific area of soil in cultivated continuous fallow with a standard plot length as 22.13 m and a standard percentage slope as 9%. The K values are usually estimated using the soil erodibility nomograph method, which uses % silt, %clay, %sand, %organic matter and soil structure and permeability classes (Wischmeier et al., 1978). However, there are lacks of % organic matter, structure and permeability class data in the soil survey data sources. Therefore, the following expression is adopted for calculations as recommended by MUSLE in the case of lack of observation data (Renard et al., 1997)

)))7101.0/659.1)(log(2

1(exp*0405.00034.0(*594.7 2+−+=g

DK

(18)

)ln(*01.0exp)( mifisummmDg −=

Where: K, soil erodibility factor (t ha h /ha Mj mm); Dg, mean geometric particle diameter (mm); Di, primary particle size fraction (%); Mi, arithmetic mean of the particle size limits of that size (mm). Slope length and steepness factor (LS) L factor is the function of slope length along with the S factor (slope steepness), and it represents the topographical factor commonly expressed as LS factor. Many researchers have used these two L and S factors as the combined LS factor. Slope length, defined as "the distance from the point of origin of overland flow to either the point where the slope decrease to the extent that deposition begins or the point where runoff enters well defined channels"(Wischmeier and Smith, 1978). S factor termed as 'slope steepness factor' is important, because it determines the velocity of the sediment runoff through water erosion. S factor is basically the function of the slope gradient.

In this study, the LS-factor is estimated from the DEM with the help of ArcView GIS software. The technique for estimating the LS-factor is proposed by Moore and Burch (1986) through Equation (19). The technique for computing LS requires a flow accumulation and the slope steepness. The flow accumulation is computed from a DEM using watershed delineation techniques. The slope steepness is computed using the DEM. Equation (19), which was

derived and used in the work for estimating LS based on flow accumulation and slope steepness. The equation is:

LS = (Flow Accumulation*Cell size/22.13)

0.4 *sin (slope/0.0896)

1.3

(19)

Cover management factor (C)

Soil erosion can most readily be controlled by managing vegetation, plant residues, and soil tillage. Erosion and runoff are markedly affected by different types of vegetative cover and by cropping systems. C reflects the effect of cropping and management practices on the soil erosion rate. The existing land cover types in the area are dry farm lands, rangelands and forests. One important surface cover existing in the study area is forest. Forests are the most important protection cover types against erosion. They can be classified into two groups as good undisturbed and poor disturbed forests. The last land use type in the study area is the rangeland grazed by sheep and cattle and it can be divided into two categories as good and degraded rangelands, which are composed of brush canopy cover, herbaceous cover and bare soil in different percentages. The C factor for rangeland also depends on the percentage of ground cover.

In the study area, C for a watershed is determined by weighting the C values of each crop and management level according to the size of area growing the crop with the same management level:

1

T

k

n

k

k

DA

DAC

C∑

==

(20) Where C is the cropping management factor for the watershed, CK

is the cropping management factor for an individual crop, k, DAk is the drainage area covered by an individual crop, k, with a particular management level, DAT is the total drainage area of the watershed and n is the number of different crops and management levels in the watershed. The land use/land cover statistics of the study area is presented in Table 1.

Conservation practice factor (P)

The P factor is the relationship between the soil loss in a soil cropped with a given support practice factor and the corresponding loss with up and down slope cultivation (Pandey et al., 2007). Practices of soil conservation (P) also have an influence on a reduction of erosion processes, therefore, soil loss values range according with practices adopted (Ruhoff et al., 2006).

4192 Afr. J. Agric. Res.

Table 2. Pxixj in the Kengir watershed, Iran.

Description P1,2 P1,3 P1,4 P2,3 P2,4 P3,4

Kengir 26/34 8/34 0/34 19/19 0/19 12/12

Table 3. KW-GIUH input parameters in the Kengir watershed, Iran.

Parameter

Kengir watershed

Order

1 2 3 4

Ni 34 20 11 1

( )mlic

3015.94 1712.22 1871.00 2572.00

( )2kmAi

242.72 328.84 379.95 388.04

POAi 0.515 0.287 0.181 0.015

( )mmSiO / 0.13 0.11 0.073 0.079

)/( mmSiC

0.042 0.083 0.017 0.003

Area(km2) 388.04

BΩ 3.5 9.25 15 20

Table 4. Soil texture and K factor in the Kengir watershed, Iran.

Soil texture Soil particles

Clay Sand Silt K factor

Loam 11 44 45 0.38

Loam 17 34 49 0.43

Loamy-silty 21 27 52 0.40

Loamy-silty 16 32 52 0.44

Loam 16 48 36 0.36

RESULTS AND DISCUSSION In this study, geomorphologic parameters of the study watershed were determined from DEM using ArcView Spatial Analyst GIS software. These extracted geomorphologic parameters were then used for the application of the KW-GIUH model. Tables 2 and 3 illustrate these GIS interface derived geomorphological parameters, which give the stream network, transitional probability and the input parameters of the KW-GIUH model for the test watershed. The average overland and channel roughness coefficients were also calculated as per the procedures detailed under materials and methods of this study. In general, the test watershed is characterized with smooth rock cut material laden channels of occasionally/frequently alternating cross-section, negligible obstruction, low vegetation and minor degree of meandering. This gives rise to channel

roughness coefficients as 0.0725, for the study water-shed. In the mean time, average overland roughness coefficient values of 0.13 for the test watershed were also computed.

Finally, using the aforementioned extracted geomophological parameters, storm-wise total runoff volumes (Q) and peak runoff rates (qp) were estimated, which required predicting sediment yield (Y) through Equation (1). For estimation of soil erodibility factor, representative soil samples from five locations in the watershed were collected and analyzed for determination of textural classes. The relative proportion of sand, silt and clay in the collected samples were given in Table 4 and based on which soil erodibility factor was estimated in t ha h /ha MJ mm using Equation (18). The spatial distribution of soil erodibility values were given in Figure 4. LS factor was derived with the help of ArcView GIS 3.2. The spatial distribution of the derived LS factor map

Arekhi et al. 4193

Figure 5. Distribution of LS factor in the Kengir watershed, Iran.

Table 5. Crop management factor for different land use/land cover classes in the Kengir watershed, Iran.

Land use C value

Densed forest 0.009

Densed rangeland 0.12

Dry farmland 0.43

Semi-densed forest 0.014

Semi-densed rangeland 0.17

low-densed rangeland 0.23

Barren land 1

was shown in Figure 5. Land use/cover map generated using Landsat satellite image of the year 2000 was adopted. C factor for different land use, were taken from the satellite based land use/land cover attribute data (Table 5) and the attribute maps. Finally, crop management factor was assigned to different land use classes (Table 5). The magnitude and the spatial distribution of crop management factor are given in Figure 6. The values for P factor are assigned to be 1.0 for the entire area, since there are no erosion control practices in the study area (Ozcan et al., 2008).

The magnitude and the spatial distribution of P are

given in Figure 7. Finally, herein, the runoff volumes and peak runoff rates were monitored at the outlet of the recording station of the Kengir watershed. The runoff factor was calculated for six storms during year 2000. Other variables in Equation (1), namely soil erodibility, topographic factor (LS), crop management and soil erosion control practice factors, were determined as per the aforementioned procedures. The average weighted values of 0.38 t ha h/ ha Mj mm, 3.26, 0.014 and 1 are thus allotted to the watershed factors of K, LS, C and P, respectively. Subsequently, all the parameters were substituted in the KW-GIUH-MUSLE model in order to derive event wise sediment yields. Further, the KW-GIUH-MUSLE model was validated by comparing the estimated sediment yields with the observed sediment yield for six storm events occurring from April to October 2000. The application results of the KW-GIUH-MUSLE model for the storms are shown in Table 6. The results of the comparative evaluation between measured and estimated sediment yield data is also presented in Figure 8. This study (Figure 8) indicates that the data points are very close and clustered around the 1:1 straight-line. Scrutinizing the results in Table 6 and Figure 8, shows that the KW-GIUH-MUSLE performs well in the prediction of storm sediment yield in the study area. The value of coefficient of determination (R

2) is about 0.99.

4194 Afr. J. Agric. Res.

Figure 6. Distribution of C factor in the Kengir watershed, Iran.

Figure 7. Distribution of P factor in the Kengir watershed, Iran.

Arekhi et al. 4195

Table 6. KW-GIUH-MUSLE results for the Kengir watershed, Iran.

No Storm date Observed sediment yield (t) Predicted sediment yield (t) Estimation error (%)

1 7/1/2000 373.56 329.53 -11.78

2 6/3/2000 13.42 10.52 -22.50

3 13/5/2000 73.21 65.52 -10.50

4 7/9/2000 1046.21 992.61 -5.12

5 29/3/2000 2730.79 2890.06 5.83

6 16/2/2000 76.85 73.65 -4.16

Figure 8. Comparison of the observed and the computed sediment yield in the Kengir watershed, Iran.

Figure 4. Distribution of soil erodibility factor in the Kengir watershed, Iran.

4196 Afr. J. Agric. Res.

The percent deviation of the storm-wise estimated sediment yield from the observed values varied in the range of -22.50 to 5.83% (Table 6).The under-prediction or over-prediction limits for the KW-GIUH-MUSLE model simulation are within 25% from the measured values for all the studied storms and are considered as the acceptable levels of accuracy for the simulations as reported by Bingner et al. (1989). Overall, the KW-GIUH-MUSLE model under-predicted the sediment yield for all the events except a storm by around 22% (Table 6). So, the results of the present investigation reconfirmed that the runoff and the sediment yields contributed by any storm, are not only a function of its antecedent moisture conditions, but also of within storm moisture conditions (Wei et al., 1998). This had a direct effect on not only the runoff, but also on the sediment yield predictions. As the SCS-CN independent method for determining rainfall excess could not account for (or mimic) these within-storm moisture changes in such conditions, therefore, this resulted into an under-prediction of the runoff and sediment yields. It could be inferred that the applied model performance efficiency for runoff and sediment yield simulations can perhaps be further improved by replacing the existing method for rainfall excess hyetograph simulations with a more objective and physical method, with the capability of mimicking even within-storm moisture changes.

The slight variation in hydrological response of the watershed in terms of sediment yield during the studied storms might be due to the spatial and temporal distribution of the rainfall and the availability of the eroded sediment throughout the watershed, which is not taken into account by KW-GIUH-MUSLE as for many other lumped and semi-lumped models. Though, the differences between the predicted and observed sediment yields from the KW-GIUH-MUSLE model contain some errors, this model can be successfully used for estimation of sediment yield in this area. From the results obtained through performance evaluation of the KW-GIUH-MUSLE application in the Kengir watershed, it could be inferred that the KW-GIUH-MUSLE does not need any modification for reliable application in the Kengir watershed. Researchers have found that runoff is a better indicator for sediment prediction than rainfall (ASCE, 1970; Williams, 1975; Foster et al., 1977; Beasley et al., 1980; Hrissanthous, 2005) for the agro-climatic condition of the study area and this has also been reported by Sadeghi (2004) for other parts of Iran. The aforestated analysis thus clearly shows that the applied model could indeed be utilized for obtaining reasonable sediment yield estimates for un-gauged/ inadequately gauged micro-watersheds. Conclusions

Soil erosion is a severe problem for many developing regions that lack adequate infrastructure to combat the

problem. Estimating sediment-transport phenomenon is essential for various purposes such as design of dams, pollutant control and development of integrated watershed management practices. A digital elevation model (DEM) of the test area was generated for delineating the watershed. The so developed DEM was then used for extracting watershed specific geomorphologic input data for the applied KW-GIUH and MUSLE based runoff and sediment yield estimating model. Both visual and objective assessments are made for evaluating the applied model performance. High coefficient of determination (0.99) indicates accurate simulation of sediment yield from the KW-GIUH-MUSLE model. The percent deviations between the sediment yield measurements and observations vary in the range of -22.50 to 5.83%.

To regionalize the results of the study, greater numbers of storm events, as well as case studies need to be considered in the future. In addition, other excess rainfall computing methods, for generating excess rainfall hyetographs, runoff hydrographs and sediment simulations for applied model, must be considered with reasonably accurate estimation of system response at the watershed scales, where scarce information exists. From the results obtained through performance evaluation of the application of the KW-GIUH-MUSLE in this watershed, it could be inferred that this model can be applied for the estimation of sediment yield in other such ungauged watersheds of Iran which have similar hydrometrological and land use conditions. Besides this, model can be used with minimum data on ungauged/inadequately gauged micro-watersheds with reasonable results.

REFERENCES

Agnese C, D’Asaro F, Giordano G (1988). Estimation of the time scale

of the Geomorphologic Instantaneous Unit Hydrograph from effective stream flow velocity. Water Res. Res., 15(6): 969-978.

ASCE (American Society of Civil Engineers) (1970). Sediment sources and sediment yields. J. Hydraulic Division. ASCE, 96(HY6): 1283-1329.

Banasik K, Walling DE (1996). Predicting sedimentgraphs for a small agricultural catchment. Nordic Hydrol., 27(4): 275-294.

Bartsch KP, Mietgroet HV, Boettinger J, Dobrowolski JP (2002). Using empirical erosion models and GIS to determine erosion at Camp William, Utah. J. Soil Water Conserv., 57(1): 29-37.

Beasley DB, Huggins LF, Monke EJ (1980). ANSWERS: A model for watershed planning. Trans. Am. Soc. Agric. Eng., 23: 938-944.

Beven k, Wood EF (1983). Catchment geomorphology and the dynamics of runoff contributing area. Hydrol. Sci. J., 65: 139-158.

Bhasker NR, Parida BP, Nayak AK (1997). Flood estimation for ungauged catchments using the GIUH. J. Water Res. Plann. Manage., 123(4): 228-238.

Bingner RL, Murphee CE, Mutchler CK (1989). Comparison of sediment yield models on various watersheds in Mississipi. Trans. Am. Soc. Agric. Eng., 32(2): 529-534.

Cambazoglu MK, Gogus M (2004). Sediment yields of basins in the Western Black sea region of Turkey. Turk. J. Eng. Environ. Sci., 28: 355-367.

Cao S, Lee KT, Ho J, Liu X, Huang E, Yang K (2010). Analysis of Runoff in Ungauged Mountain Watersheds in Sichuan, China using

Kinematic-wave-based GIUH Model, J. Mt. Sci., 7: 157-166. Cheng BLM (1982). A Study of Geomorphologic Instantaneous Unit

Hydrograph .Ph.D Thesis, Univ of IIIinois at Urbana-Champaign, Urbana, III.

Chiang S, Tachikawa Y, Takara L (2007). Hydrological Model Performance Comparison through Uncertainty Recognition and Quantification. Hydrol. Process., 21(9): 1179-1195.

Clarke RT, Chichester W (1994). Statistical Modeling in Hydrology. J. Hydrol., 170(1-4): 293-294.

Djokic D, Ye Z (1999). DEM pre-processing for efficient watershed delineation. ESRI User Conference, San Diego, CA.

Dudal R (1981). An evaluation of conservation needs. In: R.P.C.Morgan (ed.), Soil Conservation: Problems and Prospects. Wiley, New York.

Engman ET (1986). Roughness coefficients for routing surface runoff. J. Irrigation Drainage Eng., 112(1): 39-53.

ESRI (1999). ArcView GIS 3.2, Environ. Systems Res. Institute, Inc. FAO/UNEP (1994). Land degradation in south Asia: its severity causes

and effects upon the people. FAO, UNEP and UNEP project, Rome. Ferro V (1997). Further remarks on a distributed approach to sediment

delivery. Hydrol. Sci. J., 4(5): 633-647. Ferro V, Minacapilli M (1995). Sediment delivery processes at basin

scale. Hydrol. Sci. J., 40(6): 703-717 Foster GR, Meyer LD, Onstad CA (1977). A runoff erosivity factor and

variable slope length exponents for soil loss estimates. Trans. Am. Soc. Agric. Eng., 20(4): 683-687.

Goel NK, Kurothe RS, Mathur BS, Vogel RM (2000). A derived flood frequency distribution for correlated rainfall-intensity and Duration. J. Hydrol. Amsterdam, 228(1-2): 56-67.

Gupta VK, Waymire E, Wang CT (1980). A representation of an m instantaneous unit hydrograph from geomorphology. Water Res. Res., 16: 855-862.

Hann CT, Barfield BJ, Hayes JC (1996). Design hydrology and sedimentology for small catchments. Academic Press, San Diego.

Hebson C, Wood EF (1982). A derived flood frequency distribution. Water Res, 18(5): 1509-1518.

Hrissanthou V (2005). Estimate of sediment yield in a basin without sediment data. Catena, 64: 333-347.

IWRRO (Iranian Water Resources Research Organization) (1996). Annual Report of Amameh Representative Watershed. Ministry of Energy, Tehran, Iran.

Jain CX (1992). A Deterministic Gamma-type Geomorphologic Instantaneous Unit Hydrograph Based on path Types. Water Res. Res., 28(2): 479-486.

Jain MK, Kothyari UC (2001). Estimation of soil erosion and sediment yield using GIS. Hydrol. Sci. J., 45(5): 771-786.

Jain SK, Kumar S, Varghese J (2001). Estimation of soil erosion for a Himalayan watershed using a GIS technique. Water Resour. Manage., 15: 41-54

Kinnell PIA (2005). Why the universal soil loss equation and the revised version of it do not predict event erosion well. Hydrol. Process., 19: 851-854.

Kinnell PIA, Risse LM (1998). USLE-M: empirical modeling rainfall erosion through runoff and sediment concentration. Soil Sci. Soc. Am. J., 62: 1662-1672.

Kothyari UC, Jain SK (1997). Sediment yield estimation using GIS. Hydrol. Sci. J., 42(6): 833-843.

Kumar A, Kumar D (2008). Predicting Direct Runoff from Hilly Watershed Using Geomorphology and Stream-order- Law Ratios: Case Study. J. Hydrol. Eng. ASCE, 13(7): 570-576.

Kurothe RS, Goel NK, Mathur BS (1997). Derived flood frequency distribution for negatively correlated rainfall intensity and duration. Water Res. Res., 33(9): 2103-2107.

LAL R (1998). Soil erosion impact on agronomic productivity and environment quality. Crit. Rev. Plant Sci., 17: 319-464.

Lee KT, Chen NC, Chung YR (2008). Derivation of Variable IUH Corresponding to Time-varying Rainfall Intensity during Storms. Hydrol. Sci. J., 53(2): 323-327.

Lee Kwan T, Yen BC (1997). Geomorphology and Kinematic-wave-based hydrograph derivation. J. Hydraulic Eng., 123(1): 73-80. Malekian A, Mohseni Saravi M, Mahdavi M (2005). Applicability of the USDA-NRCS curve number method for runoff estimation. Iranian J. Nat. Res., 57(4): 621-633.

Arekhi et al. 4197 Masoudi M, Patwardhan AM, Gore SD (2006). Risk assessment of

water erosion for the Qareh Aghaj subbasin, southern Iran. Stochastic Environ. Res. Risk Assess., 21: 15-24.

Mishra SK, Singh VP (1999a). Another look at the SCS-CN method, J. Hydrol. Eng. ASCE, 4(3): 257-264.

Mishra SK, Singh VP (1999b). Behavior of SCS-CN method’ in C Ia*- spectrum, Proc., Int. Conf. on water, Environment, Ecology, Socio Economics, and Health, Engineering, 18-21 October, Seoul Nat. Uni., Korea.

Moore ID, Burch GJ (1986). Physical Basis of the Length-Slope Factor in the Universal Soil Loss Equation. Soil Sci. Soc. Am. J., 50: 1294-1298.

Musgrave GW (1947). The quantitative evaluation of factors in water erosion – A first approximation. J. Soil Water Conserv., 2(3): 133-138, 170.

Novotny V, Olem H (1994). Water quality: prevention, identification, and management of diffuse pollution. Wiley, New York.

Ozcan AU, Erpul G, Basaran M, Erdogan HE (2008). Use of USLE/GIS technology integrated with geostatistics to assess soil erosion risk in different land uses of Indagi Mountain Pass-Cankiri, Turkey. Environ. Geol., 53 1731-1741.

Pandey A, Chowdary VM, Mal BC (2009). Sediment Yield Modelling of an Agricultural Watershed Using MUSLE, Remote Sensing and GIS. J. Paddy Water Environ. (Springer), 7 (2): 105-113.

Pandey A, Chowdary VM, Mal BC (2007). Identification of critical erosion prone areas in the small agricultural watershed using USLE GIS and remote sensing. Water Resour. Manage., 21(4): 729-746.

Pimentel D, Elinor CT, Rada DH, Stephen R, Robert S, Eric AS, Daniel D, David R, Michael S (1976). Land Degradation: Effects on Food and Energy Resources. Science, 194: 149-155.

Regional Water Resources Organization (1999). Annual report of Kengir watershed. Ministry of energy, Tehran, Iran, 125pp.

Renard KG, Foster GR, Weesies GA, Porter JP (1991). RUSLE, Revised Universal Soil Loss Equation. J. Soil Water Conserv., 46(1): 30-33.

Renard KG, Foster GR, Weesies GA, McCool DK, Yoder DC (1997). predicting soil erosion by water: a guide to conservation planning with the Revised Universal Soil Loss Equation (RUSLE).USDA Agricultural Handbook, No 703.U.S.Gov.Print.Office, Washington, D.C.

Rodriguez-Iturbe I, Valdes JB (1979). The geomorphological structure of hydrologic response. Water Res. Res., 15(6): 1409-1420.

Ruhoff AL, Souza BSP, Giotto E, Pereira RS (2006). Avaliaçمo dos processos erosivos através da Equaçمo Universal de Perda de Solos, implementada com algoritmos em LEGAL. Geomلtica, pp. 1-9.

Sadeghi SHR, Mizuyama T (2007). Applicability of the modified universal soil loss equation for prediction of sediment yield in Khanmirza watershed, Iran. Hydrol. Sci. J., 52(5): 1068-1075.

Sadeghi SHR, Singh JK, Das G (2004). Efficiency of annual soil erosion models for storm-wise sediment prediction: a case study. Int. Agric. Eng. J. Thailand, 13: 1-14.

Shadeed S, Shaheen H, Jayyousi A (2007). GIS-based KWGIUH Hydrological Model of Semiarid Catchments: the Case of Faria Catchment, Palestine. Arabian J. Sci. Eng., 32(1C): 3-16.

Sorman AU (1995). Estimation of peak discharge using GIUH model in Saudi Arabia .J. Water Res. Plann. Manage., 121(4): 287-293.

Strahler AN (1952). Dynamic basis of geomorphology. Geological Society of America Bulletin No., 9: 923-938.

Tripathi MP, Panda RK, Pradhan S, Das RK (2001). Estimation of Sediment yield form a small watershed using MUSLE and GIS. J. Inst. Eng., 1(82): 40-45.

Tripathi MP, Panda RK, Pradhan S, Das RK (2001). Estimation of Sediment yield form a small watershed using MUSLE and GIS. J .Inst. Eng., 1(82): 40-45.

UNEP(united nation environmental programme) (1997). World Atlas of Desertification. 2nd edition Arnold London, p. 77.

Wei Y, Anthony JJ, Peter CY (1998). Identification of Improved rainfall-runoff models for an ephemeral low-yielding Australian catchment. Environ. Model. Softw., 13: 59-74. Williams JR (1975). Sediment routing for agricultural watersheds. Water Resour. Bull., 11: 965-974.

Williams JR (1975). Sediment yield prediction with Universal Equation

4198 Afr. J. Agric. Res.

using runoff energy factor. In: Present and prospective Technology for predicting sediment yields and source, Agricultural Research Service, US Department of Agriculture, 244-252.

Wischmeier WH, Smith DD (1965). Predicting rainfall-erosion losses from cropland east of Rocky Mountains: guide for selection of practices for soil and water conservation. US Department of Agriculture, Agricultural Handbook, p. 282.

Wischmeier WH, Smith DD (1978). Predicting Rainfall Erosion Losses -

A Guide to Conservation Planning. USDA handbook No. 537. U.S. Department of Agriculture in cooperation with Purdue Agricultural Experiment Station.

Wooding RA (1965). A hydraulic model for the catchment-stream problem’. II, Numerical solutions. J. Hydrol., 3: 254-267.