Catena 44 Ž2001.

305–322
www.elsevier.comrlocatercatena

A simple approach to soil loss prediction: a revised

Morgan–Morgan–Finney model
R.P.C. Morgan )
Institute of Water and EnÕironment, Cranfield UniÕersity at Silsoe, Silsoe, Bedford MK45 4DT, UK
Received 18 April 2000; received in revised form 20 August 2000; accepted 10 November 2000

Abstract

A revised version of the Morgan–Morgan–Finney model for prediction of annual soil loss by
water is presented. Changes have been made to the way soil particle detachment by raindrop
impact is simulated, which now takes account of plant canopy height and leaf drainage, and a
component has been added for soil particle detachment by flow. When tested against the same
data set used to validate the original version at the erosion plot scale, predictions made with the
revised model gave slopes of a reduced major-axis regression line closer to 1.0 when compared
with measured values. The coefficient of efficiency, for sites with measured runoff and soil loss,
increased from 0.54 to 0.65. When applied to a new data set for erosion plots in Denmark, Spain,
Greece and Nepal, very high coefficients of efficiency of 0.94 for runoff and 0.84 for soil loss
were obtained. The revised version was applied to two small catchments by dividing them into
land elements and routing annual runoff and sediment production over the land surface from one
element to another. The results indicate that, when used in this way, the model provides useful
information on the source areas of sediment, sediment delivery to streams and annual sediment
yield. q 2001 Elsevier Science B.V. All rights reserved.

Keywords: Soil erosion; Sediment yield; Erosion modelling

1. Introduction

Morgan et al. Ž1984. presented a simple empirical model for predicting annual soil
loss from field-sized areas on hillslopes. The MMF model used the concepts proposed
by Meyer and Wischmeier Ž1969. and Kirkby Ž1976. to provide a stronger physical base
than the Universal Soil Loss Equation ŽWischmeier and Smith, 1978., yet retain the

)
Tel.: q44-1525-863000; fax: q44-1525-863001.

0341-8162r01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 3 4 1 - 8 1 6 2 Ž 0 0 . 0 0 1 7 1 - 5
306 R.P.C. Morganr Catena 44 (2001) 305–322

advantages of an empirical approach regarding ease of understanding and availability of

data. The model was validated by the authors Ž1984. using erosion plot data for 67 sites
in 12 countries and then applied to simulate erosion over a 100-year period in Malaysia
under shifting cultivation. Since then, several researchers have used the model success-
fully in a wide range of environments ranging from Indonesia ŽBesler, 1987. to Nepal
ŽShrestha, 1997. and the Rocky Mountains ŽMorgan, 1985a.. De Jong and Riezebos
Ž1992. incorporated the model into a Geographical Information System from which De
Jong Ž1994. developed SEMMED ŽSoil Erosion Model for Mediterranean areas. and
`
applied it to the Bas-Vivarais area of Ardeche Province, southern France, using remote
sensing as a data source. Paracchini et al. Ž1997. applied SEMMED to the 100-km2
Timeto watershed in Sicily. A more recent summary of results with SEMMED is given
by De Jong et al. Ž1999.. The MMF model was also adapted to provide a method of
assessing erosion and evaluating different control methods on road banks and construc-
tion sites ŽCoppin and Richards, 1990..
Although the MMF model has proved simple to use and is able to give reasonable
estimates of annual runoff and erosion, some input parameters have been difficult to
determine. In particular, the top soil rooting depth ŽRD. gives problems of definition
since it describes the effective hydrological depth within which the storage of water
affects the generation of runoff. Although water storage in the soil is clearly affected by
the depth and density of the roots, it is also dependent upon the horizonation of the soil
profile, especially the depth of the A-horizon. Therefore, RD is invariably much
shallower than that normally recognised by an agriculturalist or agronomist. Morgan et
al. Ž1984. recommended values of 0.05 m for grass and 0.1 m for trees and tree crops.
Where measured runoff data are available, RD can be used as a calibrating term when
operating the model.
At the time the MMF model was developed, the soil detachability index Ž K . was not
determined experimentally for all soil types. Since then, detachability values have been
obtained for a wider range of soils for use in the European Soil Erosion Model
ŽEUROSEM; Morgan et al., 1998.. The values included in the EUROSEM User Guide
ŽMorgan et al., 1993; http:rrwww.silsoe.cranfield.ac.ukreurosemreurosem.htm. are an
order of magnitude higher than those proposed by Morgan et al. Ž1982. which were
based on rainfall simulation experiments carried out by Quansah Ž1982.. If the values for
EUROSEM are used, the MMF model will overpredict erosion. A consistent set of
values suitable for the MMF model is therefore required.
Research on rainfall interception has indicated that the exponential decay function
used in the MMF model to calculate soil particle detachment as a function of rainfall
energy is a simplification. It assumes that the intercepted rainfall does not contribute to
detachment whereas, in reality, the proportion which reaches the ground surface as leaf
drainage is capable of detaching soil particles from the soil mass, depending on the
height of fall ŽFinney, 1984; Morgan, 1985b; Styczen and Høgh-Schmidt, 1988.. An
improved description is required for the soil detachment by raindrop impact, allowing
for the effect of leaf drainage.
The MMF model simplified the erosion processes into two: detachment of soil
particles from the soil mass by raindrop impact and the transport of those particles by
runoff. The ability of rainfall to transport soil particles downslope and of runoff to
 R.P.C. Morganr Catena 44 (2001) 305–322 307

detach soil particles were ignored. Since the model was designed to evaluate erosion
where rates are likely to be accelerated by human impact, it may be reasonable to
neglect the transport of particles by raindrop impact because the rates will not be high
enough to present a problem. However, the case for neglecting the detaching power of
runoff is harder to sustain, particularly on steep slopes and where runoff becomes
channelled into rills. The model should therefore be improved by including soil particle
detachment by runoff.
The last decade has seen a change in the use of erosion models from prediction of
soil loss at a field scale to the simulation of the movement of sediment over the land
surface from its point of detachment to either sites of deposition or delivery to water
courses. The change in focus arises from increasing concern about pollution and
sedimentation problems downstream of eroded areas with less emphasis, particularly in
North America and the European Union, on the implications of erosion for long-term
soil productivity.
This paper presents a revised version of the MMF model which takes account of the
need to improve the description of the processes of erosion and the requirement of users
for better guidance on the choice of input parameter values. The revised version is tested
against the data set used to validate the original version of the model and the differences
in performance are noted. The revised version is further tested against a smaller more
recent data set. The model is then applied to two small catchments to show how it might
be used to evaluate the transport of sediment over the landscape and its delivery to water
courses.

2. The revised MMF model

The MMF model separates the soil erosion process into two phases: the water phase
and the sediment phase. The water phase determines the energy of the rainfall available
to detach soil particles from the soil mass and the volume of runoff. In the erosion
phase, rates of soil particle detachment by rainfall and runoff are determined along with
the transporting capacity of runoff. Using the procedure proposed by Meyer and
Wischmeier Ž1969., predictions of total particle detachment and transport capacity are
compared and erosion rate is equated to the lower of the two rates. Table 1 gives a list of
the input parameters needed to operate the revised version.

2.1. Estimation of rainfall energy

The procedure for calculating the energy of rainfall is revised from that in the original
version of the model to take account of the way rainfall is partitioned during interception
and the energy of the leaf drainage. The model takes the annual rainfall total Ž R; mm.
and computes the proportion Žbetween 0 and 1. which reaches the ground surface after
allowing for rainfall interception Ž A. to give the effective rainfall ŽER.:
ER s RA Ž 1.
308 R.P.C. Morganr Catena 44 (2001) 305–322

Table 1
Input parameters for the revised MMF model
Factor Parameter Definition and remarks
Rainfall R Annual or mean annual rainfall Žmm.
Rn Number of rain days per year
I Typical value for intensity of erosive rain Žmmrh.; use 10 for
temperate climates, 25 for tropical climates and 30 for strongly
seasonal climates Že.g. Mediterranean type and monsoon.
Soil MS Soil moisture content at field capacity or 1r3 bar tension Ž% wrw.
BD Bulk density of the top soil layer ŽMgrm3 .
EHD Effective hydrological depth of soil Žm.; will depend on vegetationr
crop cover, presence or absence of surface crust, presence of
impermeable layer within 0.15 m of the surface
K Soil detachability index ŽgrJ. defined as the weight of soil detached
from the soil mass per unit of rainfall energy
COH Cohesion of the surface soil ŽkPa. as measured with a torvane under
saturated conditions
Landform S Slope steepness Ž8.
Land cover A Proportion Žbetween 0 and 1. of the rainfall intercepted by the
vegetation or crop cover
Et r Eo Ratio of actual Ž Et . to potential Ž Eo . evapotranspiration
C Crop cover management factor; combines the C and P factors of the
Universal Soil Loss Equation
CC Percentage canopy cover, expressed as a proportion between 0 and 1
GC Percentage ground cover, expressed as a proportion between 0 and 1
PH Plant height Žm., representing the height from which raindrops fall
from the crop or vegetation cover to the ground surface

Typical values for A for different vegetation and crop types are given in Morgan
Ž1995..
The effective rainfall ŽER. is then split into that which reaches the ground surface as
direct throughfall ŽDT. and that which is intercepted by the plant canopy and reaches the
ground as leaf drainage ŽLD.. The split is a direct function of the percentage canopy
cover ŽCC.:

LD s ER = CC Ž 2.
DT s ER y LD Ž 3.
The kinetic energy of the direct throughfall ŽKEŽDT.; Jrm2 . is determined as a
function of the rainfall intensity Ž I; mmrh., using a typical value for the erosive rain of
the climatic region ŽTable 1.. The original version of the MMF model used the
relationship ŽWischmeier and Smith, 1978.:

KE Ž DT . s DT Ž 11.9 q 8.7log I . Ž 4.
which would be applicable to much of the USA east of the Rocky Mountains.
Alternative equations based on local rainfall energy–intensity relationships are now
available ŽTable 2. and may be used for the bracketed term in Eq. Ž4..
 R.P.C. Morganr Catena 44 (2001) 305–322 309

Table 2
Relationships between kinetic energy ŽKE; J my2 mmy1 . and rainfall intensity Ž I; mmrh.
Equation Remarks
KE s11.87q8.73log 10 I Assumes raindrop size distribution similar to that measured by
Laws and Parsons; used as the basis for rating erosivity in the
Universal Soil Loss Equation ŽWischmeier and Smith, 1978.;
suitable for North America east of the Rocky Mountains
KE s8.95q8.44log 10 I Assumes raindrop size distribution similar to that measured by
Marshall and Palmer; suitable for north-western Europe and
similar climates
KE s9.81q11.25log 10 I Developed by Zanchi and Torri Ž1980. for central Italy;
suitable for Mediterreanean-type climates
KE s 35.9Ž1y0.56ey0 .034 I . Developed by Coutinho and Tomas ´ Ž1995. in Portugal;
suitable for western Mediterranean
KE s 29.8yŽ127.5r I . Developed by Hudson Ž1965. in Zimbabwe; use for tropical climates
KE s9.81q10.60log 10 I Developed by Onaga et al. Ž1998. for Okinawa, Japan;
use for eastern Asia
KE s 29.0Ž1y0.6ey0 .04 I . Developed by Rosewell Ž1986. for New South Wales, Australia;
use for temperate southern hemisphere climates

Where available, other locally derived equations may be used in preference to the above.

The kinetic energy of the leaf drainage ŽKEŽLD.; Jrm2 . is dependent upon the height
of the plant canopy ŽPH; m. as proposed by Brandt Ž1990.:

KE Ž LD . s Ž 15.8 = PH 0.5 . y 5.87 Ž 5.

Where Eq. Ž5. yields a negative value, the energy of the leaf drainage is assumed to
be zero.
The total energy of the effective rainfall ŽKE; Jrm2 . is obtained from:
KE s KE Ž DT . q KE Ž LD . Ž 6.

2.2. Estimation of runoff

The procedure for estimating the annual runoff Ž Q; mm. remains unchanged. It is
based on the method proposed by Kirkby Ž1976. which assumes that runoff occurs when
the daily rainfall exceeds the soil moisture storage capacity Ž R c ; mm. and that daily
rainfall amounts approximate an exponential frequency distribution. The annual runoff is
obtained from:
Q s Rexp Ž yR crR o . Ž 7.
where R o s the mean rain per rain day Žmm. Ži.e. RrR n , where n s the number of rain
days in the year..
The soil moisture storage capacity is estimated from:
R c s 1000MS= BD = EHD Ž EtrEo . Ž 8.
310 R.P.C. Morganr Catena 44 (2001) 305–322

where MS s the soil moisture content at field capacity Ž% wrw., BD s the bulk density
of the soil ŽMgrm3 ., EHD s the effective hydrological depth of the soil Žm. and
EtrEo s the ratio of actual to potential evapotranspiration. The term, EHD, replaces the
rooting depth ŽRD. used in the original model and indicates the depth of soil within
which the moisture storage capacity controls the generation of runoff. It is a function of
the plant cover, which influences the depth and density of roots, and, in some instances,
the effective soil depth, for example on soils shallower than 0.1 m or where a surface
seal or crust has formed. Table 3 gives some guide values for EHD for use with the
revised MMF model. At present, the effect of different types of tillage practice on EHD
has not been evaluated. It is recommended that tillage be accounted for by adjusting the
C factor value ŽEq. Ž12. below..

2.3. Soil particle detachment by raindrop impact

In the revised MMF model, rainfall interception is allowed for in the estimation of
rainfall energy. It is therefore removed from the equation used to describe soil particle
detachment by raindrop impact Ž F; kgrm2 . which then simplifies to:
F s K = KE = 10y3 Ž 9.
where K is the erodibility of the soil ŽgrJ.. Guide values for K have been revised and
now cover a wider range of soil textures ŽTable 4..

2.4. Soil particle detachment by runoff

The revised model includes a new component to estimate the detachment of soil
particles by runoff. Based on experimental work by Quansah Ž1982., this is considered

Table 3
Recommended values for Effective Hydrological Depth ŽEHD.
Condition EHD Žm.
Bare soil without surface crust; 0.09
no impermeable barrier in top 0.2 m
Bare shallow soils on steep slopes; 0.05
crusted soils
Row crops Že.g. wheat, barley, 0.12
maize, beans, rice.
Row crops intercropped with 0.15
legumesrgrasses
Mature forest, dense secondary forest 0.20
Rubber, oil palm 0.15
Cocoa, coffee 0.12
Banana 0.18
Savannarprairie grass 0.14
Cultivated grass 0.12
Cotton 0.10
Groundnut 0.12

Where terracing is used, add 0.01 to EHD to take account of the resulting increase in water storage.
 R.P.C. Morganr Catena 44 (2001) 305–322 311

Table 4
Guide values for soil parameters
Soil type MS BD K COH
Sand 0.08 1.5 1.2 2
Loamy sand – – 0.3 2
Sandy loam 0.28 1.2 0.7 2
Loam 0.20 1.3 0.8 3
Silt – – 1.0 –
Silt loam 0.25 1.3 0.9 3
Sandy clay loam – – 0.1 3
Clay loam 0.40 1.3 0.7 10
Silty clay loam – – 0.8 9
Sandy clay – – 0.3 –
Silty clay 0.30 – 0.5 10
Clay 0.45 1.1 0.05 12

The parameters MS, BD, K and COH are defined in Table 1; where available, measured values should always
be used in preference to the guide values given above.

as a function of runoff Ž Q ., slope steepness Ž S . and the resistance of the soil Ž Z .. The
detachment by runoff Ž H; kgrm2 . is estimated from:
H s ZQ1.5 sinS Ž 1 y GC . = 10y3 Ž 10 .
where GC s percentage ground cover. The equation assumes that soil particle detach-
ment by runoff occurs only where the soil is not protected by ground cover. As a first
approximation, this seems reasonable since, where a vegetation cover is present, the
shear velocity of the flow is imparted to the plants and not to the soil.
For loose, non-cohesive soils, Z s 1.0. Based on a simplification of the work of
Rauws and Govers Ž1988. which emphasized the cohesion of the soil ŽCOH; kPa. as an
important component of its resistance to erosion:
1
Zs Ž 11 .
Ž 0.5COH .
Values of cohesion should be obtained for saturated soil using a torvane. Table 4
gives some guide values based on those used in EUROSEM ŽMorgan et al., 1993..

2.5. Transport capacity of runoff

The method for estimating transport capacity of the runoff ŽTC; kgrm2 . remains
unchanged from that used in the original version of the model, so that:
TC s CQ 2 sinS = 10y3 Ž 12 .
where C s the crop or plant cover factor, taken as equal to the product of the C and P
factors of the Universal Soil Loss Equation, and S is the slope angle Ž8.. The C factor
can be adjusted to take account of different tillage practices and levels of crop residue
retention.
 312
Table 5
Comparison of predicted annual runoff and soil loss from old and new versions of the MMF model
Site Runoff Žmm. Soil loss Žkgrm2 .
Observed Prediction Prediction Observed Prediction Prediction
old version new version old version new version
Lushoto, Tanzania
Clay soil, maizerbeans intercropping 0.2–1.0 3.74–6.85 11.93–18.98 0.001 0.001–0.002 0.005–0.014
Sandy clay loam, evergreen forest 2.6–5.7 2.59–5.96 45.57–79.39 0.001–0.003 0.000 0.002–0.005
Sandy clay loam, steep slope, evergreen forest 8.5–14.8 3.25–7.31 51.25–88.29 0.001–0.013 0.000 0.004–0.013

R.P.C. Morganr Catena 44 (2001) 305–322

Clay, steep slope, maizerbeans intercropping 0.4–0.8 3.74–6.85 1.06–11.93 0.001 0.001–0.004 0.000–0.012

Adiopodoume,´ IÕory Coast

Sandy loam, secondary tropical forest 15.0 85.9 315.5 0.001–0.02 0.003 0.0464
Sandy loam, bare ground 707–1415 1268.1 1141.8 6.9–15.0 15.5 17.5
Sandy loam, oil palm 43–172 57.8 332.5 0.001–0.05 0.02 0.77
Sandy loam, banana with mulch 11–86 102.4 384.5 0.004–0.005 0.002 0.021
Sandy loam, maize 643–1608 355.0 616.9 3.5–13.1 3.53 8.85
Sandy loam, groundnut 579–1565 452.7 730.9 5.9–12.0 7.17 3.09

Sefa, Senegal
Loam, secondary tropical forest 1.6–19.2 154.8 399.9 0.002–0.02 0.009 0.006
Loam, groundnut 130–699 370.8 723.4 0.29–1.63 1.38 2.43
Loam, cotton 15–699 429.0 645.8 0.05–1.85 1.84 4.17
Loam, mechanized maize 504 420.3 636.8 1.03 0.71 1.62
Loam, sorghum 390–683 442.5 659.9 0.33–1.24 1.57 3.48

Pong Khrai, Thailand

Clay loam, upland rice 22–32 34.4 102.3 1.4–2.4 0.07 2.20
Clay loam, upland rice, bench terraces 16–53 34.4 83.9 1.1–1.3 0.011 0.63

Marchiazza Basin, Italy

Loamy sand, bare soil with tufted grass 201–261 186.7 341.2 2.7–3.1 3.25 4.14
Loamy sand, Molinia moor grass 51–58 56.9 112.1 0.05–0.09 0.006 0.02
Loamy sand, chestnut and oak trees 36–38 48.3 92.0 0.009–0.018 0.002 0.005
Hesbaye, Belgium
Sandy loam, sugar beet nra 60.7 5.24 0.13–2.95 0.10 0.001
Sandy loam, winter wheat nra 78.6 8.66 0.045–0.10 0.10 0.003
Sandy loam, bare ground nra 415.6 84.13 0.6–8.25 12.01 0.78

Trier, Germany
Sandy loam, vines nra 5.8 25.54 0.003–0.004 0.005 0.134

Taiwan
Clay loam, citrus, clean cultivation 1268 580.2 654.4 15.64 10.25 16.34

R.P.C. Morganr Catena 44 (2001) 305–322

Clay loam, citrus, bench terracing 344 580.2 543.0 0.50 4.71 8.25
Clay loam, citrus with mulch 109 517.9 360.0 0.094–0.28 0.22 0.07
Clay loam, banana, clean cultivation 1113–1449 279.7 346.1 3.94–6.37 5.40 8.27
Clay loam, banana with mulch 189 245.6 256.5 0.009 0.042 0.015
Clay loam, banana with contour bunds 483–1029 279.7 346.1 0.11–0.39 0.54 0.22

Henderson, Zimbabwe
Clay, maize 8–61 26.7 77.7 0.2–0.3 0.013 0.106
Clay, cropped grass 8–26 18.6 60.8 0.05–0.1 0.000 0.016

Mpwapwa, Tanzania
Sandy loam, bare ground 446 212.7 389.5 14.7 2.78 6.64
Sandy loam, sorghum and millet 80–259 5.72 140.9 5.5–9.0 0.001 0.61
Sandy loam, tufted grass 8–65 2.9 109.6 0.0–0.07 0.000 0.07
Sandy loam, savanna grass 3–4 20.4 58.6 0.0 0.000 0.002

Lyamungu, Tanzania
Clay loam, coffee, clean cultivation 15–232 28.18 165.5 4.3 0.04 1.36
Clay loam, coffee, cover crops 10–98 9.57 54.3 0.4 0.002 0.015
Clay loam, coffee, contour ridges 36 28.18 165.5 0.3 0.004 0.09
Clay loam, coffee, cover crops, contour ridges 27 9.57 54.3 0.1 0.000 0.005

(continued on next page)

313
 314
Table 5 Ž continued .
Site Runoff Žmm. Soil loss Žkgrm2 .
Observed Prediction Prediction Observed Prediction Prediction
old version new version old version new version
Tuanshuangou, China
Silt loam, milletrmungbean nra 278.3 117.6–307.5 0.1–23.4 7.89 0.84–3.15
Silt loam, potato nra 250.3 219.6 43.9 7.59 2.00

R.P.C. Morganr Catena 44 (2001) 305–322

Silt loam, alfalfa nra 238.5 257.8 4.4 0.23 0.27

Malaysia
Sandy loam, oil palm 263–657 294–757 307–792 0.77–0.89 0.29–0.76 0.32–0.73
Sandy loam, bare soil 532–642 827 1346 2.93–3.39 3.73 6.56
Sandy loam, groundnut 273–328 273 636 0.64–1.01 1.04 0.74
Sandy loam, maize 365–378 340 726 0.56–0.81 1.07 1.36
Sandy loam, maize with mulch 73–80 298 638 0.04–0.06 0.06 0.15
Sandy clay loam, bare soil 688–941 829 1667 2.44–3.92 3.93 7.10
Sandy clay loam, groundnut 241–388 266 439 0.51–0.97 0.99 0.61
Sandy clay loam, cowpea 260–302 291 471 0.59–0.61 1.18 1.37
Clay loam, primary rain forest nra 181 111 0.004–0.024 0.027 0.01

Silsoe, United Kingdom

Sandy soil, bare ground 66 341 355 3.9 8.0 3.95
Sandy soil, grass 17 28 68 2.3 0.001 0.18
Sandy loam, woodland 9 2 0.65 0.001 0.000 0.000
Clay, spring barley 1 9 0.39 0.07 0.003 0.000
Clay, winter wheatrspring barley 6 5 4.73 0.05 0.001 0.001
Sandy loam, oatsrwheatrbeans 11 11 0.39 0.06 0.005 0.000
Chalk, winter wheat 5 5 6.34 0.07 0.002 0.002

For sources of measured data, see Morgan and Finney Ž1982..

 R.P.C. Morganr Catena 44 (2001) 305–322 315

2.6. Estimation of erosion

The estimates of soil particle detachment by raindrop impact and by runoff are added
together to give a total annual detachment rate. This is then compared with the annual
transport capacity and the lesser of the two values is the annual erosion rate ŽMeyer and
Wischmeier, 1969..

3. Validation of the revised MMF model with the original data set

New data input files were prepared for the 67 sites used for validating the original
version of the MMF model on erosion plots of varying sizes. Table 5 lists the sites, the
observed annual runoff and erosion rates, and the predicted values obtained with the old
and new versions of the model. Information on the sites and sources of the measured
data are contained in Morgan and Finney Ž1982..
For comparing the predicted Ž Y . with the observed Ž X . values, Morgan et al. Ž1984.
used reduced major-axis regression lines ŽKermack and Haldane, 1950; Till, 1973. to
allow for likely errors in observed values. The original version of the model produced
the following results for which coefficient of efficiency ŽCE. values ŽNash and Sutcliffe,
1970; Risse et al., 1993. have been calculated to indicate the variance from a one-to-one
prediction line:

Y s 19.776 q 0.775 X for runoff, n s 56, CE s 0.69 Ž 13 .

Y s 0.472 q 0.503 X for soil loss, n s 67, CE s 0.43 Ž 14 .
Y s y0.090 q 0.896 X for soil loss, n s 65, CE s 0.54 Ž 15 .
Eq. Ž15. is obtained after the removal of 2 years of data from the site in China where
the model failed to predict extremely high rates of erosion; the slope of the regression
line is not significantly different from unity Ž t s 1.112 - t 0.05 s 2.00. and the CE value
compares with the value of 0.58 obtained by Risse et al. Ž1993. for predicting annual
soil loss with the Universal Soil Loss Equation. Although these results indicated that the
model performed reasonably well, runoff predictions were poor where ridging and
terracing were used for erosion control or where mulching was adopted. In addition,
although the model correctly indicated very low rates of erosion Ž- 0.1 kgrm2 . where
this was the case, the predictions were often an order of magnitude out Žsee values for
the sites in Tanzania and UK in Table 5..
The results for the new version of the model are:

Y s 80.71 q 1.000 X for runoff, n s 56, CE s 0.58 Ž 16 .

Y s 0.386 q 0.527X for soil loss, n s 67, CE s 0.27 Ž 17 .
Y s 0.092 q 1.064 X for soil loss, n s 56, CE s 0.65 Ž 18 .
316 R.P.C. Morganr Catena 44 (2001) 305–322

Eq. Ž18. is based on omitting data for those sites in Belgium, Malaysia and China for
which observed runoff values are not available. The results for runoff ŽEq. Ž16.. give a

Table 6
Comparison of observed and predicted annual runoff and soil loss for revised MMF model
Site Runoff Žmm. Soil loss Žkgrm2 .
Observed Predicted Observed Predicted
Foulam, Denmark (loamy sand, 10% slope)
Rye grass cut four times per year 11.0 4.5 0.003 0.000
Spring barley sown with rye grass, 18.4 1.5 0.042 0.000
over winter, spring-ploughed
Spring barley, autumn ploughing 17.4 3.3 0.269 0.000
Winter wheat, autumn ploughing 98.4 3.3 1.279 0.000
up-and-down slope
Winter wheat, autumn ploughing 89.1 3.3 1.108 0.000
across slope
Bare soil fallow 111.6 59.2 1.087 0.365

Ødum, Denmark (sandy loam, 10% slope)

Rye grass cut four times per year 4.7 3.5 0.003 0.000
Spring barley sown with rye grass, 5.3 4.3 0.013 0.000
over winter, spring-ploughed
Spring barley, autumn ploughing 8.9 8.0 0.045 0.001
Winter wheat, autumn ploughing 20.9 8.0 0.012 0.001
up-and-down slope
Winter wheat, autumn ploughing 16.9 8.0 0.049 0.001
across slope
Bare soil fallow 46.2 83.7 0.593 0.672

El Ardal, Spain
Sandy loam, bare soil fallow, 5.9 116.7 0.009 0.948
7% slope
Loam, matorral Žmainly Rosemary, 5.0 20.8 0.007 0.018
Juniper. 28% slope
Loam, cut matorral, 28% slope 9.4 20.8 0.046 0.049
Clay loam, barley, 7% slope 10.4 51.1 0.061 0.037
Clay loam, wheat, 7% slope 10.4 42.6 0.061 0.026

Spata, Greece
Clay loam, olives, 19% slope 0.7 1.2 0.005 0.000
Clay loam, vineyard, 11% slope 3.4 1.3 0.041 0.000

Pakhribas, Nepal (sandy loam, 5% slope)

Bare soil fallow 488 675 2.575 3.347
Maizermillet intercrop, 513 453 3.667 4.106
tillage up-and-down slope
Maizermillet intercrop, 444 447 2.206 3.370
tillage across slope
Traditional maize cultivation 487 453 3.286 3.286
Maize, minimum till, mulch 390 431 1.692 0.929

´ ´
Observed data taken from Schjønning et al. Ž1995., Lopez-Bermudez ´ et al. Ž1999. and
Ž1993., Romero-Dıaz
Wouters and Shrestha Ž1986..
 R.P.C. Morganr Catena 44 (2001) 305–322 317

very good 1:1 relationship between observed and predicted values but the intercept value
is much higher than was the case with the original model, indicating an overprediction
of runoff at low values. This overprediction is the cause of the decrease in the
coefficient of efficiency. The predicted erosion, expressed by Eq. Ž18., also yields a 1:1
relationship with an intercept value close to zero and a much improved value of the
coefficient of efficiency. Overall, the revised MMF model appears to give better annual
predictions of both runoff and soil loss than the original version.

4. Validation of the revised MMF model with a new data set

A further test of the new model was carried out using data from research stations at
Foulum and Ødum, Denmark ŽSchjønning et al., 1995., El Ardal, Spain ŽLopez ´
´
Bermudez, 1993., Spata, Greece ŽRomero-Dıaz ´ et al., 1999. and Pakhribas, Nepal
ŽWouters and Shrestha, 1986. covering 24 sites in all. The list of sites with observed
annual runoff and erosion rates and the rates predicted by the model is given in Table 6.
The reduced major-axis regression lines for runoff and erosion show 1:1 relationships
between observed and predicted values and, for erosion, an intercept value close to zero:
Y s y6.204 q 1.086 X for runoff, n s 24, CE s 0.94 Ž 19 .
Y s y0.172 q 1.172 X for soil loss, n s 24, CE s 0.84 Ž 20 .
These tests, with very high values for the coefficient of efficiency, indicate that the
revised version of the model can give reasonable results in conditions ranging from the
temperate climate of northern Europe to the Mediterranean climate of southern Europe
and the tropical monsoon climate of the Himalayas.

5. Applications to small catchments

The application of the model to a catchment requires dividing the catchment into
hillslope elements which can be considered reasonably homogeneous in their soil, slope
and land cover. The elements are then arranged in a cascading sequence to represent
how runoff passes over the land surface from one element to the next. The application
depends on the assumption that it is feasible to average the water and sediment budgets
for these elements over the period of 1 year or, in some cases, an average of years where
mean annual data are used as inputs.
The routing procedure is relatively simple and operates as follows. The total runoff
Ž Q . on element Ž i . is the summation of the runoff generated on element Ž i . and that
received from the element immediately upslope. This summated runoff is used to
determine the transport capacity on element Ž i .. The annual rate of the supply of
detached material on element Ž i . is the sum of the detachment rate by rainfall and runoff
on element Ž i . and the influx of material from the element above. These estimates of
detachment and transport capacity are used to determine the annual rate of sediment
output from element Ž i . to either the next element downslope or the river system.
318 R.P.C. Morganr Catena 44 (2001) 305–322

5.1. Application to the Kulekhani Watershed, Nepal

The revised MMF model is applied to a 0.0268-km2 catchment within the Kulekhani
Watershed which lies between 1620 and 1720 m in altitude in the middle mountains of
the Himalayan Range in Nepal. Some 0.0252 km2 of the catchment is treated with
outward-sloping bench terraces on which maize, millet and beans are grown, and the
remainder of the catchment is under a poor growth of secondary broad-leaved forest
which is pruned annually for fodder and compost. The model is applied to the water year
November 1991 to October 1992 when the rainfall total was 1106 mm. The soils are silt
loams and 84% of the land area has slopes steeper than 15% ŽPerino, 1993.. The
catchment is divided into 17 land elements ŽFig. 1; Table 7..

Fig. 1. Kulekhani Watershed showing land elements Žnumbered. and direction of flow from one element to
another. Where no arrow is indicated, the element discharges directly into the river system.
 R.P.C. Morganr Catena 44 (2001) 305–322 319

Table 7
Predicted annual runoff and soil loss in the Kulekhani Watershed, Nepal, for the water year November
1991–October 1992
Land element Soil loss Žkgrm2 .
As single Allowing for Net erosion
element routing
1—1% slope, bench terraces 0.01 0.01 0.01
2—40% slope, bench terraces 0.39 0.71 0.70
3—63% slope, bench terraces 0.61 3.19 2.48
4—23% slope, bench terraces 0.22 0.22 0.22
5—40% slope, bench terraces 0.39 0.39 0.39
6—1% slope, bench terraces 0.01 0.01 0.01
7—40% slope, bench terraces 0.39 0.51 0.50
8—23% slope, bench terraces 0.22 2.20 1.69
9—1% slope, bench terraces 0.01 0.01 0.01
10—63% slope, bench terraces 0.10 0.11 0.10
11—63% slope, bench terraces 0.61 0.61 0.61
12—40% slope, forest 0.17 0.64 0.04
13—23% slope, forest 0.10 1.76 2.40
14—1% slope, bench terraces 0.01 0.01 0.01
15—63% slope, bench terraces 0.61 0.76 0.75
16—63% slope, forest 0.27 1.80 1.05
17—23% slope, forest 0.10 6.28 4.48
Observed soil loss from catchment 0.53 kgrm2
Predicted soil loss from catchment 0.64 kgrm2
Observed runoff from catchment 7.24 mm
Predicted runoff from catchment 9.22 mm

Soil loss predictions are for each element with no input of soil from the slope above Žcolumn 2. and allowing
for inputs of runoff and sediment from upslope Žcolumn 3.. Net erosion Žcolumn 4. is the result of comparing
annual output of sediment from the element with annual input from the slope above.

The results ŽTable 7. indicate that the model gives reasonable predictions of both
runoff and erosion for the catchment as a whole. A detailed analysis of the individual
elements shows that the contributors of sediment to the river system are elements 3, 13
and 17. By comparing the input and output of sediment, it is possible to determine the
net erosion or deposition on each element. Table 7 also gives the field-scale erosion rate
which would be predicted if each element was considered in isolation, i.e. with no
routing of runoff and sediment over the land surface. These data show that erosion is
highest on elements 3, 11 and 15 which are terraced slopes of 63%; in all cases, the
annual amount is less than 1.0 kgrm2 . The routing procedure implies that with the
downslope accumulation of runoff, giving increased erosion potential, the poor sec-
ondary growth of forest on elements 13 and 17 provide insufficient protection. For
example, the predicted erosion for the year on element 17 is 0.10 kgrm2 when
considered in isolation but 0.28 kgrm2 when it is placed in its catenary sequence.
Summing the erosion on the individual elements yields a gross erosion rate of 1.02
kgrm2 and a predicted sediment delivery ratio Ž0.64:1.02. of 63% which, for this size of
catchment, seems reasonable ŽRenfro, 1975..
320 R.P.C. Morganr Catena 44 (2001) 305–322

Table 8
Comparison of observed and predicted annual runoff and soil loss for the Catsop watershed in 1989
Runoff Soil loss
Žmm. Žkgrm2 .
Observed value 7.35 not measured
Predicted value Ždata on land cover 1.08 0.000006
based on crops grown.
Predicted value Ždata on land cover 12.48 0.035
seasonally weighted to allow for
periods of bare ground.

5.2. Application to the Catsop Watershed, The Netherlands

The Catsop is a 0.4585-km2 catchment in South Limburg at altitudes of 80–110 m

above sea-level, with silt loam soils and slopes of less than 10% over 84% of the land
area. Mean annual precipitation is 675 mm. The revised MMF model was applied to the
year 1989 when the land was under winter wheat, sugar beet and potatoes. The
catchment was divided into 32 land elements which effectively represent different fields
Žsee Folly et al., 1999 for the details.. The predicted runoff for the year was 1.08 mm
compared with a measured 7.35 mm. No data are available on measured erosion rates
but the predicted rate is very low at 0.6 mgrm2 . These values were obtained using input
data on land cover which is representative of the crops being grown in each field.
However, data published by De Roo Ž1993. show that four of the eight major storms in
1989 occurred in December when the land in much of the catchment was bare. Values of
the land cover parameters ŽCC, GC, EtrEo , C, PH. were therefore recalculated to give
seasonally weighted data, taking account of the likely conditions at the time of the
storms. With these recalculated values, the predicted runoff increases to a more
reasonable 12.48 mm and the erosion rate to 0.35 grm2 ŽTable 8..

6. Conclusions

The revised MMF model provides an improved description of the water erosion
processes operative on hillslopes in small catchments. The predictions of annual runoff
and soil loss at an erosion plot scale are better than those obtained with the original
version of the model, using results from the original data set as a basis for comparison.
However, neither version predicts the very high rates of erosion measured on plots in the
loess region of China. Taking results from a new data set covering north European,
Mediterranean and tropical monsoon conditions, the generally improved performance of
the model is confirmed, using the values of the slope and intercept of reduced major-axis
regression lines between predicted and observed data as criteria. In addition, the model
yields high values of the Nash and Sutcliffe coefficient of efficiency. Applications of the
model to two small catchments show that meaningful results can be obtained from
routing runoff and sediment over the landscape on an annual basis. Output from the
 R.P.C. Morganr Catena 44 (2001) 305–322 321

model can be used to identify source areas of sediment within the catchment and to
estimate sediment delivery ratios. Overall, the model is considered suitable for rapid
first-approximaton determinations of erosion rates.

References
Besler, H., 1987. Slope properties, slope processes and soil erosion risk in the tropical rain forest of
Kalimantan Timur ŽIndonesian Borneo.. Earth Surface Processes and Landforms 12, 195–204.
Brandt, C.J., 1990. Simulation of the size distribution and erosivity of raindrops and throughfall drops. Earth
Surface Processes and Landforms 15, 687–698.
Coppin, N.J., Richards, I.G., 1990. Use of Vegetation in Civil Engineering. CIRIArButterworths, London.
Coutinho, M.A., Tomas, ´ P., 1995. Characterisation of raindrop size distributions at the Vale Formoso
Experimental Erosion Centre. Catena 25, 187–197.
De Jong, S.M., 1994. Applications of reflective remote sensing for land degradation studies in a Mediterranean
environment. Nederlandse Geografische Studies No. 177.
De Jong, S.M., Riezebos, H.Th., 1992. Assessment of erosion risk using multitemporal remote sensing data
and an empirical erosion model. Department of Physical Geography, University of Utrecht.
De Jong, S.M., Paracchini, M.L., Bertolo, F., Folving, S., Megier, J., De Roo, A.P.J., 1999. Regional
assessment of soil erosion using the distributed model SEMMED and remotely sensed data. Catena 37,
291–308.
De Roo, A.P.J., 1993. Modelling surface runoff and soil erosion in catchments using Geographical Information
Systems. Nederlandse Geografische Studies 157.
Finney, H.J., 1984. The effect of crop covers on rainfall characteristics and splash detachment. Journal of
Agricultural Engineering Research 29, 337–343.
Folly, A., Quinton, J.N., Smith, R.E., 1999. Evaluation of the EUROSEM model using data from the Catsop
watershed, The Netherlands. Catena 37, 507–519.
Hudson, N.W., 1965. The influence of rainfall on the mechanics of soil erosion with particular reference to
Southern Rhodesia, MSc Thesis, University of Cape Town.
Kermack, K.A., Haldane, J.B.S., 1950. Organic correlation and allometry. Biometrika 37, 30–41.
Kirkby, M.J., 1976. Hydrogical slope models: the influence of climate. In: Derbyshire, E. ŽEd.., Geomorphol-
ogy and Climate. Wiley, London, pp. 247–267.
´ ´
Lopez-Bermudez, F., 1993. El Ardal, Murcia, Spain. In: MEDALUS 1 Final Report, pp. 433–460.
Meyer, L.D., Wischmeier, W.H., 1969. Mathematical simulation of the process of soil erosion by water.
Transactions of the American Society of Agricultural Engineers 12, 754–758 and 762.
Morgan, R.P.C., 1985a. The impact of recreation on mountain soils: towards a predictive model for soil
erosion. In: Bayfield, N.G., Barrow, G.C. ŽEds.., The Ecological Impacts of Outdoor Recreation on
Mountain Areas in Europe and North America. Rural Ecology Research Group Report, vol. 9, pp.
112–121.
Morgan, R.P.C., 1985b. Effect of corn and soybean canopy on soil detachment by rainfall. Transactions of the
American Society of Agricultural Engineers 28, 1135–1140.
Morgan, R.P.C., 1995. Soil Erosion and Conservation. Longman, Harlow.
Morgan, R.P.C., Finney, H.J., 1982. Stability of agricultural ecosystems: validation of a simple model for soil
erosion assessment. International Institute for Applied Systems Analysis Collaborative Paper CP-82-76.
Morgan, R.P.C., Morgan, D.D.V., Finney, H.J., 1982. Stability of agricultural ecosystems: documentation of a
simple model for soil erosion assessment. International Institute for Applied Systems Analysis Collabora-
tive Paper CP-82-50.
Morgan, R.P.C., Morgan, D.D.V., Finney, H.J., 1984. A predictive model for the assessment of erosion risk.
Journal of Agricultural Engineering Research 30, 245–253.
Morgan, R.P.C., Quinton, J.N., Rickson, R.J., 1993. EUROSEM: A User Guide. Silsoe College, Cranfield
University.
Morgan, R.P.C., Quinton, J.N., Smith, R.E., Govers, G., Poesen, J.W.A., Auerswald, K., Chisci, G., Torri, D.,
Styczen, M.E., 1998. The European Soil Erosion Model ŽEUROSEM.: a dynamic approach for predicting
sediment transport from fields and small catchments. Earth Surface Processes and Landforms 23, 527–544.
322 R.P.C. Morganr Catena 44 (2001) 305–322

Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models: I. Discussion of principles.
Journal of Hydrology 10, 282–290.
Onaga, K., Shirai, K., Yoshinaga, A., 1988. Rainfall erosion and how to control its effects on farmland in
Okinawa. In: Rimwanich, S. ŽEd.., Land Conservation for Future Generations. Department of Land
Development, Bangkok, pp. 627–639.
Paracchini, M.L., Minacapilli, M., Bertolo, F., Folving, S., 1997. Soil erosion modelling and coastal dynamics:
a case study from Sicily. Remote Sensing Society: Observations and Interactions. Remote Sensing Society,
Nottingham, pp. 334–339.
Perino, J.M., 1993. Paired catchment approach hydrology study within the Kulekhani and the Phewa Tal
Watersheds—results of the calibration stage 1991–1992. His Majesty’s Government of Nepal, Kathmandu,
IWMP Technical Paper No. 1r93.
Quansah, C., 1982. Laboratory experimentation for the statistical derivation of equations for soil erosion
modelling and soil conservation design. PhD Thesis, Cranfield Institute of Technology.
Rauws, G., Govers, G., 1988. Hydraulic and soil mechanical aspects of rill generation on agricultural soils.
Journal of Soil Science 39, 111–124.
Renfro, G.W., 1975. Use of erosion equations and sediment-delivery ratios for predicting sediment yield.
Present and Prospective Technology for Predicting Sediment Yields and Sources. pp. 33–45, USDA
Agricultural Research Service ARS-S-40.
Risse, L.M., Nearing, M.A., Nicks, A.D., Laflen, J.M., 1993. Error assessment in the universal soil loss
equation. Soil Science Society of America Journal 57, 825–833.
´ A., Cammeraat, L.H., Vacca, A., Kosmas, C., 1999. Soil erosion at three experimental sites in
Romero-Dıaz,
the Mediterranean. Earth Surface Processes and Landforms 24, 1243–1256.
Rosewell, C.J., 1986. Rainfall kinetic energy in eastern Australia. Journal of Climate and Applied Meteorology
25, 1695–1701.
Schjønning, P., Sibbesen, E., Hansen, A.C., Hasholt, B., Heidmann, T., Madsen, M.B., Nielsen, J.D., 1995.
Surface runoff, erosion and loss of phosphorus at two agricultural soils in Denmark—plot studies
1989–92. Danish Institute of Plant and Soil Science SP Report 14.
Shrestha, D.P., 1997. Assessment of soil erosion in the Nepalese Himalaya: a case study in Likhu Khola
Valley, Middle Mountain Region. Land Husbandry 2 Ž1., 59–80.
Styczen, M., Høgh-Schmidt, K., 1988. A new description of splash erosion in relation to raindrop size and
vegetation. In: Morgan, R.P.C., Rickson, R.J. ŽEds., Erosion Assessment and Modelling, Commission of
the European Communities Report No. EUR 10860 EN, pp. 147–198.
Till, R., 1973. The use of linear regression in geomorphology. Area 5, 303–308.
Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion losses, USDA Agricultural Research Service
Handbook 537.
Wouters, H.J., Shrestha, K., 1986. Runoff and soil loss studies in the Kulekhani Watershed. His Majesty’s
Government of Nepal, Kathmandu, DPrNEPr85r008, Field Document No. 1.
Zanchi, C., Torri, D., 1980. Evaluation of rainfall energy in central Italy. In: De Boodt, M., Gabriels, D.
ŽEds.., Assessment of Erosion. Wiley, London, pp. 133–142.