International Soil and Water Conservation Research 7 (2019) 203e225

Contents lists available at ScienceDirect

International Soil and Water Conservation Research

journal homepage: www.elsevier.com/locate/iswcr

Review Article

Using the USLE: Chances, challenges and limitations of soil erosion

modelling
Christine Alewell a, *, Pasquale Borrelli a, Katrin Meusburger b, Panos Panagos c
a
Environmental Geosciences, University of Basel, Switzerland
b
Swiss Federal Institute for Forest, Snow and Landscape Research, Birmensdorf, Switzerland
c
European Commission, Joint Research Center, Ispra, Italy

a r t i c l e i n f o a b s t r a c t

Article history: To give soils and soil degradation, which are among the most crucial threats to ecosystem stability, social
Received 10 January 2019 and political visibility, small and large scale modelling and mapping of soil erosion is inevitable. The most
Received in revised form widely used approaches during an 80year history of erosion modelling are Universal Soil Loss Equation
20 May 2019
(USLE)-type based algorithms which have been applied in 109 countries. Addressing soil erosion by
Accepted 29 May 2019
Available online 18 June 2019
water (excluding gully erosion and land sliding), we start this review with a statistical evaluation of
nearly 2,000 publications). We discuss model developments which use USLE-type equations as basis or
side modules, but we also address recent development of the single USLE parameters (R, K, LS, C, P).
Keywords:
Universal soil loss equation
Importance, aim and limitations of model validation as well as a comparison of USLE-type models with
RUSLE other erosion assessment tools are discussed. Model comparisons demonstrate that the application of
CSLE process-based physical models (e.g., WEPP or PESERA) does not necessarily result in lower uncertainties
Soil redistribution compared to more simple structured empirical models such as USLE-type algorithms. We identified four
Soil degradation key areas for future research: (i) overcoming the principally different nature of modelled (gross) versus
Water erosion measured (net) erosion rates, in coupling on-site erosion risk to runoff patterns, and depositional regime,
Model (ii) using the recent increase in spatial resolution of remote sensing data to develop process based
Review
models for large scale applications, (iii) strengthen and extend measurement and monitoring programs
to build up validation data sets, and (iv) rigorous uncertainty assessment and the application of objective
evaluation criteria to soil erosion modelling.
© 2019 International Research and Training Center on Erosion and Sedimentation and China Water and
Power Press. Production and Hosting by Elsevier B.V. This is an open access article under the CC BY-NC-
ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Background
In a world of climate and land use change, fertile soils are one of
the most essential resources to sustain humankind. In a recent
review paper in Science (Amundson et al., 2015) soil is discussed
comprehensively as THE essential resource for human security
(including climate and food security) in the 21st century with the
main threat to soils being soil erosion by wind and water ever since
humankind had started with agriculture.
To date, most of the world's soils are only in fair, poor, or very
poor condition as was stressed by the latest publications of the
United Nations, i.e. Status of the World's Soil Resources (FAO, 2015)
where soil erosion was identified as one of the major soil threats. In
* Corresponding author.
E-mail address: christine.alewell@unibas.ch (C. Alewell).
the early nineties, it was already estimated that 56% of the global
land being degraded and showed light to severe forms of soil
erosion by water (Oldeman, 1992, pp. 16e36). Since water erosion is
strongly exacerbated by the conversion of natural vegetation to
agricultural land, with nearly 40% of Earth's land currently utilized
for agricultural production (Foley, 2017), accelerated forms of soil
erosion became a widespread phenomenon representing a major
challenge to achieving the United Nations Sustainable Develop-
ment Goals (Keesstra et al., 2016).
The fact that soil erosion is a threat to one of the most essential
resources of humankind is as old as Cain and Abel (Panagos et al.,
2016c). In modern ages, Russian soil scientists were probably the
first to be successful in directing attention towards soil erosion
which resulted in governmental mitigation programs at the end of
the 19th century (Dokuchaev, 1892). Today, tools to not only map
the actual status of soil erosion, but also to test the influence of
mitigation strategies as well as management or conservations

https://doi.org/10.1016/j.iswcr.2019.05.004
2095-6339/© 2019 International Research and Training Center on Erosion and Sedimentation and China Water and Power Press. Production and Hosting by Elsevier B.V. This
is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
204 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
practices and last but not least future (climate change) scenarios are
urgently needed (Tan, Leung, Li, & Tesfa, 2018). Without the pos-
sibility to map and visualize soil erosion on large scales, soil erosion
will be out of the focus of all major environmental and agricultural
policies such as the European Union's Common Agricultural Policy
(CAP), the United Nations Sustainable Development Goals (SDGs),
the United Nations Convention to Combat Desertification (UNCCD)
and the Intergovernmental Science-Policy Platform on Biodiversity
and Ecosystem Services (IPBES). Moreover, the lack of large scale
erosion rate data sets also creates knowledge gaps in climate
change and carbon mitigation scenarios, hydrology and flood pre-
diction as well as earth science system modelling.
Modelling and prediction of soil erosion by water has a long
history with first studies published in international journals more
than seven decades ago using north American data sets (e.g.
(Bennett, 1939; Smith, 1941; Zingg, 1940b)). Many mathematical
models categorized as empirical, conceptual, physically-based or
process-oriented are available to estimate soil erosion at different
spatial and temporal scales (De Vente & Poesen, 2005; Jetten,
Govers, & Hessel, 2003; Karydas, Panagos, & Gitas, 2014; King &
Delpont, 1993; Merritt, Letcher, & Jakeman, 2003; Morgan, 2005,
pp. 365e376; Toy T. J, 2002; Vrieling, 2006; Zhang, O'Neill, & Lacey,
1996a). Most erosion models contain a mix of modules from each of
these categories. The research community has been contempora-
neously working for both, improving the applicability of complex
process-oriented models such as the Water Erosion Prediction
Project (WEPP) (Boardman, 2006; Morgan & Nearing, 2011) or the
European Soil Erosion Model (EUROSEM; Morgan et al. (1998)) as
well as updating the existing empirical approaches such as the
Universal Soil Loss Equation (USLE) (Bagarello et al., 2008, 2015;
Bagarello & Ferro, 2004; Di Stefano, Ferro, & Pampalone, 2017;
Kinnell, 2014) which not only remains attractive from a practical
point of view (Cao, Zhang, Dai, & Liang, 2015; Gessesse, Bewket, &
Bra€uning, 2015) but also gives estimates on large spatial scales (see
below, (Diodato, Borrelli, Fiener, Bellocchi, & Romano, 2017;
Doetterl, Van Oost, & Six, 2012; Van Oost et al., 2007; Yang, Kanae,
Oki, Koike, & Musiake, 2003a)).
At present, USLE and the Revised USLE (RUSLE) are by far the
most widely applied soil erosion prediction models globally and
according to (Risse, Nearing, Laflen, & Nicks, 1993) “USLE has been
used throughout the world for a variety of purposes and under
many different conditions simply because it seems to meet the
need better than any other tool available”. A Web of Science query
(http://apps.webofknowledge.com/) for the period 2007e2017
resulted in 1149 hits corresponding to the keywords “Universal Soil
Loss Equation”, “USLE”, “Revised Universal Soil Loss Equation” or
“RUSLE”, with average citations per item of 7.8 (as of August 2017).
A query with the Science Direct tool for the last 40 years resulted in
1556 studies using USLE or RUSLE with an average citation rate of
cited publications of 18.8 (studies with no citations were not
considered in this average rate, see meta-analysis in section 2.1).
The most well-known models based on USLE/RUSLE technology are
the Soil and Water Assessment Tool (SWAT) (Arnold, Srinivasan,
Muttiah, & Williams, 1998), the AGricultural Non-Point Source
Pollution Model (AGNPS) (Cronshey & Theurer, 1998; Young,
Onstad, Bosch, & Anderson, 1989), and the Water and Tillage
Erosion and Sediment Model (Watem/Sedem) (Van Rompaey,
Verstraeten, Van Oost, Govers, & Poesen, 2001) and the Chinese
Soil Loss Equation (CSLE, Liu, Zhang, and Xie (2002)) which total-
ized 1385, 311, 67 and 35 hits, respectively, in the same period (Web
of Science query 2007e2017). Modelling approaches independent
from the USLE such as the Water Erosion Prediction Model (WEPP,
Laflen, Elliot, Flanagan, Meyer, and Nearing (1997)), the Limburg
Soil Erosion Model (LISEM, (De Roo, Wesseling, & Ritsema, 1996),
the European Soil Erosion Model (EUROSEM; (Morgan et al., 1998)
and Pan European Soil Erosion Risk Assessment (PESERA, Kirkby
et al. (2008)) resulted in 220 hits (125, 32, 34 and 29, respec-
tively) of the same study period.
One of the main reasons why USLE type modelling is so widely
used throughout the world (see also Figs. 1e3 and section 2.1) is
certainly its high degree of flexibility and data accessibility, a
parsimonious parametrization, extensive scientific literature and
comparability of results allowing to adapt the model to nearly every
kind of condition and region of the world. Nevertheless, the USLE
approach is an empirical modelling approach with significant
limitations which already have been addressed in the very first
publications as there is no simulation of soil deposition (e.g.,
sedimentation) and that in most cases not enough measured data
exist to rigorously determine the single factors for all needed sit-
uations and scenarios (Wischmeier and Smith, 1965, 1978, p. 58).
The question remains, whether or not research and development of
the last 5 decades (1965 e today) was able to overcome these main
limitations, or, if not, at least improved them to such an extent that
is justifiable to apply the model algorithm to large scales and if so,
under which conditions and resolution.
The aim of this review is a thorough and rigorous evaluation of
USLE-type modelling to come to some conclusions on the sense and
non-sense of soil erosion modelling with this model concept
depending on specific circumstances and conditions. We do not
aim to provide a user guide or handbook of USLE-type modelling
but ask the interested reader to depend in this respect on the
wealth of literature starting from the original publications
(Wischmeier and Smith, 1965, 1978, p. 58) to reviews on its history
and past use (e.g. (Laflen & Flanagan, 2013; Laflen & Moldenhauer,
2003)), on special applications and further development (e.g.
consideration of runoff and event based modelling (Kinnell, 2010))
to general evaluations of model concepts and usability (Merritt
et al., 2003; Nearing, 2004; Quinton, 2013). We also do not claim
or even make an attempt to review all existing studies using USLE-
type models which is clearly beyond the scope of a manuscript in
ESR but have included a mere numerical geographical meta-
analysis to embrace the wealth of literature published on USLE-
type models quantitatively (see section 2.1).
2. Historical aspects and evolution of USLE-type modelling
The origin of USLE-type models was in the US to provide a
management decision support tool and was based upon thousands
of controlled studies on field plots and small watersheds since 1930
(Wischmeier & Smith, 1965). As with all empirical methods the
model concept is not based on process description and simulation
but rather on understanding a process, capturing the confounding
measureable parameters and delineating a mathematical algorithm
out of the relationship between these parameters and the
measured output (in this case measured eroded sediments).
As such, the USLE was defined as:
A ¼ R,K,L,S,C,P (1)
where: A (Mg ha1 yr1) is the annual average soil erosion, R (MJ
mm h1 ha1 yr1) is the rainfall-runoff erosivity factor, K (Mg h
MJ1 mm1) is the soil erodibility factor, L (dimensionless) is the
slope length factor, S (dimensionless) is the slope steepness factor,
C (dimensionless) is the land cover and management factor, P
(dimensionless) is the soil conservation or prevention practices
factor. The impact of the factors R capturing the energy and amount
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
Fig. 1. Geographic distribution per country of published papers using USLE or RUSLE modelling during the last 40 years (1977 to July 2017).
Fig. 2. Number of studies per continent and percentage of total publication number
(1977 to July 2017).
of precipitation, K accounting for the soil parameters determining
erosion potential, C describing the vegetation cover and manage-
ment as well as P delineating human management intervention is
directly related to our process understanding of soil erosion. Un-
derstanding erosional processes and driving factors, it might seem
surprising that neither runoff nor infiltrating was included in the
algorithm. However, from a pure statistical perspective the rainfall
erosivity was simply the best predictor for the measured erosion
output and attempts to include runoff even reduced the quality of
the assessment (Wischmeier, 1966; Wischmeier & Smith, 1965).
From a soil scientific perspective infiltration is not a helpful
parameter for such a modelling endeavour, because it is very prone
to measurement errors, extremely variable in soils and we will
hardly ever be able to capture infiltration at larger scales. The
implementation of the L and S are meant to capture the impact of
runoff energy which is influenced by a mix of processes and pa-
rameters. However, it has to be noted that in all these endeavours
modelling targeted at on site soil erosion risk, not at catchment
205
sediment yield.
Even though Wischmeier and Smith (1965) are generally cited
as the origin of the equation, the development of the modelling
concept actually started much earlier. The soil-loss equation was
the result of 20 years of development starting from the so called
slope-practice method (Zingg, 1940b) on which Smith (1941) added
crop and conservation-practice factors. Eventually soil erodibility,
management factors (Browning, Parish, & Glass, 1947) and the
rainfall factor (Musgrave, 1947) were added. The resulting “Mus-
grave equation” was published as a graphical solution in 1952
(Lloyd & Eley, 1952) but was further improved throughout the
1950's with an improved rainfall-erosion index, a method of eval-
uating cropping-management effects on the basis of local climatic
conditions, a quantitative soil-erodibility factor and a method of
accounting for effects of interrelations of such variables as pro-
ductivity level, crop sequence and residue management (USDA,
1961; Wischmeier, 1959; Wischmeier, 1960). Wischmeier and
Smith (1965) with an update by Wischmeier and Smith (1978, p.
58) finally put figures, tables and equations for all five single fac-
tors of the USLE together in a guide book explaining application and
use in detail. This guidebook was published by the US National
Runoff and Soil Loss Data Center, in cooperation between the
Agricultural Research Service and Purdue University, and it resulted
from statistical analysis of more than 10,000 plot-years of runoff
and soil loss data carried out in plots having a length less than or
equal to 122 m and a slope ranging from 3 to 18%. For defining the
mathematical structure of the USLE a reference condition, named
as unit plot, was used, which built on the set-up of field experi-
ments which were commonly used. The unit plot was defined as a
22.1 m long plot, with a 9% slope, maintained in a continuous
regularly tilled fallow condition with up-and-down hill tillage. The
unit plot was used to compare soil loss data collected on plots that
had different slopes, lengths, cropping and management and con-
servation practices (Wischmeier and Smith, 1965, 1978, p. 58).
The main motivation of the early studies using the USLE were to
quantify soil erosion rates and their single contributing factors in
206 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
Fig. 3. Trend of published papers (USLE/RUSLE). The period 2013e2017 covers until July 2017, the projection for the full 5 year period would approximate 600 papers (striped box).
comparison to soil-loss tolerance values and assess possible com-
binations of cropping systems and management plans for mitiga-
tion (Schwertmann, Vogl, & Kainz, 1987, pp. 1e64; Wischmeier &
Smith, 1965).
Since its first definition, USLE-type modelling was developed
further to meet the multiple needs and conditions of modelled
systems. Renard, Foster, Weesies, McCool, and Yoder (1997) stated
that the USLE for more than four decades has proven that this
technology is valuable as a conservation-planning guide in the US,
providing farmers and conservation planners with a tool to esti-
mate rates of soil erosion for different cropping systems and land
managements. Even though conceptualized and calibrated for
agricultural areas, the USLE has been soon adjusted to extend its
applicability to undisturbed land. In the early 1970's, as a result of a
meeting between the Natural Resources Conservation Service
(NRCS) and the Forest Service of the USDA, a first extension of the
USLE was made by developing a sub-factor to compute the cover-
management factors (C-factor) for woodland, permanent pasture
and rangeland (Spaeth, Pierson, Weltz, & Blackburn, 2003;
Wischmeier, 1975, pp. 118e124). The proposed method to estimate
the USLE C-factor used relationships to extrapolate three major
sub-factors, canopy, ground cover and below soil effects (Renard &
Foster, 1985), since for these areas an extensive data base of the C
factor was not available. Later, Dissmeyer and Foster (1980) pro-
posed further additions and modifications to extend the number of
sub-factors operating in woodland, including (i) amount of bare soil
(or conversely ground cover), (ii) canopy, (iii) soil reconsolidation,
(iv) high organic content, (v) fine roots, (vi) residual binding effect,
(vii) onsite storage, (viii) steps and (ix) contour tillage (agrofor-
estry). As reported by Spaeth et al. (2003) early applications of the
model in rangelands showed little agreement between estimated
and measured soil loss in both catchment (Simanton, Roger,
Herbert, & Renard, 1980) and plot experiments (Hart, 1984;
Johnson, Savabi, & Loomis, 1984). Johnson et al. (1984) suggested
that a more accurate assessment of the cover and management
conditions is needed for applications of the USLE on rangelands.
Weltz, Kidwell, and Fox (1998) stated that ‘USLE is a lumped
empirical model that does not separate factors that influence soil
erosion, such as plant growth, decomposition, infiltration, runoff,
soil detachment, or soil transport’. The increasing criticisms with
regard to the limitations of the USLE were met by an increasing
interest in both the US action agencies and the soil science research
community to update the USLE thus creating a substantial need for
a revision of the USLE which were made possible by the fast im-
provements in computation capacity.
Accordingly, the USLE was upgraded to the Revised Universal
Soil Loss Equation (RUSLE) during the 1990s (Renard et al., 1991,
1997, pp. 1e404). The basic structure of the RUSLE retains the
multiplicative form of the USLE, but it also has process-based
auxiliary components such as calculating time-variable soil erod-
ibility, plant growth, residue management, residue decomposition
and soil surface roughness as a function of physical and biological
processes. RUSLE also has updated values for erosivity (R), new
relationships for the topographical components (L and S factors)
which include ratios of rill and inter-rill erosion, consideration of
seasonality of the K factor and additional P factors for rangelands
and subsurface drainage, among other improvements (please see
section 3 for description of advancement regarding the single fac-
tors of RUSLE).
RUSLE combines the advantage of being based on the same
extensive database as was the USLE with some process-based
computations for time-varying environmental effects on the
erosional system (Nearing, 2004). However, it still has the limita-
tions in model structure that allows only for limited interactions
and inter-relationships between the basic multiplicative factors
(Nearing, 2004). As USLE-type models were designed to predict
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
long-term average annual soil loss, they have been successful to
predict event soil losses reasonably well at some geographic loca-
tions (Kinnell, 2010), but often fail to predict event erosion, which is
highly influenced by the fact that the USLE and its revisions (RUSLE)
do not consider runoff explicitly.
RUSLE2 was developed to scientifically enhance the USLE/RUSLE
equations and offer an improved tool to guide and assist erosion-
control planning (USDA, 2008). Although the fundamental empir-
ical equation scheme of the previous equations was retained,
RUSLE2 uses both empirical and process-based equations that
allow it to extend significantly beyond the original USLE structure.
In fact, RUSLE2 can be defined as a hybrid soil erosion prediction
(estimation) approach since it is a combination of the empirical,
index-based USLE and process-based equations for the detach-
ment, transport, and deposition of soil particles (USDA, 2008). The
RUSLE2 scheme allows to compute net erosion or deposition (mass/
area) for each segment in which the overland flow path is divided,
sediment load (mass/unit flow width) at the end of each segment
and at the end of the overland flow path, and sediment charac-
teristics at the detachment point and in the sediment load at the
end of each segment (USDA, 2008). Among other structural im-
provements (USDA, 2018), RUSLE2 works on a daily time step and
introduced the concept of erosivity density which resulted in sig-
nificant improvements in the calculations and mapping of rainfall
erosivity (Nearing et al., 2017). Validation of the parameters of the
original RUSLE2 was achieved with the same 10,000 plot-years data
as the original USLE.
As USLE has been frequently criticized of its empirical nature,
Ferro (2010) demonstrated that the original structure of USLE can
be theoretically obtained applying the dimensional analysis and the
self-similarity theory (Barenblatt, 1979, 1987) using the same soil
erosion representative variables and the reference condition
adopted by Wischmeier and Smith (1965). In other words, using the
factor scheme and the reference condition adopted by Wischmeier
and Smith (1965), Ferro (2010) challenges the criticism of the
empirical origin in claiming that “USLE is the subsequent logical
structure with respect to the variables used to simulate the physical
soil erosion process”. However, with this statement, we always
need to keep in mind, that USLE edriven modelling targets at the
physical soil erosion process and will thus not be capable of
simulating catchment sediment transport or yields.
Today, USLE-type modelling has been further advanced to meet
numerous special requirements and specific needs. E.g., Bagarello,
Di Stefano, Ferro, and Pampalone (2017) as well as Larson,
Lindstrom, and Schumacher (1997) adapted USLE-type models for
event based soil erosion modelling. USLE-type modelling has also
been used in all kind of extreme ecosystem types and for various
management scenarios, e.g. from volcanic soils in Chile with
Mediterranean climate (Stolpe, 2005) to the possible mitigation
impact of organic farming on soil erosion rates from mountainous
monsoonal watersheds in South Korea (Arnhold et al., 2014) or the
comparison of conventional with organic farming in northern
Bavaria (Auerswald, Kainz, & Fiener, 2003). For a discussion of ad-
vancements in model development and extension of the model
concept please see section 3.
2.1. A geographical and temporal meta-analysis of publications
addressing USLE/RUSLE models
The USLE concept, originally developed for the US agricultural
systems, has been adapted by scientists all over the world and
applied to datasets of different regions, countries and continents.
Examples would be Canada (e.g., the handbook for the application
207
of the RUSLE published by the Research Branch Agriculture and
Agri-Food Canada (Wall, Coote, Pringle, & Shelton, 2002, pp.
1e117)); Australia (Ferro & Porto, 1999); Africa (Haileslassie, Priess,
Veldkamp, Teketay, & Lesschen, 2005; Lufafa, Tenywa, Isabirye,
Majaliwa, & Woomer, 2003) and Asia (Lee & Heo, 2011;
Meusburger, Mabit, Park, Sandor, & Alewell, 2013; Xu, Miao, & Liao,
2008). In China, the USLE was adapted in substituting the C and the
P factor with three factors considering the biology, engineering and
tillage practices (Liu et al., 2002). For the application in Europe
(Schwertmann et al., 1987, pp. 1e64) was the first to rigorously
evaluate the applicability of the USLE to the different climatic
conditions and land use management systems of Bavaria (Southern
Germany). Today, USLE-type modelling is being used in Europe
from the northern Scandinavia to the Southern Mediterranean or
eastern countries like Turkey and Hungary (e.g., (Bagarello & Ferro,
2004; Ferro & Porto, 1999; Ozsoy, Aksoy, Dirim, & Tumsavas, 2012;
Podmanicky et al., 2011; Porto, 2016; Sivertun & Prange, 2003)).
Eventually USLE-type modelling was increased to continental
(Europe: Panagos et al. (2015d); China: Yue et al. (2016); Australia
Teng et al. (2016)) and finally to global scale (Borrelli et al., 2017b;
Diodato et al., 2017; Doetterl et al., 2012; Ito, 2007; Van Oost et al.,
2007; Yang, Kanae, Oki, Koike, & Musiake, 2003b).
A database of studies was developed within the frame of this
review using the Science Direct tool searching for the keywords
USLE and RUSLE (from 1977 to July 2017) with specific focus on title,
year of publication, study area, journal, paper type and keywords.
Additional 451 studies were not included as they did not propose a
specific study area or addressed theoretical issues of the model. The
author keywords of the articles have been processed by a word
cloud application to produce the most common issues addressed by
authors. In this process, we excluded self-explained keywords
(USLE, RUSLE, soil, erosion) and geographical locations. Besides the
most common keywords (GIS, model, factor, map, risk) the highest
frequency was reached by keywords such as climate, geo-statistics,
policy, crop, terraces, 137Cs, planning and nutrient (see visualization
in the graphical abstract).
The USLE/RUSLE model has been extensively used during the
last 40 years in 109 countries (Fig. 1).
The largest number of publications with the application of USLE/
RUSLE model has been found in the United States of America (274
papers), China (218 papers), Brazil (88), Italy (87), India (67), Spain
(66) Australia (50) and Turkey (43). In an analysis per continent, 519
papers (33% of total) have study sites in 32 countries of Asia (Fig. 2).
In Europe (24% of the total number of studies) USLE/RUSLE is used
in 31 countries and in Northern America and the Caribbean in 13
countries (22% of total number of studies). In Africa, 146 publica-
tions (9% of total) used USLE/RUSLE for estimating soil loss by water
in 21 countries. The meta-analysis resulted also in a wide use of the
USLE/RUSLE model in East-Southern Asia (South Korea, Malaysia,
Thailand and Indonesia), Central East Africa (Ethiopia, Kenya,
Nigeria) and the Mediterranean (Italy, Spain, Greece, Portugal,
Morocco, Tunisia, Algeria) (Fig. 1).
The 1556 papers that estimated soil loss by water erosion using
USLE/RUSLE at local/regional/national and continental scale have
been published with an average rate of 39 papers per year during
the period 1977e2017. The split into time windows of 5-years
yielded a strongly increasing trend with the highest number of
published articles (131) in 2016 (extrapolation from July 2017 to the
full year of 2017 is approximating 150 papers). A sharp increase in
publication rates started in the late 90ties. The highest relative
increase (76%) is noted in the period 2008-12 which is also ex-
pected to continue in the current period (2013-17). Estimated
published papers during the last 5 years will be more than 600
208 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225

compared to the 414 papers published in 2008-2012 (Fig. 3).
USLE/RUSLE applications were published in a wide variety of
journals. The 1556 peer reviewed papers have been published in
more than 500 Science Direct indexed journals with the highest
frequency in Catena (4.2%), followed by the Journal of Soil & Water
Conservation (3.3%) and Land Degradation and Development
(2.6%). More than 83% of the papers are classified as research arti-
cles, followed by conference papers (14%), book chapters (2%) and
review papers (1%). Finally, 1195 articles (77% of the total) were
cited with an average of 18.8 citations per article (publications with
no citations were not considered in this average rate).
3. Developments of USLE-type modelling with extended
modelling concept
Numerous advanced developments of USLE-type modelling
have been implemented during the last decades (for selections
please see Table 1). A substantial change to the USLE concept were
implemented in new modelling concepts considering runoff,
sediment transport and/or sediment delivery. An implementation
of the USLE where the rainfall erosivity is replaced by runoff vol-
ume (Modified USLE, MUSLE) resulted in a satisfying prediction of
measured sediment yield already in the mid-seventies and early
eighties (Smith, Williams, Menzel, & Coleman, 1984; Williams,
1975, pp. 244e252). The latter allows the equation to be applied
to individual storm events. Several variants of the MUSLE equation
exist and MUSLE was even integrated in GIS (e.g. in ArcMUSLE;
(Zhang, Degroote, Wolter, & Sugumaran, 2009). However, even
though runoff is considered, the restriction that USLE-modelled
and observed data represent different parameters (e.g., gross and
net erosion rates, respectively, please see box 1 and also section 5)
are not yet overcome. Another event based derivate of USLE is
USLE-M, which also includes event runoff in R, but K-factor values
are adjusted accordingly by multiplying them with the ratio of the
total value of the EI30 index to the total value of the QREI30 index
(Kinnell & Risse, 1998).

Table 1
Selected models based on the USLE concept or incorporating USLE parameters.

Model Full name Original references USLE components Additional features compared to USLE

RUSLE Revised Universal Soil Loss (Renard et al., 1991, 1997, USLE components to process-based auxiliary components (e.g., time-variable soil
Equation pp. 1e404) simulate erosion erodibility, plant growth, residue management, residue
decomposition, soil surface roughness, updated values for
erosivity (R), new relationships for L and S factors,
additional P factors
RULSE2 Revised Universal Soil Loss USDA (2008) USLE components to Net erosion and deposition, hybrid soil erosion prediction
Equation 2 simulate erosion with combination of the empirical, index-based USLE and
process-based equations for the detachment, transport, and
deposition of soil particles, calculation on a daily basis,
biomass accounting in the C factor,
Adding three dimensional components
RUSLE3D Revised Universal Soil Loss (Desmet & Govers, 1996; USLE components but Considering the flow convergence of the upslope area thus
Equation for Complex Terrain Mitasova et al., 1996; replacing the slope length improving the impact of concentrated flow on increased
Mitasova & Mitas, 1999) with contribution of erosion
upslope area
USPED Unit Stream Power Erosion and Mitasova et al. (1996) USLE components to Considering 3-D topography and prediction erosion and
Deposition simulate erosion deposition
Adding factors to consider nutrient transport, management measures or replacement of C factors to improve biological dynamic
CSLE Chinese Soil Loss Equation Liu et al. (2002) USLE but C and P factor C factor replaced by a biological conservation measures
replaced factor (B), and engineering conservation measures factor (E)
as well as a tillage conservation measures factor (T)
PERFECT Productivity, Erosion and (Littleboy et al., 1992a, Sediment yield is simulated Water balance and runoff predictions, erosion and crop
Runoff Functions to Evaluate 1992b) using MUSLE growth and crop yield; including sequences of plantings,
Conservation Techniques harvests and stubble management during fallow
G2 Geoland 2 erosion model (Karydas & Panagos, 2016; USLE but C and P factor C factor replaced by a vegetation factor retention combining
Panagos, Christos, Cristiano, replaced. land use and fractional vegetation coverage; P factor
& Ioannis, 2014a) replaced by landscape features factor quantifying the effect
of obstacles to interrupt rainfall runoff. Model erosion at
monthly step
Considering runoff, sediment transport and delivery
MUSLE Modified Universal Soil Loss (Smith et al., 1984; USLE components to prediction of sediment yield; simulation of individual storm
Equation Williams, 1975, pp. 244 simulate erosion events
e252)
SWAT Soil and Water Assessment Tool Arnold et al. (1998) Sediment delivery based on Simulation of the hydrological water balance, then using the
MUSLE runoff to simulate sediment dynamics
WATEM/SEDEM Water and Tillage Erosion and (Van Oost et al., 2000a; Van USLE components to sediment production as well as sediment transport and
Sediment Model Rompaey et al., 2001) simulate erosion pathways to derive ultimate source strengths, consideration
of tillage erosion
SEDD Sediment Delivery Distributed Ferro and Minacapilli USLE-type modelling to Gross erosion rates are coupled to a sediment delivery tool
Model (1995) calculate gross erosion to estimate sediment transport and delivery
rates
Considering runoff and sediment dynamics as well as nutrients and/or pollutants
CREAMS Chemical Runoff and Erosion (Knisel, 1980; Silburn & USLE components to Runoff modelling, erosion and sedimentation, gully erosion;
from Agricultural Management Freebairn, 1992; Silburn & simulate erosion chemistry and target non-point source pollutants
Systems model Loch, 1989) Individual storms/events as well as long term modelling
AGNPS Agricultural Non-Point Source Young et al. (1989) USLE or RUSLE used to Runoff and nutrient modelling
Model model soil erosion
EPIC Erosion Productivity Impact Williams, Renard, and Dyke USLE used to model water in addition to water erosion, hydrology, nutrient dynamics,
Calculator (1983) erosion plant growth, soil temperature tillage and economics are
simulated with physically based modules
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
Box 1
Definition of soil erosion in Universal Soil Loss Equation (USLE)-
type algorithms
Soil erosion is, strictly speaking, the detachment of soil
particles from a source site and its transport to some
depositional sink. As such, the processes of erosion and
deposition (often also addressed as sedimentation) are al-
ways coupled; both terms together might be described as
soil redistribution.
Soil redistribution encompasses soil redistribution by water
and wind as well as mass transfer of soil particles (e.g.
landslides). Soil redistribution by water can further be
differentiated into transport of single grains such as sheet
or interrill erosion, rill and gully erosion. In the modelling
concepts of all USLE-type algorithms soil erosion is defined
as: “Soil loss refers to the amount of sediment that reaches
the end of a specified area on a hillslope that is experiencing
net loss of soil by water erosion. It is expressed as a mass of
soil lost per unit area and time. Soil loss refers to net loss,
and it does not in any way include areas of the slope that
experience net deposition over the long term.” (Nearing,
Yin, Borrelli, & Polyakov, 2017). USLE-Type modelling
does not address larger rills or gully erosion (linear struc-
tures with a depth > 30 cm), but is restricted to sheet/interrill
and small rill erosion only (<several cm deep, generally
defined as structures which might be removed by
cultivation.).
For a better consideration of the terrain shape and complex
topographic conditions of the upslope area, the hillslope length
factor was replaced by the upslope contributing area in various
forms (Desmet & Govers, 1996; Mitasova, Hofierka, Zlocha, &
Iverson, 1996), an approach which was later called RUSLE3D (e.g.,
(Mitasova & Mitas, 1999; Rodriguez & Suarez, 2012)). A modified
equation for computation of the LS factor in finite difference form in
a grid cell representing a hillslope segment was derived by Desmet
and Govers (1996). A simpler, continuous form of the equation for
computation of the LS factor at a point on a hillslope was developed
by Mitasova et al. (1996). Both applications demonstrated that the
upslope area-based factor improves the impact of concentrated
flow on increased erosion. Mitasova et al. (1996) also added a 3-D
dimensional enhancement to the USLE, the Unit Stream Power
Erosion and Deposition (USPED) model which enabled to predict
erosion as well as deposition. Based on a comparison to 137Cs based
erosion/deposition measurements (Warren et al., 2005) concluded
that USPED derived estimates were more accurate and less biased
than USLE applications. However, a comparison of USPED and
RUSLE3D for a region in the Mediterranean (Bradano basin, Italy)
demonstrated for both models their feasibility to estimate the
spatial distribution of soil loss, estimates for soil erosion/deposition
at the watershed scale and a good match with the measured silting
data (Aiello, Adamo, & Canora, 2015).
A combination of RUSLE with sediment transport and delivery
processes has been achieved with the Water and Tillage Erosion
and Sediment Model (WaTEM/SEDEM (Van Oost, Govers, &
Desmet, 2000a; Van Oost, Govers, Van Muysen, & Quine, 2000b;
Van Rompaey et al., 2001)), which describes not only soil erosion
risk, e.g. risk for sediment production, but also considers sediment
transport and pathways to derive ultimate source strengths. The
spatially distributed soil erosion and sediment delivery model
WaTEM/SEDEM is a combined version of two empirically-based soil
209
erosion modes, namely WaTEM (Water and Tillage Erosion Model;
(Van Oost et al., 2000a) and SEDEM (Sediment Delivery model, Van
Rompaey et al. (2001).
A similar approach is followed with the Sediment Delivery
Distributed Modell (SEDD) where USLE-type modelling is used to
calculate spatially distributed gross erosion rates, which are then
coupled to a sediment delivery tool to estimate net sediment
transport and delivery on catchment scale (Ferro & Minacapilli,
1995; Ferro & Porto, 2000).
The above discussion clearly demonstrates that USLE e type
models have long ago left the pure empirical modelling realm and
process-based elements have been incorporated into some of the
advanced model development approaches. However, the question
arises how far this route can be followed if at the same time large
scale modelling is envisaged, needing spatially discrete and inde-
pendent units. As such, any modelling endeavor must, at the stage
of planning, clearly define envisaged aims and available data, to be
able to choose the best suited modelling concept (see also section 6,
Model Comparison).
4. State of development of the USLE factors
As stated above, the origin of the USLE was strictly empirical
with the calibration of the single factors from basic erosion plot
treatments (see section 2). However, this basic structure was sub-
stantially modified over the last decades with many process-based
auxiliary components in each of the five single factors accompanied
by integration into high resolution Geographic Information Sys-
tems (GIS) and the application of state of the art geostatistical tools.
As such, the aim of this section is to present the progress of each of
the five USLE parameters from their original definition to todays
advanced and highly developed possibilities.
4.1. R factor
The development of the R factor was one of the major im-
provements of the USLE compared to the original Musgrave equa-
tion. After analyzing c.a. 8,000 plot-years of runoff, soil loss and
rainfall, Wischmeier (1959) defined a “Rainfall Erosion Index” as a
product of the total kinetic energy of the storm and its maximum
30-min intensity (EI). Consequently, the so called iso-erodent maps
(Wischmeier & Smith, 1965) were produced from rainfall data of
about 2000 locations fairly evenly distributed over 37 States of the
US using 22-yr station rainfall records and computing the EI for
each storm. The R-factor was included as one of the inputs in the
USLE model (Wischmeier & Smith, 1978, p. 58) using a logarithmic
function between Kinetic Energy (KE) and Intensity (I) plus a con-
stant value for intensities exceeding 76 mm h1. Kinnell (1981)
proposed an exponential relationship of KE-I compared to the
logarithmic form and this change was also confirmed by Brown and
Foster (1987).
RUSLE used the proposed exponential relationship for esti-
mating the unit rainfall energy (er) based on rainfall intensity (ir):
h i
er ¼ 0:29 1  0:72eð0:05ir Þ (2)
Accordingly, the calculation of rainfall erosivity (EI30) of a single
event was based on the following equation:
X
k 
EI30 ¼ er vr (3)
I30
r¼1
With vr the rainfall volume (mm) during the rth time period of a
rainfall event divided in k-parts. I30 is the maximum 30-min rainfall
intensity (mm h1). Finally, the mean annual erosivity sums the
210 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
erosive events during a given period and is divided by the number
of years. In RUSLE2 a slight modification of unit rainfall energy
function (Yin, Xie, Liu, & Nearing, 2015) was proposed but its use
has been mostly limited to the United States. Equation (3) is mostly
used in R-factor estimations elsewhere in the world.
The varying relationship between rainfall kinetic energy (KE)
and intensity (I) have been addressed in a recent study by Angulo-
Martinez, Begueria, and Kysely (2016). In the absence of KE mea-
surements (e.g., measuring raindrop size and velocity with dis-
drometers) KE is usually estimated with empirical equations from
measured rainfall intensity. Angulo-Martinez et al. (2016) evalu-
ated 14 different KE-I equations to estimate the 1 min KE and event
total KE and compared these results with 821 observed rainfall
events recorded by an optical disdrometer. They concluded that (i)
empirical relationships performed well, when complete events
were considered but performed poorly for within-event variation
(1 min resolution) and (ii) to use local measured data or local ki-
netic energy equations. In a recent global R factor assessment 97.7%
of the calculated R-factors stations were based on the original
equation of Brown and Foster (1987) and many of those were
supported by local observation and validation (Panagos et al.,
2017b).
The impact of snow and snowmelt was not considered in R-
factor equations but Schwertmann et al. (1987, pp. 1e64) suggested
a very rough approximation in adding 1/10 of the precipitation for
those month, where significant soil movement due to snow melt is
observed. Recent approaches subtracted the precipitation amount
that falls below 0  C (Meusburger, Steel, Panagos, Montanarella, &
Alewell, 2012). The later approach considers that snowfall does
not exert erosive energy, but neglects the subsequent process of
snow melt that may trigger soil erosion.
An open point of discussion regarding equation (3) is the rain
drop size. In tropical areas such as Ethiopian highlands the algo-
rithm for unit rainfall energy (equation (2)) may underestimate the
rainfall erosivity due to large drop sizes in those areas (Nyssen et al.,
2005).
In the erosion studies applying USLE/RUSLE in late 1990s and
early 2000s, many scientists have either applied simplistic multi-
plications for estimating erosivity (R-factor ¼ 1.3 * precipitation) or
have proposed empirical erosivity equations such as the Fournier
Index (FI) and its modification (MFI) by Arnoldus (1980). The lack of
stations with sub-hourly data for long-periods and measured
rainfall intensity forced scientists to develop simple empirical
equations correlating R-factor with available daily/monthly/annu-
ally rainfall data (Aronica & Ferro, 1997; Bonilla & Vidal, 2011;
Diodato, Knight, & Bellocchi, 2013; Loureiro & Coutinho, 2001).
Recently, the availability of high temporal resolution data
(hourly, sub-hourly) for long periods and the development of geo-
statistical algorithms for spatial interpolations contributed to
regional, national and continental erosivity maps in Spain (Angulo-
Martinez, Lopez-Vicente, Vicente-Serrano, & Begueria, 2009),
Switzerland (Meusburger et al., 2012; Schmidt, Alewell, Panagos, &
Meusburger, 2016), Italy (Borrelli, Diodato, & Panagos, 2016a), Ko-
rea (Risal et al., 2016), Iran (Sadeghi, Zabihi, Vafakhah, & Hazbavi,
2017), Brazil (Oliveira, Wendland, & Nearing, 2013) and the Euro-
pean Union (Panagos et al., 2015a).
A group of scientists, working with R-factor data all over the
world, has collected data on rainfall erosivity from 3,625 meteo-
rological stations in 63 countries and established the Global Rainfall
Erosivity Database (GloReDa) developing a global erosivity map
(Panagos et al., 2017a).
4.2. K factor
The K factor is statistically related to the soil properties that
influence erodibility by water which are those that affect the
infiltration rate, permeability, and total water capacity, as well as
those that might influence the dispersion, splashing, abrasion, and
transporting forces of the rainfall and runoff. The soil erodibility
factor K was at first an empirical value determined from 20 years of
data on experimental plots from 23 major soil types from the US
kept fallow for at least 2 years with all other factors kept constant
(Wischmeier & Smith, 1965) and soil erodibility was regarded as
the amount of soil loss per unit erosive force with K equal to A/R.
Since the direct measurement of the K-value requires the estab-
lishment and maintenance of natural runoff plots over lengthy,
expensive observation periods at various locations, numerous at-
tempts have been made to simplify the technique and to establish
estimators for soil erodibility calculation from readily available soil
property data and standard profile description (Wischmeier,
Johnson, & Cross, 1971).
The nomograph considering the four most crucial soil parame-
ters (particle-size, percent organic matter, soil structure and soil
permeability) was suggested to derive the K-factor (Wischmeier
et al., 1971). Later, an approximation equation of the nomograph
was developed (Wischmeier & Smith, 1978, p. 58):
K ¼ 2.77*105*M1,14*(12- a)þ 0.043 (b- 2) þ 0.033 (4- c) (4)
where M ¼ the particle-size parameter, which equals % silt (0.1-
0.002 mm) times the quantity (with the quantity defined as 100 - %
clay), a ¼ percent organic matter, b ¼ the soil-structure code used in
soil classification, and c ¼ the profile-permeability class. However,
Wischmeier and Smith (1978, p. 58) state that this equation did not
describe the entire nomograph and differs for soils having high silt
content, low erodibility or high organic matter content.
(Auerswald, Fiener, Martin, & Elhaus, 2014) developed a set of
equations mimicking the original K factor nomograph. The de-
viations from the original equation were tested on 19055 soils
obtained during soil surveys throughout Germany. The set of
equations proposed resulted in deviations compared to the classical
K factor equation in more than 50% of all cases. Thus, the authors
recommend using this set of equations, which describe the
nomograph even beyond the limitations of the classical equation,
but they restrict that predictions may be far from “perfect” since
these equations are ignoring seasonality or interaction with
climate. Generally, the nomograph should only be used if there is
not a locally derived relationship or measured data that would also
account for seasonal changes in the K-factor.
Seasonal effects on the K-Factor due to freezing and thawing
processes (accompanied by effects on shear strength) or on
compaction (e.g., by rainfall during winter/fall or by life stock
trampling during summer), and subsequent release processes have
been discussed (Renard et al., 1991; Renard & Ferreira, 1993).
However, none of these seasonal processes and effects can easily be
captured for inclusion in the K factor. Kinnell (2010) reviewed
different approaches to assess the seasonality of the K-factor.
However, none of these approaches include the hardly measurable
influencing interactions and effects (e.g., climate influences and
seasonality of freeze-thaw, compaction by life stock trampling,
weathering, human management activities,) simultaneously for a
proper process-oriented modeling (Leitinger, Tasser, Newesely,
Obojes, & Tappeiner, 2010; Pineiro, Paruelo, Oesterheld, &
Jobbagy, 2010; Vannoppen, Vanmaercke, De Baets, & Poesen,
2015). Furthermore, the divergence of seasonal K-factors to an
annual K-factor is poorly discussed in the literature (e.g.(Wall,
Dickinson, Rudra, & Coote, 1988)). In the RUSLE2 User's Reference
Guide (Foster et al., 2008) it is even stated that no statistical evi-
dence exists for an inconsistency of soil erodibility over time.
Another important parameter affecting the erodibility is stone
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
content. Originally the effect of surface stones was not included in
the K factor equation, instead (Wischmeier & Smith, 1978, p. 58)
suggested to consider the effect of surface stones in the C factor. The
effect of stone cover on soil loss is scale dependent. At the meso-
scale stones and rock fragments on the soil surface may have
ambivalent effects while at the micro and macroscale a decrease in
sediment yield is predominant (Poesen, Torri, & Bunte, 1994). For
the macroplot scale, Poesen et al. (1994) developed a soil erodibility
reduction factor expressed as an exponential decay function based
on a world-wide compilation of experimental field data. Imple-
menting the effect of stoniness, reductions of 20-40% of the soil
erodibility were found to be particularly significant in the Medi-
terranean (Panagos, Meusburger, Ballabio, Borrelli, & Alewell,
2014b).
Today, soil surveys together with large scale data bases such as
ISRIC SoilGrids database at a 250 m spatial resolution (Hengl et al.,
2014) and remote sensing give new opportunities for mapping soil
erodibility in space and time (Panagos et al., 2014b; Wang, Fang,
Teng, & Yu, 2016). Several studies already explored the suitability
of hyper-spectral reflectance from soil surfaces to assess the
chemical and physical properties of soil (Ben-Dor & Banin, 1995;
Luleva, van der Werff, Jetten, & van der Meer, 2011; Rossel &
Behrens, 2010; Shepherd & Walsh, 2002; Stenberg & Rossel,
2010; Wang et al., 2016).
It has to be noted, that the underlying pedotransfer functions to
derive soil erodibility estimations are essentially all based on
American soil-erosion databases, which raises the obvious question
about their applicability in other geographical locations. Conse-
quently several studies aimed to determine the best suited soil
erodibility estimation for specific scales, regions or conditions
(Hussein, Kariem, & Othman, 2007; Ro €mkens, Poesen, & Wang,
1988; Torri, Poesen, & Borselli, 1997; Wang, Zheng, & Wang,
2012; Wawer & Nowocien, 2006; Zhang et al., 2004a, 2008).
However, hardly any study came up with an ultimate conclusion on
the suitability of different erodibility estimations, because of a lack
of measured plot data (Wang et al., 2016). As such, some new
studies, and databases incorporating measured K-values from long-
term observations of natural runoff plots are of utmost interest.
4.3. LS factor
In USLE models the effect of topography is usually considered by
Fig. 4. Dependency of the S factor on inclination (slope steepness) for different parametrizations.
211
the factor slope length (L) and slope steepness (S). The original
value of S was derived empirically as S ¼ (0.43 þ 0.30s þ 0.043s2)
*6.613 with s ¼ slope gradient in percent (Smith & Wischmeier,
1957; Wischmeier & Smith, 1965). Previous parametrizations
(Musgrave, 1947; Smith & Whitt, 1948; Zingg, 1940b) and many
posteriori derivatives (Fig. 4) of the original empirical equation
with different curve shapes of S against slope angle have been
developed since then (Foster, 1982, pp. 17e35; Liu, Nearing, & Risse,
1994; McCool, Brown, Foster, Mutchler, & Meyer, 1987; Nearing,
1997; Renard et al., 1997, pp. 1e404). All equations allow the
assessment of the S-factor up to a slope angle of 100%, even though
empirical measurements only exist up to a slope angle of 55% (Liu
et al., 1994). According to our knowledge only one study explored
soil loss up to slope angles of 90% using rainfall experiments
(Tresch, 2014, p. 132).
Wischmeier and Smith (1978) defined the slope length (L) as:
“the distance from the point of origin of the surface flow to the
point where each slope gradient (S) decreases enough for the
beginning of deposition or when the flow comes to concentrate in a
defined channel”. As such, the L factor is defined as the ratio of field
soil loss to that from a 22 m slope and the value of L may be
expressed as (l/22.1)m, where l is field slope length in meters and
m is a factor that varies with slope gradient in the USLE and the
ratio of rill to interrill erosion in the RUSLE.
Soil loss is much less sensitive to changes in slope length than to
changes in slope steepness (McCool et al., 1987). Most measured
slope lengths are less than 120 m and generally do not exceed 300
m (McCool, Foster, & Weesies, 1997). Nevertheless, it has to be
stated, that little research has been conducted to assess the relation
between slope length and soil loss. Moreover, the existing findings
are not ambiguous as with increasing slope length soil erosion was
observed to decrease (Joel, Messing, Seguel, & Casanova, 2002;
Kara, Şensoy, & Bolat, 2010; van de Giesen, Stomph, & de Ridder,
2005; Xu et al., 2009; Yair & Raz-Yassif, 2004), increase (Rejman
& Brodowski, 2005; Wischmeier & Smith, 1958; Zingg, 1940a) or
having no remarkable effect (Wischmeier & Smith, 1958).
The USLE was firstly developed to predict soil loss on uniform
slopes and fields (Wischmeier, 1976; Wischmeier and Smith, 1965,
1978, p. 58). But already in 1974, Foster and Wischmeier (1974)
developed a method to divide an irregular slope into a number of
uniform segments and as such accounting for the effect of the
shape of the slope on soil loss opening the doors to watershed scale
212 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
USLE applications (Griffin, Beasley, Fletcher, & Foster, 1988;
Williams & Berndt, 1977).
However, these manual methods were demanding and by
replacing plots with watersheds the slope length loses significance
for predicting overland flow and soil loss, since in reality, surface
water flows converge and diverge across the landscape (Desmet &
Govers, 1996). Thus, it was proposed that the unit contributing area
which is the upslope drainage area per unit of contour length
should substitute slope length (Desmet & Govers, 1996). The latter
approach did not only account for segment length or grid cell size,
but also for the position of the segment or cell within the landscape
and consequently the LS factor was not one-dimensional anymore
(Kinnell, 2008; Zhang et al., 2013). The first physically based slope-
length factor was developed by Moore and Burch (1986). They
concluded that the theoretical and empirical slope-length factors
are equivalent and thus the USLE slope-length factor is indeed a
measure of the sediment transport capacity of runoff from the
landscape. However, the USLE slope-length factor fails to fully ac-
count for the hydrological processes that affect runoff and erosion
(Moore & Burch, 1986).
A multiple flow algorithm is implemented by MUSLE, RUSLE,
EPIC and G2 models (Karydas et al., 2014). Winchell, Jackson,
Wadley, and Srinivasan (2008) further improved the method and
compared several variations of the GIS approach. A second algo-
rithm was associated with the development of the USPED model by
Mitasova et al. (1996). The greatest limitation of all these methods
is (i) the absence of an algorithm for predicting deposition
(Mitasova, Hofierka, Zlocha, & Iverson, 1997; Winchell et al., 2008)
and ii) the neglect of the variability of infiltration along the flow
path. The flow path and cumulative cell length-based method (FCL)
proposed by Dunn and Hickey (1998) and Hickey (2000) and
advanced by Van Remortel, Hamilton, and Hickey (2001) and Van
Remortel, Maichle, and Hickey (2004) who tried to overcome this
disadvantage. In 2013, Zhang et al. (2013) presented an algorithm
that merges both approaches to account for flow convergence
based on the contributing area as well as slope cutoff conditions
and thus accomplished an important step towards net erosion
estimation.
To conclude, increased availability of gridded digital elevation
data, often referred to as DEMs, improved the estimation of the LS
factor. The availability of high resolution DEM (Europe 25m but in
certain countries much higher (e.g., Switzerland DEM 2m)) has
outpaced method development of flow algorithms (Seibert &
McGlynn, 2007), which need to define the contributing area.
Currently, four algorithms exist to calculate the upslope contrib-
uting area: 1) single-direction flow algorithm, D8 (e.g., O'Callaghan
and Mark (1984), p. 2) the multiple flow direction algorithm (MD8,
Quinn, Beven, Chevallier, and Planchon (1991)), 3) infinite possible
single-direction flow pathways (D∞), and 4) triangular multiple
flow direction algorithm (MD∞, Seibert and McGlynn (2007)). Ac-
cording to our knowledge comparison of the different flow algo-
rithm to mapped erosion features is still missing.
Another main research gap related to the availability of high
resolution DEM is the dependence of USLE-type estimates on grid
size of the DEM. Soil erosion estimates were found to decrease with
decreasing resolution (Mondal et al., 2017). Consequently erosion
estimates may be biased depending on the resolution of the DEM.
To investigate resolution dependence seems even more pending in
the light of drone derived DEM with high horizontal resolutions of
5 cm (Peralta et al., 2017).
4.4. C factor
The cover-management factor (C) in USLE-type equations
measures the combined effect of all interrelated cover and
management measures (Wischmeier & Smith, 1965). In other
words, it reflects the effects of biomass cover and soil-disturbing
activities. It is expressed as the ratio of soil loss from land under
specified conditions to the corresponding loss from clean-tilled,
continuous fallow (given by the product of R$K$L$S). The soil loss
ratio (SLR) is the ratio of the soil loss from the non-bare fallow
surface and the soil loss from the bare fallow surface over a given
period of time. Accordingly, it is inversely proportional to the soil
surface cover which intercepts raindrops and hinders surface
runoff by slowing it down.
As nearly all crops can be grown continuously or in rotations
with varying sequences, the soil loss ratios can vary in cropped land
over time as they are a function of canopy, ground cover, roughness,
soil biomass and consolidation change. The C factor depends on the
particular stage of vegetation and vegetation cover development at
the time the rain event occurs, but also on the prior land use con-
ditions. For instance, plant residues can be removed or left on the
field, incorporated near the surface or ploughed under, chopped or
omitted by the harvesting operation. All these different measures
can have different effects on soil loss. Originally five crop stages
were defined including rough fallow, seedling, establishment,
growing and maturing crop and residue or stubble. From 10,000
plot years of runoff and soil-loss data assembled from 47 research
stations in 24 states of the US empirical tables were created taking
the crop stage, vegetation type, management measures and their
timing, and rainfall intensity EI in the specific periods into
consideration (Wischmeier & Smith, 1965). In RUSLE, the compu-
tation of the C factor follows a scheme proposed by Laflen, Foster,
and Onstad (1985) and (Weltz, Renard, & Simanton, 1987, pp.
104e111) and tables of SLR are no longer provided. Five sub-factors
are multiplied together to yield soil loss ratios (prior-land-use
(PLU), canopy-cover (CC), surface-cover (SC), surface-roughness
(SR) and soil-moisture (SM)). Individual values of SLR are
computed in time intervals within which the five sub-factors
remain constant. Subsequently, the SLR values are weighted by
the fraction of rainfall erosivity (EI) of the coinciding time period (i)
and combined to achieve an overall C factor value using
X
n
C¼ ðSLRi $ EIi Þ = EIt (5)
i¼1
Although it is possible to compute C values for a wide set of
tillage techniques and crop rotations within RUSLE, a large number
of input data are required (Gabriels, Ghekiere, Schiettecatte, &
Rottiers, 2003). With the advent of GIS-based modelling and the
ambitions to assess the potential impact on soil erosion over larger
areas, the C factor, as with most other factors within RUSLE, has
experienced a process of computation transformation and simpli-
fication. At catchment- and regional-scale, most of the input pa-
rameters for the RUSLE C sub-factors became hard, if not impossible
to assess and quantify given that the model simulation is a specific
field plot. In recent years, the methods most commonly employed
to compute the C factor and set large-scale RUSLE-based modelling
are simple attributions of literature C-factor values to land use
maps without further land sub-classification (Bakker et al. (2008);
Ma €rker et al. (2008), among others) or satellite remote sensing
approaches ((Borrelli, Panagos, Ma €rker, Modugno, & Schütt, 2017;
De Jong, 1994; Scho € nbrodt, Saumer, Behrens, Seeber, & Scholten,
2010; Van der Knijff, Jones, & Montanarella, 1999), among others).
With increasing calls for more knowledge on soil erosion on macro-
regional scale, research on C factor based on remote sensing tech-
niques has become a widely applied method across the globe
(Vrieling, 2006; Zhang, Zhang, Liu, Qiao, & Hu, 2011). While these
applications often inadequately represent the management aspect
of the C factor, the gained insights nevertheless provide a valuable
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
contribute to improve the understanding of the soil erosion po-
tential of a given region, especially when considering monthly or
sub-monthly land use/land cover change. Most recent, regional-
scale C factor computation approaches have moved towards
spatially and spatiotemporally descriptions using vegetation based
indices for land use classifications (Meusburger, Konz, Schaub, &
Alewell, 2010b; Mo € ller, Gerstmann, & Gla €ßer, 2014, pp.
5072e5075; Panagos et al., 2014c; Schmidt, Alewell, & Meusburger,
2018; Scho €nbrodt et al., 2010; Teng et al., 2016) or weighted average
operations combining crop statistics with remote sensing and GIS
modelling techniques (Borrelli, Panagos, Langhammer, Apostol, &
Schütt, 2016b; Panagos et al., 2015b). A geo-statistical approach
based on weighted average operations was recently introduced to
thoroughly incorporate the extent, types, spatial distribution of
global croplands, and the effects of the different regional cropping
systems into a global RUSLE-based soil erosion model (Diodato
et al., 2017). In these recent approaches, the C factor reflects the
relative effectiveness of the soil and crop management systems in a
given region in terms of their ability to reduce soil loss.
4.5. P factor
As in the case of the C factor, the support practice factor (P) in
the RUSLE-type equations is based on the concept of deviation from
a standard. It expresses the ratio of soil loss in a field with specific
support practice to the loss of soil under conditions of straight-row
farming up and down the slope. The support practices affecting the
erosion process by modifying the amount and dynamics of runoff
are considered in the P factor. For cultivated land supporting
practices generally considered as P factor are contour tillage/
planting, strip-cropping, terrace farming systems and stabilized
water ways because improved tillage practices such as sod-based
rotations, fertility treatments, and greater quantities of crop resi-
dues left on the field are already part of the C factor (Wischmeier &
Smith, 1965). The overall P factor is the product of the individual
support practices. In the original paper of Wischmeier and Smith
(1965) empirical P values to consider contouring were provided
but the researchers recommended to consult further expert
judgement to adapt the P value for specific field conditions espe-
cially when small gullies or rills exist or one of the above mentioned
measures was adapted. Within RUSLE, a list of P values obtained
from experimental data supplemented by analytical experiments is
provided, offering a wide range of support practice conditions.
Nevertheless, according to (McCool et al., 1987) the values of the P
factor are the least reliable of the USLE/RUSLE factors as the effec-
tiveness observed in field studies conducted on given slopes
showed wide ranges of reductions.
Typical P values range from about 0.2 for reverse-slope bench
terraces to 1.0 where there are no erosion control practices
(Wischmeier & Smith, 1978, p. 58). The lower the P value, the more
successfully the support practice or the combination thereof pro-
motes the deposition of soil particles. From a conservation planning
perspective, the deposition of the soil that is close to the source is
preferable but the effectiveness of the support practice depends on
the characteristics of the agricultural area (e.g., the slope gradient).
With the transition from field-to GIS-based modelling, the lack of
spatial information to compute the P factor implies that in the vast
majority of the catchment-to regional-scale GIS-based models the
crucial effect of support practices on soil erosion has been omitted.
However, recent large-scale studies proposed alternative ap-
proaches that are able to provide relative estimations of the P factor
and assess the possible effects of conservation policy (Panagos
et al., 2015c; Yue et al., 2016). Panagos et al. (2015c) considers the
latest policy developments in the Common Agricultural Policy, and
applies the rules set by Member States for contour farming over a
213
certain slope. Furthermore, the impact of stone walls and grass
margins was modelled using the more than 226,000 observations
from the Land use/cover area frame statistical survey (LUCAS)
carried out in 2012 in the European Union (Panagos et al., 2015c).
5. Model validation
5.1. General thoughts on model validation
‘Validation’ refers to the testing of the model output to confirm
the results that should be produced to reality (Fishman & Kiviat,
1968). Ecosystem modelers suggested that a ‘valid’ model of a
biological system would necessarily be an exact copy of the system
(Oreskes, Shraderfrechette, & Belitz, 1994). Alewell and
Manderscheid (1998) argued that a direct validation of a model
would prove that this model is an exact description of the modeled
system, the model output and the system's output matching under
any circumstance. As such, model validation would only be possible
for thermodynamically closed systems (Oreskes et al., 1994), where
all parameters and processes are described and measureable (e.g.
within a mathematical proof). Following this line of thinking, a
model describing a thermodynamically open system cannot be
validated.
A very different definition of validation derives from the
meaning of the Latin word validus (¼ strong, healthy, powerful) and
was used by Martin (1996). In that sense, a model is validated, if it
has been proven itself as a strong political or economic tool (Martin,
1996), e.g. putting a legislation into action or setting up a man-
agement plan. Foster, Yoder, Weesies, and Toy (2001), in presenting
the RUSLE2 concept, also connects model success to its usefulness,
when stating that a model which provides perfect erosion esti-
mates but which somehow leads to a poor conservation planning
decision has failed to meet its objectives. Indubitable, we have left
the realm of natural science with these lines of thinking. However,
following the recent critical debate of soil erosion modelling, the
definition of Martin (1996) or the logic of Foster et al. (2001) has
been implicitly applied. E.g., one of the main concluding arguments
to dismiss the Pan-European modelling of Panagos et al. (2015d)
was, that “erosion rate figures can be used by politicians/policy
makers to advocate measures that may be misguided” or that
“modelled information can be used to suggest that policies to
mitigate erosion are successfully tackling soil loss across Europe”
(Evans & Boardman, 2016a; Fiener & Auerswald, 2016).
If we accept models as tools in natural science, we can use them
to test hypothesis on process understanding, relative differences
between systems or potential development over time. Unlike a
mathematical proof, a scientific theory or hypothesis is empirical,
and is always open to falsification if new evidence is presented. A
given model which describes a specific system significantly better
might be declared the ‘valid’ model compared to another model
which might be rejected and the term ‘valid’ in this sense is then
used that any model that could not be proven invalid would be a
valid model for the system (Alewell & Manderscheid, 1998).
There is general agreement in modelling that the calibration
dataset must be independent from any dataset which is used later
to validate the model, and that if the same dataset is used for both it
should be no surprise that the model is a perfect predictor (Nearing,
2004). Split sample approaches, in which the available data is
separated into a calibration set and a separate validation set, are
usually the solution to this problem.
For the open systems described in ecosystem modelling it was
suggested that a model can only be validated, if it was not cali-
brated (Sverdrup, Warfvinge, Blake, & Goulding, 1995), implying
that measured input parameters implemented in the model will be
directly compared to measured target values without any inverse
214 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
process step. Interestingly, soil erosion modelling using empirical
models, as much as it has been criticized in the past, commonly
follows this most rigorous of all approaches: e.g., USLE-type model
applications derive the input parameters from measured data and,
without the process of calibration, compare modelled soil erosion
maps directly with measured data.
5.2. Validating USLE-type models with measured data
One major obstacle in validating the (R)USLE with measured
data is that (R)USLE provides gross erosion rates while most
monitoring programs and measurements (except for experimental
approaches) provide net erosion rates (e.g. on-site measurements
as the budget of erosion and deposition of sediments or even off-
site sediment yields in waters). Thus, and as long as USLE is not
connected to sediment transport models (which is mostly not
operational on large scales, see section 3), modelled and observed
data represent different fluxes. Or, as (Trimble & Crosson, 2000)
stated, USLE only presumes to predict the amount of soil moved on
a field (e.g., gross erosion), not necessarily the amount of soil
moved from a field (e.g., net erosion). As such, a RUSLE modelling
was not successfully validated when compared to sediment yields
of 454 small catchments with various land covers and uses across
the United States, Puerto Rico, U.S. Virign Island and Guam (Tan
et al., 2018). Instead, models that use the power function of
runoff, shear stress or stream power such as the Morgan, the tRIBS-
Erosion and the GUEST model were more successful to target
catchment sediment yields (Tan et al., 2018).
Nearing (2004) state that observations will always be closer to
the truth than modelling and must remain the most important
component of scientific investigation. However, Garcia-Ruiz et al.
(2015) observed a correlation of measured erosion rates with size
of the study area, measurement method and duration of experi-
ment or observation period in a worldwide meta-analysis of 4000
sites. Furthermore, there are many regions of the world being un-
derrepresented and erosion measurements lack long term obser-
vations (Garcia-Ruiz et al., 2015). Thus, it has to be taken into
account that observations, especially due to the highly heteroge-
neous nature of soil erosion and the tendency of erosion mea-
surements being very prone to errors (Garcia-Ruiz et al., 2015),
might also be biased in regional distribution or time periods chosen
by the researchers or due to high systematic errors or random
uncertainty (Rüttimann, Schaub, Prasuhn, and Rüegg (1995), see
section 5.2). As such, quality and potential biases of modelled and
measured data has to be equally scrutinized, when model valida-
tion with measured data is aimed at.
One reason of high modelling uncertainty frequently addressed
in the literature is certainly that erosion measurements itself are
connected to considerable uncertainty. Rüttimann et al. (1995) re-
ported a statistical analysis of data from four sites, each with five to
six reported treatments. Each treatment had three replications.
Reported coefficients of variation of soil loss within treatments
ranged from 3.4% to 173.2%, with the ploughed sites with higher soil
losses having a tendency to lower coefficient of variation between
3.4 and 71.4%. Work carried out by Wendt, Alberts, and Hjelmfelt
(1986) using 40 erosion plots, all with the same treatment in
Minnesota found that coefficients of variation for the 25 storms
ranged from 18% to 91%. The larger erosion events again showed
less variability. Very little of the variability could be attributed to
measured plot properties, and the plots did not perform in the
same manner relative to each other in subsequent events (Wendt
et al., 1986). Nearing, Govers, and Norton (1999) studied erosion
variability using data from replicated soil-loss plots from the Uni-
versal Soil Loss Equation (USLE) database. Data from replicated plot
pairs for 2061 storms, 797 annual erosion measurements and 53
multi-year erosion totals were used. Values of the coefficient of
variability ranged from nearly 150% for a soil loss of 1 Mg ha1yr1
to as low as 18% or less for soil-loss values greater than 100 Mg
ha1yr1 (Nearing et al., 1999).
As such, trying to validate USLE-type modelling with measured
data, needs to take the high variability of soil erosion measure-
ments into account. In addition, it has been noted all throughout
the history of using the USLE that there is a striking lack of data to
rigorously determine the single factors for all needed regions, sit-
uations and scenarios (Auerswald et al., 2003; Wischmeier and
Smith, 1965, 1978, p. 58). Many plot studies measuring soil
erosion rates investigated bare fallow plots as a baseline reference
which hardly ever exist in reality. While the focus of measurements
is towards medium sloped (e.g. > 3% < 18%) arable soils, flat areas or
steeper slopes >20% under grassland or forest used are hardly ever
investigated (Bonta & Sutton, 1983; Borst & McCall, 1945; Fan, 1987;
Kilinc & Richardson, 1973; Liu et al., 1994; Tresch, 2014, p. 132;
Zhang, Wang, & Yang, 2014). Furthermore, measurement plots
were either located in areas with relatively large rainfall erosivity or
experimental treatment with high irrigation rates and intensities
was used. So, all in all, there is bias in the available measurement
data towards regions or situations with conditions prone to erosion
(Auerswald et al., 2003; Wischmeier and Smith, 1965, 1978, p. 58).
Nevertheless, ultimately we need a validation of soil erosion
modelling to be able to attribute the relative importance of sedi-
ment source strength to specific kinds of land use and management
as well as other triggering factors (e.g. climatic events, topography,
and vegetation cover (Alewell, Birkholz, Meusburger, Wildhaber, &
Mabit, 2016; Meusburger, Banninger, & Alewell, 2010a;
Meusburger et al., 2010b; Panagos et al., 2015d; Scheurer, Alewell,
Banninger, & Burkhardt-Holm, 2009; Schindler Wildhaber,
Michel, Burkhardt-Holm, Baenninger, & Alewell, 2012). As such,
erosion models have been compared extensively to measured data
sets.
Risse et al. (1993) applied the USLE to 1700 plot years of data
from 208 natural runoff plots. Annual values of measured soil loss
averaged 35.1 Mg ha1 with an average magnitude of prediction
error of 21.3 Mg ha1, or approximately 60% of the mean (Risse
et al., 1993). Rapp (1994, pp. 1e95) followed the latter approach
and determined the error associated with RUSLE from 21 US sites
representing 1704 years of measurements from 206 plots. The
average annual magnitude of error was 11.7 Mg ha1 and the model
efficiency coefficient was 0.73, while the prediction of the 1638
individual annual erosion values had an average magnitude of error
of 20.8 Mg ha1 and a model efficiency of 0.58 (Rapp, 1994, pp.
1e95). Zhang, Nearing, Risse, and McGregor (1996b) applied the
Water Erosion Prediction Project (WEPP) computer simulation
model to 290 annual values and obtained an average of 21.8 Mg
ha1for the measured soil loss, with an average magnitude of
prediction error of 13.4 Mg ha1, or approximately 61% of the mean.
In all cases the relative errors tended to be greater for the lower
soil-loss values with an over prediction on plots with low erosion
rates and under prediction on plots with high erosion rates. Even
though deviations from modelled and measured data might seem
substantial in absolute numbers, the deviation between modelled
and observed erosion rates is not larger than the variability within
measured data.
Kinnell (2010) in evaluating data from Rosewell (1993) also
concludes that even though the RUSLE works well for a number of
agricultural conditions in New South Wales, Australia, on a wider
geographic area and larger variety of agricultural systems, there is a
tendency for the USLE and the RUSLE to over-predict low average
annual soil losses and to under-predict high average annual soil
losses. Zhang, Nearing, Risse, and McGregor (1996c) also discussed
an over prediction of small soil losses as well an under-prediction of
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
large soil losses for USLE approaches as well as the WEPP model.
On-site erosion measurements are not available on large scale,
but Cerdan et al. (2010) compiled an extensive database from
literature on short to medium-term erosion rates as measured on
erosion plots in Europe under natural rainfall conditions. Panagos
et al. (2014c) compared the soil erosion estimates of this
compiled plot data base (Cerdan et al., 2010) with modelled values
from the Pan-European Soil Erosion Risk Assessment (PESERA)
model (Kirkby et al., 2008) and the European Environment Infor-
mation and Observation Network for soil (EIONET-SOIL) database of
Europe (which are regional to national USLE-type model applica-
tions). The mean values of soil loss reported by the national in-
stitutes (EIONET-SOIL) are larger than the PESERA estimates, with
the main differences being for sloping land (>2 ) and for the land
cover type forest and heterogeneous agricultural land cover
(Panagos et al., 2014c). A graphical illustration of this comparison
results in a deviation of both modelling approaches from the plot
based measurements at the high end of modelled soil losses (Fig. 5).
Cerdan et al. (2010) warned of extrapolations from local mea-
surements to regions and in addition the comparison might illus-
trate the above discussed principally different nature of net erosion
rates based on plot measurements versus larger scale modelled
gross rates. Furthermore, there is no straightforward conclusion
from these comparisons of modelled and measured data: are the
models invalidated by the measured data or is the measured data
biased and/or not reflecting large spatial scale and long temporal
average situations?
As the direct measurement of soil erosion on plots implies long
term monitoring all year round, the indirect assessment using
fallout radionuclides (FRN) as tracers of soil erosion has been used.
Meusburger et al. (2010b) with an adapted C factor implemented
from fractional vegetation cover map derived from QuickBird im-
agery, validated USLE derived erosion rates for hot spots of erosion
in Alpine grasslands against FRN derived erosion rates (20.1 Mg
ha1yr1 versus 16 Mg ha1yr1 for 137Cs based versus modelled
rates, respectively). RUSLE was also successfully validated at sites of
South Korea, with the 137Cs method ranging from 0.9 Mg ha1yr1
to 7 Mg ha1yr1 and RUSLE (considering the steep slopes of up to
40 and the erosive monsoon events (R factor of 6600
MJmmha1 h1 yr1)) yielding 0.02 Mg ha1yr1 to 5.1 Mg
Fig. 5. Erosion estimates from plot based measurements (Cerdan et al., 2010) versus
modelled erosion rates from PESERA (Kirkby et al., 2008), USLE-type models (EIONET
Data Base; (Panagos et al., 2014c)) and RUSLE2015 (Panagos et al., 2015d).
215
ha1yr1 (Meusburger et al., 2013). Thus, even though the FRN
method also determines net erosion rates while USLE-type
modelling results in gross erosion estimates, the promising re-
sults of these studies might corroborate the assumptions that
comparisons may be applicable over long time periods of several
decades which are evaluated by the FRN methods. However,
Haciyakupoglu, Hizal, Gokbulak, and Kiziltas (2010) compared 137Cs
based erosion rates to USLE assessment in the Turkish Omerli
catchment and found substantial deviations between the two
methods (higher ULSE rates by a factor of 2-3). The latter over-
estimation of soil losses by USLE compared to 137Cs based erosion
rates was also noted by Warren et al. (2005), but was overcome
when USPED was applied, which considers net erosion rates (see
section 3) and might thus be better comparable to 137Cs based
erosion rates.
Regardless of the latter promising FRN results the question may
be asked or, nevertheless, has to be asked: are the model pre-
dictions ‘good enough’? And what is actually ’good enough‘ or what
are the reference values considering the high uncertainty and
biases of measured data? Even though Bagarello et al. (2012) con-
cludes that soil loss estimates are considered reasonably accurate
for most practical purposes, especially for wide spatial scale ap-
plications, when the errors of the predictions do not exceed a factor
of two or three, erosion modelling remains assailable as long as
there is no general agreement on the ’good enough’.
Nearing (2004) concluded that model validation is not just a
matter of comparing measured to modelled data, one must also ask
the question: “How variable is nature?” We would like to add, that
in bidding farewell to the idea of accurately predicting absolute
values with models but rather concentrating on the prediction of
relative differences, trends over times and systems reactions to
processes and management practices, we can use models as tools to
learn about the modelled systems and their reactions. In this con-
ceptual approach, modelling in general and large-scale modelling
specifically will per se not aim at an accurate prediction of point
measurements, but will test hypotheses on process understanding,
relative spatial and temporal variations, scenario development and
controlling factors (Oreskes et al., 1994). In an agreement with the
latter and the above discussion, many modelers accept that full
model validation is a logical impossibility and Morton and Su'arez
(2001), Refsgaard and Storm (1996) as well as Senarath, Ogden,
Downer, and Sharif (2000) suggest that in most practical contexts
the term ‘model’ should be thought of as being synonymous with
‘theory’ or ‘hypothesis’, with the added implication that they are
being confronted and evaluated with data. An interesting example
of such ’hypothesis testing‘ is a study in northern Bavaria where a
comparison of conventional with organic farming practices found
measured soil losses to be about an order of magnitude lower than
RUSLE based values due to best management practices adopted on
both farms (Auerswald et al., 2003). However, modelled and
measured values pointed to the same conclusion of organic agri-
culture resulting in 15% less erosion on arable land despite its
localization in areas with substantially higher erosion susceptibility
(Auerswald et al., 2003). Garcia-Ruiz et al. (2015) concluded from a
global meta-analysis that “Geomorphologists should be more
interested in processes, causes, internal relationships, and the role
of stream channels than in obtaining absolute erosion rates, the
interpretation of which can lead to numerous errors”. However, as
USLE-type modelling is often used by land scape managers and
stake holders, it is important to communicate that these modelling
endeavors should not be used to predict absolute erosion rates (e.g.
for comparison to legal limit values) but rather to optimize catch-
ment management and land use in delineating best practice
techniques.
With the acceptance that at least at large scales we should not
216 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
be solely focused on an accurate prediction of absolute values but
rather see models as a tool to test hypotheses of relative differences
between systems, trends over time, systems reactions to driving
factors and management practices, we should, however, not
generally exclude the use of models to assess soil erosion magni-
tudes at least at smaller scales with the aim to assess sustainability
of management. An example would be a study by Alewell, Egli, and
Meusburger (2015) where 137Cs as well as 239þ240Pu derived soil
erosion estimates were compared with USLE estimates from
Meusburger et al. (2010b) and confronted with soil formation rates
in Swiss Alpine grasslands at plot scale. The conclusion from the
latter was, that measured and modelled soil erosion rates compare
relatively well but that they are in any case considerably higher
than soil formation rates and that thus land management cannot
claim to be sustainable.
6. Model comparisons
6.1. Model choice and suitability
A complete overview in a recent review on water erosion
models identified 82 models (Karydas et al., 2014). As stated above
an Web of Science query (http://apps.isiknowledge.com/) for the
period 2007e2017 resulted in 1149 hits corresponding to the key-
words “Universal Soil Loss Equation”, “USLE”, “Revised Universal
Soil Loss Equation” or “RUSLE”. In comparison, most popular ap-
proaches independent of the USLE technology, such as WEPP
(Laflen et al., 1997), LISEM (De Roo et al., 1996), EUROSEM (Morgan
et al., 1998) and PESERA (Kirkby et al., 2008) totalized 254 hits.
Needless to say, that the fact that USLE-type models are the most
commonly used erosion tools globally, does not rank them auto-
matically as the best available technique and the suitability of
model concept needs to be proven for each specific case.
In the choice of the soil erosion model to be applied, and
congruent with the discussion of section 5, it has to be taken into
account that it is not always necessary to calculate the exact erosion
rate for a particular situation (Nearing, 2013, pp. 365e378), but
rather to compare among different situations or time periods.
Choosing which model to apply and model development might
become a matter of what type of information is available as input
data. As an example, Nearing (2013, pp. 365e378) noted that most
applications of WEPP were developed in the United States because
of the availability of soil, climate and crop information for those
areas. Another important criterion for the choice of a soil erosion
model is the question of spatial and temporal scale of available
input parameters as well as envisaged model output. As most
erosion parameters are scale-dependent, data collected on plot
scale are not always appropriate for calibrating or validating
models (see also section 5) and upscaling of models through data
aggregation may cause biases (Karydas et al., 2014).
In today's modelling world, many of the current generation of
erosion models are getting more complex, with ever more pro-
cesses being incorporated into them (Quinton, 2013). However,
even the most complex model of erosion and sedimentation pro-
cesses can still only be an approximate representation of reality and
not reality itself. The problems with complex models are twofold:
(i) unavailability of the complex set of input parameters especially
at large temporal and spatial scales and (ii) over parameterization
which results in a non-uniqueness of fit and/or insensitivity of
model output to some of the input parameters. From a mathe-
matical view point the calibration of ecosystem models in general
are hardly ever unique and the non-uniqueness of fit increases with
increasing numbers of parameters (Alewell & Manderscheid, 1998).
Quinton (2013) as well as Veihe and Quinton (2000) and Veihe,
Quinton, and Poesen (2000) showed that many of EUROSEM's input
parameters (e.g. hydrological as well as vegetation parameters) did
not invoke a sensitive response from the model. A similar pattern
can be observed for WEPP in the work of Brazier, Beven, Freer, and
Rowan (2000). Quinton (2013), in discussing the uncertainty of
complex erosion models like EUROSEM and WEPP, suggests that
this uncertainty is reason enough to describe the erosion system
with simpler models than those that are currently in vogue. One
reason for the popularity of USLE-type modelling is certainly that it
is a good compromise between applicability in terms of required
input data and relatively good reliability of obtainable soil loss es-
timates (Risse et al., 1993). As such, it has become the standard
technique of many soil conservation workers (Morgan, 2005, pp.
365e376). However, with remote sensing techniques improving
the availability of high resolution data immensely in recent years
(dynamic monitoring of vegetation development as well as esti-
mating dynamic of climate variables), the way is paved for process
based techniques in the future needing high resolution spatial
input data.
The choice of a soil erosion-prediction tool has to be dependent
on the spatial and temporal scale of the intended model applica-
tion, as the question of scale is crucial in choosing the right
modelling approach. Process-oriented models require an applica-
tion of the used equations at a given spatial scale, ranging from plot
to basin, and at event temporal scale. Event and at-point or small
scale (hillslope) process based models are not suitable to be applied
for simulating soil loss on a wide region not only because of the lack
of required input data but also because of the required indepen-
dency of spatially discrete units. At a large spatial scale, the area has
to be discretized using, for example, a square grid subdivision
(raster scheme), choosing a mesh size consistent with the scale of
the original model deduction. Using a raster scheme applied to the
USLE model corresponds to hypothesize that each cell is indepen-
dent of the others with respect to soil loss. Even though this meets
the original requirements postulated by Wischmeier and Smith
(1965) that for larger scale applications the area needs to be
“broken down into a series of tracts having relatively homogeneous
land use and treatment.”, it requires independency of each of these
units. In other words, at large spatial scales such as a region or even
a global perspective, a simple index-based model able to calculate
an average soil loss at annual or mean annual scale, allows
computing the involved factors only using spatially distributed
input values. However, the cells cannot be assumed to be inde-
pendent of each other when sediment delivery processes have to be
modeled (e.g., at basin scale). In this case, the USLE scheme has
been applied by coupling it with a mathematical operator
expressing the hillslope transport efficiency (Ferro, 1997; Ferro &
Porto, 2000).
6.2. Direct model comparison used implicitly as model validation
Model comparison has been used implicitly as a way of model
validation. However, for a rigorous model comparison the question
of scale is decisive (see also section 7). In a European application of
the models PESERA and SENSOR both models compared well,
however when ‘validated’ to the national scale USLE derived
erosion map of Hungary, the European applications of SENSOR and
PESERA underestimated (or USLE overestimated) erosion rates
considerably (Fig. 6a; (Podmanicky et al., 2011)). The original
thought of this comparison was, to validate the European applica-
tions of SENSOR and PESERA with the higher resolution soil erosion
rates from Hungary derived with USLE (Podmanicky et al., 2011).
However, if model type and scale are changed simultaneously, it
remains unsolved, whether the deviation originates from scaling
inconsistencies or from model deviations or even failures.
A comparison of national scale USLE-type modelling (EIONET
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
Fig. 6. Model comparison of soil loss data (Mg ha1yr1) calculated with SENSOR and
PESERA calibrated to European scale with USLE estimates applied to the national data
of Hungary (data of Podmanicky et al. (2011), Fig. 6a) and of national USLE-type
modelling (EIONET data base, Panagos et al. (2014c)) to PESERA (Kirkby et al., 2008)
and RUSLE2015 (Panagos et al., 2015d), Fig. 6b).
data base, Panagos et al. (2014c)) with the European scale appli-
cation of RUSLE2015 (Panagos et al., 2015d) or PESERA (Kirkby et al.,
2008) resulted in a relatively good agreement for low values but
high deviations of the models at the higher end of soil losses
(Fig. 6b).
Stolpe (2005) compared RUSLE, WEPP and EPIC at the Medi-
terranean climate of Chile for their consistency with measured
erosion rates (2*20 m plots) and those calculated using the previ-
ously calibrated USLE. Their conclusion was that WEPP permitted
the most adequate parameterization for both climate and volcanic
soils and that WEPP erosion rates were in best agreement with
measured rates as well as with calibrated USLE values (Fig. 7).
Nearing (2004) in comparing USLE and WEPP in two case studies,
concluded that USLE can provide certain information which WEPP
simply cannot provide because of the restrictions of model
complexity, but that the WEPP model can be used in a way where
only the complex model interactions will provide the information
needed regarding system response.
Kinnell (2010) criticizes that WEPP ignores any variation in the
kinetic energy per unit quantity of rain, even though it is dependent
on rainfall intensity data. These intensities are generated by the
stochastic daily parameter weather generator CLIGEN (Nicks, Lane,
& Gander, 1995), but the meteorological data required to parame-
terize CLIGEN are not available in many countries outside the USA.
The ability of WEPP to predict soil losses in the USA has been found
217
Fig. 7. Erosion estimates at the Mediterranean climate of Chile from plot based mea-
surements versus modelled rates from USLE, RUSLE, EPIC and WEPP (all data from
Stolpe (2005) with two different parameterizations of the WEPP model).
to be not better than either the USLE or the RUSLE (Kinnell, 2010).
USLE-type application has been criticized in complex topo-
graphic terrain, because the LS factor does not consider terrain
shape of the upslope area (Desmet & Govers, 1996; Mitasova et al.,
1996). Mitasova et al. (1996) compared the USLE LS factor and the
topographic erosion/deposition index incorporated in the so called
unit stream power based approach USPED and concluded, that the
latter is more appropriate for landscape erosion modelling, espe-
cially when erosion risk and deposition potential needs to be
evaluated.
Using more than 1600 plot years of data from natural
rainfallerunoff plots in the USA, Tiwari, Risse, and Nearing (2000)
observed that for average annual soil losses, the model efficiency
of WEPP determined using the (Nash & Sutcliffe, 1970) factor was
0.71, whereas it was 0.80 and 0.72 for the USLE and the RUSLE,
respectively. However, as the WEPP model derives its strength from
being a process based model and has the capability to predict
spatial and temporal distribution of soil loss and deposition, it has
to be considered in the latter study that the data set was biased
towards the USLE (all uniform natural runoff plots), the parameters
used in the USLE have undergone more refinement, and the WEPP
model was not calibrated at all (Tiwari et al., 2000). WEPP, USLE and
RUSLE were observed to over-predict low and under-predict high
annual soil losses by Kinnell (2010). However, the latter was not
confirmed for the above presented data from Europe (Fig. 7).
Favis-Mortlock (1997) drew together the results from models
simulating annual soil loss for a field site in the USA. The models
ranged from simple models, such as CSEP (Kirkby and Cox, 1995)
and those based on the USLE (Wischmeier & Smith, 1978, p. 58)
such as EPIC (Williams et al., 1983) and GLEAMS (Leonard, Knisel, &
Still, 1987), to the more process based and complex WEPP model.
Favis-Mortlock's results show no significant difference between the
performances of the models. Morgan and Nearing (2011) compared
the performance of the USLE, RUSLE and WEPP using 1700 plot
years of data from 208 natural runoff plots. Once again it became
clear that the more complex model, in this case WEPP, performs no
better than the empirical, five parameter USLE (Morgan & Nearing,
2011).
Using the USLE basically comes down to testing the hypothesis,
that only a small number of parameters control the response of an
erosion model and the remainder become largely inconsequential
(Quinton, 2013). As such, model uncertainty does not seem to be
218 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
smaller in complex, process based models compared to simpler,
empirical models. Quinton (2013) when discussing several erosion
model studies increasing the complexity of models and comparing
it to the original simpler version, concludes that from a scientific
point of view, the inclusion of more processes and parameters
make the models a better representation of reality, but it did not
make the models better in predicting erosion rates. However,
stating this, we have to keep in mind that as noted above, USLE-
type modelling will always be less successful when models are
compared to catchment sediment yields or off-site measurements
in freshwaters (see discussion in section 5.2 and as good example
(Tan et al., 2018)).
To conclude, there is not one ”superior” model suitable for all
kind of situations, temporal and spatial scales as well as data res-
olution. This basically leaves soil erosion modellers with two pos-
sibilities: (i) accept that there are no adequate tools to model soil
erosion by water and give up modelling until modelling and the
required input data sets are improved or (ii) recollect what was
discussed in section 5: At least at large scales we should bid fare-
well to the idea of accurately predicting absolute values with
models but should rather aim at the prediction of relative differ-
ences, trends over times and systems reactions to driving factors
and management practices and use models as tools to learn about
the modelled systems and their reactions. In this conceptual
approach, modelling in general and large-scale modelling specif-
ically will be understood as a tool to test hypotheses on process
understanding, relative spatial and temporal variations, scenario
development and controlling factors (Oreskes et al., 1994).
7. Upscaling: from plot to global and from event to long term
averages
Application of environmental modelling always involves four
different scales: the geographic scale of a research area, the tem-
poral scale related to the time period of research, the measurement
scale of parameters (input data resolution) and the model scale
referring to both temporal and spatial scales when a model was
established (Zhang, Drage, & Wainwright, 2004b).
The crucial importance of temporal scale becomes obvious
when considering that over and under-estimation of modelled
erosion rates might be connected merely to the length in obser-
vational data: when the observation period is too short to catch the
long-term average, observations either miss out on extreme events,
or might by chance catch an accumulation of extreme events. The
USLE, designed to predict long-term average, will then of course
overestimate or underestimate these short measurement periods,
respectively.
Regarding the spatial scale, the original USLE model and its
factors were developed for plot scale of 22.1 m length and thus its
validity for large scale application has been discussed in recent
debates (Auerswald et al., 2015; Evans, 2013; Evans and Boardman,
2016a, 2016b; Fiener & Auerswald, 2016; Panagos et al., 2016a,
2016b; Panagos et al., 2015e; Schwertmann et al., 1987, pp. 1e64).
Schwertmann et al. (1987, pp. 1e64) when transferring the model
from the US to sites in Bavaria (Southern Germany) warned that the
model approach is limited to plot scale and other equations should
be used for catchments scale applications. However, Wischmeier
and Smith (1965) already stated there is principally no logical
reason why it should not be correct to “break down the drainage
area into a series of tracts having relatively homogeneous land use
and treatment. The erosion equation is then used to approximate
the average annual rate of soil movement from each tract.” Large
scale modelling using GIS tools is basically doing nothing else,
when approximating average annual rates of soil movement for
single tracts (¼pixels) presenting the result of each of these pixels
as a graphical summary in maps. With a spatially discrete model-
ling approach as the USLE-type algorithm uncertainty is per se in-
dependent from scale, because the uncertainty within each single
discrete pixel is independent from the number of these pixels. The
uncertainty is merely coupled to the resolution of the modelling
approach (e.g. size of pixels) and the connected heterogeneity
within these pixels. However, limitation of the modelling capacity
might be connected to scale, as demands on input data and
computational capacity is of course increasing with increase in
resolution and scale.
Again, as discussed above, when aiming at large scales and
considering erosion risk of (large) catchments, nations, continents
or even global scales, often an misunderstanding evolves during
this transfer from small to large scale and USLE-type modelling
output is implicitly compared to sediment yield and sediment
transport capacities. As such, evaluating large-scale model appli-
cation of USLE-type modelling needs the repeated reminder that
on-site erosion risk is simulated and not off-site sediment yield and
transport.
While modelling on catchment scale nowadays operates on a
resolution of 2 m (depending on DEM resolution, see section 4), the
latest large scale modelling assessments achieved a spatial reso-
lution of 100 m for European (Panagos et al., 2015d) and 250 m for
global scale (Borrelli et al., 2017; Diodato et al., 2017). The latter
approaches use considerably coarser resolution than the original
USLE-type plot of 22.1 m length. However, variability does not
necessarily increase with increase in scale. Even though it might be
expected that measured soil loss variability should decrease with
increasing plot size, Rüttimann et al. (1995) could not confirm an
effect of plot size on increase in variability but rather noticed a
striking absence of any effect of an increase in plot size on mea-
surement variability or uncertainty.
So one question to be answered is, how homogenous the model
units (‘tracts’ or better pixels) are and how valid the approach is in
terms of assuming homogenous R, K, LS, C and P factors within each
pixel. As spatial data availability is increasing continuously in res-
olution and remote sensing tools shedding more and more light on
small scale heterogeneity, the above question might eventually
dissipate in the future. In this, there is no principle difference be-
tween a global scale modelling from a national scale modelling and
not even from large catchments: it always comes down to pixel size
and data quality versus the system's heterogeneity.
Kinnell (2010) warns at applying the USLE/RUSLE model to
extrapolate data collected at the plot scale to areas much larger area
than the field sized areas for which it was originally designed. He
argues that at the hillslope scale, spatial variability in soil and
vegetation result in spatial variations in runoff, so that the
modelling approach would need to consider those spatial varia-
tions in runoff, which models like WEPP and EUROSEM have an
advantage over the USLE, RUSLE and RUSLE2 because they include
explicit consideration of runoff (Kinnell, 2010). Again, the implicit
transfer from on-site soil erosion risk to off-site sediment yields
might have slipped into the discussion. However, the problem with
increasing variability in vegetation cover at hill slope (or even
larger) scales might at least partially be overcome by advanced
approaches of considering high resolution spatial input data from
remote sensing. One example would be a study of Meusburger et al.
(2010b) who implemented spatial patterns of fractional vegetation
cover from Quickbird imagery into the C factor map and achieved
(i) a good agreement of spatial patterns of soil erosion predicted
with USLE compared to observed remote sensing patterns and (ii)
erosion rates which are at least in the same order of magnitude
compared to 137Cs derived erosion rates (for hot spots of erosion
20.1 Mg ha1yr1 versus 16 Mg ha1yr1 for 137Cs based and USLE
derived soil losses, respectively). The latter was neither possible
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
with an USLE application with uniform C factor rates not consid-
ering the spatial heterogeneous distribution of vegetation cover
with PESERA where the same spatially distributed vegetation cover
was considered but spatial soil loss patterns could not be captured
adequately and soil losses on hot spots were considerably under-
estimated (2.9 Mg ha1yr1 compared to the above mentioned
137
Cs based estimates (Meusburger et al., 2010b)). Schmidt, Alewell,
and Meusburger (2019) followed the latter approach and combined
the spatio-temporal heterogeneity from the C factor map with the
spatio-temporal heterogeneity of the R factor to a soil erosion risk
map of Switzerland with monthly temporal and 100 m spatial
resolution.
8. Limitations of the USLE concept and the practical need to
model soil erosion
Recent studies continue to question the use of models to predict
erosion potential, and argue that instead there is a need for more
field-based assessment and monitoring of water erosion (Evans,
2013). Modelling cannot be an alternative to measurement and
monitoring but might be a powerful tool in understanding obser-
vations and in developing and testing theories. Understanding
processes, regulating parameters, trends and differences between
spatially or temporally differing data is very often the ultimate goal
in science and models serve as tools towards that end. As such,
models can be understood as a tool for understanding, as well as
tools for simulation and prediction, a virtual laboratory which
might as a positive aspect serve as an integrator within and be-
tween disciplines, as it brings together data, observations and
knowledge of different fields (Nearing, 2004). However, as models
are also used as means of communicating science and the results of
science, there is always the danger, that what they communicate is
weak or even wrong and these weaknesses might not be apparent
to the model user, or the user of the model output (Nearing, 2004).
Trimble and Crosson (2000) criticized the lack of prediction of
sediment delivery ratios by USLE and its failure to predict gully
erosion, which led them to conclude for the US that “the limitations
of the USLE (…) are such that we do not seem to have a truly
informed idea of how much soil erosion is occurring in this country,
let alone of the processes of sediment movement and deposition.”
As discussed above, USLE-type models were not designed to predict
gully erosion nor sediment delivery ratios and modelers, model
users and stakeholders should of course be aware of underlying
model concepts and parameters, the above discussed limitations
and the possible evaluation and interpretation of the model output.
Regarding the latter, the simplicity of the USLE-concept might be
regarded as an advantage as each of the five USLE parameters might
be evaluated separately even by non-expert stakeholders
increasing thus transparency and objectiveness of evaluation.
The lack of quantifying gully erosion or stream bank erosion has
been frequently addressed as a lack or even general failure of USLE-
type modelling (Belyaev, Wallbrink, Golosov, Murray, & Sidorchuk,
2005; Evans and Boardman, 2016a, 2016b; Quinton, 2013; Trimble
& Crosson, 2000). Gully erosion and stream bank erosion involve
complex and highly heterogeneous hydrological and soil erosion
processes of channel formations difficult to model. We would like
to state that the USLE was never meant to address gully, wind or
stream bank erosion, as is clearly stated in the very first publica-
tions of USLE (Wischmeier and Smith, 1965, 1978, p. 58). It would be
like discrediting a butterfly monitoring tool for missing out on
important bird species.
USLE-type modelling is frequently discussed to be developed for
US type soils and systems and thus questioned to be applicable for
other regions (see above). However, as long as model parameters
are carefully considered and adapted to site or region specific
219
climate and rain patterns, vegetation cover and management, as
well as the specific soil characteristics, US type systems and soils
are not a case sui generis and soil erosion from US soils is regulated
by the same factors and processes as everywhere else in the world.
Long term investigations of Schwertmann et al. (1987, pp. 1e64)
already supported the latter assumption three decades ago.
Recent studies confirmed that when appropriately parametrized,
uncertainty of USLE-type modelling all around the globe is not
bigger than within the US (Kinnell, 2010; Meusburger et al., 2010b,
2013; Stolpe, 2005; Yue et al., 2016).
Apart from the discussion whether or not modelled soil erosion
estimates are actually ‘good enough’ from a scientific perspective
and the question what actually ‘good enough’ exactly is, we would
like to raise another issue: the practical need of soil erosion
modelling.
Without modelling there will be no large scale erosion assess-
ments. Without large scale erosion assessments, soil erosion as one
of the most serious threats to soils will not be taken into consid-
eration by any major environmental and agricultural policy pro-
grams, as well as integrated land surface models (LSMs). For
example, the voluntary Guidelines for Sustainable Soil Manage-
ment of Food and Agriculture Organization (FAO, 2017) identify soil
erosion by water and wind as the most significant threat to global
soils and the ecosystem services they provide. At global scale, the
Sustainable Development Goals (SDGs) include a sub-indicator on
estimated soil erosion by water. Europe's Common Agricultural
Policy sets requirements to protect utilized agricultural areas
against erosion, which established a framework of standards that
aim, among others, to preventing soil erosion. The Good Agricul-
tural and Environmental conditions (GAEC) implemented in the
European Common Agricultural Policy reform (Borrelli et al., 2016c)
introduced the prevention of soil erosion and higher residues
restitution, increased soil cover and lowered tillage disturbance
which are among the most effective mitigation practices in Euro-
pean arable lands. All these policy programs can not consider soil
erosion if there are no tools for large scale modelling and mapping
available.
Tolerable soil erosion rates all over the world are given between
0.1 and 1 mm yr1 (for overview of all studies see Li, Du, Wu, and
Liu (2009)). For sake of simplicity, these mm rates might be con-
verted with a bulk density of 1 t m3 which is typical for the loose
soil structure of surface soils and would result in soil loss rates of
1e10 Mg ha1yr1. For conditions prevalent in Europe, the upper
limit of soil formation has been estimated to be approximately
1.4 Mg ha1 yr1 while the lower limit is given as 0.3 Mg ha1 yr1
(Alewell et al., 2015; Verheijen, Jones, Rickson, & Smith, 2009). As
tolerable soil erosion rates should not exceed soil formation rates,
many countries in the world are actually performing non-
sustainable agriculture. Without large-scale soil erosion mapping
we do not have tools to address this problem and thus no possi-
bilities to demand mitigation actions.
9. Conclusions and outlook
Soil loss occurs not because of any lack of knowledge on how to
protect soils, but a lack in policy governance (Montanarella, 2015;
Panagos et al., 2016b). The lack in policy governance is due to a lack
in research, knowledge and dissemination on spatial distribution of
soil erosion and sediment source attribution from water and wind
erosion. As such, soil erosion modelling is crucially needed on both
small and large spatial scales for planning, management and policy
measures. The USLE-type models have been extensively used and
modified during the last decades reaching the ceiling of improve-
ment in recent years. Recent advancements in remote sensing,
release of climatic datasets, greater availability of earth
220 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
observations data, and increased processing of big datasets are the
promising factors for more dynamic and process-based modelling
approaches in the near future. Soil erosion modeling moves to-
wards more dynamic approaches and can use the field parcels as
objects due to availability of highly accurate spatial and climatic
data for monitoring intra-annual variability of climate, vegetation
and management practices (Borrelli, Meusburger, Ballabio,
Panagos, Alewell, 2018).
However, improving interpretation of the soil erosion processes
at different spatial and temporal scales and testing or developing
physically oriented modeling approaches is not only of socio-
economic and political importance but has also an obvious scien-
tific aim of process understanding and general prediction of
ecosystem dynamics (as soil erosion is an important process of
ecosystem stability and dynamic, e.g., erosion rates in climate
change modelling of carbon dynamics, soil nutrient balance or
pollution dissemination prediction).
From our literature evaluation, we conclude that at today's state
of knowledge USLE-type models have no higher uncertainty than
more complex based models provided modelling targets at on-site
soil erosion risk (compared to off-site sediment transport and
yield). Simultaneously, they can be a choice with regard to data
availability concerns and/or ease of usage and interpretation. This
has to be considered with care in case of increased data input
availability in the near future which will allow modelling erosional
process in a more dynamic and potentially process based way.
Obtaining accurate and reliable soil loss estimates using
spatially distributed models on wide regions depends on both the
resolution (vertical and horizontal) of the input topographic in-
formation and the quality of the land-use input data. USLE-type
models are a compromise for wide region application when both
a mesh-size comparable with the original developed slope-length
scale is applied and information on rainfall, soil and land-use sys-
tems is available.
The meta-analysis of published articles using USLE-type
modelling to estimate soil loss by water erosion from local to
continental scale showed an increasing trend especially in the last
20 years. The models are widely used on all continents and within
109 countries. The long-tradition of articles published mainly in the
US in the early ‘80s and ‘90s, has been continued in Europe,
Australia and Southern America and more recently increasingly in
Asia and Africa with scientists addressing also emerging research
topics connected to soil erosion such as climate change, nutrients
balance, freshwater pollution, land use and management and pro-
tection measures.
As discussed above, soil erosion modelling is, like any other
ecosystem model, only a representation of reality and not reality
itself, and is thus prone to deviate from reality and predicted un-
certainty might be more or less substantial. As such, for each
modelling endeavor a deep understanding of underlying processes,
assumptions and model structure is necessary to avoid misinter-
pretation (e.g. in case of USLE-Type modelling the conceptual
confusion between on-site gross erosion risk and off-site net
sediment yields). However, the above discussed validation at-
tempts showed that when appropriately parameterized soil loss
estimates with USLE-type models are within the order of magni-
tude compared to measured soil loss rates and relative differences
between areas, management scenarios and time periods seem to be
captured correctly in many studies. Based on this, the modelling
approach can be recommended to be used to target risk areas,
scenario calculation for conservation tools or other management
plans, spatial comparison of regions, countries and continents and
temporal comparisons of differing time periods. At the current state
of knowledge, USLE-type models are the only model algorithms
which have been used at large scales (e.g. continental, global), with
the exception of the pan-European application of the PESERA
model (Kirkby et al., 2008).
As measured soil loss data itself can be object to high un-
certainties, future research should target reducing uncertainties in
modelling and measurement. Comprehensive and well planned
measurement campaigns at small as well as large scale are crucially
needed for soil erosion modelling validation as well as deepening
our process understanding. As such, it is recommended to invest on
monitoring campaigns and experimental designs towards more
dynamic modelling. In addition, comparison of large scale model-
ling with local or regional studies might help to narrow possible
pitfalls, inconsistencies and uncertainty of modelling. A possible
future model validation approach might be the comparison of
continental scale modeling with local modelling data sets, which
might give important insights into processes, trends and driving
factors of soil erosion. In the upcoming Land Use/Cover Area frame
statistical Survey Soil (LUCAS Soil) data collection campaign, sur-
veyors will make a qualitative assessment of soil erosion in >26,000
locations in Europe (Orgiazzi, Ballabio, Panagos, Jones, &
Fernandez-Ugalde, 2018). As LUCAS is repeated every 3 years, this
has the potential to evolve into a monitoring program, which can
eventually be used to validate modeling on European scale.
Validation of the USLE-type applications are usually inexplicitly
carried out in the most rigorous of possible validation procedures as
the model algorithms are per se not calibrated, but the final out-
comes are compared to measured data.
Despite the large number of soil erosion models available for the
prediction of both runoff and soil loss at a variety of scales, little
quantification is made of uncertainty and error associated with
model output. The latter seems crucially important to not only
advance todays understanding but also the acceptance of soil
erosion modelling outputs on larger scales. It might be suggested
that future applications should use objective modelling criteria for
evaluation (e.g. as has been suggested by Alewell and
Manderscheid (1998) or Brazier et al. (2000)). However, the latter
would not solve the paradigm of the principally different nature of
modelled (gross) versus measured (net) erosion rates by USLE-type
models. As such, parallel to increasing research efforts in soil
erosion measurement and investing in rigorous application of
evaluation criteria, the path where soil erosion is connected to
topography, runoff and flow path dynamic as well as depositional
regime should be further followed and solutions should be found
for large-scale applications.
References
Aiello, A., Adamo, M., & Canora, F. (2015). Remote sensing and GIS to assess soil
erosion with RUSLE3D and USPED at river basin scale in southern Italy. Catena,
131, 174e185.
Alewell, C., Birkholz, A., Meusburger, K., Wildhaber, Y. S., & Mabit, L. (2016).
Quantitative sediment source attribution with compound-specific isotope
analysis in a C3 plant-dominated catchment (central Switzerland). Bio-
geosciences, 13(5), 1587e1596.
Alewell, C., Egli, M., & Meusburger, K. (2015). An attempt to estimate tolerable soil
erosion rates by matching soil formation with denudation in Alpine grasslands.
Journal of Soils and Sediments, 15(6), 1383e1399.
Alewell, C., & Manderscheid, B. (1998). Use of objective criteria for the assessment
of biogeochemical ecosystem models. Ecological Modelling, 107(2), 213e224.
Amundson, R., Berhe, A. A., Hopmans, J. W., Olson, C., Sztein, A. E., & Sparks, D. L.
(2015). Soil and human security in the 21st century. Science, 348(6235), 6.
Angulo-Martinez, M., Begueria, S., & Kysely, J. (2016). Use of disdrometer data to
evaluate the relationship of rainfall kinetic energy and intensity (KE-I). Science
of the Total Environment, 568, 83e94.
Angulo-Martinez, M., Lopez-Vicente, M., Vicente-Serrano, S. M., & Begueria, S.
(2009). Mapping rainfall erosivity at a regional scale: A comparison of inter-
polation methods in the Ebro basin (NE Spain). Hydrology and Earth System
Sciences, 13(10), 1907e1920.
Arnhold, S., Lindner, S., Lee, B., Martin, E., Kettering, J., Nguyen, T. T., et al. (2014).
Conventional and organic farming: Soil erosion and conservation potential for
row crop cultivation. Geoderma, 219, 89e105.
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
Arnold, J. G., Srinivasan, R., Muttiah, R. S., & Williams, J. R. (1998). Large area hy-
drologic modeling and assesment Part I: Model development. JAWRA Journal of
the American Water Resources Association, 34(1), 73e89.
Arnoldus, H. M. J. (1980). An approximation of the rainfall factor in the universal soil
loss equation. In M. De Boodts, & D. Gabriels (Eds.), Assessments of erosion (pp.
127e132). John Wiley and Sons Ltd.
Aronica, G., & Ferro, V. (1997). Rainfall erosivity over the Calabrian region. Hydro-
logical Sciences Journal-Journal Des Sciences Hydrologiques, 42(1), 35e48.
Auerswald, K., Fiener, P., Gomez, J. A., Govers, G., Quinton, J. N., & Strauss, P. (2015).
Comment on "Rainfall erosivity in Europe" by Panagos et al. (Sci. Total Environ.,
511, 801-814, 2015). Science of the Total Environment, 532, 849e852.
Auerswald, K., Fiener, P., Martin, W., & Elhaus, D. (2014). Use and misuse of the K
factor equation in soil erosion modeling: An alternative equation for deter-
mining USLE nomograph soil erodibility values. Catena, 118, 220e225.
Auerswald, K., Kainz, M., & Fiener, P. (2003). Soil erosion potential of organic versus
conventional farming evaluated by USLE modelling of cropping statistics for
agricultural districts in Bavaria. Soil Use and Management, 19(4), 305e311.
Bagarello, V., Di Piazza, G. V., Ferro, V., & Giordano, G. (2008). Predicting unit plot
soil loss in Sicily, south Italy. Hydrological Processes, 22(5), 586e595.
Bagarello, V., Di Stefano, V., Ferro, V., Giordano, G., Iovino, M., & Pampalone, V.
(2012). Estimating the USLE soil erodibility factor in Sicily, South Italy. Applied
Engineering in Agriculture, 28(2), 199e206.
Bagarello, V., Di Stefano, C., Ferro, V., & Pampalone, V. (2017). Predicting maximum
annual values of event soil loss by USLE-type models. Catena, 155, 10e19.
Bagarello, V., & Ferro, V. (2004). Plot-scale measurement of soil erosion at the
experimental area of Sparacia (southern Italy). Hydrological Processes, 18(1),
141e157.
Bagarello, V., Ferro, V., & Pampalone, V. (2015). A new version of the USLE-MM for
predicting bare plot soil loss at the Sparacia (South Italy) experimental site.
Hydrological Processes, 29(19), 4210e4219.
Bakker, M. M., Govers, G., van Doorn, A., Quetier, F., Chouvardas, D., & Rounsevell, M.
(2008). The response of soil erosion and sediment export to land-use change in
four areas of Europe: The importance of landscape pattern. Geomorphology, 98,
213e226.
Barenblatt, G. I. (1979). Similarity, self-similarity and intermediate asymptotics. New
York: Consultants Bureau.
Barenblatt, G. I. (1987). Dimensional analysis (Amsterdam).
Belyaev, V. R., Wallbrink, P. J., Golosov, V. N., Murray, A. S., & Sidorchuk, A. Y. (2005).
A comparison of methods for evaluating soil redistribution in the severely
eroded Stavropol region, southern European Russia. Geomorphology, 65(3e4),
173e193.
Ben-Dor, E., & Banin, A. (1995). Near-infrared analysis as a rapid method to
simultaneously evaluate several soil properties. Soil Science Society of America
Journal, 59(2), 364e372.
Bennett, H. H. (1939). A permanent loss to new england soil erosion resulting from
the hurricane. Geographical Review, 29(2), 196e204.
Boardman, J. (2006). Soil erosion science: Reflections on the limitations of current
approaches. Catena, 68(2e3), 73e86.
Bonilla, C. A., & Vidal, K. L. (2011). Rainfall erosivity in Central Chile. Journal of
Hydrology, 410(1e2), 126e133.
Bonta, J. V., & Sutton, P. (1983). Erosion and reclamation plots: Research on the
hydroogy and water quality of watersheds subjected to surface mining. Bureau of
Mines, U.S: Dept. Interior.
Borrelli, P., Diodato, N., & Panagos, P. (2016a). Rainfall erosivity in Italy: A national
scale spatio-temporal assessment. International Journal of Digital Earth, 9(9).
Borrelli, P., Meusburger, K., Ballabio, C., Panagos, P., & Alewell, C. (2018). Object-
oriented soil erosion modelling: A possible paradigm shift from potential to
actual risk assessments in agricultural environments. Land Degradation &
Development, 29(4), 1270e1281.
Borrelli, P., Panagos, P., Langhammer, J., Apostol, B., & Schütt, B. (2016b). Assessment
of the cover changes and the soil loss potential in European forestland: First
approach to derive indicators to capture the ecological impacts on soil-related
forest ecosystems. Ecological Indicators, 60.
Borrelli, P., Panagos, P., Ma €rker, M., Modugno, S., & Schütt, B. (2017). Assessment of
the impacts of clear-cutting on soil loss by water erosion in Italian forests: First
comprehensive monitoring and modelling approach. Catena, 149, 770e781.
Borrelli, P., Paustian, K., Panagos, P., Jones, A., Schütt, B., & Lugato, E. (2016c). Effect
of good agricultural and environmental conditions on erosion and soil organic
carbon balance: A national case study. Land Use Policy, 50, 408e421.
Borrelli, P., Robinson, D. A., Fleischer, L. R., Lugato, E., Ballabio, C., Alewell, C.,
Meusburger, K., Modugno, S., Schütt, B., Ferro, V., Bagarello, V., Oost, K. V.,
Montanarella, L., & Panagos, P. (2017b). An assessment of the global impact of
21st century land use change on soil erosion. Nature Communications, 8(1),
2013.
Borst, H. L., & McCall, A. G. (1945). Investigations in erosion control and the recla-
mation of eroded land at the Northwest Appalachian conservation experiment
station. OH: Zanesville.
Brazier, R. E., Beven, K. J., Freer, J., & Rowan, J. S. (2000). Equifinality and uncertainty
in physically based soil erosion models: Application of the GLUE methodology
to WEPPethe water erosion prediction projectefor sites in the UK and USA.
Earth Surface Processes and Landforms, 25(8), 825e845.
Brown, L. C., & Foster, G. R. (1987). Storm erosivity using idealized intensity dis-
tributions. Transactions of the Asae, 30(2), 379e386.
Browning, G. M., Parish, C. L., & Glass, J. A. (1947). A method for determining the use
and limitation of rotation and conservation practices in control of soil erosion in
221
Iowa. American Society of Agronomy Journal, 39, 65e73.
Cao, L., Zhang, K., Dai, H., & Liang, Y. (2015). Modeling interrill erosion on unpaved
roads in the loess plateau of China. Land Degradation and Development, 26(8),
825e832.
Cerdan, O., Govers, G., Le Bissonnais, Y., Van Oost, K., Poesen, J., Saby, N., et al. (2010).
Rates and spatial variations of soil erosion in Europe: A study based on erosion
plot data. Studies on Large Volume Landslides, 122(1e2), 167e177.
Cronshey, R. G., & Theurer, F. D. (1998). AnnAGNPS-non-point pollutant loading model
(Las Vegas).
De Jong, S. M. (1994). Derivation of vegetative variables from a landsat tm image for
modelling soil erosion. Earth Surface Processes and Landforms, 19, 165e178.
De Roo, A. P. J., Wesseling, C. G., & Ritsema, C. J. (1996). Lisem: A single-event
physically based hydrological and soil erosion model for drainage basins. I:
Theory, input and output. Hydrological Processes, 10(8), 1107e1117.
De Vente, J., & Poesen, J. (2005). Predicting soil erosion and sediment yield at the
basin scale: Scale issues and semi-quantitative models. Earth Science Reviews,
71(1e2), 95.
Desmet, P., & Govers, G. (1996). A GIs procedure for automatically calculating the
USLE LS factor on topographically complex landscape units. Journal of Soil and
Water Conservation, 51(5), 427e433.
Di Stefano, C., Ferro, V., & Pampalone, V. (2017). Testing the USLE-M family of
models at the Sparacia experimental site in South Italy. Journal of Hydrologic
Engineering, 22(8), 05017012-05017012.
Diodato, N., Borrelli, P., Fiener, P., Bellocchi, G., & Romano, N. (2017). Discovering
historical rainfall erosivity with a parsimonious approach: A case study in
Western Germany. Journal of Hydrology, 544.
Diodato, N., Knight, J., & Bellocchi, G. (2013). Reduced complexity model for
assessing patterns of rainfall erosivity in Africa. Global and Planetary Change,
100, 183e193.
Dissmeyer, G. E., & Foster, G. R. (1980). A guide for predicting sheet and rill erosion on
forest land.
Doetterl, S., Van Oost, K., & Six, J. (2012). Towards constraining the magnitude of
global agricultural sediment and soil organic carbon fluxes. Earth Surface Pro-
cesses and Landforms, 37(6), 642e655.
Dokuchaev, V. V. (1892). Nashi stepi prezde i teper(Our steppes before and recently).
Russia(in Russian). Sankt_Peterburg: Published House Evdokimov.
Dunn, M., & Hickey, R. (1998). The effect of slope algorithms on slope estimates
within a GIS. Cartography, 27(1), 9e15.
Evans, R. (2013). Assessment and monitoring of accelerated water erosion of
cultivated land - when will reality be acknowledged? Soil Use and Management,
29(1), 105e118.
Evans, R., & Boardman, J. (2016a). The new assessment of soil loss by water erosion
in Europe. Panagos P. et al., 2015 Environmental Science & Policy 54, 438-447-A
response. Environmental Science & Policy, 58, 11e15.
Evans, R., & Boardman, J. (2016b). A reply to panagos et al., 2016 (Environmental
science & policy 59 (2016) 53-57. Environmental Science & Policy, 60, 63e68.
Fan, J. C. (1987). Measurements of erosion on highway slopes and of the universal soil
loss erosion equation. West Lafayette: Purdue Univ. (Ind).
FAO. (2015). GLADIS - global land degradation information system.
FAO. (2017). Voluntary Guidelines for sustainable soil management food and agricul-
ture organization of the united nations. http://www.fao.org/3/a-bl813e.pdfhttp://
www.fao.org/3/a-bl813e.pdf (Rome, Italy).
Favis-Mortlock, D. T. (1997). Validation of field scale soil erosion models using
common data sets. In J. Boardman, & D. T. F.-M (Eds.), Modeling soil erosion by
WaterNATO ASI series (pp. 89e127). Berlin: Springer-Verlag.
Ferro, V. (1997). Further remarks on a distributed approach to sediment delivery.
Hydrological Sciences Journal, 42(5), 633e647.
Ferro, V. (2010). Deducing the USLE mathematical structure by dimensional analysis
and self-similarity theory. Biosystems Engineering, 106(2), 216e220.
Ferro, V., & Minacapilli, M. (1995). Sediment delivery processes at basin-scale.
Hydrological Sciences Journal-Journal Des Sciences Hydrologiques, 40(6), 703e717.
Ferro, V., & Porto, P. (1999). A comparative study of rainfall erosivity estimation for
southern Italy and southeastern Australia. Hydrological Sciences Journal-Journal
Des Sciences Hydrologiques, 44(1), 3e24.
Ferro, V., & Porto, P. (2000). Sediment delivery distributed (SEDD) model. Journal of
Hydrologic Engineering, 5(4), 411e422.
Fiener, P., & Auerswald, K. (2016). Comment on "The new assessment of soil loss by
water erosion in Europe" by Panagos et al. (Environmental Science & Policy 54
(2015) 438-447). Environmental Science & Policy, 57, 140e142.
Fishman, G. S., & Kiviat, P. J. (1968). The statistics of discrete event simulation.
Simulation, 10, 185e191.
Foley, J. (2017). Living by the lessons of the planet. Science, 356(6335), 251e252.
Foster, G. R. (1982). Relation of Usle Factors to Erosion on Rangeland. Science and
Education Administration Publications. ARM(NW-26).
Foster, G. R., & Wischmeier, W. H. (1974). Evaluating irregular slopes for soil loss
prediction. Transactions of the Asae, 17(2), 305e309.
Foster, G. R., Yoder, D. C., Weesies, G. A., McCool, D. K., McGregor, K. C., &
Bingner, R. L. (2008). Draft User's Guide, Revised Universal Soil Loss Equation
Version 2 (RUSLE-2). Washington, DC: USDA-Agricultural Research Service.
Foster, G. R., Yoder, D. C., Weesies, G. A., & Toy, T. J. (2001). The design philosophy
behind RUSLE2: Evolution of an empirical model. In Soil erosion research for the
21st century, proceedings (pp. 95e98).
Gabriels, D., Ghekiere, G., Schiettecatte, W., & Rottiers, I. (2003). Assessment of USLE
cover-management C-factors for 40 crop rotation systems on arable farms in
the Kemmelbeek watershed, Belgium. Soil and Tillage Research, 74, 47e53.
222 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
Garcia-Ruiz, J. M., Begueria, S., Nadal-Romero, E., Gonzalez-Hidalgo, J. C., Lana-
Renault, N., & Sanjuan, Y. (2015). A meta-analysis of soil erosion rates across the
world. Geomorphology, 239, 160e173.
Gessesse, B., Bewket, W., & Bra €uning, A. (2015). Model based characterization and
monitoring of runoff and soil erosion in response to land use/land cover
changes in the. Land Degradation & Development, 26, 711e724. February 2014.
van de Giesen, N., Stomph, T. J., & de Ridder, N. (2005). Surface runoff scale effects in
west African watersheds: Modeling and management options. Agricultural
Water Management, 72(2), 109e130.
Griffin, M. L., Beasley, D. B., Fletcher, J. J., & Foster, G. R. (1988). Estimating soil loss
on topographically non-uniform field and farm units. Journal of Soil and Water
Conservation, 43(4), 326e331.
Haciyakupoglu, S., Hizal, A., Gokbulak, F., & Kiziltas, S. (2010). A comparison of soil
losses estimated using fallout radionuclides and the USLE for the Omerli
watershed, Turkey. In M. Zlatic (Ed.), Global change: Challenges for soil man-
agement. Advances in geoecology (pp. 175e183).
Haileslassie, A., Priess, J., Veldkamp, E., Teketay, D., & Lesschen, J. P. (2005).
Assessment of soil nutrient depletion and its spatial variability on smallholders'
mixed farming systems in Ethiopia using partial versus full nutrient balances.
Agriculture Ecosystems & Environment, 108(1), 1e16.
Hart, G. E. (1984). Erosion from simulated rainfall on mountain rangeland in Utah.
Journal of Soil and Water Conservation, 39, 330e334.
Hengl, T., de Jesus, J. M., MacMillan, R. A., Batjes, N. H., Heuvelink, G. B. M.,
Ribeiro, E., et al. (2014). SoilGrids1km-Global soil information based on auto-
mated mapping. PLoS One, 9(8), 17.
Hickey, R. (2000). Slope angle and slope length solutions for GIS. Cartography, 29(1),
1e8.
Hussein, M. H., Kariem, T. H., & Othman, A. K. (2007). Predicting soil erodibility in
northern Iraq using natural runoff plot data. Soil & Tillage Research, 94(1),
220e228.
Ito, A. (2007). Simulated impacts of climate and land-cover change on soil erosion
and implication for the carbon cycle, 1901 to 2100. Geophysical Research Letters,
34(9).
Jetten, V., Govers, G., & Hessel, R. (2003). Erosion models: Quality of spatial pre-
dictions. Hydrological Processes, 17, 887e900.
Joel, A., Messing, I., Seguel, O., & Casanova, M. (2002). Measurement of surface water
runoff from plots of two different sizes. Hydrological Processes, 16(7),
1467e1478.
Johnson, C. W., Savabi, M. R., & Loomis, S. A. (1984). Rangeland erosion measure-
ments for the USLE. Transactions of the ASAE, 27, 1313e1320.
Kara, O., Şensoy, H., & Bolat, I. (2010). Slope length effects on microbial biomass and
activity of eroded sediments. Journal of Soils and Sediments, 10(3), 434e439.
Karydas, C. G., & Panagos, P. (2016). Modelling monthly soil losses and sediment
yields in Cyprus. International Journal of Digital Earth, 9(8), 766e787.
Karydas, C. G., Panagos, P., & Gitas, I. Z. (2014). A classification of water erosion
models according to their geospatial characteristics. International Journal of
Digital Earth, 7(3), 229e250.
Keesstra, S. D., Bouma, J., Wallinga, J., Tittonell, P., Smith, P., Cerda , A., et al. (2016).
The significance of soils and soil science towards realization of the United
Nations Sustainable Development Goals. SOIL, 2(2), 111e128.
Kilinc, M., & Richardson, E. V. (1973). Mechanics of soil erosion from overland flow
generated by simulated rainfall. Ft. Collins: Colorado State Univ.
King, C., & Delpont, G. (1993). Spatial assessment of erosion: Contribution of remote
sensing, a review. Remote Sensing Reviews, 7(3e4), 223e232.
Kinnell, P. I. A. (1981). Rainfall intensity kinetic-energy relationships for soil loss
prediction. Soil Science Society of America Journal, 45(1), 153e155.
Kinnell, P. I. A. (2008). Sediment delivery from hillslopes and the universal soil loss
equation: Some perceptions and misconceptions. Hydrological Processes, 22(16),
3168e3175.
Kinnell, P. I. A. (2010). Event soil loss, runoff and the universal soil loss equation
family of models: A review. Journal of Hydrology, 385(1e4), 384e397.
Kinnell, P. I. A. (2014). Applying the QREI30 index within the USLE modelling
environment. Hydrological Processes, 28(3), 591e598.
Kinnell, P. I. A., & Risse, L. M. (1998). USLE-M: Empirical modeling rainfall erosion
through runoff and sediment concentration. Soil Science Society of America
Journal, 62(6), 1667e1672.
Kirkby, M. J., & Cox, N. J. (1995). A climatic index for soil-erosion potential (csep)
including seasonal and vegetation factors. Catena, 25(1-4), 333e352.
Kirkby, M. J., Irvine, B. J., Jones, R. J. A., Govers, G., Boer, M., Cerdan, O., et al. (2008).
The PESERA coarse scale erosion model for Europe. I. - model rationale and
implementation. European Journal of Soil Science, 59(6), 1293e1306.
Knisel, W. G. (1980). Creams: A field scale model for chemicals, runoff and erosion from
agricultural management systems; conservation research report No. 26 (pp.
1e640). US Departement of Agriculture.
Laflen, J. M., Elliot, W. J., Flanagan, D. C., Meyer, C. R., & Nearing, M.a. (1997). WEPP-
predicting water erosion using a process-based model. Journal of Soil and Water
Conservation, 52, 96.
Laflen, J. M., & Flanagan, D. C. (2013). The development of U. S. soil erosion pre-
diction and modeling. International Soil and Water Conservation Research, 1(2),
1e11.
Laflen, J. M., Foster, G. R., & Onstad, C. A. (1985). Simulation of individual- storm soil
loss for modeling the impact of soil erosion on crop productivity. Soil Erosion
and Conservation, 285e295.
Laflen, J. M., & Moldenhauer, W. C. (2003). Pioneering soil erosion prediction e the
USLE story. Bejing, China: World Association of Soil and Water Conservation.
Larson, W. E., Lindstrom, M. J., & Schumacher, T. E. (1997). The role of severe storms
in soil erosion: A problem needing consideration. Journal of Soil and Water
Conservation, 52, 90e95.
Lee, J. H., & Heo, J. H. (2011). Evaluation of estimation methods for rainfall erosivity
based on annual precipitation in Korea. Journal of Hydrology, 409(1e2), 30e48.
Leitinger, G., Tasser, E., Newesely, C., Obojes, N., & Tappeiner, U. (2010). Seasonal
dynamics of surface runoff in mountain grassland ecosystems differing in land
use. Journal of Hydrology, 385(1e4), 95e104.
Leonard, R. A., Knisel, W. G., & Still, D. A. (1987). GLEAMS: Groundwater loading
effects of agricultural management systems. Transactions of the ASAE, 30(5),
1403e1418.
Li, L., Du, S., Wu, L., & Liu, G. (2009). An overview of soil loss tolerance. Catena, 78(2),
93e99.
Littleboy, M., Freebairn, D. M., Hammer, G. L., & Silburn, D. M. (1992a). Impact of
soil-erosion on production in cropping systems .2. simulation of production and
erosion risks for a wheat cropping system. Australian Journal of Soil Research,
30(5), 775e788.
Littleboy, M., Silburn, D. M., Freebairn, D. M., Woodruff, D. R., Hammer, G. L., &
Leslie, J. K. (1992b). Impact of soil-erosion on production in cropping systems .1.
development and validation of a simulation-model. Australian Journal of Soil
Research, 30(5), 757e774.
Liu, B. Y., Nearing, M. A., & Risse, L. M. (1994). Slope gradient effects on soil loss for
steep slopes. Transactions of the Asae, 37(6), 1835e1840.
Liu, B. Y., Zhang, K. L., & Xie, Y. (2002). An empirical soil loss equation. In Proc. of 12th
ISCO. Beijing: Tsinghua Press.
Lloyd, C. H., & Eley, G. W. (1952). Graphical solution of probable soil loss formula for
northeastern region. Journal of Soil and Water Conservation, 7, 189e191.
Loureiro, N. D., & Coutinho, M. D. (2001). A new procedure to estimate the RUSLE
EI30 index, based on monthly rainfall data and applied to the Algarve region,
Portugal. Journal of Hydrology, 250(1e4), 12e18.
Lufafa, A., Tenywa, M. M., Isabirye, M., Majaliwa, M. J. G., & Woomer, P. L. (2003).
Prediction of soil erosion in a Lake Victoria basin catchment using a GIS-based
Universal Soil Loss model. Agricultural Systems, 76(3), 883e894.
Luleva, M. I., van der Werff, H., Jetten, V., & van der Meer, F. (2011). Can infrared
spectroscopy Be used to measure change in potassium nitrate concentration as
a proxy for soil particle movement? Sensors (Basel, Switzerland), 11(4),
4188e4206.
Ma €rker, M., Angeli, L., Bottai, L., Costantini, R., Ferrari, R., Innocenti, L., et al. (2008).
Assessment of land degradation susceptibility by scenario analysis: A case study
in southern Tuscany, Italy. Geomorphology, 93(1e2), 120e129.
Martin, P. H. (1996). Physics of stamp-collecting? Thoughts on ecosystem model
design. Science of the Total Environment, 183(1e2), 7e15.
McCool, D. K., Brown, L. C., Foster, G. R., Mutchler, C. K., & Meyer, L. D. (1987).
Revised slope steepness factor for the universal soil loss equation. Transactions
of the Asae, 30(5), 1387e1396.
McCool, D., Foster, G., & Weesies, G. (1997). Slope length and steepness factors (LS).
In K. Renard, G. Foster, G. Weesies, D. McCool, & D. Yoder (Eds.), Predicting soil
erosion by water: A guide to conservation planning with the revised universal soil
loss equation (RUSLE) (pp. 101e141). USA: Department of Agriculture, Agricul-
tural Research Service, Washington, District of Columbia.
Merritt, W. S., Letcher, R. A., & Jakeman, A. J. (2003). A review of erosion and
sediment transport models. Environmental Modelling & Software, 18(8e9),
761e799.
Meusburger, K., Banninger, D., & Alewell, C. (2010a). Estimating vegetation
parameter for soil erosion assessment in an alpine catchment by means of
QuickBird imagery. International Journal of Applied Earth Observation and Geo-
information, 12(3), 201e207.
Meusburger, K., Konz, N., Schaub, M., & Alewell, C. (2010b). Soil erosion modelled
with USLE and PESERA using QuickBird derived vegetation parameters in an
alpine catchment. International Journal of Applied Earth Observation and Geo-
information, 12(3), 208e215.
Meusburger, K., Mabit, L., Park, J. H., Sandor, T., & Alewell, C. (2013). Combined use of
stable isotopes and fallout radionuclides as soil erosion indicators in a forested
mountain site, South Korea. Biogeosciences, 10(8), 5627e5638.
Meusburger, K., Steel, A., Panagos, P., Montanarella, L., & Alewell, C. (2012). Spatial
and temporal variability of rainfall erosivity factor for Switzerland. Hydrology
and Earth System Sciences, 16.
Mitasova, H., Hofierka, J., Zlocha, M., & Iverson, L. R. (1996). Modelling topographic
potential for erosion and deposition using GIS. International Journal of
Geographical Information Systems, 10(5), 629e641.
Mitasova, H., Hofierka, J., Zlocha, M., & Iverson, L. (1997). Reply to comment by
Desmet and Govers. International Journal of Geographical Information Science,
11(6), 611e618.
Mitasova, H., & Mitas, L. (1999). Modeling soil detachmentwith RUSLE 3D using GIS.
University of Illinois at Urbana-Champaign. date accessed February 2018 http://
skagit.meas.ncsu.edu/~helena/gmslab/erosion/usle.html.
Mo €ller, M., Gerstmann, H., & Gla €ßer, C. (2014). Dyna C : A framework for the dynamic
derivation of the usle C -factor using high temporal multi-spectral satellite imagery.
Mondal, A., Khare, D., Kundu, S., Mukherjee, S., Mukhopadhyay, A., & Mondal, S.
(2017). Uncertainty of soil erosion modelling using open source high resolution
and aggregated DEMs. Geoscience Frontiers, 8(3), 425e436.
Montanarella, L. (2015). Govern our soils. Nature, 528(7580), 32e33.
Moore, I. D., & Burch, G. J. (1986). Physical basis of the length-slope factor in the
universal soil loss equation. Soil Science Society of America Journal, 50(5),
1294e1298.
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
Morgan, R. P. C. (2005). Soil loss conservation: Problems and prospects. Chichester:
John Wiley & Sons.
Morgan, R. P. C., & Nearing, M. A. (2011). Handbook of Erosion Modelling (pp. 1e401).
John Wiley and Sons.
Morgan, R. P. C., Quinton, J. N., Smith, R. E., Govers, G., Poesen, J. W. A., Auerswald, K.,
et al. (1998). The European soil erosion model (EUROSEM): A process-based
approach for predicting soil loss from fields and small catchments. , 527-544.
Earth Surface Processes and Landforms, 23(6), 527e544.
Morton, A., & Súarez, M. (2001). Kinds of models. In M. G. Anderson, & P. D. Bates
(Eds.), Model validation: Perspectives in hydrological science (pp. 11e21). Chi-
chester: John Wiley & Sons.
Musgrave, G. W. (1947). The quantitative evaluation of factors in water erosion, a
first approximation. Journal of Soil and Water Conservation, 2, 133e138.
Nash, J. E., & Sutcliffe, J. E. (1970). River flow forecasting through conceptual models.
Part 1 - a discussion of principles. Journal of Hydrology, 10, 282e290.
Nearing, M. A. (1997). A single, continuous function for slope steepness influence on
soil loss. Soil Science Society of America Journal, 61(3), 917e919.
Nearing, M. (2004). Soil erosion and conservation. In J. Wainwright, & M. Mulligan
(Eds.), Environmental modelling: Finding simplicity in complexity (pp. 277e290).
John Wiley & Sons, Ltd.
Nearing, M. A. (2013). Soil erosion and conservation.
Nearing, M. A., Govers, G., & Norton, L. D. (1999). Variability in soil erosion data from
replicated plots. Soil Science Society of America Journal, 63(6), 1829.
Nearing, M. A., Yin, S. Q., Borrelli, P., & Polyakov, V. O. (2017). Rainfall erosivity: An
historical review. Catena, 157.
Nicks, A. D., Lane, L. J., & Gander, G. A. (1995). Weather generator (chapter 2). In
D. C. Flanagan, & M. A. Nearing (Eds.), USDA water erosion prediction project:
Hillslope profile and watershed model documentation- NSERL report. USDA-ARS
National Soil Erosion Research Laboratory.
Nyssen, J., Vandenreyken, H., Poesen, J., Moeyersons, J., Deckers, J., Haile, M., et al.
(2005). Rainfall erosivity and variability in the northern Ethiopian highlands.
Journal of Hydrology, 311(1e4), 172e187.
O'Callaghan, J. F., & Mark, D. M. (1984). The extraction of drainage networks from
digital elevation data. Computer Vision, Graphics, and Image Processing, 28(3),
323e344.
Oldeman, L. R. (1992). Global extent of soil degradation. ISRIC Bi-annual report 1991-
1992.
Oliveira, P. T. S., Wendland, E., & Nearing, M. A. (2013). Rainfall erosivity in Brazil: A
review. Catena, 100, 139e147.
Oreskes, N., Shraderfrechette, K., & Belitz, K. (1994). Verification, validation, and
confirmation of numerical-models in the earth-sciences. Science, 263(5147),
641e646.
ndez-Ugalde, O. (2018). LUCAS
Orgiazzi, A., Ballabio, C., Panagos, P., Jones, A., & Ferna
soil, the largest expandable soil dataset for Europe: A review. European Journal
of Soil Science, 69(1), 140e153.
Ozsoy, G., Aksoy, E., Dirim, M. S., & Tumsavas, Z. (2012). Determination of soil
erosion risk in the mustafakemalpasa river basin, Turkey, using the revised
universal soil loss equation, geographic information system, and remote
sensing. Environmental Management, 50(4), 679e694.
Panagos, P., Ballabio, C., Borrelli, P., Meusburger, K., Klik, A., Rousseva, S., et al.
(2015a). Rainfall erosivity in Europe. Science of the Total Environment, 511,
801e814.
Panagos, P., Borrelli, P., Meusburger, K., Alewell, C., Lugato, E., & Montanarella, L.
(2015b). Estimating the soil erosion cover-management factor at the European
scale. Land Use Policy, 48, 38e50.
Panagos, P., Borrelli, P., Meusburger, K., van der Zanden, E. H., Poesen, J., & Alewell, C.
(2015c). Modelling the effect of support practices (P-factor) on the reduction of
soil erosion by water at European scale. Environmental Science and Policy, 51.
Panagos, P., Borrelli, P., Meusburger, K., Yu, B., Klik, A., Jae Lim, K., et al. (2017a).
Global rainfall erosivity assessment based on high-temporal resolution rainfall
records. Scientific Reports, 7(1), 4175.
Panagos, P., Borrelli, P., Meusburger, K., Yu, B., Klik, A., Lim, k.j., et al. (2017b). Global
rainfall erosivity assessment based on high-temporal resolution rainfall records.
Scientific Reports. In review(February): 1-12.
Panagos, P., Borrelli, P., Poesen, J., Ballabio, C., Lugato, E., Meusburger, K., et al.
(2015d). The new assessment of soil loss by water erosion in Europe. Environ-
mental Science & Policy, 54, 438e447.
Panagos, P., Borrelli, P., Poesen, J., Meusburger, K., Ballabio, C., Lugato, E., et al.
(2016b). Reply to the comment on "The new assessment of soil loss by water
erosion in Europe" by Fiener & Auerswald. Environmental Science & Policy, 57,
143e150.
Panagos, P., Borrelli, P., Poesen, J., Meusburger, K., Ballabio, C., Lugato, E.,
Montanarella, L., & Alewell, C. (2016a). Reply to "The new assessment of soil loss
by water erosion in Europe. In P. Panagos, et al. (Eds.), 2015 Environ. Sci. Policy 54
(Vol. 59, pp. 438e447). Environmental Science & Policy. A response" by Evans
and Boardman Environ. Sci. Policy 58, 11-15. 53-57.
Panagos, P., Christos, K., Cristiano, B., & Ioannis, G. (2014a). Seasonal monitoring of
soil erosion at regional scale: An application of the G2 model in Crete focusing
on agricultural land uses. International Journal of Applied Earth Observation and
Geoinformation, 27, 147e155.
Panagos, P., Imeson, A., Meusburger, K., Borrelli, P., Poesen, J., & Alewell, C. (2016c).
Soil conservation in Europe: Wish or reality? Land Degradation & Development,
27(6), 1547e1551.
Panagos, P., Meusburger, K., Ballabio, C., Borrelli, P., & Alewell, C. (2014b). Soil
erodibility in Europe: A high-resolution dataset based on LUCAS. Science of the
223
Total Environment, 479e480(1), 189e200.
Panagos, P., Meusburger, K., Ballabio, C., Borrelli, P., Begueria, S., Klik, A., et al.
(2015e). Reply to the comment on "Rainfall erosivity in Europe" by Auerswald
et al. Science of the Total Environment, 532, 853e857.
Panagos, P., Meusburger, K., Van Liedekerke, M., Alewell, C., Hiederer, R., &
Montanarella, L. (2014c). Assessing soil erosion in Europe based on data
collected through a European network. Soil Science and Plant Nutrition, 60(1),
15e29.
Peralta, C., Beriain, E., Iglesias, C., Marangunic, C., Marangunic, A., & Casassa, G.
(2017). Ground topography validation of a high-resolution DEM and ortho-
photos obtained with a commercial UAS. In 2017 first IEEE international sym-
posium of geoscience and remote sensing (GRSS-Chile) (Vol. 4, p. 4).
Pineiro, G., Paruelo, J. M., Oesterheld, M., & Jobbagy, E. G. (2010). Pathways of
grazing effects on soil organic carbon and nitrogen. Rangeland Ecology &
Management, 63(1), 109e119.
Podmanicky, L., Balazs, K., Belenyesi, M., Centeri, C., Kristof, D., & Kohlheb, N. (2011).
Modelling soil quality changes in Europe. An impact assessment of land use
change on soil quality in Europe. Ecological Indicators, 11(1), 4e15.
Poesen, J. W., Torri, D., & Bunte, K. (1994). Effects of rock fragments on soil-erosion
by water at different spatial scales - a review. Catena, 23(1e2), 141e166.
Porto, P. (2016). Exploring the effect of different time resolutions to calculate the
rainfall erosivity factor R in Calabria, southern Italy. Hydrological Processes,
30(10), 1551e1562.
Quinn, P., Beven, K., Chevallier, P., & Planchon, O. (1991). The prediction of hillslope
flow paths for distributed hydrological modelling using digital terrain models.
Hydrological Processes, 5(1), 59e79.
Quinton, J. (2013). Erosion and sediment transport. In J. Wainwright, & M. Mulligan
(Eds.), Environmental modelling: Finding simplicity in complexity (pp. 187e196).
John Wiley & Sons, Ltd.
Rapp, J. F. (1994). Error assessment of the revised universal soil loss equation using
natural runoff plot data, Master Thesis. Arizona: The University of Arizona.
Refsgaard, J. C., & Storm, B. (1996). Construction, calibration and validation of hy-
drological models. In M. B. Abbott, & J. C. Refsgaard (Eds.), Distributed hydro-
logical modelling (pp. 41e45). Dordrecht: Reidel.
Rejman, J., & Brodowski, R. (2005). Rill characteristics and sediment transport as a
function of slope length during a storm event on loess soil. Earth Surface Pro-
cesses and Landforms, 30(2), 231e239.
Renard, K. G., & Ferreira, V. A. (1993). Rusle model description and database
sensitivity. Journal of Environmental Quality, 22(3), 458e466.
Renard, K. G., & Foster, G. R. (1985). Managing rangeland soil resources: The uni-
versal soil loss equation. Rangelands, 7(3), 118e122.
Renard, K., Foster, G., Weesies, G., McCool, D., & Yoder, D. (1997). Predicting soil
erosion by water: A guide to conservation planning with the revised universal soil
loss equation (RUSLE). Agriculture handbook 703. Washington, District of
Columbia, USA: Department of Agriculture, Agricultural Research Service.
Renard, K. G., Foster, G. R., Weesies, G. A., & Porter, J. P. (1991). Rusle - revised
universal soil loss equation. Journal of Soil and Water Conservation, 46(1), 30e33.
Risal, A., Bhattarai, R., Kum, D., Park, Y. S., Yang, J. E., & Lim, K. J. (2016). Application
of Web ERosivity module (WERM) for estimation of annual and monthly R
factor in Korea. Catena, 147, 225e237.
Risse, L. M., Nearing, M. A., Laflen, J. M., & Nicks, A. D. (1993). Error assessment in
the universal soil loss equation. Soil Science Society of America Journal, 57(3),
825.
Rodriguez, J. L. G., & Suarez, M. C. G. (2012). Methodology for estimating the
topographic factor LS of RUSLE3D and USPED using GIS. Geomorphology, 175,
98e106.
€ mkens, M., Poesen, J., & Wang, J. (1988). Relationship between the USLE soil
Ro
erodibility factor and soil properties. Proceedings, 371e385.
Rosewell, C. J. (1993). SOILOSS: A program to assist in the selection of management
practices to reduce erosion. Technical handbook (Vol. 11). Soil Conservation Ser-
vice of New South Wales.
Rossel, R. A. V., & Behrens, T. (2010). Using data mining to model and interpret soil
diffuse reflectance spectra. Geoderma, 158(1), 46e54.
Rüttimann, M., Schaub, D., Prasuhn, V., & Rüegg, W. (1995). Measurement of runoff
and soil erosion on regularly cultivated fields in Switzerland d some critical
considerations. Catena, 25(1), 127e139.
Sadeghi, S. H., Zabihi, M., Vafakhah, M., & Hazbavi, Z. (2017). Spatiotemporal
mapping of rainfall erosivity index for different return periods in Iran. Natural
Hazards, 87(1), 35e56.
Scheurer, K., Alewell, C., Banninger, D., & Burkhardt-Holm, P. (2009). Climate and
land-use changes affecting river sediment and brown trout in alpine countries-
a review. Environmental Science and Pollution Research, 16(2), 232e242.
Schindler Wildhaber, Y., Michel, C., Burkhardt-Holm, P., Baenninger, D., & Alewell, C.
(2012). Measurement of spatial and temporal fine sediment dynamics in a small
river. Hydrology and Earth System Sciences, 16(5), 1501e1515.
Schmidt, S., Alewell, C., & Meusburger, K. (2018). Mapping spatio-temporal dy-
namics of the cover and management factor (C-factor) for grasslands in
Switzerland. Remote Sensing of Environment, 211, 89e104.
Schmidt, S., Alewell, C., & Meusburger, K. (2019). Monthly RUSLE soil erosion risk of
Swiss grasslands. Journal of Maps, 1e10.
Schmidt, S., Alewell, C., Panagos, P., & Meusburger, K. (2016). Regionalization of
monthly rainfall erosivity patterns in Switzerland. Hydrology and Earth System
Sciences, 20(10), 4359e4373.
Scho€nbrodt, S., Saumer, P., Behrens, T., Seeber, C., & Scholten, T. (2010). Assessing the
USLE crop and management factor C for soil erosion modeling in a large
224 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
mountainous watershed in Central China. Journal of Earth Science, 21, 835e845.
Schwertmann, U., Vogl, W., & Kainz, M. (1987). Bodenerosion durch Wasser -
vorhersage des Abtrags und Bewertung von Gegenmaßnahmen. Stuttgart: Verlag
Eugen Ulmer.
Seibert, J., & McGlynn, B. L. (2007). A new triangular multiple flow direction algo-
rithm for computing upslope areas from gridded digital elevation models.
Water Resources Research, 43(4).
Senarath, S. U. S., Ogden, F., Downer, C. W., & Sharif, H. O. (2000). On the calibration
and verification of two-dimensional, distributed, Hortonian, continuous
watershed models. Water Resources Research, 36, 1495e1510.
Shepherd, K. D., & Walsh, M. G. (2002). Development of reflectance spectral libraries
for characterization of soil properties. Soil Science Society of America Journal,
66(3), 988e998.
Silburn, D. M., & Freebairn, D. M. (1992). Evaluations of the creams model .3.
simulation of the hydrology of vertisols. Australian Journal of Soil Research,
30(5), 547e564.
Silburn, D. M., & Loch, R. J. (1989). Evaluation of the creams model .1. sensitivity
analysis of the soil-erosion sedimentation component for aggregated clay soils.
Australian Journal of Soil Research, 27(3), 545e561.
Simanton, J., Roger, H. B., Herbert, B., & Renard, K. G. (1980). Application of the USLE
to southwestern rangelands. Hydrology and water resources in Arizona and the
Southwest.
Sivertun, A., & Prange, L. (2003). Non-point source critical area analysis in the
Gisselo watershed using GIS. Environmental Modelling & Software, 18(10),
887e898.
Smith, D. D. (1941). Interpretation of soil conservation data for field use. Agricultural
Engineering, 22, 173e175.
Smith, D. D., & Whitt, D. M. (1948). Evaluating soil losses from field areas. Agri-
cultural Engineering, 29, 394e396.
Smith, S. J., Williams, J. R., Menzel, R. G., & Coleman, G. A. (1984). Prediction of
sediment yield from southern plains grasslands with the modified universal soil
loss equation. Journal of Range Management, 37(4), 295e297.
Smith, D. D., & Wischmeier, W. H. (1957). Factors affecting sheet and rill erosion.
American Geophysical Union Transaction, 38, 889e896.
Spaeth, K. E., Pierson, F. B., Weltz, M.a., & Blackburn, W. H. (2003). Evaluation of
USLE and RUSLE estimated soil loss on rangeland. Journal of Range Managment,
56, 234e246.
Stenberg, B., & Rossel, R. A. V. (2010). Diffuse reflectance spectroscopy for high-
resolution soil sensing. In R. A. Viscarra Rossel, A. B. McBratney, & B. Minasny
(Eds.), Proximal soil sensing (pp. 29e47). Dordrecht: Springer Netherlands.
Stolpe, N. B. (2005). A comparison of the RUSLE, EPIC and WEPP erosion models as
calibrated to climate and soil of south-central Chile. Acta Agriculturae Scandi-
navica Section B Soil and Plant Science, 55(1), 2e8.
Sverdrup, H., Warfvinge, P., Blake, L., & Goulding, K. (1995). Modeling recent and
historic soil data from the rothamsted-experimental-station, UK using safe.
Agriculture Ecosystems & Environment, 53(2), 161e177.
Tan, Z., Leung, L. R., Li, H.-Y., & Tesfa, T. (2018). Modeling sediment yield in land
surface and earth system models: Model comparison, development, and eval-
uation. Journal of Advances in Modeling Earth Systems, 0(0).
Teng, H. F., Rossel, R. A. V., Shi, Z., Behrens, T., Chappell, A., & Bui, E. (2016).
Assimilating satellite imagery and visible-near infrared spectroscopy to model
and map soil loss by water erosion in Australia. Environmental Modelling &
Software, 77, 156e167.
Tiwari, A. K., Risse, L. M., & Nearing, M. A. (2000). Evaluation of WEPP and its
comparison with USLE and RUSLE. Transactions of the Asae, 43(5), 1129e1135.
Torri, D., Poesen, J., & Borselli, L. (1997). Predictability and uncertainty of the soil
erodibility factor using a global dataset. Catena, 31(1), 1e22.
Toy, T. J., F, G. R., & R, K. G. (2002). Soil erosion: Processes, prediction, measurement,
and control. New York: John Wiley & Sons, Inc.
Tresch, S. (2014). Determination of the S-factor in steep slopes. Master Thesis,. Basel:
University Basel.
Trimble, S. W., & Crosson, P. (2000). U.S. Soil erosion rates–myth and reality. Science,
289(5477), 248e250.
USDA. (1961). US agricultural research service. a universal equation for predicting
rainfall-erosion losses. U.S. Dept. Agr. ARS, 22e66, 1e11.
USDA. (2008). Revised universal soil loss equation version 2 (RUSLE2). Science docu-
mentation. Washington, D.C: USDA-Agricultural Research Service.
USDA. (2018). Revised universal soil loss equation 2 - download RUSLE2. https://www.
ars.usda.gov/southeast-area/oxford-ms/national-sedimentation-laboratory/
watershed-physical-processes-research/research/rusle2/revised-universal-soil-
loss-equation-2-download-rusle2/.
Van Oost, K., Govers, G., & Desmet, P. (2000a). Evaluating the effects of changes in
landscape structure on soil erosion by water and tillage. Landscape Ecology,
15(6), 577e589.
Van Oost, K., Govers, G., Van Muysen, W., & Quine, T. A. (2000b). Modeling trans-
location and dispersion of soil constituents by tillage on sloping land. Soil Sci-
ence Society of America Journal, 64(5), 1733e1739.
Van Oost, K., Quine, T. A., Govers, G., De Gryze, S., Six, J., Harden, J. W., et al. (2007).
The impact of agricultural soil erosion on the global carbon cycle. Science, 318,
626e629.
Van Remortel, R. D., Hamilton, M. E., & Hickey, R. J. (2001). Estimating the LS factor
for RUSLE through iterative slope length processing of digital elevation data
within arclnfo grid. Cartography, 30(1), 27e35.
Van Remortel, R. D., Maichle, R. W., & Hickey, R. J. (2004). Computing the LS factor
for the Revised Universal Soil Loss Equation through array-based slope
processing of digital elevation data using a Cþþ executable. Computers &
Geosciences, 30(9e10), 1043e1053.
Van Rompaey, A. J. J., Verstraeten, G., Van Oost, K., Govers, G., & Poesen, J. (2001).
Modelling mean annual sediment yield using a distributed approach. Earth
Surface Processes and Landforms, 26(11), 1221e1236.
Van der Knijff, J., Jones, R. R. J., & Montanarella, L. (1999). Soil erosion risk assessment
in Italy. Luxemburg: Off. Publ. Eur. Communities EUR 19022.
Vannoppen, W., Vanmaercke, M., De Baets, S., & Poesen, J. (2015). A review of the
mechanical effects of plant roots on concentrated flow erosion rates. Earth-
Science Reviews, 150, 666e678.
Veihe, A., & Quinton, J. (2000). Sensitivity analysis of EUROSEM using Monte Carlo
simulation I: Hydrological, soil and vegetation parameters. Hydrological Pro-
cesses, 14(5), 915e926.
Veihe, A., Quinton, J., & Poesen, J. (2000). Sensitivity analysis of EUROSEM using
Monte Carlo simulation II: The effect of rills and rock fragments. Hydrological
Processes, 14(5), 927e939.
Verheijen, F. G. A., Jones, R. J. A., Rickson, R. J., & Smith, C. J. (2009). Tolerable versus
actual soil erosion rates in Europe. Earth-Science Reviews, 94(1e4), 23e38.
Vrieling, A. (2006). Satellite remote sensing for water erosion assessment: A review.
Catena, 65(1), 2e18.
Wall, G. J., Coote, D. R., Pringle, E. A., & Shelton, I. J. (2002). RUSLEFAC d revised
universal soil loss equation for application in Canada: A handbook for estimating
soil loss from water erosion in Canada. ECORC contribution number 02-92, AAFC
contribution number AAFC/AAC2244E. Ottawa, Ontario: Research Branch Agri-
culture and Agri-Food Canada.
Wall, G. J., Dickinson, W. T., Rudra, R. P., & Coote, D. R. (1988). Seasonal soil erod-
ibility variation in southwestern ontario. Canadian Journal of Soil Science, 68(2),
417e424.
Wang, G. Q., Fang, Q. Q., Teng, Y. G., & Yu, J. S. (2016). Determination of the factors
governing soil erodibility using hyperspectral visible and near-infrared reflec-
tance spectroscopy. International Journal of Applied Earth Observation and Geo-
information, 53, 48e63.
Wang, B., Zheng, F., & Wang, Y. (2012). Adaptability analysis on soil erodibility
models in typical thin layer black soil area of Northeast China. Nongye Gong-
cheng Xuebao/Transactions of the Chinese Society of Agricultural Engineering,
28(6), 126e131.
Warren, S. D., Mitasova, H., Hohmann, M. G., Landsberger, S., Iskander, F. Y.,
Ruzycki, T. S., et al. (2005). Validation of a 3-D enhancement of the Universal
Soil Loss Equation for prediction of soil erosion and sediment deposition. Ca-
tena, 64(2e3), 281e296.
Wawer, R., & Nowocien, E. (2006). Cyfrowa mapa wystepowania erozji wodnej
powierzchniowej obszaru Polski z wykorzystaniem CORINE Land Cover 2000.
Rozdzielczosc przestrzenna 500 m. Roczniki Akademii Rolniczej w Poznaniu.
Rolnictwo, 65, 207e213.
Weltz, M. A., Kidwell, M., & Fox, H. D. (1998). Influence of abiotic and biotic factors
in measuring and modeling soil erosion on rangelands: State of knowledge.
Journal of Range Management, 51, 482e495.
Weltz, M. A., Renard, K. G., & Simanton, J. R. (1987). Revised universal soil loss
equation for western rangelands, General technical report RM-Rocky Mountain
Forest and Range Experiment Station. USA): US Department of Agriculture, Forest
Service.
Wendt, R. C., Alberts, E. E., & Hjelmfelt, A. T. (1986). Variability of runoff and soil loss
from fallow experimental plots. Soil Science Society of America Journal, 50(3),
730e736.
Williams, J. R. (1975). Sediment-yield prediction with universal equation using runoff
energy factor, present and prospective technology for predicting sediment yield and
sources. US Department of Agriculture.
Williams, J. R., & Berndt, H. D. (1977). Sediment yield prediction based on watershed
hydrology. Transactions of the Asae, 20(6), 1100e1104.
Williams, J. R., Renard, K. G., & Dyke, P. T. (1983). Epic: A new method for assessing
erosion's effect on soil productivity. Journal of Soil and Water Conservation,
38(5), 381e383.
Winchell, M., Jackson, S., Wadley, A., & Srinivasan, R. (2008). Extension and vali-
dation of a geographic information system-based method for calculating the
Revised Universal Soil Loss Equation length-slope factor for erosion risk as-
sessments in large watersheds. Journal of Soil and Water Conservation, 63(3),
105e111.
Wischmeier, W. H. (1959). A rainfall erosion index for a universal SOIL-loss equa-
tion. Soil Science Society of America Proceedings, 23, 246e249.
Wischmeier, W. H. (1960). Cropping-management factor evaluations for a universal
soil-loss equation. Soil Science Society of America Proceedings, 24, 322e326.
Wischmeier, W. H. (1966). Relation of field-plot runoff to management and physical
Factors1. Soil Science Society of America Journal, 30(2), 272e277.
Wischmeier, W. H. (1975). Estimating the soil loss equations cover and management
factor for undisturbed areas, Present and prospective technology for predicting
sediment yields and sources. Oxford.
Wischmeier, W. H. (1976). Use and misuse of the universal soil loss equation. Journal
of Soil and Water Conservation, 3, 5e9.
Wischmeier, W. H., Johnson, C. B., & Cross, B. V. (1971). A soil erodibility nomograph
for farmland and construction sites. Journal of Soil and Water Conservation, 26,
189e193.
Wischmeier, W. H., & Smith, D. D. (1958). Evaluation of factors in the soil-loss
equation. Agricultural Engineering, 39(8), 458e462.
Wischmeier, W. H., & Smith, D. D. (1965). Predicting rainfall-erosion losses from
cropland east of the Rocky Mountains. Agriculture Handbook (Vol 282).
 C. Alewell et al. / International Soil and Water Conservation Research 7 (2019) 203e225
Washington DC: US Department of Agriculture.
Wischmeier, W. H., & Smith, D. D. (1978). Predicting rainfall erosion losses - a guide to
conservation planning, agriculture handbook No. 537. USDA/Science and Educa-
tion Administration, US. Govt. Printing Office, Washington D.C.
Xu, X. L., Liu, W., Kong, Y. P., Zhang, K. L., Yu, B., & Chen, J. D. (2009). Runoff and
water erosion on road side-slopes: Effects of rainfall characteristics and slope
length. Transportation Research Part D: Transport and Environment, 14(7),
497e501.
Xu, X., Miao, C., & Liao, Y. (2008). Quantitative remote sensing study on regional soil
erosion. Wuhan University Journal of Natural Sciences, 13(3), 377e384.
Yair, A., & Raz-Yassif, N. (2004). Hydrological processes in a small arid catchment:
Scale effects of rainfall and slope length. Geomorphology, 61(1e2), 155e169.
Yang, D., Kanae, S., Oki, T., Koike, T., & Musiake, K. (2003a). Global potential soil
erosion with reference to land use and climate changes. Hydrological Processes,
17(14), 2913e2928.
Yang, D. W., Kanae, S., Oki, T., Koike, T., & Musiake, K. (2003b). Global potential soil
erosion with reference to land use and climate changes. Hydrological Processes,
17(14), 2913e2928.
Yin, S., Xie, Y., Liu, B., & Nearing, M. A. (2015). Rainfall erosivity estimation based on
rainfall data collected over a range of temporal resolutions. Hydrology and Earth
System Sciences, 19(10), 4113e4126.
Young, R. A., Onstad, C. A., Bosch, D. D., & Anderson, W. P. (1989). AGNPS - A
nonpoint-source pollution model for evaluating agricultural watersheds. Jour-
nal of Soil and Water Conservation, 44(2), 168e173.
Yue, Y., Ni, J. R., Ciais, P., Piao, S. L., Wang, T., Huang, M. T., et al. (2016). Lateral
transport of soil carbon and land-atmosphere CO2 flux induced by water
erosion in China. Proceedings of the National Academy of Sciences of the United
States of America, 113(24), 6617e6622.
Zhang, Y., Degroote, J., Wolter, C., & Sugumaran, R. (2009). Integration of modified
universal soil loss equation (musle) into a gis framework to assess soil erosion
225
risk. Land Degradation & Development, 20(1), 84e91.
Zhang, X., Drage, N. A., & Wainwright, J. (2004b). Scaling issues in environmental
modelling. In J. Wainwright, & M. Mulligan (Eds.), Environmental modelling:
Finding simplicity in complexity (pp. 319e334). John Wiley & Sons, Ltd.
Zhang, K., Li, S., Peng, W., & Yu, B. (2004a). Erodibility of agricultural soils on the
loess plateau of China. Soil and Tillage Research, 76(2), 157e165.
Zhang, X. C., Nearing, M. A., Risse, L. M., & McGregor, K. C. (1996b). Evaluation of
runoff and soil loss predictions using natural runoff plot data. Transactions of the
American Society of Agricultural Engineers, 39(3), 855e863.
Zhang, X. C., Nearing, M. A., Risse, L. M., & McGregor, K. C. (1996c). Evaluation of
WEPP runoff and soil loss predictions using natural runoff plot data. Trans-
actions of the Asae, 39(3), 855e863.
Zhang, L., O'Neill, A. L., & Lacey, S. (1996a). Modelling approaches to the prediction
of soil erosion in catchments. Environmental Software, 11(1), 123e133.
Zhang, K. L., Shu, A. P., Xu, X. L., Yang, Q. K., & Yu, B. (2008). Soil erodibility and its
estimation for agricultural soils in China. Journal of Arid Environments, 72(6),
1002e1011.
Zhang, F. B., Wang, Z. L., & Yang, M. Y. (2014). Validating and improving interrill
erosion equations. PLoS One, 9(2), 11.
Zhang, H., Yang, Q., Li, R., Liu, Q., Moore, D., He, P., et al. (2013). Extension of a GIS
procedure for calculating the RUSLE equation LS factor. Computers & Geo-
sciences, 52, 177e188.
Zhang, W., Zhang, Z., Liu, F., Qiao, Z., & Hu, S. (2011). Estimation of the USLE cover
and management factor C using satellite remote sensing: A review. In Pro-
ceedings - 2011 19th international conference on geoinformatics, geoinformatics
2011.
Zingg, A. W. (1940a). Degree and length of land slope as it affects soil loss in runoff.
Agricultural Engineering, 21(2), 59e64.
Zingg, A. W. (1940b). Degree and length of land slope as it affects soil loss in runoff.
Agricultural Engineering, 21, 59e64.