Faculteit Bio-ingenieurswetenschappen Academiejaar 2015 – 2016 Soil erosion risk mapping using RUSLE in Rwanda Simon De Taeye Promotor: Prof. Dr. Ir. Ann Verdoodt Tutor: Nick Ryken Masterproef voorgedragen tot het behalen van de graad van Master in de bio-ingenieurswetenschappen: Milieutechnologie
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Faculteit Bio-ingenieurswetenschappen
Academiejaar 2015 – 2016
Soil erosion risk mapping using RUSLE in Rwanda
Simon De Taeye Promotor: Prof. Dr. Ir. Ann Verdoodt Tutor: Nick Ryken
Masterproef voorgedragen tot het behalen van de graad van
Master in de bio-ingenieurswetenschappen: Milieutechnologie
Acknowledgements
The drop of water hollows out the stone, not through its force but by falling very persistently.
Which may seem as a very convenient quote to start a dissertation about erosion serves in
this context more as a metaphorical description on the making of this thesis. It was a long and
bumpy road and I hope that everyone fully realizes how much I’ve appreciated their help along
the way, despite my occasional shortcomings when it comes down to expressing emotion.
First of all, I would like to thank Prof. Dr. ir. Ann Verdoodt who gave me a shot at the topic
regardless my unconventional background in environmental technology. I’m very grateful for
the time you put in proofreading my text, organizing meetings and providing constructive
comments from the very start until the finish line. Many thanks also to Nick Ryken for the
guidance and useful information. Both contributed greatly in raising the bar scientifically and
holding me on topic.
My first experience in Sub-Saharan Africa wouldn’t be as wonderful without Jules Rutebuka,
Aline and Olive. Each one of you guided me during my stay and I’m very grateful for the
conversations, dinners, Kinyarwanda lessons, gospel concerts and discussions. They expanded
my horizon more than I ever could imagine. Special thanks goes also to Dr. Desire Kagabo and
his family whom provided so much more than just a roof during my stay in Kigali, I’m still
amazed by the great display of boundless hospitality. Murakoze cyane.
This thesis wouldn’t have existed without the support from VLIR-UOS. I would also like to
address the numerous (anonymous) people populating internet forums, help from this
community deepened my knowledge of Excel, guided me through Access and threw lifelines
when I risked drowning in GIS technologies.
Deze thesis vormt de zwanenzang van mijn illustere academische carrière. Graag draag ik hem
dan ook op aan mijn ouders Joost De Taeye en Ludwine Bekemans die mij altijd heel hard en
op allerlei manieren hebben gesteund. Bedankt voor alle kansen die jullie me hebben gegeven
en de vrijheid die jullie mij lieten om mijn eigen (transcontinentale) weg te volgen. Ongeacht
de wateren die ik in de toekomst zal bevaren weet ik dat er altijd een veilige en warme
thuishaven op me wacht.
Table of Contents
List of Abbreviations ....................................................................................... i
Summary .........................................................................................................ii
Samenvatting ................................................................................................. iii
Introduction ............................................................................................ 1
Background & problem statement ............................................................................. 1
Research aim & objectives ......................................................................................... 2
Literature review ..................................................................................... 4
Soil erosion process .................................................................................................... 4
Factors influencing water erosion .............................................................................. 4
2.2.1 Rainfall ............................................................................................................................... 4
2.2.2 Soil ..................................................................................................................................... 5
2.2.3 Topography ........................................................................................................................ 6
2.2.4 Vegetation ......................................................................................................................... 6
2.2.5 Management ..................................................................................................................... 7
Soil erosion models ..................................................................................................... 7
2.3.1 Types of erosion models & model choice.......................................................................... 8
2.3.2 Revised Universal Soil Loss Equation ............................................................................... 11
Overview RUSLE factors ........................................................................................... 13
2.4.1 Rainfall erosivity factor R ................................................................................................. 13
2.4.2 Soil erodibility factor K .................................................................................................... 15
2.4.3 Topographic factor LS ...................................................................................................... 20
Soil erosion in Rwanda ............................................................................................. 26
2.5.1 Physical environment ...................................................................................................... 26
2.5.2 Former research on soil loss assessment in Rwanda ...................................................... 27
2.5.3 Soil conservation strategies in Rwanda ........................................................................... 28
Materials and Methods .......................................................................... 30
Study area and overview methodology .................................................................... 30
Watershed delineation ............................................................................................. 31
R-factor ..................................................................................................................... 31
3.3.1 Collecting relevant data ................................................................................................... 31
3.3.2 Evaluate existing or new regression equations ............................................................... 33
K-factor ..................................................................................................................... 35
3.4.1 Soil maps: Rwanda Soil Information System ................................................................... 35
3.4.2 Soil series present in watershed ...................................................................................... 37
3.4.3 K-factor estimation models ............................................................................................. 38
3.4.4 Transformation from soil series to soil units ................................................................... 44
LS-factor ................................................................................................................... 45
Results & Discussion ............................................................................... 46
Watershed delineation ............................................................................................. 46
R-factor ..................................................................................................................... 48
4.2.1 Estimations based on annual precipitation and MFI ....................................................... 48
4.2.2 Calibrating own equations ............................................................................................... 49
4.2.3 Final R-value for watershed ............................................................................................. 50
4.2.4 Data reliability ................................................................................................................. 51
K-factor ..................................................................................................................... 52
4.3.1 SESA approach based on parent material ....................................................................... 52
4.3.2 RUSLE approaches: nomograph vs Dg-model ................................................................. 53
4.3.3 Nomograph approach vs algorithm Borselli et al. (2012)................................................ 54
4.3.4 Corrections to global erodibility models ......................................................................... 54
4.3.5 Soil erodibility map .......................................................................................................... 55
LS-factor ................................................................................................................... 57
4.4.1 Comparison with SESA report.......................................................................................... 57
4.4.2 LS-factor results ............................................................................................................... 60
Potential erosion risk map for Tangata watershed .................................................. 61
Conclusions ............................................................................................ 63
References .............................................................................................. 65
Annex I: Description of soil series present in Tangata watershed .................. 77
Annex II: Soil erodibility classification Kassam et al. (1992) ........................... 79
i
List of Abbreviations
C Cover and management factor (R)USLE
CDF Cumulative Distribution Function
DEM Digital Elevation Model
Dg Geometric mean particle diameter
EI30 Rainfall kinetic energy times maximal 30min intensity
FAO Food and Agricultural Organization of the United Nations
GDP Gross domestic product
GIS Geographic Information System
K Soil erodibility factor in (R)USLE
L Slope length factor in (R)USLE
LS Topography factor in (R)USLE
MININFRA Ministry of Infrastructure
MFI Modified Fournier Index
OM Organic Matter
P Support practice factor in (R)USLE
PDF Probability Density Function
PSTA III Third Strategic Plan for the Transformation of Agriculture in Rwanda
R Rainfall erosivity factor in (R)USLE
RUSLE Revised Universal Soil Loss Equation
S Slope-steepness factor in (R)USLE
SESA Service des Enquêtes et des Statistiques Agricoles
SOC Soil Organic Carbon
UCA Unit Contributing Area
USLE Universal Soil Loss Equation
ii
Summary
Soil erosion forms a prominent threat to the agricultural development of Rwanda, in which
steep slopes prevail and the ever increasing population pressurize smallholder family farms.
Erosion models such as RUSLE can serve as a tool for policy makers to select the most
appropriate strategies for combatting soil erosion. In this context, recent government plans
have stated the ambition to recalibrate erosion models. The objective formulated in this
dissertation is to use RUSLE for the production of a potential soil erosion risk map of a
catchment in Northern Rwanda.
Mapping potential soil erosion risk involves three factors: rainfall erosivity, soil erodibility and
topography. Numerous approaches exist to estimate each factor. The main challenge in this
thesis consisted of selecting a suitable and reproducible methodology for establishing reliable
input values for each factor. For rainfall erosivity existing equations published in Moore (1979)
and Vrieling et al. (2010) are compared with measured data from Ryumugabe and Berding
(1992). Also new regression equations are developed based on monthly precipitation records
in Ryumugabe and Berding (1992). Soil erodibility is estimated by five different approaches.
Both the universal nomograph and Dg-model are applied on data extracted from the Rwanda
soil information system. An estimation approach developed for Kenia described in Kassam et
al. (1992), erodibility values based on parent material from SESA (1986) and a new procedure
for estimating probable erodibility values described in Borselli et al. (2012) are also applied.
The topographic factor is determined using GIS techniques. Also here to commonly used
techniques are compared to each other.
The estimated rainfall erosivity value was 3521 MJ mm-1 ha-1 yr-1, the soil erodibility factor
hovered around 0.015 ton ha hr ha-1 MJ-1 mm-1 and the average topography value was 13. All
of these values are in line with other published research. To improve the application of the
RUSLE model in Rwanda, more measured soil erodibility values are needed and the available
rainfall records need to be reviewed.
iii
Samenvatting
Bodemerosie bedreigt een verdere ontwikkeling van de landbouw in Rwanda, waar steile
hellingen grossieren en een toenemende bevolking de landbouwproductie onder druk zet.
Erosiemodellen zoals RUSLE kunnen gebruikt worden als een middel om de meest gepaste
strategieën te selecteren om bodemerosie te bestrijden. Recente overheidsplanen hebben
dan ook de ambitie geformuleerd om erosiemodellen te herkalibreren. Het doel van deze
thesis was om RUSLE te gebruiken voor de productie van een potentiële erosiekaart voor een
stroomgebied in Noord Rwanda.
Drie factoren zijn belangrijk wanneer potentiële erosie wordt berekend: de regenval
erosiviteit, de gevoeligheid van de bodem en de topografie. Om elk van deze factoren te
berekenen werden verschillende aanpakken toegepast. De uitdaging van deze thesis was om
de beste methodologie te selecteren om elke input factor te schatten. Voor regenval werden
bestaande vergelijkingen gepubliceerd in Moore (1979) en Vrieling et al. (2010) vergeleken
met data uit Ryumugabe and Berding (1992). Ook nieuwe regressievergelijkingen werden
ontwikkeld gebaseerd op maandelijkse regenval uit Ryumugabe and Berding (1992). De
bodemgevoeligheid werd geschat met vijf verschillende methodologieën. Zowel het
universeel nomogram als het Dg-model werden toegepast op de data beschikbaar in het
bodem informatiesysteem van Rwanda. Daarnaast werd nog een procedure ontwikkeld voor
Kenia beschreven in Kassam et al. (1992) toegepast, naast waarden op basis van het
moedermateriaal uit SESA (1986) en een nieuw protocol om waarden te schatten uit Borselli
et al. (2012). Om de topografie te bepalen werden GIS technieken toegepast. Ook hier werden
gangbare technieken met elkaar vergeleken.
De geschatte regenvalerosiviteit was 3521 MJ mm-1 ha-1 yr-1, de bodemgevoeligheidsfactor
schommelde rond 0.015 ton ha hr ha-1 MJ-1 mm-1 en de gemiddelde topografische factor was
13. Al deze waarden liggen in de lijn van eerder gepubliceerd onderzoek. Om de toepassing
van het RUSLE model in Rwanda te verbeteren zijn meer gemeten waardes nodig rond
bodemgevoeligheid en moeten beschikbare regenvaldata herzien worden.
Introduction
1
Introduction
Background & problem statement
Rwanda, land of a thousand hills, is a small land-locked country located in central Africa
characterized by its mountainous topography. The last decades, the population has grown
rapidly up to 471 people per km², which makes Rwanda the most densely populated country
in Africa (Worldbank, 2016). The ever-increasing population leads to a substantial decline in
the ratio of arable land to number of people (Jayne et al., 2010). Rural densities up to 700
people per km² have caused small-holder farms to cultivate even the steepest slopes, half of
all agricultural fields are located on slopes greater than 18% (Bidogeza et al., 2009). These
conditions, where steep slopes and abundant rainfall prevail, induce severe soil loss by
erosion. Rivers in Rwanda are brown-red colored due to soil loss by rainfall. About 40% of
Rwanda’s land is classified by the FAO as having a very high erosion risk, and another 37%
requires soil retention measures before cultivation, which leaves only 23% of the cultivated
land more or less free from erosion risk (Minagri, 2009). Agriculture accounts for a third of
Rwanda’s GDP, covers 90% of the national food needs and provides livelihood for more than
80% of all households (Rwanda Development Board, 2016), which signify that soil erosion
forms a prominent risk to the general development of Rwanda.
Addressing this treat, one of the primary policy objectives formulated in the third Strategic
Plan for the Transformation of Agriculture in Rwanda (PSTA III) is the implementation of soil
conservation programs at watershed level throughout the country. The establishment of a
successful soil protection and management program requires a useful and reliable tool to
identify risk areas and quantify the magnitude of the problem. The PSTA III report states that
there is a general lack of information on erosion rates. In this context, a major sub target of
PSTA III is the recalibration of erosion models (Minagri, 2013). Soil loss models are routinely
employed at watershed level for predicting soil losses and mapping soil erosion risks (Fu et al.,
2005; Gelagay and Minale, 2016; Luliro et al., 2013; Mukashema, 2007; Schiettecatte et al.,
2008). The empirical nature of the equations used in such models requests, apart from reliable
input data, a thorough calibration stage before it can be properly applied.
Introduction
2
The Universal Soil Loss Equation (USLE) developed in the USA has gained widespread
application as the standard technique for soil conservation workers (Kinnell, 2010; Lane et al.,
1992; Morgan, 2005; Renard et al., 2010; Wischmeier and Smith, 1965). For Rwanda, the
Service des Enquêtes et des Statistiques Agricoles (SESA) published the first soil loss estimates
with USLE (SESA, 1986). For the calibration, 100 selected study sites covering different agro-
ecological zones were monitored with erosion plots for a one-year period (1983-1984).
Besides the strikingly low observed erosion losses (König, 1994; Moeyersons, 1991), this study
also suffered from shortcomings in rainfall erosivity estimates and a deficiency of adequate
soil data. Due to the limitations of the initial USLE model numerous revisions and
modifications continue to appear. The application of the USLE framework in countries outside
the USA has resulted in the publication of numerous relationships estimating input factors for
various regions. Besides enhancing the universal character by integrating data obtained from
non-American soils, the scale at which the model is applied has vastly expanded. The
application of the model has surpassed the field-level scale from which it originated and can
now be upgraded for application on 2-dimensional landscapes and regional scales (de Vente
et al., 2013; Kinnell, 2001).
Research aim & objectives
The general aim of this thesis is to create a spatially distributed potential soil erosion risk map
for a single catchment called Tangata. The Revised Universal Soil Loss Equation (RUSLE) will
be used in a GIS environment and the modelling protocol followed is summarized in figure 1.
The related central research question is: Taken into account more recent data, regression
equations and revisions of the USLE model, how can we improve the use of soil erosion models
in Rwanda and establish a reproducible methodology for estimating potential soil loss in a
watershed?
Each factor that defines the potential erosion risk (rainfall, soil & topography) generates
separate research questions:
1. Which relationships can be applied to link readily available data such as monthly or
yearly precipitation to rainfall erosivity? How reliable is the available climatic data?
Introduction
3
2. How can the data stored in the Rwandan soil information system best be used to
estimate soil erodibility values?
3. Can we upscale the RUSLE from field scale application and transpose it to a GIS
environment for the application on watershed level? What are the effects for the
topography factor?
To encounter these questions, following objectives have been defined:
1. Collect rainfall erosivity values, precipitation records and select the optimal
relationship that connects both.
2. Try different universal and local approaches on estimating soil erodibility values,
analyze estimations relative to each other and discuss results .
3. Discuss, apply and compare GIS methodologies that estimate the topographic factor
and compare to data from field measurements.
Figure 1: The general protocol followed in this thesis
Problem statement: Soil erosion in
Rwanda risks agricultural development
Objective: Create a potential soil erosion risk map.
The outcome serves as a useful tool for policy makers
combatting erosion.
Requirements: local data on
rainfall, soil and topography
Project
definition
Transformation to GIS data
A = R * K * LS
Model setup: RUSLE
Functional evaluation: The best approach to
calculate each parameter is selected
Results & Discussion
Calibration
& Validation
Additional climatic data
Data collection
Literature review
4
Literature review
Soil erosion process
Soil erosion by water is a two-phase process consisting of the detachment of individual soil
particles from the soil mass and their transport by erosive agents such as running water and
wind (Morgan, 2005; Pimental et al., 1995). The process is driven by kinetic energy
transmitted from rainfall to the soil and can take various forms depending on the pathways
taken by the flow of water in its movement over the ground surface. Sheet erosion, now
frequently called interrill erosion, involves the more-or-less even removal of layers of soil from
an entire segment of sloping land, and is by far the least conspicuous. Rill erosion results from
the concentration in surface depressions of water that subsequently flows downslope along
paths of least resistance thus forming microchannels or rills. In gully erosion the incisions have
become very large and the gullies have progressed so deeply and extensively that the land
cannot be used for normal cultivation (El-Swaify et al., 1982). Rills are distinguished from
gullies by having a critical cross-sectional area larger than 929 cm², the square foot criterion
(Poesen et al., 2003). The focus here will be exclusively on modelling rill and interrill erosion
by water.
Factors influencing water erosion
2.2.1 Rainfall
Rainfall initiates the process of erosion by provoking soil detachment and transport directly
by raindrop splash or through the contribution of rain to runoff (El-Swaify et al., 1982; Morgan,
2005). There is a clear consensus that the rainfall erosivity, defined as the potential ability of
rainfall to detach soil particles and transport sediment, is correlated more closely to rainfall
intensity rather than to total rainfall amount (Hudson, 1971; Kinnell, 2010; van Dijk et al.,
2002). As stated in the previous paragraph the soil erosion process can be viewed as a
transmission of kinetic energy. Both raindrop velocity and mass, the two key ingredients
defining kinetic energy, are positively correlated with intensity (Gunn and Kinzer, 1949; van
Dijk et al., 2002). Related to this, rainfall intensity needs to top a certain threshold value before
Literature review
5
it can initiate soil erosion (Wischmeier and Smith, 1978). These elements show why a long,
slow rain and a shorter rain at much higher intensity, may have the same rainfall amount but
differ entirely in terms of rainfall erosivity. The different rainfall indices used to quantify
rainfall erosivity are discussed in paragraph 2.4.1.
2.2.2 Soil
The likelihood that detachment and entanglement of soil particles occur depends not only on
rainfall characteristics but also on the structural characteristics of the soil. Soil erodibility is
defined by Morgan (2005) as the resistance of soil to both detachment and transport. It is
dependent on parameters such as soil texture, aggregate stability, shear strength and
infiltration capacity. For each of these factors general principles encompass their relationship
with soil erodibility. For texture, soils that are high in silt and low in clay are the most erodible
(Wischmeier and Mannering, 1969). Lower aggregate stability increases the ease for raindrops
to pulverize soil structure enabling faster entrainment (Amézketa, 1999; Bryan, 1968). Shear
strength is an important parameter to determine soil erodibility under concentrated flow (Al-
Durrah and Bradford, 1982; Bradford et al., 1992; Rauws and Govers, 1988). Infiltration
capacity determines which fraction of total rainfall will contribute to overland flow (Horton,
1945) and is influenced by texture, organic matter content, pore size and aggregate stability
(Morgan, 2005).
The soil characteristics listed are highly internally connected. In the context of soil erosion
modelling, fixing a soil erodibility value based on readily available soil properties can be
troublesome. If the stability of microaggregates is considered as the decisive estimator of soil
erodibility, it can be related to the main structuring components i.e. clay content, cation
exchange capacity (CEC) and soil organic carbon (SOC). Identifying and modelling the
dominant factor controlling the aggregation dynamic is complicated since the mechanisms
involved vary for different soil units, textures or soil pH ranges and are susceptible to
externalities such as climate or biological activity (Bronick and Lal, 2005; Six et al., 2004).
Polyvalent cations in the soil form bridges between clay and organic anions (Tisdall, 1996). The
role of SOC as a binding agent is more important on soils deficient of other structuring
components, which explains the decreasing importance of SOC for increasing clay content (Nill
Literature review
6
et al., 1996; Wischmeier and Mannering, 1969). With increasing clay content, clay mineralogy
becomes more important that the amount (Bronick and Lal, 2005). Soils dominated by swelling
clays are characterized by low aggregate stability, whereas oxides and kaolin clays are
responsible for highly stable aggregation (Six et al., 2004).
A mathematical relationship for soil erodibility enveloping all those contributing elements
would be fairly complex and highly impractical. However, successful efforts have been put into
constructing equations that tackle this complexity based on the internal connectivity of soil
characteristics and allow reliable (local) soil erodibility estimates based on a limited amount
of soil properties (see paragraph 2.4.2, page 15). However, care must always be taken in
extrapolating such relationships to areas that differ from the region from which the
relationship originated.
2.2.3 Topography
Evidently, mountainous regions are more prone to soil erosion. The surface slope affects the
erosion process in various ways: an increasing slope gradient has a positive effect on splash
detachment (Torri and Poesen, 1992) while infiltration rate decreases with increasing slope
angle (Fox et al., 1997). A higher slope gradient creates a higher flow velocity which causes
more detachment and transport of soil particles (Fox and Bryan, 1999). In general, soil loss
increases exponentially with slope steepness for tropical soils (El-Swaify, 1997). Similar to the
slope gradient, longer slope lengths allow higher runoff velocities influencing interrill erosion
rates (Chaplot and Le Bissonais, 2003). The shape of a slope affects soil loss as well, the
average soil loss from a convex slope can easily be 30% greater than an uniform slope with
the same average steepness (Renard et al., 1997).
2.2.4 Vegetation
Vegetation has several properties making it a useful tool for reducing soil erosion. First of all,
a direct mechanical protection of the soil surface is provided by the canopy and litter covers
that intercept rainfall. This reduces the kinetic energy of water that reaches the soil, causing
a lower detachment of soil particles (Bochet et al., 1998; Zuazo et al., 2006). Secondly, there
is an indirect improvement of the soil physical and chemical properties by the incorporation
Literature review
7
of organic matter (Aranda and Oyonarte, 2005). For the protection against rill and gully
erosion plant roots are at least as important as vegetation cover (Gyssels et al., 2005). Plant
roots have a mechanical effect on soil strength: by penetrating the soil mass, roots reinforce
the soil and increase soil shear strength (De Baets et al., 2006). Vegetation can act as a physical
barrier as well, altering sediment flow at the soil surface. The way vegetation is spatially
distributed along slopes can be an important factor for decreasing sediment runoff (Calvo-
Cases et al., 2003; Zuazo and Pleguezuelo, 2008).
The effects of vegetation on runoff are not always beneficial. Raindrops intercepted by the
canopy may coalesce on the leaves to form larger drops which are more erosive (Brandt,
1989). Stemflow, intercepted rainfall that flows down and runs off the base of a plant, was
identified by Keen et al. (2010) as a major contributor to the movement of soil from under
macadamia trees in New South Wales.
2.2.5 Management
Even in ancient times, farmers discovered that shaping lands in certain ways, such as contour
planting and terracing, was necessary for sustained agricultural production (El-Swaify et al.,
1982). Management techniques can work in two ways: human enforced mismanagement is a
major cause of erosion (Lal, 2001; Oldeman, 1993). Soil erosion control measures
implemented can take various forms and have variable effectiveness. Physical structures
include bench terraces or infiltration ditches. Biological anti-erosion measures envelop hedges
perpendicular to the slope or continuous plant cover. Water conservation techniques can
make a huge difference, in a study conducted in southern Rwanda, the annual soil loss under
alley-cropping treatments ranged from 1 to 5 t h-1 yr-1 in the fourth year of the experiment,
while those under local farmers’ practices were as high as 30-50 t h-1 yr-1, with a maximum
observed of 111 t h-1 yr-1 (König, 1992).
Soil erosion models
Three reasons can be distinguished for erosion modelling (Lal, 1994): first of all erosion models
can be used as predictive tools for assessing soil loss for conservation planning, soil erosion
inventories or regulation. Secondly, models can predict where and when erosion is occurring,
Literature review
8
thus helping the conservation planner target efforts to reduce erosion. Finally, models can be
used as tools for getting a sharpened understanding of the erosion process and for setting
research priorities.
2.3.1 Types of erosion models & model choice
In general, erosion models fall into three main categories depending on the physical processes
simulated by the model, the model algorithms describing these processes and the data
dependence of the model (Merritt et al., 2003). The first category consists of empirical or
statistical/metric models that use an extended database to identify significant relationships
between input variables and soil loss. Empirical models are based primarily on observation
and inductive logic. The applicability is generally limited to those conditions for which the
parameters have been calibrated (Lal, 1994).
Secondly there are the conceptual models, they lie somewhere between physically-based
models and empirical models, and are based on spatially lumped forms of water and sediment
continuity equations. Conceptual models tend to include a general description of catchment
processes, without including the specific details of process interactions, which would require
detailed catchment information (Merritt et al., 2003). Parameter values for conceptual models
have typically been obtained through calibration against observed data, such as stream
discharge and concentration measurements (Sorooshian, 1991).
Physically based models, the third category, are intended to represent the essential
mechanisms controlling erosion. Physically-based models are built on the solution of
fundamental physical equations enveloping the laws of conservation of mass and energy
(Morgan, 2005). The power of physically-based models is that they represent a synthesis of
the individual components which affect erosion, including the complex interactions between
various factors and their spatial and temporal variability (Lal, 1994). Physically based models
also have drawbacks: in theory, the parameters used are measurable and known. In practice,
the large number of parameters involved and the heterogeneity of important characteristics,
particularly in catchments, means that often these parameters must be calibrated against
observed data (Merritt et al., 2003). Equations governing the processes in physics-based
Literature review
9
models are derived at a small scale and under very specific physical conditions. In practice,
these equations are regularly used at much greater scales and under different physical
conditions, which may lead to fundamental problems in the application (Beven, 1989).
The distinction between model types is not sharp and can be somewhat subjective. Models
are likely to contain a mix of modules from each of these categories (Merritt et al., 2003). An
overview of various erosion models currently operative is given in table 1, some of the models
listed can be considered more as a framework or modelling technique rather than a firm set
of inputs, fixed equations and outputs.
In the context of this thesis, there is not enough data available to allow the usage of a
physically based model. Considering the data poor conditions and main objective of mapping
potential erosion risk, RUSLE is the most favorable option to contemplate with our goals. The
general structure of the (R)USLE model has allowed soil scientist worldwide to adapt, modify
and calibrate specific relationships for the required input factors. Publications in which the
RUSLE model has been used for tropical watersheds (Angima et al., 2003; Gelagay and Minale,
2016; Luliro et al., 2013; Millward and Mersey, 1999; Schiettecatte et al., 2008), or that
present alternative ways to estimate input factors for data-poor regions (Angulo-Martínez and
Beguería, 2009; Arnoldus, 1980; Declercq and Poesen, 1992; Renard and Freimund, 1994; Yu
and Rosewell, 1996a) are omnipresent, making RUSLE a highly flexible and appropriate model
for application on a Rwandan watershed.
Literature review
10
Table 1: Category, data requirement level, description and origin of some currently applied erosion models
name model type data
needs Short description
Developed in
reference
USLE empirical low User-friendly model that predicts
average annual soil loss on field scale by multiplying six factors
USA
Wischmeier and Smith
(1965)
RUSLE empirical low Revision on the calculation of all six USLE
factors USA
Renard et al. (1997)
USLE-M empirical low Update of USLE for application on
catchment scale and for estimating event soil loss
USA Kinnell (2001)
MUSLE empirical low Modification of USLE to model sediment
yield USA
Williams (1975)
WEPP Physical High Daily simulation model that predicts soil
erosion and sediment delivery USA
Laflen et al. (1991)
EUROSEM Physical High Simulates sediment transport and
deposition for single storms Europe
Morgan et al. (1998)
EPIC Empirical High Modification of USLE, simulates erosion
and its impact on soil productivity USA
Sharpley and Williams (1990)
WATEM Empirical Moderate
Simulates the effect of changes in landscape structure on water and tillage
erosion, the water erosion modelling part is a modification of RUSLE
Belgium Van Oost et
al. (2000)
SLEMSA Empirical Low Estimates sheet erosion from arable
land, sub-models needs to be constructed to suit a specific area
Southern Africa
Stocking and Elwell (1982)
GUEST Physical High Predicts soil losses at plot scale. Mainly used to determines the soil erodibility
parameter Australia
Yu et al. (1997)
LISEM Physical High
Based on EUROSEM, simulates runoff and erosion from single rainstorms in a
GIS environment for agricultural catchments
The Netherlands
De Roo and Jetten (1999)
SedNet Conceptual Moderate Estimates sediment deposition from
hillslopes, gullies and riverbanks into a river network
Australia Prosser et al.
(2001)
Literature review
11
2.3.2 Revised Universal Soil Loss Equation
2.3.2.1 Introduction
RUSLE is a computerization and update of the 1978 version of USLE, from which it retains the
model framework (Renard et al., 2010). The USLE was derived from and tested on data from
experimental stations representing over 10,000 years of records for all regions of the USA east
of the Rocky Mountains (Lane et al., 1992; Morgan, 2005; Wischmeier and Smith, 1965). The
erosion model was designed to predict the longtime average soil losses associated with sheet
and rill erosion from agricultural fields in specified cropping and management systems. Since
the procedure is based on six factors that envelop all influencing erosion factors the USLE
model is believed to be applicable wherever numerical values of its factors are available
(Wischmeier and Smith, 1978).
2.3.2.2 Formula
The soil loss equation is (Renard et al., 1997):
𝐴 = 𝑅 ∗ 𝐾 ∗ 𝐿 ∗ 𝑆 ∗ 𝐶 ∗ 𝑃 (1)
Where A is the computed soil loss per unit area (ton ha-1 yr-1), R is the rainfall erosivity factor
(MJ mm hr-1 ha-1 yr-1), K is the soil erodibility factor (ton ha hr ha-1 MJ-1 mm-1), L stands for the
slope length factor (-), S for the slope-steepness factor (-), C represents the cover and
management factor (-) and P is the support practice factor (-).
2.3.2.3 Principles and proper use of the model
a) The six parameters
Although equation 1 is commonly seen in the literature, the model works mathematically in
two steps. The (R)USLE is based on the unit plot concept, where the unit plot is defined as
tilled bare fallow area, 22.1 m long on a 9% slope with cultivation up and down the plot
(Kinnell, 2010; Renard et al., 1997; Wischmeier and Smith, 1965). Only R and K have units, and
the product of those two factors computes soil loss for the unit plot, as a second step the
other factors are used to extrapolate the calculated soil loss from the unit plot to the actual
Literature review
12
situation. The L,S,C & P factor are mathematically forced to take on values of 1.0 for the unit
plot (Kinnell, 2008). A C-factor of 0.43 means that a particular cropping management system
is 43% as erodible as the unit plot condition, bare plots are assigned a C-value of 1 (Lane et al.,
1992). This research only concerns potential erosion risk assessment, so the C- and P-factor
won’t be discussed.
b) Proper use
The main strength of RUSLE is its user-friendliness due to both the simplicity of the equation
and the availability of parameter values (Loch and Rosewell, 1992). The low input data
requirements have highly contributed to its popularity (Merritt et al., 2003). The RUSLE is
universal only insofar as it identifies all the elements that determine the magnitude of soil loss
due to rill and interrill erosion. It is not useful, nor was it intended, for estimating losses caused
by other forms, such as gully erosion (El-Swaify et al., 1982). The USLE is also not
recommended and designed for prediction of specific soil loss events or soil losses in short
term (Kinnell, 2010; Wischmeier and Smith, 1978). The model tends to over-predict small
annual soil losses and under-predict large annual soil losses (Risse et al., 1993).
Literature review
13
Overview RUSLE factors
2.4.1 Rainfall erosivity factor R
2.4.1.1 Determining R
The research data used by Wischmeier and Smith for establishing the USLE indicated that
when factors other than rainfall are held constant, storm soil losses from cultivated fields are
directly proportional to a rainstorm parameter identified as the El. El is an abbreviation for
energy times intensity, and the term should not be considered simply as an energy parameter
(Wischmeier and Smith, 1965, 1978). The E factor is calculated from rainfall intensity–kinetic
energy relationships. Equation 2 shows the original relationship used to express the total
energy of a rainstorm.
{𝑒 = 0.119 + 0.0873 log(𝑖) 𝑖 ≤ 76 𝑚𝑚/ℎ
𝑒 = 0.283 𝑖 > 76 𝑚𝑚/ℎ
(2)
Where e is kinetic energy of rainfall (MJ ha-1 mm-1) and i is the rainfall intensity in mm h-1. A
limit was imposed on i since median drop size does not continue to increase when intensities
exceed 76mm h-1. In the RUSLE-model, equation 2 is replaced by (Renard et al., 1997):
𝑒 = 0.29[1 − 0.72exp (−0.05𝑖)] (3)
The storm energy indicates the volume of rainfall and runoff, but a long, slow rain may have
the same E value as a shorter rain at much higher intensity. Therefore the energy is multiplied
with an intensity factor I30, this factor is defined as the maximum 30 min intensity and
indicates the prolonged-peak rates of detachment and runoff (Wischmeier and Smith, 1978).
The final R-factor is calculated by equation 4 (Renard et al., 1997):
𝑅 =
∑ (𝐸𝐼30)𝑖𝑗𝑖=1
𝑁 (4)
Where (EI30)i = total storm kinetic energy multiplied with max 30 min intensity for storm i, j is
the total amount of storms in an N year period. A break between storms is defined as 6h or
more with less than 1.3 mm of precipitation. Rains less than 13 mm are omitted as insignificant
unless the maximum 15 min intensity exceeds 24 mm/h (Wischmeier and Smith, 1978).
Literature review
14
2.4.1.2 Estimating R-values
High-resolution rainfall measurements are not always readily available. The general approach
used to estimate R-factor values for areas without data is as follows (Renard and Freimund,
1994): first of all R-factor values are calculated by the prescribed method (equation 4) for
stations with recording rain gages. Secondly a relation is established between the calculated
R-values and more readily available types of precipitation data (i.e. daily, monthly or annual
rainfall amounts), thirdly this relation is extrapolated and R-values are estimated to locations
without long term detailed intensity precipitation data. Numerous regression equations have
been produced following this procedure and some of them are presented in section 3.3.2.
Since relationships between rainfall amount on one side and rainfall intensity and erodibility
on the other side are highly location dependent, care must be taken not to extrapolate the
derived equations for regions others than the ones for which they were derived.
2.4.1.3 Comments regarding R-factor
The use of kinetic energy as a rainfall erosivity parameter is questioned by Goebes et al.
(2014), where the suggestion is made to use rainfall momentum (mass x velocity) rather than
kinetic energy as a substrate-independent erosivity predictor. The relationships presented
between kinetic energy and rainfall intensity is based on measurements at single locations in
the US, i.e. Washington D.C. for equation 2 (Wischmeier and Smith, 1965) and Holly Springs,
Mississippi for equation 3 (Brown and Foster, 1987). A general review of these equations on
applicability revealed overestimates of rainfall energy for regions experiencing strong oceanic
influence or at high elevation. Contrarily, rainfall energy may be higher than estimated for
semi-arid to sub-humid locations (van Dijk et al., 2002).
Related, the strong correlation between soil loss and the EI30 parameter plus the threshold
value for erosion initiation at 13mm were observed while developing the USLE-model. The
legitimacy of EI30 as the most accurate erosivity estimator is entirely based on American soil
loss data. The multiplication of energy and intensity has no valid physical basis which, together
with the arbitrary character of the 30 min time period, undermines the universal character of
Literature review
15
the EI30 parameter as the best suitable rainfall erosivity predictor. Other erosivity indices have
been used for connecting soil loss to rainfall (Kiassari et al., 2012).
The RUSLE framework doesn’t allow interaction between the six input factors, which induces
that the timing of erosive rains with respect to crop cover is not considered. For East-African
conditions, Moore (1979) remarked that mainly for this reason the correlation coefficient
between a rainfall erosivity parameter and soil loss is unlikely to be higher than 0.7.
2.4.2 Soil erodibility factor K
2.4.2.1 Determining K
The soil erodibility factor, K, is a quantitative value preferably experimentally determined on
natural rainfall plots or rainfall simulation plots. For a particular soil, it is the rate of soil loss
per erosion index unit as measured on a unit plot. A unit plot is 22.1 long, with a uniform
lengthwise slope of 9 percent, in continuous fallow, tilled up and down the slope. When all of
these conditions are met, L, S, C, and P each equal 1.0, and K equals A/El (Wischmeier and
Smith, 1978).
2.4.2.2 Estimating K
Because the direct measurement of the K-value requires the establishment and maintenance
of natural runoff plots for long observation periods at various locations, numerous attempts
have been made to simplify the costly technique and develop equations to calculate soil
erodibility values from readily available soil properties (Bryan, 1968; Wang et al., 2012).
One of the most widely used and frequently cited relationships for estimating K-factor values
is the soil-erodibility nomograph (Renard et al., 1997; Wischmeier et al., 1971). The algebraic
approximation of this nomograph, only valid when the silt and very fine sand fraction does
not exceed 70%, is given by equation 5 (Renard et al., 1997).
𝐾 = (1
7.59) ∗ [2.1 ∗ 10−4 ∗ (12 − 𝑂𝑀) ∗ 𝑀1.14 + 3.25 ∗ (𝑆𝑡 − 2) + 2.5 ∗ (𝑝′ − 3)]/100 (5)
Literature review
16
Where K is the soil erodibility in t ha h ha-1 MJ-1 mm-1, M equals the silt and very fine sand
fraction multiplied with silt and total sand fraction expressed as percentages. St is the soil
structure class, p’ is a permeability class and OM is the soil organic content in percentage. In
RUSLE2, the second computerized version of RUSLE, equation 5 was modified for an improved
estimation of erodibility for soils rich in clay or sand. The structure sub factor (which is the
second term) was altered to 3.25*(2-St) (Foster, 2005). Increasing computer accessibility
combined with cumbersome reading of the nomograph proceeded in a general use of
equation 5, neglecting the cases requiring the nomograph (Auerswald et al., 2014).
Alternatively, RUSLE includes an approach for estimating K-values based on geometric mean
particle diameter (Dg) calculated as:
𝐷𝑔(𝑚𝑚) = 𝑒∑ 𝑓𝑖∗𝑙𝑛(
𝑑𝑖+𝑑𝑖−12
) (6)
For which fi equals the weight fraction of interval i and di refers to the upper diameter
boundary of interval i. Data from both natural and simulated rainfall studies worldwide was
used to connect Dg with a K-factor, the regression equation is given by equation 7 (Renard et
al., 1997):
𝐾 = 0.0034 + 0.0405 ∗ exp [−0.5 ∗ (log(𝐷𝑔) + 1.659
0.7101)
2
] (7)
Soil properties affecting soil erodibility are many and varied (see paragraph 2.2.2). Several
erosion mechanisms are operating at the same time, invoking a difficult relationship between
specific soil properties and soil erodibility. The implicit assumptions on which the soil
erodibility concept is built, i.e. firstly that the K-factor is valid for all erosion processes,
secondly that it can be estimated by a few, usually physical soil properties and thirdly that it
remains constant over time, are questionable (Bryan et al., 1989). In fact, the properties that
dominate erosional response such as aggregation and shear strength are changing over short
and long-term cycles of varying magnitude and predictability (Bryan, 2000). A highly dynamic
parameter such as soil water content correlates with shear strength since water weakens the
bonds between soil particles (Fredlund et al., 1996). In the same context and similar to the R-
Literature review
17
factor, the framework of the model separates soil erodibility, soil management and vegetation
while there is significant interaction among these parameters that affects soil erodibility. To
accommodate seasonal variance, an overall K-value can be calculated by equation 8, which
takes a weighted average proportional to rainfall erosivity (Renard et al., 1997):
𝐾 = ( ∑ 𝐸𝐼𝑖 ∗ 𝐾𝑖
100
𝑖=1
)/100 (8)
Where the Ki parameter corresponds to the soil erodibility measured during a fixed time
interval (typically 15 days) and EIi accounts for the relative percentage of rainfall erosivity
during the same period.
Secondly the universal character of the developed equations is controversial. The widely
applied nomograph relationship is derived from rainfall-simulation data from 55 midwestern,
mostly (81%) medium-textured, surface soils (Renard et al., 1997). El-Swaify et al. (1982)
criticized the use of the mainland U. S. -based nomographs to predict the erodibility of tropical
soils and provided preliminary data supporting the need for different predictive parameters
to estimate tropical soil erodibility.
2.4.2.3 Estimating K-factor for tropical soils
El-Swaify and Dangler (1976) developed a relationship for tropical soils of volcanic origin which
has been applied for Kenya (Angima et al., 2003).
𝐾 = (−0.03970 + 0.00311𝑥1 + 0.00043𝑥2 + 0.00185𝑥3 + 0.00258𝑥4
− 0.00823𝑥5)/7.59 (9)
Where x1 is the unstable aggregate size fraction % less than 0.250 mm, x2 is the product of %
modified silt (0.002-0.1 mm) and % modified sand (0.1-2 mm), x3 is the % base saturation, x4
is the % silt fraction (0.002-0.050 mm) and x5 is the modified sand fraction (0.1-5 mm) in
percent. The relationship was based on volcanic soils in Hawaii and should be considered only
for similar soils.
Literature review
18
For Ethiopia, Hurni (1985) developed a method for quickly estimating soil erodibility based on
soil color (table 2) which is still used today (Bewket and Teferi, 2009; Gelagay and Minale,
2016).
Table 2: Soil erodibility estimation based on color for Ethiopia (Hurni, 1985)
Soil color Soil reference group K value (t ha h ha-1 MJ-1 mm-1)
Black Andosols, Vertisols, etc. 0.02
Brown Cambisols, Phaeozem, Regosols, Luvisols, etc.
0.026
Red Lixisols, Nitisols, Alisols, etc 0.033
Yellow Fluvisols, Xerosols, etc 0.040
Roose (1977) observed satisfactory results estimating the K-factor with the USLE-nomograph
in West-Africa for clay soils that are predominantly kaolinitic, but added to his conclusions
that one must be very cautions for the application of the nomograph on Vertisols. For Nigeria,
researchers observed a very poor correspondence between measured erodibility and
estimations based on the nomograph (De Vleeschauwer et al., 1978; Vanelslande et al., 1984).
Erodibility measurements from 28 tropical soils from Cameroon and Nigeria were used to
establish a protocol for estimating soil erodibility on tropical soils starting from the
nomograph (Nill et al., 1996). Three different soil groups were distinguished; the first group
contains soils with a much higher erodibility than estimated by the nomograph. They are
characterized with a low soil bulk density and are prone to surface sealing, the volcanic soils
in El-Swaify and Dangler (1976) described by equation 9 correspond to the first category. For
the second group erodibility values calculated with the nomograph agreed well with measured
erodibility, about half of the soils belonged to this category. The third group consist of soils
that tend to be more clayey and richer in OM than the other two soil groups. Soil loss behavior
is not determined by surface sealing but by the permeability of the profiles, only sequence of
several storms saturating these soils have a thorough erosive impact (Nill et al., 1996).
Equations based on bulk density, silt content, OM, pH and amount of air-dry aggregates were
developed to classify each soil in one of the three groups. The final equations to estimate soil
erodibility for the different groups are:
Literature review
19
Group 1: 𝐾 = 2.3 ∗ 𝐾𝑛𝑜𝑚 + 0.12 (10a)
Group 2: 𝐾 = 1.1 ∗ 𝐾𝑛𝑜𝑚 (11b)
Group 3: 𝐾 = 0.03 ∗ 𝐾𝑛𝑜𝑚 + 0.006
(12c)
Where Knom equals the soil erodibility value estimated by the nomograph. Wang et al. (2012)
compared measured values for tropical soils in China with predicted values obtained from the
USLE-nomograph and Dg-approaches. The existing approaches overestimated soil erodibility
for almost all 51 natural runoff plots. Averagely, the values estimated by the USLE-nomograph
were almost two times the values measured. Satisfying results (R²=0.67) for soil erodibility
estimation were obtained with following newly calibrated relationship:
𝐾 = 0.0364 − 0.0013 [ln (𝑂𝑀
𝐷𝑔) − 5.6706]
2
− 0.015 ∗ exp [−28.9589(log(𝐷𝑔) + 1.827)²] (13)
Where OM is the soil organic matter content in %. The relationship between soil erodibility
and OM is ambigous. As already explained in paragraph 2.2.2, organic matter relates primarily
yet not exclusively to aggregates stability as one of the most important and well-known
stabilizing agents in soil (Le Bissonais, 1995). Soils with less than 2% OM by weight are highly
erodible (Fullen and Catt, 2004), as organic matter increases above the 2 per cent critical
threshold the relationship between decreasing organic matter content and growing soil
erodibility is not linear (Barthès et al., 1999) and can vary depending on the origin of the
organic material (Ekwue et al., 1993). The high interaction with other soil properties and
related complexity doesn’t allow simple extrapolation.
Torri et al. (1997) attepted to link a global dataset of longterm measured K-values from 596
soils to readily available and universal parameters (Dg, OM, clay content) and concluded that
no universal relationship of any predictive value could be developed, which contradicts the
Dg-equation in RUSLE. Alternatively a protocol was developed to establish lower and upper
bounds for K-values based on different Dg ranges and clay fraction. Different than in RUSLE,
Dg is defined as:
Literature review
20
𝐷𝑔,𝑡𝑜𝑟𝑟𝑖 = ∑ 𝑓𝑖ln (√𝑑𝑖𝑑𝑖−1)
𝑖
(14)
The minimal and maximal K-values for the two lowest Dg classes are presented in figure 2.
Figure 2: Approach for establishing minima and maxima for two Dg categories according to Torri et al. (1997)
Singh and Khera (2010) observed that nor the nomograph neither Torri et al. (1997) could be
used to estimate soil erodibility values for Indian conditions.
2.4.3 Topographic factor LS
2.4.3.1 Slope length factor L
The effective slope length is the distance from the point of origin of overland flow to the point
where either the slope decreases enough to allow deposition or the runoff water enters a
well-defined channel (Wischmeier and Smith, 1978). More questions and concerns have been
expressed over the L-factor than any of the USLE factors (Renard et al., 1991), different users
choose different slope lengths for similar situations (McCool et al., 1995). In the (R)USLE
model, the L factor is given by
𝐿 = (𝜆/22.1 )𝑚 (15)
Where λ equals the slope length in meter, the number 22.1 is obtained from the unit plot
length and m is a variable slope-length exponent. The slope length λ is not the distance parallel
0
0,02
0,04
0,06
0,08
0,1
0,12
0
0,02
0,04
0,06
0,08
0,1
0,12
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8
K-v
alu
e (
ton
ha
h/h
a M
J m
m)
Clay fraction
Estimating K-ranges based on Clay content and Dg
Dg < -2,5
-2,5 < Dg < -2,25
Literature review
21
to the soil surface but the horizontal projection. In USLE, m equals 0.5 if the slope exceeds 5%,
0.4 on slopes of 3.5 to 4.5 percent, 0.3 on slopes of 1 to 3 percent, and 0.2 for slopes lower
than 1% (Wischmeier and Smith, 1978).
In the RUSLE (Renard et al., 1997), m is related to the ratio β of rill erosion to interrill erosion
by the following equations:
𝑚 = 𝛽/(1 + 𝛽) (16) with
𝛽 = (
𝑠𝑖𝑛 𝜃
0.0896)/[3.0(𝑠𝑖𝑛𝜃)0.8 + 0.56]
(17)
Where θ is the slope angle.
Slopes up to 18% were used to develop USLE and RUSLE, Liu et al. (2000) conducted research
on natural runoff plots with slope steepness increasing from 20 to 60% and concluded that for
steep slopes the USLE equation was a better fit to describe the relationship between slope
length and soil loss then the RUSLE equations. The best results were obtained when the
exponent m was fixed at 0.44. Figure 3 shows a comparison between the original USLE-
approach (black graphs) in assessing an L-factor as a function of slope length and more
recently established equations. For steeper slopes, RUSLE values start diverting from to the
other approaches.
Literature review
22
Figure 3: L-factor as a function of slope length
2.4.3.2 Slope steepness factor S
The S factor in USLE, based on natural runoff plots from 3 to 18%, is given by
𝑆 = 64.41 𝑠𝑖𝑛2𝜃 + 4.56 𝜃 + 0.065 (18)
Because this equation over-predicts soil losses from high-gradient slopes (Kinnell, 2010), its
replaced in RUSLE by (Renard et al., 1997):
𝑆 = 10.8 sin 𝜃 + 0.03 𝑠 < 9%
𝑆 = 16.8 sin 𝜃 − 0.5 𝑠 ≥ 9%
(19)
For slopes greater than approximately 22%, Liu et al. (1994) found an underestimation of the
slope factor with RUSLE. They analyzed data from three sites located in the Yellow River loess
plateau of China for slopes up to 55% and presented an alternative equation:
𝑆 = 21.91 sin 𝜃 − 0.96 (20)
0
0,5
1
1,5
2
2,5
3
0 10 20 30 40 50 60 70 80 90 100
L-fa
cto
r
Slope length λ (m)
Comparison L-factor estimations
USLE (>5%)
USLE (4-5%)
USLE (1-3%)
Literature review
23
Equation 18, 19 and 20 were updates to predict the S-factor for ever increasing slope ranges,
in an attempt to making a single continuous function of sin θ consistent with all previously
established relationships Nearing (1997) developed following equation:
𝑆 = −1.5 +
17
1 + 𝑒2.3−6.1𝑠𝑖𝑛𝜃
(21)
The S-factor in function of slope steepness for different approaches is given in Figure 4, which
shows that the USLE equation contrasts greatly with the other approaches for steeper slopes.
Figure 4: S -factor in function of slope steepness for different approaches
2.4.3.3 Calculating LS values on a catchment scale
For RUSLE applications on a catchment scale, the subfactors L and S are often considered
together as one topography factor LS which encloses length, steepness and shape. For GIS
applications, slope length calculations are often the most problematic (Hickey, 2000) and have
limited the use of RUSLE at landscape scales (Van Remortel et al., 2001). RUSLE has undergone
major modifications and upgrades to implement topographic complexity centering around
0
5
10
15
20
25
30
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
S fa
cto
r
slope steepness (%)
Comparison S-factor estimations
USLE
RUSLE
Liu et al (1994)
Nearing (1997)
(sin θ/0,0896)^1,3
Literature review
24
three major methods: unit contributing area, unit stream power theory or grid cumulative
length.
Foster and Wischmeier (1974) were the first to develop a procedure to account for slope
shape by dividing an irregular slope into a limited number of uniform segments. For a two-
dimensional application, Desmet and Govers (1996) adopted the structure of the formula, but
replaced slope length with the concept of unit contributing area (UCA), i.e. the upslope
drainage area per unit of contour length:
𝐿𝑖,𝑗 =𝐴𝑆𝑖,𝑗−𝑜𝑢𝑡
𝑚+1 − 𝐴𝑆,𝑗−𝑖𝑛𝑚+1
(𝐴𝑆𝑖,𝑗−𝑜𝑢𝑡 − 𝐴𝑆𝑖,𝑗−𝑖𝑛)(22.13)𝑚
(22)
Where Li,j is the slope length factor from the grid cell with coordinates (i,j), ASi,j-out is the unit
contributing area at the outlet of the grid cell with coordinates (i,j) (m²/m), and Asi,j-in equals
unit contributing area at the inlet of the grid cell with coordinates (i,j) (m²/m). Since this
approach incorporates flow convergence and slope shape, it can differ considerably from the
manually measured distance to an upland border of a field. For the calculation of upslope area
the distinction can be made between single flow algorithms (all flow from one cell follows the
steepest descent and goes into one of the eight neighboring cells) or multiple flow algorithms,
which allow the outgoing flow to be poured out in several receiving cells (Freeman, 1991).
An alternative approach is based on the unit stream power method (Moore and Burch, 1986).
The general equation in a sediment transport model has the following form (Mitasova et al.,
1996):
𝑞𝑠 = 𝜙𝑞𝑚(𝑠𝑖𝑛𝛽)𝑛𝑖𝛿(1 −𝜏0
𝜏) (23)
Where qs is the sediment flux (kg m-1 s-1), q is the water flux (m³ m-1 s-1), β is the slope angle, i
is the rainfall intensity (m s-1), τ0, τ are the critical shear stress and shear stress (Pa) and φ, m,
n, δ and ε are experimental or physically based coefficients. The influence of terrain is
incorporated in the term qm(sinβ)n. q, the water flux, can be rewritten as the product of
upslope contributing area per unit contour width, As (m²/m) and rainfall excess rate (m/s).
Literature review
25
After separating rainfall from topography and transformation to a dimensionless form
conform unit plot conventions the LS factor can be expressed as (Moore and Wilson, 1992):
𝐿𝑆 = (𝐴𝑠
22.13)
𝑚
(sin (𝜃)
0.0896)
𝑛
(24)
Equation 24 is an excessively used (Erdogan et al., 2007; Gelagay and Minale, 2016; Lee, 2004;
Prasannakumar et al., 2012) approach to establish the combined LS factor. The m and n are
exponents varying from 0.4 to 0.56 and 1.2 to 1.3 respectively, and can be set to match the
equations for L and S discussed earlier (see figure 4). The greatest limitation of equation 24 is
that slope breaks aren’t considered since an algorithm for predicting zones of deposition is
lacking.
A third option is a cumulative grid-based method. The protocol consists of two steps and
follows a single flow algorithm. Firstly, non-cumulative slope length is determined for each
cell and equals the cell resolution if flow comes from a cell in a cardinal direction, 1.4 times
the cell resolution otherwise or zero if the cell is located on a high point (Van Remortel et al.,
2001). Secondly, the non-cumulative slope segments are sewed together. If two or more cells
pour into one cell the one with the highest slope length is withhold, which is a different
interpretation of concavity than the approaches using contributing area. A mechanism for
detecting areas of deposition is included by defining a cutoff slope angle (Van Remortel et al.,
2004). The change in slope along the flow direction is monitored and zones of deposition are
identified when the relative decrease in slope exceeds the cutoff slope angle, which is a user
defined constant from 0 to 1. Available literature (Hickey, 2000) suggests a cutoff slope around
0.5 (meaning that deposition occurs, and slope length is reset to zero, when the maximal
downhill slope in one cell is less than half of the downhill slope in the previous cell). For slope
gradients <5%, a cutoff slope around 0.7 is suggested (Claessens et al., 2008). Liu et al. (2011)
observed that slope length values are most accurately calculated with the grid cumulative
method, for which the slope lengths were approximately 50% less than the values obtained
with the UCA-method. A limitation of the grid cumulative approach is that emerging
concentrated channel networks are not considered (Zhang et al., 2013).
Literature review
26
Soil erosion in Rwanda
2.5.1 Physical environment
Rwanda is a small landlocked country with a total area of 26 338 km², located between 1°04’
and 2°51’ southern latitude and between 28°53’ and 30°53’ eastern longitude. The
neighboring countries are Uganda in the north, Tanzania in the east, Burundi in the south and
the Democratic Republic of the Congo in the west. Thanks to its high altitude, ranging from
970 to 4500m, Rwanda enjoys a mild sub-equatorial climate despite its location very close to
the equator (Verdoodt and Van Ranst, 2006a) . Mean temperature is relatively stable during
the year hovering around 20 °C without significant seasonal variation (Mukashema, 2007).
Two rain seasons can be distinguished, the first one from February to May and the second one
from mid-September to mid-December accounting for respectively 40% and 27% of the total
annual rainfall (Imerzoukene and Van Ranst, 2001). The amount of rain is strongly dependent
of the elevation, the highlands receive up to 2000 mm annually whereas in the south-east
lowlands the rainfall drops below 1000 mm (Verdoodt and Van Ranst, 2006a, b). Based on
rainfall and temperature, Rwanda can be divided in 10 different agro climatic zones (Verdoodt,
2003).
The topography of Rwanda changes from west to east and is shown together with the agro-
climatic zones in figure 5. The Congo – Nile watershed divide in the west is characterized by
steep slopes and sharp peaks with the highest elevations situated in the volcano zone up in
the North (Verdoodt and Van Ranst, 2006a). In the center of the country the central plateau
dominates, forming the landscape of thousand hills which are characterized with elongated
hills and rounded tops. The Eastern part of the country can be a considered as a transition
zone towards the peneplaine of eastern Africa, with Lake Victoria as center (Deprez, 2001).
Literature review
27
Figure 5: elevation, agro-climatic zones and location of Tangata watershed in Rwanda
2.5.2 Former research on soil loss assessment in Rwanda
By authors’ knowledge, the systematic research on soil erosion in Rwanda is launched by
Moeyersons in 1977 by installing erosion pins and later on catchpits on hills near Butare in the
Southern province (Moeyersons, 1991). He estimated the soil loss for Rwaza in function of its
destination: 0-5 t/ha/y for fallow land inaccessible for livestock, 30 t/ha/y if it’s intensively
grazed and up to 120 t/ha/y for cultivated parcels (Moeyersons, 1991).
The first erosion plots were installed by Wassmer near Kibuye, Western Province. He observed
on four Wischmeier-type runoff plots with slope 60% a soil loss of the order of 240 t/ha in a
period of 6 months, but added that soil cover could greatly reduce soil loss (< 6t/ha) (König,
1994; Wassmer, 1981).
Acknowledging the need to develop planning strategies for soil conservation, the government
started a project in 1983 on estimating the soil loss for the different regions in Rwanda. The
research project was carried out by ‘le Service des Enquêtes et des Statistiques Agricoles’
(SESA) and data was processed in collaboration with Professor Lewis of Clark University (SESA,
1986). Soil loss was measured by installing Gerlach troughs at the bottom of 100 sample fields
Literature review
28
spread over Rwanda, and data was collected for one agricultural year, 1984. The maximum
soil loss measured on a plot was 28 t/ha/y. These data were used to calibrate the USLE and
the model was applied using information obtained from a national agricultural survey. The
highest estimate of soil loss was 143 t/ha/y (Lewis et al., 1988). Using a soil loss tolerance
value of 15t/ha/y, the assessment was made that 4 out of the 10 districts risked a declination
of soil productivity. Lewis emphasized the positive effect of the widespread cultivation of
banana and the prevailing agricultural practice of intercropping providing good groundcover
throughout the rainy seasons. A description and evaluation of the formulas used to assess the
rainfall and K-factor in USLE is given in the material & methods section.
The Gerlach-type sediment traps used in the SESA study had a maximum retention capacity of
2 kg and were emptied only once per week (SESA, 1986). A replication study with the same
sediment traps in the Ruhengeri prefecture, northwest Rwanda, was abandoned prematurely
because the trap capacity was too small to contain the sediment displaced after most storms
(Byers, 1990). The methodological problems have led to an initial underestimation of the soil
erosion risk (König, 1994; Moeyersons, 1991).
Nyamulinda (1991) observed soil losses between 35 and 240 t/ha/y for parcels without anti-
erosion structures in the Gakenke district, Northern Province. König (1994) estimated the soil
loss on steep slopes in Rwanda in the range of 100 t/ha, while Roose and Ndayizigiye (1997)
observed an annual soil loss in the Southern Province from 450 t/ha for bare plots to 27 t/ha
on cropped and hedged plots. Kagabo et al. (2013) found a mean annual soil loss of 40 t/ha
on plots in the Northern Province and observed a reduction of 57% when grass strips and
infiltration ditches were added. In general, literature for Rwanda shows a large spectrum of
observed and estimated soil loss values depending on the specific region, soil cover and anti-
erosion strategies.
2.5.3 Soil conservation strategies in Rwanda
Bench and progressive terraces are considered to be the most important erosion control
techniques in recent (post 1994) history of soil conservation for Rwanda (Bizoza, 2014).
Systematic efforts controlling soil erosion control have been put into practice since the 1930s
Literature review
29
(Bizoza and De Graaff, 2012). In 1947, the colonial administration made the creation of
infiltration ditches and the planting of grass trees obligatory for all land holders (Rushemuka
et al., 2014). The imposed strategy was poorly accepted and many erosion control structures
were abandoned after Independence (1962) as they required a lot of labor and maintenance
without immediately increasing crop yield (Roose and Ndayizigiye, 1997; Rushemuka et al.,
2014). Radical or bench terraces were introduced around 1972 at Kisaro, in the North of
Rwanda (Mupenzi et al., 2012). Alternative to a physical construction, also living hedges with
or without infiltration ditches are installed parallel to contour lines so progressive terraces can
form naturally (Kagabo et al., 2013). After 1994 the use of bench terraces has expanded
throughout the country with the extensive support from (non)governmental programs; the
active participation of farmers to the programs was more related to cash incentives and food
rations than with a general concern with regard to soil loss (Bizoza, 2014; Bizoza and De
Graaff, 2012). Whereas some recently constructed terraces are used effectively, others were
soon neglected (Rushemuka et al., 2014).
Materials and Methods
30
Materials and Methods
Study area and overview methodology
The Tangata watershed is located in the Buberuka highlands, North of Rwanda with a latitude
of 1°31’S and longitude of 29°51 around 2000m above sea level (figure 5). The location was
selected in compliance with the research conducted by drs. Jules Rutebuka, whom installed
several erosion plots within the watershed. The main objective is the production of a potential
soil erosion risk map for the Tangata watershed; the most important data source was formed
by the Rwandan Soil Information System. A schematic summary of the processes followed to
achieve the objective starting from the consulted data is given in figure 6.
Figure 6: Overview with consulted data sources and consecutive steps to calculate each factor
Measured
R-values
in
Rwanda Lab analysis
Taxonomy
Location
Soil maps Soil profile
database
Mapping unit: association
of soil series
Rwandan Soil Information System
Topography
DEM
Fill sinks
Flow direction
Slope
Flow
accumulation
Slope length
LS-factor
Examine soil profiles taken for each
soil series, omit profiles taken in
forestland or savannah
Calculate K-value for each profile
K-value soil series = average K-
value of profiles member of the soil
series
K-factor
Daily
rainfall
stations
Existing
literature
regarding
R-factor
estimation
for (East)
Africa
Additional climatic data
Evaluate
performance
of existing
equations Calibrate
new
equation
Extrapolate equations to
daily rainfall records in
watershed
Watershed
delineation
R-factor
Transform K-value for each soil
serie to K-value for each map unit Potential soil erosion risk map
Materials and Methods
31
Watershed delineation
The Digital Elevation Model (DEM) originates from digitized topographic maps with 25m
contour lines at scale 1:50000 and was developed at a resolution of 30m. The procedure for
extracting a watershed from the DEM consist of three stages: first the depressions in the DEM
were omitted, secondly a channel network is generated and thirdly the boundaries of the
watershed were defined.
The algorithms used for setting a flow direction based on elevation records always use a
gradient measured outwards from a grid cell. A problem that arises is that virtually all DEM
files contain flat or depression pixels, where the maximum outward gradient is zero or
negative (Pan et al., 2012). The flow direction for such pixels cannot be determined from
adjacent cells, which eventually results in an output of disconnected stream-flow patterns.
Consequently, a sink removal step is essential for proper hydraulic processing. The algorithm
developed by Wang and Liu (2006) was employed for handling surface depressions.
Determining the flow direction is based on ‘the eight neighbor principle’, where the steepest
descent from each grid cell to a neighboring cell is calculated. A number from zero to seven is
given to each cell depending on the flow direction (0=North, 1=Northeast, …, 7=Northwest).
From the flow direction a catchment area can be established by multiplying the number of
cells draining into a cell with the area of each cell, i.e. 900m². Finally, a channel network can
be generated by setting a threshold value for the flow accumulation area needed for channel
initiation (threshold used=20 000m²). The boundaries of the watershed are then defined as
the total area for which water flows into a selected pour point.
R-factor
3.3.1 Collecting relevant data
Long-term average monthly and annual R-values, determined following the EI30 approach
(equation 2 was applied for converting rainfall intensity to kinetic energy), are provided by
Ryumugabe and Berding (1992) for 5 stations covering the period 1980-1989. R-values (MJ
mm/hr ha yr) together with precipitation amount (mm) are presented in table 3.
Materials and Methods
32
Table 3: Rainfall (mm) and R (MJ mm/hr ha yr) for different stations in Rwanda (Ryumugabe and Berding, 1992)
Station Jan Feb Mar Apr May Jun July Aug Sep Oct Nov Dec total
Kigali Aeroport
Rainfall 61.8 93.0 92.9 183.4 82.2 13.0 5.0 27.6 69.4 100.9 120.6 79.4 929.2
R 256 525 368 842 393 77 26 158 375 441 572 228 4261
Kamembe Rainfall 123.5 120.3 157.8 144.3 80.2 27.6 6.4 43.8 99.7 144.8 170.6 110.6 1229.6
R 546 352 779 380 173 94 4 128 245 372 343 373 3789
Butare Rainfall 84.7 80.8 103.5 198.9 88.5 16 4.7 28.9 63.3 80.6 102.8 77 929.9
R 530 262 248 665 224 72 6 84 209 37.9 267 290 3236
Ruhengeri Rainfall 48.7 73.6 126.8 164.5 145.5 39.2 18.0 52.0 112.8 150.3 134.9 68.7 1135
R 96 127 337 426 420 147 59 139 267 470 444 173 3105
Gisenyi Rainfall 60.1 66.2 102.6 128.9 88.9 33.1 11.3 37.9 92.5 92.5 109.5 72.3 895.8
R 283 128 2710 550 346 95 0 117 344 181 321 99 2735
Average Rainfall 96.4 101.8 121.5 164.1 98.6 27.0 10.2 36.9 89.4 118.5 135.3 88.4 1088.1
R 497 450 444 571 339 113 25 157 31.2 426 406 269 4009
Daily rainfall records can be extracted from a database maintained by the Rwanda
Meteorology Agency under the ministry of Infrastructure (MININFRA). This database consists
of records from 200 locations in Rwanda, the observing stations are a mix of automatic
observers, recordings in the framework of specific scientific research and efforts from
volunteers. The timeframe for which records are available is highly dependent on the station,
some contain long-term data and go back to 1907, others registered daily rainfall for only a
couple of years. Data records for the period 1990 – 2000 are scarce. In this context the most
relevant station is Rwerere Colline, which is located in the watershed and has daily rainfall
records from 1962 until 1980. Figure 7 shows data extracted from the rainfall station present
is the watershed.
Materials and Methods
33
Figure 7: rainfall characteristics in watershed obtained from the daily rainfall station, Rwerere Colline
Two procedures are explored for establishing an R-value for the Tangata watershed. Firstly,
existing equations can be combined with the daily rainfall registered in the watershed. Table
3 can be used to select the equation with the best performance. Alternatively, attempts can
be made to link the measured R-values with the monthly rainfall precipitation also given in
table 3.
3.3.2 Evaluate existing or new regression equations
Determining a scorings factor for the equations was done by establishing a coefficient of
determination R², defined by equation 27.
SST = ∑(𝐸𝑗 − ��)²
𝑗
(25)
SSE = ∑(𝐸𝑗 − E𝑗)²
𝑗
(26)
𝑅2 = 1 −𝑆𝑆𝐸
𝑆𝑆𝑇
(27)
Where Ej are the measured rainfall erosivity values from Ryumugabe and Berding, �� is the
average of measured values and E𝑗 are the estimated rainfall values from precipitation data.
New regression equations are developed by minimizing R².
0
25
50
75
100
125
150
175
200
225
Mo
nth
ly p
reci
pit
ati
on
Precipitation records Rwerere Colline
General propertiesAltitude station: 2312mdays with rain: 51%days with precipitation >13mm: 8%
Average precipitation valuesMonthly = 103,31 mm/monthYearly = 1239,68 mm/year
Materials and Methods
34
For East Africa, Moore (1979) developed relationships between the KE>25 parameter
(Hudson, 1971) and mean annual rainfall for the coastal area, inland zones less and higher
than 1250m, and the Ugandan plateau based on rainfall records from stations in Kenya,
Tanzania and Uganda. These equations are summarized in table 4. A regression equation
developed from Kenyan data was furthermore used to convert KE>25 into R.
Table 4: Zones, regression equations and correlation coefficients from Moore (1979)
Equation R² # stations
Inland < 1250m R = 17.02 ∗ (0.029 ∗ (11.36 ∗ 𝑃𝑎 − 701) − 26) (28a) 0.962 7
Inland > 1250m R = 17.02 ∗ (0.029 ∗ (3.96 ∗ 𝑃𝑎 + 3122) − 26) (28b) 0.554 12
Uganda Plateau R = 17.02 ∗ (0.029 ∗ (16.58 ∗ 𝑃𝑎 − 6963) − 26) (28c) 0.918 11
Where R = rainfall erosivity (MJ mm/h ha yr) and Pa equals average annual precipitation (mm).
In 1986, SESA used these equations to estimate R within Rwanda, yet, by applying equation
28b to estimate the rainfall erosivity in Rwanda for altitudes lower than 1250m, and equation
28c for altitudes higher than 1250m.
Kassam et al. (1992) used following set of equations to estimate rainfall erosivity for Kenya.
𝑅 = 117.6 ∗ 1.00105𝑃𝑎 (Pa < 2000mm) (29a)
𝑅 = 0.5 ∗ 𝑃𝑎 (Pa > 2000mm) (29b)
Equation 29b is identical to the relationship found by Roose for West-Africa, and has proven
to perform poorly for stations in mountainous zones (Roose, 1976, 1977). Another commonly
used estimator for rainfall erosivity is the Modified Fournier Index (MFI) developed by
Arnoldus (1977):
𝑀𝐹𝐼 =∑ 𝑝𝑖
212𝑖=1
𝑃
(30)
Where pi is the average monthly precipitation, and P is average annual precipitation.
Materials and Methods
35
Vrieling et al. (2010) attempted to map rainfall erosivity for Africa and obtained a good
correlation (R²=0.84) using a linear relationship between MFI and R, expressed in equation 31.
(Kabanza et al., 2013)
𝑅 = 50.7 ∗ 𝑀𝐹𝐼 − 1405 (31)
K-factor
3.4.1 Soil maps: Rwanda Soil Information System
The soil information system of Rwanda is a computerized database containing spatial and
numerical data of Rwanda from a soil survey that started in 1981 and ended in 1994. The
spatial database consists of 43 semi-detailed soil map sheets at a scale of 1:50 000,
representing the hydrography, physiography, road network and administrative units.
Numerical data consist of about 2000 soil profiles with the related physico-chemical analytical
data (Imerzoukene and Van Ranst, 2001; Verdoodt and Van Ranst, 2006a). The Tangata
catchment is located on soil map sheet 3 (Kirambo). Soils are classified into soil series,
characterized by a unique combination of parent material, profile development, texture, and
soil depth (upland), or internal drainage (lowland). The soil mapping units are generally
associations of soil series. For each soil series, at least one soil profile has been described and
analyzed. Figure 8 shows parent material, soil texture & soil depth, soil classification name
according to Soil Taxonomy (Soil Survey Staff, 1975) and soil series name of the dominant soil
series for the Tangata watershed. The general protocol followed to establish a K-value is to
estimate a K-factor for each soil profile, disregarding the soil profiles taken in forestland or
savanne, and calculate an average K of the soil profiles that are grouped within each soil series.
The soil series present in the watersed are listed in table 5 together with the number of soil
profiles in the database belonging to that particular soil series. A description of each soil series
is included in Annex I. The table includes also information on parent material, dominant soil
texture, soil unit (FAO, 1990), average Dg as defined by equation 6, average OM, average %
coarse fragments and average fraction silt and very fine sand (0.002-0.1mm). Figure 9 shows
the distribution of clay, sand and silt for the soil series present is the watershed.
Materials and Methods
36
Figure 8: Parent material, texture and depth, soil classification name according to Soil Taxonomy and soil series
name of the dominant soil series present in the watershed
Materials and Methods
37
3.4.2 Soil series present in watershed
Table 5: Soil series present in watershed, no of profile description in database, parent material, texture, unit, OM and particle size information
Soil series No of
profiles parent material
dominant texture
soil unit (FAO, 1990) Dg
(mm) St.
dev. fsilt+vfSa
(%)
St. dev.
f>2mm (%)
St. dev.
OM (%)
St. dev.
AKAZI 11 Shale clay Dystric Regosols / Dystric Leptosols 0,0214 0,0097 43,7 19,5 37,2 25,2 4,1 4
BUJUMU 11 shale (sec. quartzite) sandy clay loam Dystric Regosols / Dystric Leptosols 0,0252 0,0145 48,6 17,4 37,15 29 4,6 3,4
CYARUGIRA 1 Alluvium clay Terric Histosols 0,0028 - 26,2 - 0 - 20 -
FUMBA 6 Shale (sec. quartzite) sandy clay Haplic (Humic) Acrisols 0,0207 0,0053 20,3 9,5 0 0 6,2 2,5
GIHIMBI 6 Quartzite (sec. shale) sandy clay loam Dystric Cambisols / Dystric
Regosols 0,0198 0,0138 48,9 15,6 27,19 23,4 4,1 1,7
GITABA 9 Shale silty clay Haplic Acrisols / Ferralic Cambisols 0,0147 0,0088 35,6 19,1 5,53 10,1 5,2 3,5
KABIRA 20 Quartzite (sec shale) clay Humic Acrisols (Sombric) 0,013 0,0079 33,7 15,1 9,87 16,6 3,7 1,5
KAYUMBU 12 Shale clay Humic Acrisols / Humic Ferralsols 0,0161 0,008 37,7 9,8 10,11 15,4 4,4 2,1
MUGOZI 8 Shale (sec. quartzite) mixed Humic Dystric Cambisols 0,0273 0,0193 44,1 12,9 30,84 24,8 4,5 2,8
MWOGO 13 Quartzite sandy loam Dystric Regosols / Dystric Leptosols 0,0478 0,0343 41,1 18 22,73 26,9 5,8 4,2
NSIBO 50 Shale clay Haplic Ferralsols / Haplic (Humic)
Acrisols 0,0165 0,0129 36,9 15,7 2,51 7,8 6,3 3,2
RUKO 6 Colluvium & alluvium mixed Dystric (Humic) Cambisols / Haplic
(Humic) Alisols 0,0481 0,0662 40,3 15,1 5,31 6,9 7,1 7,4
RUMULI 8 Alluvium clay loam Umbric Gleysols 0,022 0,0163 49,8 15 0 0 6,9 4,8
RUNABA 3 Shale (sec. quartzite) clay Humic (Dystric) Cambisols 0,0149 0,0098 29,8 6,1 10,34 15,3 7,6 2,4
RUTABO 2 Quartzite (sec.
Micaschist) sandy clay loam Humic Ferralsols / Umbric Regosol 0,026 0,0065 18,3 3,3 51,48 21,4 3,7 0,3
SHANGO 6 Shale sandy clay Humic Alisols / Humic Acrisols 0,0224 0,0342 33,5 10,4 48,07 31,3 8,1 4,5
Materials and Methods
38
3.4.3 K-factor estimation models
Since estimating soil erodibility factors for tropic soils is a delicate process, several approaches
are used. Firstly, the approaches discussed in RUSLE, i.e. the conventional nomograph and the
Dg-model are used. Measured K-values from Rwanda are not available, however the SESA
report estimates soil erodibility values based on parent material. Thirdly, a method
established for Kenia based on soil unit and texture is applied (Kassam et al., 1992) and lastly
minimal and maximal K-values are determined following a protocol based on Dg, OM and
coarse fragment fraction described in Borselli et al. (2012).
Figure 9: distribution of sand, clay and silt fractions in the profiles from the soil series present in the watershed
3.4.3.1 RUSLE: Dg-model & nomograph
Both the Dg-model (equation 7) and (simplified) nomograph will be applied on the dataset.
The granulometric data available in the soil profile database is presented in table 6. Following
the example of Declercq and Poesen (1992) only the properties of the upper 30cm are
withhold and a depth-weighted average is taken from the obtained K-values of each horizon
in the top soil.
Sand (%)
10% ---
20% ---
30% ---
40% ---
50% ---
60% ---
70% ---
80% ---
90% ---
Materials and Methods
39
Table 6: size classifications used in the soil profile database
name Clay fine silt Coarse silt Very fine
sand Fine sand
Medium sand Coarse sand
Very coarse sand
Boundaries (mm)
<0.002 0.002-0.02
0.02-0.05 0.05-0.1 0.1-
0.250 0.250-0.5 0.5-1 1-2
a) Nomograph
A simplified version of the nomograph approach is used in which the first K-value estimation
is withheld, without considering permeability and soil structure classes. The structure factor
included in equation 5 does not refer to the actual structure present at soil surface but to the
structure after 2 years of bare fallow, which makes it hard to estimate in this context. For
permeability no relationship connecting soil texture to permeability was found to have
confirmed validity for tropical soils, therefore also this parameter was excluded from the
calculations. A procedure consisting of four steps described in Auerswald et al. (2014), which
is fundamentally an expansion of equation 5, is followed to mimic the nomograph:
1) 𝐾1 = 2.77 ∗ 10−6 ∗ 𝑀1.14
𝐾1 = 1.75 ∗ 10−6 ∗ 𝑀1.14 + 0.00024 ∗ 𝑓𝑆𝑖+𝑣𝑓𝑆𝑎 + 0.016
For fSi+vfSa <70%
For fSi+vfSa >70%
2) 𝐾2 = (12 − 𝑓𝑂𝑀)/10
𝐾2 = 0.8
For fOM < 4%
For fOM > 4%
3) 𝐾3 = 𝐾1 ∗ 𝐾2
𝐾3 = 0.091 − 0.034 ∗ 𝐾1 ∗ 𝐾2 + 0.179 ∗ (𝐾1 ∗ 𝐾2)² + 0.048 ∗ 𝐾1 ∗ 𝐾2
For K1 * K2 > 0.02
For K1 * K2 < 0.02
4) 𝐾 = 𝐾3
𝐾 = 𝐾3 ∗ (1.1 ∗ exp(−0.024 ∗ 𝑓𝑟𝑓) − 0.06)
For frf <1.5%
For frf >1.5%
Where K is the soil erodibility value (t ha h/ha MJ mm), M represents percentage silt and very
fine sand multiplied with percentage silt and total sand, fSi+vfSA equals the percentage silt and
very find sand , fom is the %OM and frf is the percentage coarse fragments. 1.72 is used as a
conversion factor to calculate soil organic matter from organic carbon percentage.
Materials and Methods
40
b) Dg-model
Equations 6 and 7 are applied to estimate K-values according to the Dg-model. The Dg-model
discussed in RUSLE was calibrated only considering soils with less than 10% rock fragments
(>2mm, weight percentage). As 20% of the horizons analyzed contained rock fragments
exceeding 10%, a similar correction as in Mati et al. (2000) for the protective effects of coarse
fragments in topsoil is included, based on the equation described by van den Berg (1992):
𝐹𝑐𝑜𝑎𝑟 = 1.026 − 0.025 ∗ 𝑓𝑟𝑓 + 2.534 ∗ 10−4 ∗ 𝑓𝑟𝑓2 − 1.02 ∗ 10−6 ∗ 𝑓𝑟𝑓³ (32)
Where Fcoar is the correction factor to be multiplied with the uncorrected K and COAR is the
weight fraction of coarse fragments in the first horizon. The Dg-model defined by equation 7
together with a histogram of the Dg-values retrieved from the soil profile analyses is displayed
in figure 10.
Materials and Methods
41
Figure 10: Histogram of Dg-values in dataset together with the model that estimates K from Dg described in
Renard et al. (1997)
3.4.3.2 K-value based on parent material
SESA (1986) used parent material as a soil erodibility indicator and assigned K-values according
to table 7. Contrasting to the other approaches discussed, these estimations were specifically
made for Rwanda.
Table 7: Estimated K-values based on parent material (SESA, 1986)
Parent material K-value (t ha h/ha MJ mm)
Alluvial material 0.02634
Basaltic rocks 0.015804
Colluvium material 0.028974
Granite 0.02634
Volcanic rock 0.015804
Quartzite & secondary influence of shale 0.03706
Shale and sec. influence of quartzite 0.019755
0
2
4
6
8
10
12
14
0
0,005
0,01
0,015
0,02
0,025
0,03
0,035
0,04
0,045
0,05
-3 -2,5 -2 -1,5 -1 -0,5 0
K-v
alu
e (t
ha
h h
a-1
MJ-1
mm
-1)
log (Dg)
Histogram and Dg-model
Materials and Methods
42
3.4.3.3 Kassam et al. (1992): Taxonomy and texture
Kassam et al. (1992) developed a method for estimating K-values for soils in Kenya.
Observations from the nomograph were modified to account for the behavior of groups of
soils which appeared to be more erodible than the nomograph indicates. To extrapolate the
collected K-values to data scarce regions, soil unit (FAO, 1990) and soil texture were used as
erosivity estimators. Each combination connects with a certain soil erodibility range. Table 11
and 12, included in Annex II, depict the estimation protocol. If the percentage of stone and
gravel present exceeds 25% or 50%, the obtained K-values were multiplied with 0.7 and 0.4,
respectively.
The texture and soil unit that characterizes each soil series present in the watershed can be
used as inputs for the approach of Kassam et al. (1992). In this context the database of soil
profile descriptions includes remarks on stone cover. For most soil series no stones were
observed, however for Bujumu, Rutabo and Shango a clear majority of soil profile descriptions
indicated the presence of stones, so a correction factor of 0.7 was applied. For Shango the
observations even noted a stone cover presence exceeding 60% so a correction factor of 0.4
was applied.
3.4.3.4 Determining minimal and maximal values
As already discussed in paragraph 2.4.2.3, Torri et al. (1997) developed a methodology based
on a global dataset for estimating minimal and maximal K-values centered around accessible
and universal textural parameters such as Dg, OM, rock fragment cover and clay content.
Borselli et al. (2012) copied the objectives and database described in Torri et al. (1997) but
divided the global dataset in climatic subcategories according to findings specified in Salvador
Sanchis et al. (2008). Soils corresponding to cool climates (conferring to the Köppen–Geiger
climate classification) are separated from soils located in warm climates, which are
characterized with lower erodibility values. Additional to climate, rock fragment cover (frf) was
also used for subset delineation: within each climate a threshold value of 10% rock fragment
content by mass is applied which organizes the global dataset into four different subsets.
Materials and Methods
43
Four soil parameters are considered: Dg, OM, percentage rock fragment content (frf) and the
logarithm of geometric standard deviation of Dg (Sg). Unlike RUSLE the particle size of each
class used for calculating Dg isn’t represented by its arithmetic mean diameter but by equation
33.
𝐷𝑔,𝐵𝑜𝑟𝑠𝑒𝑙𝑙𝑖 = ∑ 𝑓𝑖log10 (√𝑑𝑖𝑑𝑖−1)
𝑖
(33)
Sg is defined as:
𝑆𝑔 = √∑ 𝑓𝑖[𝑙𝑜𝑔10√𝑑𝑖𝑑𝑖−1 − 𝐷𝑔]²
𝑖
(34)
Essentially, the algorithm compares the given input variables to the existing measured K-
values present in the subsets and creates a cumulative distribution function (CDF) which
estimates probable values of soil erodibility. The regression functions are described in Borselli
et al. (2012) and the methodology is implemented in a program called KUERY. An example of
KUERY output for three soil profiles from three different soil series is given in figure 11.
Materials and Methods
44
Figure 11: Estimated K-probable values for different soil profiles according to Borselli et al. (2012)
In this context, the soil granulometric data from each soil profile, together with rock fragment
content and OM%, are fed to the program and an average maximal and median value are
calculated for each soil series.
3.4.4 Transformation from soil series to soil units
The soil mapping units present in the watershed are associations of different soil series. For a
proper interpretation of the mapping units, the distinction should be made between
associated soil series and inclusions. Both account for the occupation of several soil series in
one mapping unit, however associations indicate a much more equally proportioned presence
between a dominant and secondary soil series whereas inclusions signify an occurrence of less
than 10% for the given soil series in the soil mapping unit. To transform the K-values obtained
for the soil series into one K-factor for every soil unit, a weighted average is taken depending
on the amount and type of soil series present. For example, the K-value of a soil unit complex
consisting of two associated series Akazi (dominant soil series) & Mwogo (secondary soil
0
0,02
0,04
0,06
0,08
0,1
0,12
0,14
0
0,1
0,2
0,3
0,4
0,5
0,6
0,7
0,8
0,9
1
0 0,005 0,01 0,015 0,02 0,025
PD
F (d
ott
ed l
ine)
CD
F (f
ull
lin
e)
K (ton ha hr/ha MJ mm)
Probability fuctions produced by KUERY
profile with Dg (eq.33) = -1.86, Sg (eq.34) = 1.32, Organic Matter = 5.66%, coarse fragments = 0%
(catogorized with Nsibo soil series)
profile with Dg (eq.33) = -1.85, Sg (eq.34) = 1.40, Organic Matter = 3.5%, coarse fragments = 30,9%
(catogorized with Rutabo soil series)
profile with Dg (eq.33) = -2.304, Sg (eq.34) = 1.362, Organic Matter = 7.2%, coarse fragments = 0%
(catogorized with Kayumbu series)
Materials and Methods
45
series) and one inclusion Gihimbi will be calculated as: Kunit = 0.55*KAKAZI + 0.35*KMWOGO+
0.1*KGIHIMBI.
LS-factor
Slope and slope length information are extracted separately rather than use integrated terrain
and drainage metrics that lump these two parameters together. This way each factor can be
assessed independently. Similar as for watershed delineation, a sink removal step is executed
previous to all hydraulic related processing steps. The slope of each pixel is determined by
taking the maximal downhill slope i.e. divide height difference by horizontal distance for all
neighboring cells and keep the maximal value. The formula established by Nearing (1997),
equation 21, is used to derive the S-factor from the slope angle.
Slope length was determined with two different methods discussed in section 2.4.3.3. Firstly
the unit contributing area was calculated and used in the unit stream appraoch (formula 24).
Uplsope flow area was derived with the multiple flow direction algorithm developed in
Freeman (1991). Alternatively, a cumulative grid-based algorithm implemented in SAGA-GIS
was executed. The cutoff slope coefficient used was 0.5 (Claessens et al., 2008). The
transformation of slope length λ to L-factor was done by the equation established by Liu et al.
(2000):
𝐿 = (𝜆/22.1 )0.44 (35)
Results & Discussion
46
Results & Discussion
Watershed delineation
The maps produced during the delineation protocol and a 3D model of the watershed are
shown in figure 12 and 13. The area of the watershed is 13.6 km².
Figure 12: different stages during the delineation of the Tangata watershed
Results & Discussion
47
Figure 13: 3D model of Tangata watershed with satellite images from Google earth including the generated channel network, pour point used and the location of the weather station
daily rainfall station
Results & Discussion
48
R-factor
4.2.1 Estimations based on annual precipitation and MFI
Figure 14 & 15 show the measured R-values using the EI30 procedure together with the
estimates following the approaches discussed in paragraph 3.3.2. All five stations referred to
in Ryumugabe and Berding (1992) are located on an altitude exceeding 1250m. Regardless of
the estimation parameter used (annual precipitation or MFI), Kigali deviates from the general
trend and inhibits the creation of a linear equation with a satisfying R² value. Based on this
data, the equation developed by Moore for inland stations seems most appropriate for
estimating R-values, which has a R² value of 0.74 if Kigali is left out of consideration.
Figure 14: R-values from Ryumugabe and Berding (1992) together with regression equations based on annual
rainfall
KigaliKamembe
Butare
RuhengeriGisenyi
2000
2500
3000
3500
4000
4500
5000
5500
6000
6500
7000
800 850 900 950 1000 1050 1100 1150 1200 1250 1300
R (
MJ
mm
/h h
a yr
)
Pa (mm)
Evaluation regression equations based on Pa
Kassam et al. (1992)
Moore (1979) Inland < 1250m
Moore (1979) Uganda Plateau
Moore (1979) Inland > 1250m
Results & Discussion
49
Figure 15:Measured values from Ryumugabe and Berding (1992) together with the rainfall erosivity estimation
approach based on MFI discussed in Vrieling et al. (2010)
Similar to Yu and Rosewell (1996b), the conclusion can be drawn that the MFI is not inherently
superior to annual rainfall as a rainfall erosivity estimator. Rather on the contrary, determining
MFI values requires more data and its values are climatologically harder to interpret. Plotting
the measured R-factor as a function of annual rainfall also demonstrates the inland variability
of rainstorms in terms of size and intensity: some stations have similar annual precipitation
but a significant higher R-factor and vice versa. Even for a country as small as Rwanda, the
topography and related climatic variability doesn’t allow the establishment of a nationwide
relationship between mean annual precipitation (or MFI) and rainfall erosivity, which is
somewhat in line with the comments made in Wischmeier and Smith (1978): “Where
adequate rainfall intensity data aren’t available, the erosion index cannot be estimated solely
from annual precipitation data (…) a given annual rainfall will indicate only a broad range of
possible values of the local erosion index”.
4.2.2 Calibrating own equations
Next to evaluating existing approaches, new equations can be calibrated. If monthly rainfall
amount and R-values are plotted (figure 16), a much more linear correlation is observed.
Kigali
Kamembe
ButareRuhengeri
Gisenyi
1500
2000
2500
3000
3500
4000
4500
5000
5500
6000
80 90 100 110 120 130 140
R (
MJ
mm
/h h
a y
r)
MFI
Evaluation MFI and R
Results & Discussion
50
Regression equations are calculated based on all data (dotted line) and for two stations
individually: Kigali (the outlier in previous plots) and Ruhengeri (the station closest to the
Tangata watershed). Equation 36 links the monthly R-factor (Rm) values with monthly
precipitation (Pm) for Ruhengeri.
𝑅𝑚 = 2.9356 ∗ 𝑃𝑚 − 18.91 (36)
The R² value improves greatly if only one station is considered, which relates to the national
variation in rainstorm properties discussed earlier: the steeper slope for the line connecting
the points for Kigali demonstrates that for the same rainfall amount, rain erosivity is much
higher for Kigali compared to Ruhengeri or Kamembe.
Figure 16: monthly R-values and monthly rainfall amount
4.2.3 Final R-value for watershed
When equation 36 is used on the monthly precipitation data available for the watershed a
final value of 3412 MJ mm hr-1 ha-1 yr-1 is obtained. Equation 28b described in Moore (1979)
returns an erosivity value of 3521 MJ mm hr-1 ha-1 yr-1. The estimated erosivity for the Tangata
watershed corresponds to the erosivity values calculated for central Hawaii or New York in
0
100
200
300
400
500
600
700
800
900
1000
0 20 40 60 80 100 120 140 160 180 200
R-f
acto
r (M
J m
m/h
ha
mo
nth
)
Monthly rainfall (mm)
Monthly rainfall and R-factorKigali
Kamembe
Butare
Ruhengeri
Gisenyi
Results & Discussion
51
the original USLE handbook (Wischmeier and Smith, 1978) and is four times the R-factor
established for Flanders (Notebaert et al., 2006). It’s about half the value measured for Central
Kenia (Angima et al., 2003) or estimated for South Eastern Tanzania (Kabanza et al., 2013). For
Eastern Uganda, Jiang et al. (2014) estimated rainfall erosivity in the range of 900 to 2800 MJ
mm ha-1 hr-1 yr-1. The measured values presented in Vrieling et al. (2010) acknowledge that an
erosivity value of 3521 MJ mm hr-1 ha-1 yr-1 is common for Sub-Saharan Africa.
The rainfall records from the watershed (period: 1960-1982), the equations discussed in
Moore (1979) and the values presented in Ryumugabe and Berding (1992) all relate to the
period before 1990 and could be outdated. The available meteorological data suggests a
declining trend in recent precipitation for Rwanda (Habiyaremye et al., 2012). However not
enough records exist to evaluate changes in heavy rainfall events since 1990 (McSweeny,
2011). The quantity and quality of observed data, see next paragraph, limits the capacity to
draw firm conclusions concerning rainfall evolutions or the applicability of rainfall records
measured three decades ago.
4.2.4 Data reliability
A comparison of the rainfall data presented in Ryumugabe and Berding (1992) and the daily
rainfall database provided by MINAGRI reveals that the numbers don’t always comply. Table
8 compares neighboring stations from the two consulted datasets. Strikingly, stations
assumed to be identical such as Kigali airport are stored with different geographic coordinates
and contrasting precipitation records in both datasets. This nonconformity is a major
complication. Obviously, the protocol of establishing and extrapolating a relationship between
R and monthly rainfall assumes matching procedures for data collection, registration and
conservation.
Results & Discussion
52
Table 8: Comparison of stations with R-values to stations with daily rainfall data
Station Long Latitude Altitude
annual rainfall (mm)
Δx (km)
R-value Kigali aero 1°58’S 30°8’E 1480 929.2 3.3
Daily rainfall Kigali aero 1°57’S 30°6’E 1490 1048.12
R-value Kamembe 2°28’S 28°54’E 1591 1229.6 11.7
Daily rainfall Mibirizi 2°34’S 28°57’E 1750 1410.17
R-value Butare 2°36’S 29°44’E 1768 929.9 2.7
Daily rainfall Butare aero 2°22’S 29°10’E 1760 1261.04
R-value Ruhengeri 1°30’S 29°38’E 1878 1135 3.3
Daily rainfall Ruhengeri aero 1°28’S 29°23’E 1878 1227.74
R-value Gisenyi 1°40’S 29°15’E 1554 895.8 1
Daily rainfall Gisenyi aero 1°24’S 29°9’E 1554 1169.67
Analyzing rainfall data from other sources such as the climatologic database included in the
Rwanda soil information system could only uncover a further lack of consistency between
different datasets. Since one of the objectives of this thesis includes an assessment on data
reliability, the incoherent rainfall records are a conclusion in itself.
K-factor
4.3.1 SESA approach based on parent material
The approach based on the SESA-report assigns similar K-values to soil series with a
completely different texture (Mwogo vs Shango or Akazi), which unveils that this approach
not truly meets the objective of determining accurately spatially varied soil erodibility values,
but serves more as a provider of guideline values. For most soil series the parent material
consisted of shale, for which a value of K-value around 0.02 ton ha hr/ha MJ mm was
estimated.
Figure 17 displays estimated K-values for every soil series for the other approaches together
with the standard deviation for the approaches based on averages of multiple soil profiles.
General properties for each soil series were listed on page 37. The number of soil profiles
analyzed for each series is indicated at the bottom of the graphs. No measured K-values are
available for Rwanda so the different soil erodibility estimations can only be considered
relative to each other.
Results & Discussion
53
Figure 17: K-factor estimates for different approaches
4.3.2 RUSLE approaches: nomograph vs Dg-model
The estimated values with the Dg-model are consistently larger than the estimations with the
nomograph and surpass in most cases the average maximum values estimated by the protocol
described in Borselli et al. (2012) . Despite the fact that the dataset for which the nomograph
was developed contained solely American soils with significant higher K-values averaging
around 0.05 ton ha h/ha MJ mm (Renard et al., 1997; Zhang et al., 2016), the nomograph
generally complies better with the range of K-values locally estimated in the SESA report or
the range established with Borselli et al. (2012). This observation brands the nomograph as
the preferred RUSLE option in this context. This doesn’t necessarily prove the superiority of
the nomograph approach over a Dg-model: Wang et al. (2012) and Declercq and Poesen
(1992) observed better estimations with the latter model. Wang et al. (2012) observed higher
estimations for the nomograph compared to a Dg-model. An important distinction however
with the results displayed in figure 17 is that in both cases the equation relating Dg to K was
developed based on a more local dataset.
11 11 1 6 6 9 20 12 8 13 50 6 8 3 2 60
0,01
0,02
0,03
0,04
0,05
0
0,01
0,02
0,03
0,04
0,05
K-fa
ctor
(ton
ha
hr/h
a M
J m
m)
Results of different approaches for estimating soil erodibility
Kassam et al. (1992) RUSLE: Dg-model RUSLE: Nomograph Median Borselli et al (2012) Range Borselli et al. (2012)
Results & Discussion
54
4.3.3 Nomograph approach vs algorithm Borselli et al. (2012)
There is no general trend concerning the magnitude of the estimated values when the
nomograph is compared to the median values extracted from the algorithm described in
Borselli et al. (2012). However the standard deviation within soil series does contrast for the
two approaches. This opens the exploration of alternative way of scoring the approaches, i.e.
by performing an analysis of variance (ANOVA) test. The criteria used for grouping soils into
soil series (parent material, profile development, texture and soil depth) aren’t centered
around topsoil conditions, however still the listed properties influence to some extent soil
erodibility. This implies that average calculated K-value for at least some series (μseries) must
contrast. Consequently the performance of both estimations can be assessed by studying the
variance within and between soil series. The null hypothesis to be challenged by both
approaches is that the average erodibility of each soil series is equal, i.e. μAkazi = μBujumu =
μCyarugira = … = μShango. The alternative hypothesis is that at least one soil series has a different
average erodibility. The obtained p-values after performing an ANOVA test are 0.092 for the
nomograph and 0.000485 for Borselli et al. (2012). On a 5% significance level the nomograph
approach doesn’t reject the null hypothesis that all soil series have equal erodibility. This
ANOVA test doesn’t automatically imply that Borselli et al. (2012) provides better estimates
than the nomograph for tropic soils, it’s merely an indicator that the nomograph estimations
on this database are more robust and don’t cover minor differences between soil series when
the soil profiles are analyzed. The methodology established in Borselli et al. (2012) was
designed to estimate probable values of soil erodibility as shown in figure 11, extracting a
median value to estimate erodible is a too simplistic use of the algorithms. Therefore, the
values obtained from the nomograph are withhold for mapping potential erosion. However
with some corrections.
4.3.4 Corrections to global erodibility models
If the global estimation approaches based on mostly textural information are compared to the
classification system developed for Kenia (Kassam et al., 1992), the erodibility of some soil
groups seems to be in a different range. Specifically the erodibility values estimated by Kassam
et al. (1992) for clayey acrisols and ferralsols (Kabira, Kayumbu, Nsibo) don’t seem to follow
Results & Discussion
55
the universal erodibility equations. On clayey Acric Ferralsols in South Eastern Tanzania an
erodibility value of 0.009 ton ha hr ha-1 MJ-1 mm-1 was measured (Kabanza et al., 2013).
Considering that these soils seem to correspond to the third category in the classification
system described by Nill et al. (1996), see paragraph 2.4.2.3 on page 17, a soil erodibility value
around 0.01 ton ha hr ha-1 MJ-1 mm-1 is perceived as more realistic. The estimates made by the
nomograph for Rutabo and Shango are low due to the high percentage of coarse fragments
(around 50%). The correction factor for stoniness added for both RUSLE approaches, discussed
on pages 39 and 40, is based on figure 6 page 19 of the original USLE guidebook (Wischmeier
and Smith, 1978). Coarse fragments are accounted the same way effect as mulch or canopy
cover. The data driven algorithm described in Borselli et al. (2012) does show lower output
values for soil profiles with high coarse fragments percentages, however the reducing effect
is not as strong as described by equation 32 or the 4th equation from Auerswald et al. (2014).
4.3.5 Soil erodibility map
The K-values estimated for each soil serie are based on the values obtained from the
nomograph with exceptions for Kabira, Kayumbu and Nsibo, for which an erodibility value of
0.01 ton ha hr ha-1 MJ-1 mm-1 is estimated. For Rutabo and Shango the median value obtained
from the algorithm described in Borselli et al. (2012) is withheld since the high stoniness seems
to result in underestimations when nomograph and adjacent correction formula is applied for
those soil series. The final soil erodibility values are shown in table 9, the soil erodibility map
after transformation to soil units is given in figure 18.
Results & Discussion
56
Table 9: final K-values estimated for soil series present in Tangata catchment
Soil series K-value (ton ha hr ha-1
MJ-1 mm-1)
Akazi 0,0151
Bujumu 0,0214
Cyarugira 0,0099
Fumba 0,0117
Gihimbi 0,0175
Gitaba 0,0159
Kabira 0,01
Kayumbu 0,01
Mugozi 0,0133
Mwogo 0,0167
Nsibo 0,01
Ruko 0,0194
Rumuli 0,0262
Runaba 0,0103
Rutabo 0,0093
Shango 0,014
The estimated erodibility factor hovers around 0.01-0.02 ton ha hr ha-1 MJ-1 mm-1 and is
strikingly low compared to K-factors registered from other continents (Torri et al., 1997). The
estimated values are in line with measured values for tropical conditions (Nill et al., 1996;
Wang et al., 2012) and recent estimations made for East-Africa (table 10).
Results & Discussion
57
Table 10: Recent published K-value estimations for East Africa
Estimated K-value range (ton ha hr ha-1 MJ-1 mm-1)
Region reference
0.016 Central Kenyan higlands Angima et al. (2003) 0.009 & 0.014* South Eastern Tanzania Kabanza et al. (2013)
0.019-0.04 Western Kenya Cohen et al. (2005) 0.02-0.036 Uganda Jiang et al. (2014)
0.013-0.035 Southern Kenya Mati et al. (2000)
*measured values
Figure 18: soil erodibility for Tangata watershed
LS-factor
4.4.1 Comparison with SESA report
The generated slope gradient and slope length (grid-based method) for the watershed are
presented in figure 19. SESA (1986) published results regarding slope length and steepness
from a measuring campaign involving 9.662 agricultural fields spread throughout Rwanda.
Results & Discussion
58
Slope steepness was averagely 13.2 degrees, with maxima exceeding 45 degrees. Measured
slope length varied between 2 and 226m, with an average of 24m.
The GIS method applied on the watershed calculated a slope gradient varying from 0 to 38
degrees, with an average of 19 degrees and a standard deviation of 8.76. Field surveys in the
area demonstrated that the maximal slopes in the watershed can exceed 50 degrees, which
indicates an underestimation of the calculated gradient. The 30m resolution of the DEM
smoothens to some extent the topographic reality, detailed topographic information is lost
during the rasterization process. However, in the context of mapping potential soil erosion
risk, figure 4 revealed that a small underestimation of slope steepness doesn’t drastically
impact the eventual S-factor.
Figure 19: Calculated slope gradient and slope length (from the grid based method) of Tangata catchment
GIS-methods applied in our watershed obtained higher slope lengths, a histogram with the
obtained results for 0 to 300m is shown in figure 20, note that the Y-values for the UCA
histogram are 10 times smaller and that there is an excessively long tale. 30% of the calculated
values are larger than 300m. To maxima in the elevation map, the UCA-method assigns a value
equal to the cell resolution, which explains why this method doesn’t produce values lower
than 30 m²/m. The incapacity of this method to identify deposition areas proves to be a big
limitation. For long hills with variable downward slope the algorithm keeps summing up
Results & Discussion
59
contributing flow cells, which results in extreme outliers (the maximal UCA-value is around
420 000 m²/m). RUSLE is designed to estimate average yearly soil loss as a result of rill and
sheet erosion, if slope lengths exceeding 300m do occur, the erosion type changes and the
empirical formula’s developed in RUSLE don’t apply, consequently the obtained results have
no fundamental significance. The cumulative grid-based method has an average slope length
of 132m. Since the method is based on a single flow algorithm the calculated cell-length values
are often multiples of the cell resolution, or 1.4 times the cell resolution.
Figure 20: distribution of slope length factors
The obtained average slope length (132m) is still a lot higher than the numbers mentioned in
the SESA report (24m), or other articles that estimated the average parcel length for Rwanda
around 20m (König, 1994). The higher values obtained can be attributed to several factors.
Firstly, the interpretation of slope length is completely different when field measurements are
compared to GIS-technologies, especially when no parcel map is available. GIS doesn’t
consider parcel boundaries, hedges, ridges, drains, roads, houses, walking paths or other
micro-relief related features that normally could block water flow. Since Rwanda is dominated
by small family farms, the average farm size is 0.8ha (Bizimana et al., 2004), these features are
important. Secondly, slope length is only reset to zero if deposition occurs, the GIS-method
doesn’t recognize concentrated flow through channels which results in a overestimation of
the actual slope length. Lastly, as discussed in Desmet and Govers (1996) manual slope length
0
0,005
0,01
0,015
0,02
0,025
0
0,05
0,1
0,15
0,2
0,25
0 10
20
30
40
50
60
70
80
90
10
0
11
0
12
0
13
0
14
0
15
0
16
0
17
0
18
0
19
0
20
0
21
0
22
0
23
0
24
0
25
0
26
0
27
0
28
0
29
0
30
0
Occ
ura
nce
UC
A v
alu
es
Occ
ura
nce
gri
d-b
ase
d m
eth
od
slope length/unit contributing area (m)
Histogram of slope length factors
cumulative grid-based method
unit contributing area
Results & Discussion
60
determination doesn’t consider convergence which can result in significantly lower values
when compared to GIS techniques.
4.4.2 LS-factor results
Figure 21 maps the LS-factor in the watershed.
Figure 21: LS-factor for watershed
The mean LS-value is 13, with a standard deviation of 14. The maximal value calculated is 98,
90% of the obtained values are lower than 30. These values are high but not exceptional. For
a catchment situated in Ugandan highlands, LS-values varied from 0 to 184 or from 0 to 95
depending on the applied method (Jiang et al., 2014). Other mountainous catchment show LS
values up to 300 (Millward and Mersey, 1999), 53 (Dabral et al., 2008) or 109 (Gelagay and
Minale, 2016). High LS-values relate much more to steep slopes compared than to high slope
lengths. The average L-factor is 1.7, with a maximum value of 6.9. For the S-factor, the
maximum value is 15, with an average of 6.8, proving that S has a bigger impact. This is in line
with the comments made in Renard et al. (1991), where it is pointed out that the attention
Results & Discussion
61
given to the L factor is not always warranted since soil loss is less sensitive to slope length
that to any other factor. The frequently cited statement that a 10% error in slope length results
in a 5% error in computed soil loss whereas a 10% error in slope steepness results in a 20%
error in computed soil loss (Renard et al., 1991) proves to have some validity, see figure 22.
However, caution must be always be taken since figure 22 also shows that high slope length
values can have a significant impact on the final outcome of the RUSLE model and possibly
result in an overestimation of the erosion risk.
Figure 22: relative increase of computed erosion risk with increasing slope values
For the Tangata catchment, both factor have a stabilizing effect towards each other, i.e.
large slope lengths occur mostly in combination with low slope gradients, whereas steep
slopes are in practice often shortened in length with the installation of hedges or other
interventions preventing an exceedingly high slope length.
Potential erosion risk map for Tangata watershed
The potential erosion risk map is obtained by multiplying the final values of the calculated
factors. Since the rainfall erosivity is a fixed constant for the catchment, the spatial variance
of the obtained map follows very much the trends established by the slope map. The average
potential erosion risk value is 593 ton ha-1 yr-1 but the values are highly variable (st. dev. is
682). If a soil tolerance value of 15 ton ha-1 yr-1 is assumed, a short estimate shows that the
product of the other factors (C & P) have to be around 0.02. Measured C-values for different
crops in Rwanda go from 0.02 (Coffee) over 0.22 (potatoes) up to 0.45 (Sorghum) indicating
that for most areas extra conservation strategies are necessary to prevent soil degradation.
104% 114%142% 163%
315%
219%
slope length slope gradient
Re
lati
ve in
cre
ase
of p
ote
tial
so
il ri
sk Sensitivity of output to changes in slope gradient
and slope length
average value + 10% average value + one stand. dev maximal value
Results & Discussion
62
As stated before the potential erosion risk is highly spatially variable so the right crop choice
and conservation strategy will be location specific and is out of the scope of this dissertation.
Figure 23: Final potential soil erosion risk map for Tangata watershed
Conclusions
63
Conclusions
This dissertation explores the best approach to estimate rainfall, soil and topographic factors
needed for the application of RUSLE in Rwanda. It refines estimate approaches made in SESA
rapport, pinpoints fields for improvement and provides values each factor. The most suitable
methodology for estimating rainfall erosivity in Northern Rwanda is by utilizing the equations
described in Moore (1979). Only yearly precipitation is required as erosivity estimator and it
produces satisfactory results when compared to most of the measured values mentioned in
Ryumugabe and Berding (1992). However, care must be taken when extrapolating the robust
equations to more eastern zones in Rwanda, the subset delineation based on elevation
doesn’t seem to fully cover the inland climatic variability of Rwanda which may lead to an
underestimation of erosivity values for less mountainous zones. Considering rainfall data a
thorough evaluation on the reliability of all available data is required since different data
sources produce contrasting precipitation records. In general the available equations and
measured erosivity values do have an outdated character.
Considering soil erodibility the nomograph produces lower and more realistic values
compared to the Dg-model. More measured values for tropical soils in central or eastern Africa
are needed to improve estimations made by the nomograph. Data gathered in Nill et al.
(1996), Wang et al. (2012) and El-Swaify and Dangler (1976) already prove that the nomograph
approach has limited applicability for tropical soils and that modifications or alternative
equations are needed. Data provided in Kabanza et al. (2013) indicate that clayey Acric
Ferralsols seem to have lower erodibility than predicted by the nomograph. A recent protocol
for estimating soil erodibility discussed in Borselli et al. (2012) produced promising results,
different than the other universal approaches the estimation protocol does consider climate.
In this context the algorithm was used to estimate erodibility for soil profiles characterized
with high coarse fragment fractions. However also this data driven approach lacks
fundamental information on east-African soils to be fully considered as reliable.
Conclusions
64
Using GIS technologies to estimate the topographic factor results in higher slope length factors
, often exceeding 300m. Especially methodologies that don’t identify deposition areas
produce excessively large slope length values. When the potential soil erosion risk is mapped,
it follows mostly the topography, steep slopes have the highest erosion risk. The obtained
values indicate a need for conservation strategies.
References
65
References
Al-Durrah, M.M., Bradford, J.M., 1982. The Mechanism of Raindrop Splash on Soil Surfaces. Soil Science Society of America Journal 46, 1086-1090.
Amézketa, E., 1999. Soil Aggregate Stability: A Review. Journal of Sustainable Agriculture 14, 83-151.
Angima, S.D., Stott, D.E., O’Neill, M.K., Ong, C.K., Weesies, G.A., 2003. Soil erosion prediction using RUSLE for central Kenyan highland conditions. Agriculture, Ecosystems & Environment 97, 295-308.
Angulo-Martínez, M., Beguería, S., 2009. Estimating rainfall erosivity from daily precipitation records: A comparison among methods using data from the Ebro Basin (NE Spain). Journal of Hydrology 379, 111-121.
Aranda, V., Oyonarte, C., 2005. Effect of vegetation with different evolution degree on soil organic matter in a semi-arid environment (Cabo de Gata-Níjar Natural Park, SE Spain). Journal of Arid Environments 62, 631-647.
Arnoldus, H.M.J., 1977. Methodolgy used to determine the maximum potential average annual soil loss due to sheet and rill erosion in Morocco. FAO SOil Bulletins 34, 39-51.
Arnoldus, H.M.J., 1980. An approximation of the rainfall factor in the Universal Soil Loss equation Wiley, Chichester, UK.
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 determining USLE nomograph soil erodibility values. Catena 118, 220-225.
Barthès, B., Albrecht, A., Asseline, J., De Noni, G., Roose, E., 1999. Relationship Between Soil Erodibility and Topsoil Aggregate Stability or Carbon Content in a Cultivated Mediterranean Highland (Aveyron, France). Communications in Soil Science and Plant Analysis 30, 1929-1938.
Beven, K., 1989. Changing ideas in hydrology - the case of physically-based models. Journal of Hydrology 105, 157-172.
Bewket, W., Teferi, E., 2009. Assessment of soil erosion hazard and prioritization for treatment at the watershed level: Case study in the Chemoga watershed, Blue Nile basin, Ethiopia. Land Degradation & Development 20, 609-622.
Bidogeza, J.C., Berentsen, P.B.M., De Graaff, J., Oude Lansink, A.G.J.M., 2009. A typology of farm households for the Umutara Province in Rwanda. Food Security 1, 321-335.
References
66
Bizimana, C., Nieuwoudt, W.L., Ferrer, S.R.D., 2004. Farm size, land fragmentation and economic efficiency in Southern Rwanda. Agrekon 43, 244-262.
Bizoza, A.R., 2014. Three-Stage Analysis of the Adoption of Soil and Water Conservation in the Highlands of Rwanda. Land Degradation & Development 25, 360-372.
Bizoza, A.R., De Graaff, J., 2012. Financial cost-benefit analysis of bench terraces in Rwanda. Land Degradation & Development 23, 103-115.
Board, R.D., 2016. Agriculture Overview.
Bochet, E., Rubio, J.L., Poesen, J., 1998. Relative efficiency of three representative matorral species in reducing water erosion at the microscale in a semi-arid climate (Valencia, Spain). Geomorphology 23, 139-150.
Borselli, L., Torri, D., Poesen, J., Iaquinta, P., 2012. A robust algorithm for estimating probable values of soil erodibility in different climates. Catena 97, 85-94.
Bradford, J.M., Truman, C.C., Huang, C., 1992. Comparison of Three Measures of Resistance of Soil Surface Seals to Raindrop Splash. Soil technology 5, 47-56.
Brandt, S.C.J., 1989. The size distribution of throughfall drops under vegetation canopies. Catena 16, 507-524.
Bronick, C.J., Lal, R., 2005. Soil structure and management: a review. Geoderma 124, 3-22.
Brown, L.C., Foster, G.R., 1987. Storm erosivity idealized intenstiy distributions. Transactions of the ASAE 30, 379-386.
Bryan, R.B., 1968. The development, use and efficiency of indices of soil erodibility. Geoderma 2, 5-26.
Bryan, R.B., 2000. Soil erodibility and processes of water erosion on hillslope. Geomorphology 32, 385-415.
Bryan, R.B., Govers, G., Poesen, J., 1989. The concept of soil erodibility and some problems or assessment and application. Catena 16, 392-412.
Byers, A., 1990. Preliminary results of the RRAM Project soil loss and erosion control trials, Rwanda, 1987-1988. IAHS Publication 192, 94-102.
Calvo-Cases, A., Boix-Fayos, C., Imeson, A.C., 2003. Runoff generation, sediment movement and soil water behaviour on calcareous (limestone) slopes of some Mediterranean environments in southeast Spain. Geomorphology 50, 269-291.
Chaplot, V.A.M., Le Bissonais, Y., 2003. Runoff Features for Interrill Erosion at Different Rainfall Intensities, Slope Lengths, and Gradients in a n Agricultural Loessial Hillslope. Soil Science Society of America Journal 67, 544-581.
References
67
Claessens, L., Van Breugel, P., Notenbaert, A., Herrero, M., Van De Steeg, J., 2008. Modelling potential soil erosion in East Africa in the context of climate change, Sediment Dynamics in Changings Environments, Christchurch, New Zealand.
Cohen, M.J., Shepherd, K.D., Walsh, M.G., 2005. Empirical reformulation of the universal soil loss equation for erosion risk assessment in a tropical watershed. Geoderma 124, 235-252.
Dabral, P.P., Baithuri, N., Pandey, A., 2008. Soil Erosion Assessment in a Hilly Catchment of North Eastern India Using USLE, GIS and Remote Sensing. Water Resources Management 22, 1783-1798.
De Baets, S., Poesen, J., Gyssels, G., Knapen, A., 2006. Effects of grass roots on the erodibility of topsoils during concentrated flow. Geomorphology 76, 54-67.
De Roo, A.P.J., Jetten, V.G., 1999. Calibrating and validating the LISEM model for two data sets from the Netherlands and South Africa. Catena 37, 477-493.
de Vente, J., Poesen, J., Verstraeten, G., Govers, G., Vanmaercke, M., Van Rompaey, A., Arabkhedri, M., Boix-Fayos, C., 2013. Predicting soil erosion and sediment yield at regional scales: Where do we stand? Earth-Science Reviews 127, 16-29.
De Vleeschauwer, D., Lal, R., De Boodt, M., 1978. Comparison of detachability indices in relation to soil erodivility for some important Nigerian soils. Pedologie 18, 5-20.
Declercq, K., Poesen, J., 1992. Evaluation of two models to calculate the soil erodibility factor K. Pedologie 42, 149-169.
Deprez, C., 2001. Assessment of the Vulnerbaility for Some Aspects of Land Degredation in Rwanda. University of Ghent, Ghent, p. 106.
Desmet, P.J.J., 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, 427-433.
Ekwue, E.I., Ohu, J.O., Wakawq, I.H., 1993. Effects of incorporating two organic materials at varying levels on splash detachment of some soils from Borno state, Nigeria. Earth Surface Processes and Landforms 18, 399-406.
El-Swaify, S.A., 1997. Factors affecting soil erosion hazards and conservation needs for tropical steeplands. Soil Technology 11, 3-16.
El-Swaify, S.A., Dangler, E.W., 1976. Erodibilites of selected tropical soils in relation to structural and hydrologic parameters, Soil erosion: prediction and control. Soil Conservation Society America, Ankeny, Iowa, pp. 105-114.
El-Swaify, S.A., Dangler, E.W., Armstrong, C.L., 1982. Soil Erosion by Water in the Tropics. College of tropical agriculture and human resources, Honolulu, Hawaii.
References
68
Erdogan, E.H., Erpul, G., Bayramin, I., 2007. Use of USLE/GIS methodology for predicting soil loss in a semiarid agricultural watershed. Environ Monit Assess 131, 153-161.
Foster, G.R., 2005. Draft: Science Documentation. Revised Universal Soil Loss Equation version 2 (RUSLE2) USDA-Agricultural Research Service, Washington D.C.
Foster, G.R., Wischmeier, W.H., 1974. Evaluating Irregular Slopes for Soil Loss Prediction. Transactions of the ASAE 17, 305-309.
Fox, D.M., Bryan, R.B., 1999. The relationship of soil loss by interrill erosion to slope gradient. Catena 38, 211-222.
Fox, D.M., Bryan, R.B., Price, A.G., 1997. The influence of slope angle on final infiltration rate for interrill conditions. Geoderma 80, 181-194.
Fredlund, D.G., Xing, A., Fredlund, M.D., Barbour, S.L., 1996. The Relationship of the Unsaturated Soil Shear Strength Function to the Soil-Water Characteristic Curve. Canadian Geotechnical Journal 33, 440-448.
Freeman, T.G., 1991. Calculating catchment area with divergent flow based on a regular grid. Computers & Geosciences 17, 413-422.
Fu, B.J., Zhao, W.W., Chen, L.D., Zhang, Q.J., Lü, Y.H., Gulinck, H., Poesen, J., 2005. Assessment of soil erosion at large watershed scale using RUSLE and GIS: a case study in the Loess Plateau of China. Land Degradation & Development 16, 73-85.
Fullen, M.A., Catt, J.A., 2004. Soil Management: Problems and Solutions. Routledge.
Gelagay, H.S., Minale, A.S., 2016. Soil loss estimation using GIS and Remote sensing techniques: A case of Koga watershed, Northwestern Ethiopia. International Soil and Water Conservation Research.
Goebes, P., Seitz, S., Geißler, C., Lassu, T., Peters, P., Seeger, M., Nadrowski, K., Scholten, T., 2014. Momentum or kinetic energy – How do substrate properties influence the calculation of rainfall erosivity? Journal of Hydrology 517, 310-316.
Gunn, R., Kinzer, G.D., 1949. The terminal velocity of fall for water droplets in stagnant air. Journal of Meteorology 6, 243-248.
Gyssels, G., Poesen, J., Bochet, E., Li, Y., 2005. Impact of plant roots on the resistance of soils to erosion by water: a review. Progress in Physical Geography 29, 189-217.
Habiyaremye, G., Jairu, N.D., Mupenzi, J.d.l.P., Ngamije, J., Baragahoranye, I., Karangwa, A., 2012. Statistical Analysis of Climatic Variables and Prediction Outlook in Rwanda. East African Journal of Science and Technology 1, 2012.
Hickey, R., 2000. Slope Angle and Slope Length Solutions for GIS. Cartography 29, 1-8.
References
69
Horton, R.E., 1945. Erosional development of streams and their drainage basins; hydrophysical approach to quantitative morphology. Bulletin of the geological society of America 56, 275-370.
Hudson, N.W., 1971. Soil Conservation, Batsford, London.
Hurni, H., 1985. Soil conservation manual for Ethiopia, in: Agriculutre, M.o. (Ed.), Addis Ababa.
Imerzoukene, S., Van Ranst, E., 2001. Une banque de données pédologiques et son S.I.G. pour une nouvelle politique agricole au Rwanda. Mededelingen der zittingen, koninklijke academie voor overzeese wetenschappen, Brussel 47, 299-325.
Jayne, T.S., Mather, D., Mghenyi, E., 2010. Principal Challenges Confronting Smallholder Agriculture in Sub-Saharan Africa. World Development 38, 1384-1398.
Jiang, B., Bamutaze, Y., Pilesjö, P., 2014. Climate change and land degradation in Africa: a case study in the Mount Elgon region, Uganda. Geo-spatial Information Science 17, 39-53.
Kabanza, A.K., Dondeyne, S., Kimaro, D.N., Kafiriti, E., Poesen, J., Deckers, J.A., 2013. Effectiveness of soil conservation measures in two contrasting landscape units of South Eastern Tanzania. Zeitschrift für Geomorphologie 57, 269-288.
Kagabo, D.M., Stroosnijder, L., Visser, S.M., Moore, D., 2013. Soil erosion, soil fertility and crop yield on slow-forming terraces in the highlands of Buberuka, Rwanda. Soil and Tillage Research 128, 23-29.
Kassam, A.H., van Velthuizen, H.T., Mitchell, A.J.B., Fischer, G.W., Shah, M.M., 1992. Agro-ecological land resources assessment for agricultural development planning: A case study of Kenya Resources Database and land productivity - Technical Annex 2, in: Analysis, L.a.W.D.D.F.a.I.I.f.A.S. (Ed.), Rome, p. 23.
Keen, B., Cox, J., Morris, S., Dalby, T., 2010. Stemflow runoff contributes to soil erosion at the base of macadamia trees, 19th World Congress of Soil Scinece, Soil Solutions for a Changing World.
Kiassari, E.M., Nikkami, D., Mahdian, M.H., Pazira, E., 2012. Investigating rainfall erosivity indices in arid and semiarid climates of Iran. Turkish Journal of Agriculture and Forestry 36, 365-378.
Kinnell, P.I.A., 2001. The USLE-M and Modeling erosion Within Catchments, in: Stott, D.E., Mohtar, R.H., Steinhardt, G.C. (Eds.), Sustaining the Global Farm. Selected papers from the 10th International Soil Conservation Organization Meeting Perdue University, pp. 924-928.
Kinnell, P.I.A., 2008. Sediment delivery from hillslopes and the Universal Soil Loss Equation: some perceptions and misconceptions. Hydrological Processes 22, 3168-3175.
Kinnell, P.I.A., 2010. Event soil loss, runoff and the Universal Soil Loss Equation family of models: A review. Journal of Hydrology 385, 384-397.
References
70
König, D., 1992. The Potential of Agroforestry Methods for Erosion Control in Rwanda. Soil technology 5, 167-176.
König, D., 1994. Degradation et Erosion des Sols au Rwanda. Cahiers d'Outre-mer 47, 35-48.
Laflen, J.M., Lane, L.J., Foster, G.R., 1991. WEPP, a New Generation of Erosion Prediction Technology. Journal of soil and water conservation 46, 34-38.
Lal, R., 1994. Soil erosion research methods. Soil and Water Conservation Society.
Lal, R., 2001. Soil degradation by erosion. Land Degradation & Development 12, 519-539.
Lane, L.J., Renard, K.G., Foster, G.R., Laflen, J.M., 1992. Development and Application of Modern Soil Erosion Prediction Technology - The USDA Experience. Australian Journal of Soil Research 30, 893-912.
Le Bissonais, Y., 1995. Soil Characteristics and Aggregate Stability, in: Agassi (Ed.), Soil Erosion, Conservation, and Rehabilitation. CRC Press, Boca Raton, pp. 41-60.
Lee, S., 2004. Soil erosion assessment and its verification using the Universal Soil Loss Equation and Geographic Information System: a case study at Boun, Korea. Environmental Geology 45, 457-465.
Lewis, A.L., Clay, D.C., Dejaegher, M.J., 1988. Soil loss, agriculture, and conservation in Rwanda: toward sound strategies for soil manegement. journal of soil and water conservation 43, 418-421.
Liu, B.Y., Nearing, M.A., Risse, L.M., 1994. Slope Gradient Effectes on Soil Loss for Steep Slopes. Transactions of the ASEA 37, 1835-1840.
Liu, B.Y., Nearing, M.A., Shi, P.J., Jia, Z.W., 2000. Slope Lenght Effects on Soil Loss for Steep Slopen. Soil Science Society of America Journal 64, 1759-1763.
Liu, H., Kiesel, J., Hörmann, G., Fohrer, N., 2011. Effects of DEM horizontal resolution and methods on calculating the slope length factor in gently rolling landscapes. Catena 87, 368-375.
Loch, R.J., Rosewell, C.J., 1992. Laboratory Methods for Measurement of Soil Erodibilities (K Factors) for the Universal Soil Loss Equation. Australian Journal of Soil Research 30, 233-248.
Luliro, N.D., Tenywa, J.S., Majaliwa, M.J.G., 2013. Adaptation of RUSLE to Model Erosion Risk in a Watershed with Terrain Heterogenety. International Journal of Advanced Earth Science and Engineering 2, 93-107.
Mati, B.M., Morgan, R.P.C., Gichuki, F.N., Quinton, J.N., Brewer, T.R., Liniger, H.P., 2000. Assessment of erosion hazard with the USLE and GIS: A case study of the Upper Ewaso Ng’iro North basin of Kenya. International Journal of Applied Earth Observation and Geoinformation 2, 78-86.
References
71
McCool, D.K., Foster, G.R., Renard, K.G., Yoder, D.C., Weesies, G.A., 1995. The Revised Universal Soil Loss Equation, Dod Interagency Workshop on Technologies to Address Soil Erosion on DoD Lands, , San Antonia, TX, pp. 195-202.
McSweeny, R., 2011. Rwanda's Climate: Observations and Projections, appendix E, in: Meegan, C. (Ed.). Smith School of Enterprise and the Environment, Oxford, U.K.
Merritt, W.S., Letcher, R.A., Jakeman, A.J., 2003. A review of erosion and sediment transport models. Environmental Modelling & Software 18, 761-799.
Millward, A.A., Mersey, J.E., 1999. Adapting the RUSLE to model soil erosion potential in a mountainous tropical watershed. Catena 38, 109-129.
Minagri, 2009. Strategic Plan for the Transformation of Agriculture in Rwanda - Phase II Ministry of Agriculture and Animal Resources, Kigali.
Minagri, 2013. Strategic Plan for the Transformation of Agriculture in Rwanda - Phase III. Ministry of Agriculture and Animal Resources, Kigali, pp. 13-20.
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, 629-641.
Moeyersons, J., 1991. La recherche géomorphologique au Rwanda. Bulletin de la Société géographique de Liège 27, 49-68.
Moore, D., Wilson, J.P., 1992. Length-slope factors for the Revised Universal Soil Loss Equation: Simplified method of estimation. Journal of soil and water conservation 47, 423-428.
Moore, I.D., Burch, G.J., 1986. Sediment Transport Capacity of Sheet and Rill Flow: Application of Unit Stream Power Theory. Water Resources Research 22, 1350-1360.
Moore, T.R., 1979. Rainfall Erosivity in East Africa. Geografiska Annaler. Series A, Physical Geography 61, 147-156.
Morgan, R.P.C., 2005. Soil Erosion and Conservation. Blackwell publishing, Oxford, UK.
Morgan, R.P.C., Quinton, J.N., Smith, R.E., Govers, G., Poesen, J.W.A., Auerswald, K., Chisci, G., Torri, D., Styczen, M.E., 1998. The European Soil Erosion Model (EUROSEM): a dynamic approach for predicting sediment transport from fields and small catchments. Earth Surface Processes and Landforms 23, 527-544.
Mukashema, A., 2007. Mapping and Modelling Landscape-based Soil Fertility Change in Relation to Human Induction. Case study: GISHWATI Watershed of the Rwandan highlands, International Institute for geo-information science and earth observation ITC University of Twente, Enschede, The Netherlands.
References
72
Mupenzi, J.d.l.P., Lanhai, L., Jiwen, G., Habumugisha, J.d.D., Habiyaremye, G., Ngamije, J., Innocent, B., 2012. Radical Terraces in Rwanda. East African Journal of Science and Technology 1, 53-58.
Nearing, M.A., 1997. A single, continuous function for slope steepness influence on soil loss. Soil Science Society of America Journal 61, 917-919.
Nill, D., Scwertmann, U., Sabel-Koschella, U., Bernhard, M., Breuer, J., 1996. Soil Erosion by Water in Africa - Principles, Prediction and Protection. Deutsche Gesellschat für Technische Zusammenartbeit (GTZ), Eschborn.
Notebaert, B., Govers, G., Verstraeten, G., Van Oost, K., Ruysschaert, G., Poesen, J., Van Rompaey, A., 2006. Verfijnde erosiekaart Vlaanderen: eindrapport, Studie uitgevoerd in opdracht van het Ministerie van de Vlammse Gemeensschap, ANIMAL.
Nyamulinda, V., 1991. Erosion sous cultures de versants et transports solides dans les bassins de drainage des hautes terres de Ruhengeri au Rwanda. Bulletin-Réseau Erosion 11, 38-63.
Oldeman, L.R., 1993. Global extent of Soil Degradation, Bi-annual report 1991 -1992, International Soil Reference and Information Centre.
Pan, F., Stieglitz, M., McKane, R.B., 2012. An algorithm for treating flat areas and depressions in digital elevation models using linear interpolation. Water Resources Research 48, 1-13.
Pimental, D., Harvey, C., Resusodarmo, P., Sinclaier, K., Kurz, D., McNair, M., Crist, S., Shpritz, L., Fitton, L., Saffouri, R., Blair, R., 1995. Evironmental and Economic Costs of Soil Erosion and Conservation Benefits. Science-AAAS-Weekly Paper Edition 267, 1117-1122.
Poesen, J., Nachtergaele, J., Verstraeten, G., Valentin, C., 2003. Gully erosion and environmental change: importance and research needs. Catena 50, 91-133.
Prasannakumar, V., Vijith, H., Abinod, S., Geetha, N., 2012. Estimation of soil erosion risk within a small mountainous sub-watershed in Kerala, India, using Revised Universal Soil Loss Equation (RUSLE) and geo-information technology. Geoscience Frontiers 3, 209-215.
Prosser, I.P.Y., W.J., Rustomji, P., Hughes, A.O., Moran, C.J., 2001. A Model of River Sediment Budgets as an Element of River Health Assessment, Proceedings of the International Congress on Modelling and Simulation (MODSIM'2001), pp. 10-13.
Rauws, G., Govers, G., 1988. Hydraulic and soil mechanical aspects of rill generation on agricultural soils. Journal of Soil Science 39, 111-124.
Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., Yoder, D.C., 1997. Predicting Soil Erosion by Water: A Guide to Conservation Planning with the Revised Universal Soil Loss equation (RUSLE). U.S. Department of Agriculture.
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 6, 30-33.
References
73
Renard, K.G., Freimund, J.R., 1994. Using monthly precipitation data to estimate the R-factor in the revised USLE. Journal of Hydrology 157, 287-306.
Renard, K.G., Yoder, D.C., Lightle, D.T., Dabney, S.M., 2010. Universal Soil Loss Equation and Revised Universal Soil Loss Equation, in: Morgan, R.P.C., Nearing, M.A. (Eds.), Handbook of Erosion Modelling. John Wiley & Sons, Chichester, UK.
Risse, L.M., Nearing, M.A., Nicks, A.D., Laflen, J.M., 1993. Error Assessment in the Universal Soil Loss Equation. Soil Science Society of America Journal 57, 825-833.
Roose, E., 1976. Use of the universal soil loss equation to predict erosion in West Africa. Soil erosion: prediction and control 21, 60-74.
Roose, E., 1977. Erosion et ruissellement en Afrique de l'ouest, vinght années de mesures en petites parcelles expérimentales, in: l'O.R.S.T.O.M, T.e.d.d. (Ed.), Paris.
Roose, E., Ndayizigiye, F., 1997. Agroforestry, water and soil fertility management to fight erosion in tropical mountains of Rwanda. Soil technology 11, 109-119.
Rushemuka, P.N., Bock, L., Mowo, J.G., 2014. Soil science and agricultural development in Rwanda: state of the art. A review. Biotechnologie, Agronomie, Société et Environnement 18, 142-154.
Ryumugabe, J.B., Berding, F.R., 1992. Variabilité de l'indice d'agressivité des pluies au Rwanda. Bulletin Réseay Erosion 12, 113-119.
Salvador Sanchis, M.P., Torri, D., Borselli, L., Poesen, J., 2008. Climate effects on soil erodibility. Earth Surface Processes and Landforms 33, 1082-1097.
Schiettecatte, W., D´hondt, L., Cornelis, W.M., Acosta, M.L., Leal, Z., Lauwers, N., Almoza, Y., Alonso, G.R., Díaz, J., Ruíz, M., Gabriels, D., 2008. Influence of landuse on soil erosion risk in the Cuyaguateje watershed (Cuba). Catena 74, 1-12.
SESA, s.d.e.e.d.s.a., 1986. Pertes de terre dues a l'erosionn, in: Minagri (Ed.), Ministere de l'agriculture, de l'elevage et des forets, pp. 1-22.
Sharpley, A.N., Williams, J.R., 1990. EPIC - Erosion/Productivity Impact Calculator: 1, Model documentation, in: Agriculture, U.S.D.o. (Ed.), p. 1768.
Singh, M.J., Khera, K.L., 2010. Evaluation and estimation of soil erodibility by different techniques and their relationships, 19th World Congr. Soil Science: Soil Solutions for a Changing World. , pp. 1-6.
Six, J., Bossuyt, H., Degryze, S., Denef, K., 2004. A history of research on the link between (micro)aggregates, soil biota, and soil organic matter dynamics. Soil and Tillage Research 79, 7-31.
Soil Survey Staff, 1975. Soil Taxonomy. SCS-USDA, Washington DC.
References
74
Sorooshian, S., 1991. Parameter Estimation, Model Identification, and Model Validation: Conceptual-Type Models, in: Bowles, D.S., O'Connell, P.E. (Eds.), Recent Advances in the Modeling of Hydrologic Systems, pp. 443-467.
Stocking, M., Elwell, H., 1982. Developing a simple yet practical method of soil-loss estimation. Tropical Agriculture 59, 43-48.
Tisdall, J.M., 1996. Formation of soil aggregates and accumulation of soil organic matter, in: Carter, M.R., Steward, B.A. (Eds.), Structure and Organic Matter Storage in Agricultural Sols CRC Press, Boca Raton, FL, pp. 57-96.
Torri, D., Poesen, J., 1992. The Effect of Soil Surface Slope on Raindrop Detachment. Catena 19, 561-578.
Torri, D., Poesen, J., Borselli, L., 1997. Predictability and uncertainty of the soil erodibility factor using a global dataset. Catena 31, 1-22.
van den Berg, M., 1992. SWEAP A Computer program for water erosion assessment applied to SOTER. International Society of Soil Science.
van Dijk, A.I.J.M., Bruijnzeel, L.A., Rosewell, C.J., 2002. Rainfall intensity - kinetic energy relationships: a critical literature appraisal. Journal of Hydrology 261, 1-23.
Van Oost, K., Govers, G., Desmet, P.J.J., 2000. Evaluating the effects of changes in landscape structure on soil erosion by water and tillage. Landscape Ecology 15, 577-589.
Van Remortel, R., Hamilton, M., Hickey, R., 2001. Estimating the LS factor for RUSLE through iterative slope length processing of digital elevation data. Cartography 30, 27-35.
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 C++ executable. Computers & Geosciences 30, 1043-1053.
Vanelslande, A., Rousseau, P., Lal, R., Gabriels, D., Ghuham, B.S., 1984. Testing the apllicability of a soil erodibility nomogram for some tropical soils. IAHS Publication 144, 463-473.
Verdoodt, A., 2003. Elaboration and Application of an Adjusted Agricultural Land Evaluation Model for Rwanda Volume I, Faculty of Agricultural and applied biological sciences. UGhent, Ghent, p. 561.
Verdoodt, A., Van Ranst, E., 2006a. Environmental assessment tools for multi-scale land resources information systems - A case study of Rwanda. Agriculture, Ecosystems & Environment 114, 170-184.
Verdoodt, A., Van Ranst, E., 2006b. The soil information system of Rwanda: A useful tool to identify guidelines towards sustainable land management. Afrika focus 19, 69-92.
References
75
Vrieling, A., Sterk, G., de Jong, S.M., 2010. Satellite-based estimation of rainfall erosivity for Africa. Journal of Hydrology 395, 235-241.
Wang, B., Zheng, F., Römkens, M.J.M., 2012. Comparison of soil erodibility factors in USLE, RUSLE2, EPIC and Dg models based on a Chinese soil erodibility database. Acta Agriculturae Scandinavica, Section B - Soil & Plant Science 63, 69-79.
Wang, L., Liu, H., 2006. An efficient method for identifying and filling surface depressions in digital elevation models for hydrologic analysis and modelling. International Journal of Geographical Information Science 20, 193-213.
Wassmer, P., 1981. Recherches géomorphologigues au Rwanda. Etude Préfecture de Kibuye. Université Louis Pasteur, Strasbourg, p. 157.
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 Agriculutre, Washington, 244-252.
Wischmeier, W.H., Johnson, C.B., Cross, B.V., 1971. Soil Erodibility nomograph for farmland and construction sites. Soil & Water Conservation Journal 26, 189-193.
Wischmeier, W.H., Mannering, J.V., 1969. Relation of Soil Properties to its Erodibility. Soil Science Society of America Journal 33, 131-137.
Wischmeier, W.H., Smith, D.D., 1965. Predicting rainfall-erosion losses from cropland east of the rocky mountains. U.S Department of Agriculture.
Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion losses: a guide to conservation planning.
Worldbank, 2016. Indicators: Population density, Food and Agriculture Organization and World Bank population estimates, retrieved from: http://data.worldbank.org/indicator/EN.POP.DNST.
Yu, B., Rose, C.W., Ciesiolka, C.A.A., Coughlan, K.J., Fentie, B., 1997. Toward a framework for runoff and soil loss prediction using GUEST technology. Australian Journal of Soil Research 35, 11914-11212.
Yu, B., Rosewell, C.J., 1996a. An assessment of a daily rainfall erosivity model for New South Wales. Australian Journal of Soil Research 34, 139-152.
Yu, B., Rosewell, C.J., 1996b. Technical notes: A Robust Estimator of the R-factor for the Universal Soil Loss Equation. Transactions of the ASEA 39, 559-561.
Zhang, H., Yang, Q., Li, R., Liu, Q., Moore, D., He, P., Ritsema, C.J., Geissen, V., 2013. Extension of a GIS procedure for calculating the RUSLE equation LS factor. Computers & Geosciences 52, 177-188.
References
76
Zhang, K., Lian, L., Zhang, Z., 2016. Reliability of soil erodibility estimation in areas outside the US: a comparison of erodibility for main agricultural soils in the US and China. Environmental Earth Sciences 75.
Zuazo, V.H.D., Martínez, J.R.F., Pleguezuelo, C.R.R., Raya, A.M., Rodríguez, B.C., 2006. Soil-erosion and runoff prevention by plant covers in a mountainous area (se spain): Implications for sustainable agriculture. The Environmentalist 26, 309-319.
Zuazo, V.H.D., Pleguezuelo, C.R.R., 2008. Soil-erosion and runoff prevention by plant covers. A review. Agronomy for Sustainable Development 28, 65-86.
Annex I: Description of soil series present in Tangata watershed
77
Annex I: Description of soil series present in Tangata watershed
Name Description FAO, 199)
AKAZI
The Akazi soil series is part of the ‘Fine clayey mixed, isothermic Lithic Humitropepts’ (Soil Taxonomoy, 1975). Soils in this series have derived from schists. They are clayey, with typically more than 25% silt, yellow and well drained. They are shallow and minimally developed, with presence of an entic horizon. The limited soil depth (<50cm) is due to the presence of saprolit and fresh parent material.
Dystric Regosols / Dystric Leptosols
CYARUGIRA
The Cyarugira soil series is part of the ‘Euic, isohyperthermic Fluvaquentic Tropohemists’ (Soil Taxonomy, 1975). This soil type consists of hemic or fibric organic material mixed with alluvial material. Drainage is poor and soil depth is not limited by the presence of coarse fragments.
Terric Histosol
FUMBA
The Fumba soil series is part of the ‘Clayey, mixed isothermic Orthoxic Tropohumults’ (Soil Taxonomy, 1975). This series groups soils developed from shale and quartzite formations. It are sandy clay, yellow, well drained and highly weathered soils. Soil depth is not limited by the presence of coarse fragments.
Haplic (Humic) Acrisols
GIHIMBI
The Gihimbi soil series is part of the ‘Loamy-skeletal, mixed, isothermic Typic Dystropepts’ (Soil Taxonomy, 1975). This series groups soils developed from shale and quartzite formations. The texture is sandy clay loam. They are yellow, well drained and moderately weathered. Soil depth ranges between 50 and 100cm.
Dystric Cambisols / Dystric Regosols
KABIRA
The Kabira soil series belong to the ‘Clayey, kaolinitic, isothermic Humoxic Sombrihumults’ (Soil Survey Staff, 1975). This soil series groups soils developed from schists. It are well drained, deep, strongly weathered, red, clayey soils. Soil depth is not limited by the presence of coarse fragments
Humic Acrisols (Sombric)
MWOGO
The Mwogo soil series are part of the ‘Loamy-skeletal, mixed, isothermic Lithic Troporthents’ (Soil Taxonomy, 1975). This soil series groups soils developed from schists & quartzite. The texture is sandy loam with >25% (very) fine sand. It are yellow, well drained and low weathered soils with a limited soil depth (< 50cm)
Dystric Regosols / Dystric Leptosols
NSIBO
The Nsibo soil series belong to the ‘Clayey, mixed, isothermic Typic Tropohumults’ (Soil Taxonomy, 1975). The soils are formed from schists and are clayey (with generally more than 25% silt), yellow, well drained and highly weathered. Soil depth is not limited by the presence of coarse fragments.
Haplic Ferralsols / Haplic (Humic) Acrisols
RUKO
The Ruko soil series belong to the ‘Coarse-loamy, mixed, isohyperthermic Fluventic Humitropepts’ (Soil Taxonomy, 1975). Soils formed from colluvial and alluvial material. Loamy, yellow, not perfectly drained soils. Soil depth is not limited by the presence of coarse fragments.
Dystric (Humic) Cambisols / Haplic (Humic) Alisols
RUMULI The Rumuli soil series are part of ‘Fine, mixed, nonacid, isohyperthermic Aeric Tropaquepts’ (Soil Taxonomy, 1975). Soils formed from alluvial material. Soil texture is clay loam. Red, poorly drained soils.
Umbric Gleysols
Annex I: Description of soil series present in Tangata watershed
78
SHANGO
The Shango soil series are part of ‘Clayey-skeletal, kaolinitic, isohyperthermic Sombriorthox (Soil Taxonomy, 1975). Soils in the soil series have derived from schists. Sandy clay, red, well drained and
highly weathered. Limited soil depth (<50 cm by laterite)
Humic Alisols / Humic Acrisols
MUGOZI The Mugozi soil series are part of the ‘Typic Humitropept’ (Soil Taxonomy, 1975). Soils derived from schist, low weathered soils.
Humic Dystric Cambisols
KAYUMBU
The Kayumbu soil series are part of the ‘Clayey over clayey-skeletal, kaolinitic, isothermic Humoxic Tropohumults’ (Soil Taxonomy, 1975). Soils developed from schists. Soils belonging to this soil series are clayey, red, well drained and highly weathered. Quartz material limit soil depth.
Humic Acrisols / Humic Ferralsols
GITABA
The Gitaba soil series are part of the ‘Fine-silty, mixed, isothermic Oxic Dystropepts’ (Soil Taxonomy, 1975). Soils formed from schist material. The texture is silt loam (with in general more than 25% silt). Red, well drained and moderately weathered soil. Soil depth is not limited by the presence of coarse fragments.
Haplic Acrisols / Ferralic Cambisols
GIHIMBI
The Gihimbi soil series is part of the ‘Loamy-skeletal, mixed, isothermic Typic Dystropepts’ (Soil Taxonomy, 1975). The dominant soil texture is sandy clay loam. They are yellow, well drained and moderately weathered. Soil depth is between 50 and 100 cm and limited by fresh parent material.
Dystric Cambisols / Dystric Regosols
BUJUMU
The Bujumu soil series is part of the ‘Loamy-skeletal, mixed, nonacid, isothermic Lithic Troporthents (Soil Taxonomy, 1975). Soils developed out of shale. The soils are sandy clay loam with in general more than 25% very fine sand. Yellow, well drained and minimally developed, with presence of an entic horizon. The limited soil depth (<50cm) is due to the presence of saprolit and fresh parent material.
Dystric Regosols / Dystric Leptosols
Annex II: Soil erodibility classification Kassam et al. (1992)
79
Annex II: Soil erodibility classification Kassam et al. (1992)
Table 11: Soil erodibilitv classification of soil units by soil texture
Soil unit
Soil texture class
Sand Loamy sand
Sandy loam
Loam Clay loam
Sandy clay loam
Sandy clay
Clay Silty clay
Silty clay loam
Silty loam
Silt
Acrisol Humic - - - - 3 2 1 1 - - - -
Other - - 4 - 4 3 2 3 - - - -
Cambisol Humic - - - - 3 2 1 1 - - - -
Other - - 4 5 4 3 2 3 4 5 - -
Chernozem - - - - 4 3 2 3 - - - -
Rendiza - - - - - - 4 5 - - - -
Ferralsol Humic - - - - - - 1 1 - - - -
Nitohumic - - - - - - 1 1 - - - -
Other - - 4 - - 3 2 2 - - - -
Gleysol Humic - - - - - - 2 2 - - - -
Mollic - - - - - - 2 2 - - - -
Other - - - - - - 4 3 - - - -
Phaeozem - - - - 3 3 1 2 - 4 - -
Lithosol - - 5 5 4 3 2 3 - - 6 -
Fluvisol - - 4 5 4 3 2 3 - 4 6 -
Kastanozem - - - - 4 - - - - - - -
Luvisol - - 3 3 3 3 2 3 - - - -
Greyzem - - - - - - 2 3 - - - -
Nitosol Andohumic - - - - - - - 4 - - - -
Other - - - - - - 3 3 - - - -
Histosol - - - 4 - - - 2 - - - -
Arenosol 3 3 4 - - - - - - - - -
Regosol Andocalcaric 3 - 5 5 4 - - - - - - 7
Other - - 4 5 4 3 2 3 - - - -
Solonetz - - 5 - 4 4 3 4 - - 6 -
Andosol - - - 5 4 3 2 3 4 5 - -
Ranker - - 4 5 4 - - 3 - - - -
Vertisol - - - - - 5 - 5 - - - -
Planosol Humic - - - - 3 - - 2 - - - -
Others - 5 5 5 4 4 3 4 - 5 - -
Xerosol/Yermosol - - 4 5 4 4 3 3 - - 6 -
Solonchak - - 5 6 5 4 3 4 5 - 7 -
Ironstones soil - - 4 - 4 3 - 3 - - - -
Annex II: Soil erodibility classification Kassam et al. (1992)
80
Table 12: K-value classes from Kassam et al. (1992)
Erodibility class K (t ha h/ha MJ mm)
1 0.005268
2 0.014487
3 0.023706
4 0.036876
5 0.055314
6 0.07902
7 0.10536
Table 13: applying Kassam et al on soil series present in watershed
Soil series parent material dominant
texture soil unit (FAO, 1990)
Stone cover correction
Erodibility class
AKAZI Shale Clay Dystric Regosols / Dystric
Leptosols -
3
BUJUMU shale (sec. quartzite) sandy clay
loam Dystric Regosols / Dystric
Leptosols 0.7
3
CYARUGIRA Alluvium Clay Terric Histosols - 2
FUMBA Shale (sec. quartzite) sandy clay Haplic (Humic) Acrisols - 2
GIHIMBI Quartzite (sec. shale) sandy clay
loam Dystric Cambisols / Dystric
Regosols -
3
GITABA Shale silty clay Haplic Acrisols / Ferralic Cambisols - 4
KABIRA Quartzite (sec shale) clay Humic Acrisols (Sombric) - 1
KAYUMBU Shale clay Humic Acrisols / Humic Ferralsols - 1
MUGOZI Shale (sec. quartzite) mixed Humic Dystric Cambisols - 3.5
MWOGO Quartzite sandy loam Dystric Regosols / Dystric
Leptosols -
4
NSIBO Shale clay Haplic Ferralsols / Haplic (Humic)
Acrisols -
1
RUKO Colluvium & alluvium mixed Dystric (Humic) Cambisols / Haplic
(Humic) Alisols -
2.5
RUMULI Alluvium clay loam Umbric Gleysols - 3.5
RUNABA Shale (sec. quartzite) clay Humic (Dystric) Cambisols 1
RUTABO Quartzite (sec.
Micaschist) sandy clay
loam Humic Ferralsols / Umbric Regosol 0.7
1
SHANGO Shale sandy clay Humic Alisols / Humic Acrisols 0.4 1
Related Documents
