You are on page 1of 12

Journal of Environmental Management 277 (2021) 111369

Contents lists available at ScienceDirect

Journal of Environmental Management


journal homepage: http://www.elsevier.com/locate/jenvman

Research article

Effectiveness of terracing techniques for controlling soil erosion by water


in Rwanda
Jules Rutebuka, PhD a, b, *, Aline Munyeshuli Uwimanzi, Msc a, Olive Nkundwakazi, Msc a,
Desire Mbarushimana Kagabo, PhD c, Jean Jacques Muhinda Mbonigaba, PhD, Professor d,
Pieter Vermeir a, Ann Verdoodt, Professor a
a
Ghent University, Faculty of Bioscience Engineering, Department of Environment, Coupure Links 653, 9000, Gent, Belgium
b
Rwanda Agriculture and Animal Resources Development Board (RAB), Department of Land Husbandry, Irrigation Research and Technology Transfer, P.O. Box 5016,
Huye, Rwanda
c
International Center for Tropical Agriculture (CIAT), Kigali, Rwanda
d
Association for Strengthening Agricultural Research in Eastern and Central Africa (ASARECA), P.O.Box 765 Mpigi Road, Entebbe, Uganda

A R T I C L E I N F O A B S T R A C T

Keywords: Despite long-standing efforts in terracing, limited field-based evidence of its effectiveness as implemented within
Agro-ecology rural farming systems of humid tropical regions, such as Rwanda, is available. This study aimed to reveal
Farmers’ fields regional differences in effectiveness of two widely used terracing techniques. Traditional slope farming (NP) was
Runoff
compared to bench (BT) and farmers’ based progressive terraces (PT) in terms of runoff, soil losses, and topsoil
Soil-loss
Terraces
fertility in two contrasting agro-ecological zones, the Eastern Plateau (Murehe) and Buberuka Highlands (Tan­
Topsoil fertility gata). During four consecutive rainy seasons, event-based data were collected using erosion plots (5 m width x
22.2 m length). Effectiveness indices of both terracing systems, as well as (R)USLE P-factor values, were
calculated. The annual average soil losses under NP ranged from 4.71 ± 5.02 ton ha− 1 to 46.01 ± 7.28 ton ha− 1
in Murehe (14% slope gradient) and Tangata (43% slope gradient), respectively. Bench terracing clearly out­
performed the farmer-based progressive terrace at both locations, leading to negligible soil losses. In terms of
runoff reduction, an effectiveness of 70 and 85% respectively, was observed at Murehe and Tangata. The
effectiveness of PT reached 52% for runoff control and 93% for soil loss control at Tangata, thereby confirming
its huge potential as erosion control measure, even in mountainous areas. In the hilly landscape of Murehe, the
runoff generated by PT – in some years - can exceed that under traditional farming, while the measure reduced
soil losses by half on average. Associated USLE P-factors varied between seasons with an annual average values of
0.001–0.02 for BT, and 0.07 to 0.55 for PT at Tangata and Murehe, respectively. These variations in performance
by site and terracing system also resulted in differences in topsoil chemical fertility, with BT generally out­
performing both PT and NP at Tangata. At Murehe, PT showed a significantly lower chemical fertility compared
to BT and NP. Poor quality risers explained the overall lower performance of PT at Murehe. The study thus
confirmed the huge potential of (bench) terraces to sustainably reduce soil erosion rates when established within
an integrated approach, paying attention to correct installation and fertility-supporting agronomic practices.
More attention should be given to riser installation (e.g. distance) and maintenance of PT. Adoption of these
erosion control measures can be recommended to similar agro-ecological zones for sustainably protecting the
lands while mitigating or adapting the effects of climate change.

1. Introduction problems (Lal, 1995). This problem has attracted the attention of many
scientists specialized in different aspects of erosion (Bocco and Napo­
Soil erosion by water is one of the major problems that faces sub- letano, 2017; Moges and Taye, 2017; Wang et al., 2017) as it negatively
Saharan Africa, resulting in severe economic and environmental affects human livelihood (Ighodaro et al., 2013). Soils provide a range of

* Corresponding author. Rwanda Agriculture and Animal Resources Development Board (RAB). Huye PO Box 5016, Rwanda.
E-mail addresses: jules.rutebuka@rab.gov.rw, rujub2@gmail.com (J. Rutebuka), uwimanzia@gmail.com (A. Munyeshuli Uwimanzi), dadolive50@yahoo.fr
(O. Nkundwakazi), j.mbonigaba-muhinda@asareca.org (J.J.M. Mbonigaba), pieter.vermeir@ugent.be (P. Vermeir), ann.verdoodt@ugent.be (A. Verdoodt).

https://doi.org/10.1016/j.jenvman.2020.111369
Received 28 April 2020; Received in revised form 5 August 2020; Accepted 7 September 2020
Available online 25 September 2020
0301-4797/© 2020 Elsevier Ltd. All rights reserved.
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

key functions, including production of food, storage of organic matter, differentiated with respect to specific agro-ecological conditions and
and cycling of water and nutrients; and around 99% of human food landscape properties (Clay and Dejaegher, 1987). During the last
comes from land. Thus, soil degradation through erosion has an impact decade, these initiatives strongly shifted to bench (radical) terraces as an
on the ability of soils to perform this range of vital functions (Morgan, effective way to fight soil erosion and to improve land management
1995). (Bizoza, 2011; Fleskens and Stroosnijder, 2007), with the ultimate aim
The highest water erosion rates are observed in tropical regions with to create a productive environment resilient to climate change (Bizoza
averages ranging from 30 to 40 t ha− 1 yr− 1 compared to the much lower and de Graaff, 2012; Government of Rwanda, 2009). On other hand,
rates observed in the United States and Europe, averaging at about 17 t farmers developed their own techniques to protect their lands against
ha− 1 yr− 1 (Cerdan et al., 2010). Soil erosion can be a serious issue when erosion, including locally made terraces referred to as progressive ter­
soil erosion rates are higher than soil formation rates, resulting in races by gradually creating risers, digging trenches and planting agro­
long-term loss of soil depth, and therefore reducing water holding ca­ forestry trees/shrubs along the contour lines, adopting grass strips alone
pacity, nutrient storage and overall soil productivity (Li et al., 2009). or combining them with infiltration ditches (Kagabo et al., 2013; Gov­
Land degradation by water erosion remains a key challenge in most ernment of Rwanda, 2009).
humid highlands in East Africa whereby soil and associated nutrient Despite existing research on the dynamics of runoff-sediment gen­
losses have been estimated to affect 50% of the total arable land pro­ eration and associated nutrient losses from agricultural lands (Gabiri
ductivity (Dass et al., 2011; Gabiri et al., 2015; Henao and Baanante, et al., 2015; Henao and Baanante, 1999; Ovuka, 2000), most studies in
1999; Ovuka, 2000). Land degradation in Rwanda is mainly character­ Rwanda and similar regions have been conducted in research-managed
ized by soil erosion, declining soil fertility and acidification, and is stations. Only Ethiopia embedded several erosion studies in rural
driven by unsustainable land use practices on steep slopes with no or farming systems (Adimassu et al., 2017). Additional efforts are therefore
inappropriate soil conservation measures (Roose and Ndayizigiye, needed to understand the effectiveness of both traditional and
1997). Productivity decline is accelerated by soil loss (Roose and farmer-developed soil erosion control measures at field scale. In addition
Ndayizigiye, 1997), especially in the highlands of Rwanda (Clay and to the focus on farmers’ fields, studies covering diverse climatic, topo­
Lewis, 1996; Lewis and Nyamulinda, 1996; Steiner, 1998). The erosion graphic, edaphic and agricultural management conditions are needed to
problems occur mainly in the Congo Nile Ridge in the west, the Volcanic provide more comprehensive insights on the performance of the selected
Range mountain areas in the north-west and the Buberuka Highlands in terracing techniques, in order to inform policy makers and the farming
the north, regions characterized by steep slopes and aggressive rainfall. community on the best practices while mitigating climate change
About 40% of Rwanda’s land has very high erosion risk (REMA, 2008). hazards.
According to the Stockholm Environment Institute (2009), soil erosion This paper therefore aimed to evaluate the effectiveness of two
in Rwanda results in a loss of 1.4 million tons of soil per year, equivalent common terracing techniques (bench and progressive terraces) on
to an economic loss of US $ 34.3 thousand, corresponding to about 2% of erosion control within two contrasting agro-ecological zones of Rwanda
the national GDP. The soil erosion rate in cropland, which occupies 56% i.e. Buberuka Highlands (Tangata watershed) and Eastern Plateau
of the national land area, has been estimated at 421 t ha− 1yr− 1 using the (Murehe watershed). The evaluation focused on short-term monitoring
RUSLE model and is responsible for about 95% of the national soil losses of (1) runoff and (2) soil losses using erosion plots located in both wa­
(Karamage et al., 2016). Ninety percent of the domestic cropland is on tersheds. This provided farmers’ field-based evidence of the effective­
slopes ranging from 5% to 55% (MINAGRI, 2013). Rwandan farmers ness of these widely used terracing techniques to inform local land users
depend primarily on seasonal crops grown twice a year, hence exposing and regional decision makers. Opportunities and challenges in terracing
soils to erosion. Generally, the farm size per household is less than 0.5 ha related to soil fertility status were revealed as well. Finally, the study
and it is expected to decrease in coming years, contributing to an intense also allowed generating preliminary values of the (R)USLE P-factor, for
pressure on natural resources (Bidogeza et al., 2009). The population further application in erosion modelling.
increase has driven the Rwandans to inhabit and cultivate also the
marginal lands (Nahayo et al., 2016; REMA, 2010). Research by the 2. Materials and methods
National Agricultural Research Institute (ISAR) reported soil losses
ranging from 35 to 246 t ha− 1 yr− 1, with most research stations 2.1. Study area
measuring over 100 t ha− 1 yr− 1, using Wischmeier erosion plots (König,
1993; Roose and Ndayizigiye, 1997). 2.1.1. Representativity of the selected watersheds
Therefore, appropriate management practices are needed to reduce The watersheds of Tangata and Murehe are representative in terms of
soil erosion and maintain soil productivity (Moradi et al., 2015). climate, soil and topography for the agro-ecological zones (AEZs) of
Effective erosion control measures should be based on reducing slope Buberuka Highlands and Eastern Plateau, respectively (Fig. 1).
length to minimize runoff, soil and associated nutrient losses, reducing Murehe watershed, located in the Eastern Plateau at 1400–1700 m
soil surface crust formation and increasing infiltration rates (Adimassu altitude, is subjected to a tropical semi-arid climate with annual mean
et al., 2017; Gabiri et al., 2015) while restoring soil fertility and vege­ temperature of 20.2 ◦ C and about 787 mm annual rainfall. The dominant
tation cover. Research and policy instruments have been initiated in that slope gradients range from 0 to 18% while Haplic Ferralsols (Karama
matter. In the same research stations by ISAR, the effectiveness of soil series) (Van Ranst et al., 2000a) dominate the soilscape. About 50%
different erosion control systems was tested (König, 1993; Roose and of cropland in Murehe watershed is bench terraced, the rest being
Ndayizigiye, 1997). Recently, Kagabo et al. (2013) reported a reduction non-protected farmers’ fields.
in average annual soil loss from 42 to 18 t ha− 1 when changing Located in the Buberuka Highlands AEZ, Tangata is found at a higher
non-protected fields to fields including grass strips combined with altitude ranging from 1980 to 2400 m, and experiences a higher annual
infiltration ditches on farmers’ land in the Buberuka Highlands. Soil and rainfall (1069 mm) and lower average temperature (15.9 ◦ C). Slopes are
water conservation strategies ranging from ‘biological’ or ‘vegetative’ generally steeper than 27% and Humic Acrisols (Kabira soil series) (Van
methods to ‘physical’ or ‘mechanical’ methods, have been initiated in Ranst et al., 2000b) are the dominant soils. Contouring and strip crop­
combination or individually (De Graaff, 1996; Hurni et al., 2008). ping is the most extensive (area extent of 68%) erosion control measure
Addition of organic matter and mineral fertilizers has been recom­ within the Tangata watershed, while BT occupied only 7% of the total
mended to promote long-term sustainable land management (König, area (Figure S1).
1993; MINAGRI, 2014). The Government provided technical guidance
for soil and water conservation interventions (REMA, 2010), counter­
acting erosion and sustaining production. These interventions have been

2
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

2.1.2. Biophysical characterization and farming system of the experimental Table 1


sites Location and biophysical characterisation of the experimental fields.
The experimental sites were installed at mid-slope landscape posi­ Watershed Coordinates Altitude Natural slope Soil
tions within six representative, farmer managed fields. Table 1 sum­ (m) gradient (%) classification
marises their main biophysical characteristics. Table 1S provides the Tangata 1.51510◦ S 2068 43 Umbric Acrisol
description and analytical data of the soil profiles investigated near the 29.85319◦ E
plots. The Acrisol at Tangata has a clay loam topsoil, covering an illuvial Murehe 2.01118◦ S 1626 14 Umbric Vetic
Bt horizon (40–50% clay, Table 1S). In contrast, the entire profile 30.28787◦ E Ferralsol

(Table 1S) of the Ferralsol at Murehe is clayey.


The bimodal rainfall pattern characterizing Rwanda gives rise to two NP plots were installed at 14 and 43% slope gradients in Murehe and
agricultural seasons. Traditionally, season A runs from September to Tangata respectively (Table 1). Terraces had been installed 4–6 years
December (SOND), which matches with the period of the so-called short ago. BT were constructed through government funding, while PT were
rains. Agricultural season B is typically taking from March to June, generally farmer-made. BT plots had strong and stable risers with Pen­
exploiting the long rains from March to May (MAM). During previous nisetum clandestinum (Kikuyu grass) at the embankments. PT developed
years, the farming system at all sites had been as uniform as possible. It by progressive sedimentation of soil behind living grass strips and were
consisted of a rotational cropping system of maize in season A and characterised by weak and unstable (especially at Murehe) risers with
common (Murehe) or climbing (Tangata) beans in season B. Land Penissetum purpureum (napier grass) (Figures S2 and S3). Risers were
preparation, planting and weeding were done in accordance with the highest at Tangata given its steeper slopes (Table 2).
best farmers’ practice in Rwanda. Maize crops were planted at an inter-
and intra-row spacing of 30 cm and 75 cm, respectively. Beans were 2.2.2. Erosion plots
planted with a row spacing of 20 cm and within-row spacing of 40 cm. The erosion plots were 22.2 m long in the flow direction and 5 m
The farmers applied integrated soil fertilisation using both farmyard wide, matching the standard USLE plot size (Wischmeier and Smith,
manure and inorganic fertilisers. Maize fields received with DAP (100 1978; Renard et al., 1997). In case intervals for BT and PT were shorter
kg ha− 1) and a top dressing of urea (50 kg ha− 1), while NPK (100 kg than the plot length, several risers and terrace intervals were combined
ha− 1) was applied on bean fields. Farmers cultivate their non-terraced in one erosion plot (Table 2; Figure S4). The collection gutters were
and progressively terraced lands along the slope direction (upward), installed right below the lowest riser. The plots were bounded with
while bench terraced land is tilled across slope direction at both sites. removable metal sheets on the upper and lateral sides to allow uniform
farmer’s practice management (Figure S4). The metal sheets were tall
2.2. Experimental design enough to prevent overtopping and inserted deep enough to prevent
water leakage towards or out of the plot.
2.2.1. Treatments Runoff collection gutters connected with tanks via PVC tubes were
The experiment included three different treatments: (1) traditional established and fixed in the ground at the base of each bounded plot.
slope farming practice, referred to as non-protected plots (NP), (2) so Several collection water tanks were connected to each other to be able to
called radical or bench terracing (BT), and (3) slow forming or pro­ capture total runoff water volumes of large events. The water tanks used
gressive terracing (PT). for runoff collection under NP and PT plots had a capacity of about 183

Fig. 1. Location of the Murehe and Tangata experimental sites in the Rwandan agro-ecological zones of the Eastern Plateau and Buberuka Highlands, respectively.

3
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

Table 2
Physical characteristics of the bench and progressive terraces at both study sites.
Site Terrace type Interterrace Risers

Gradient (%) Length (m) Height (m) Stabilisation Number

Murehe Progressive 1.3 22 1.4 Napier grass (unstable) 1


Bench (leveled) 0.0 10–11 1.4 Kikuyu grass 2
Tangata Progressive 36.0 <8 2.0 Napier grass 3
Bench (reversed) 1.3 4–5.5 2.8 Kikuyu grass 4

dm3 (54 cm diameter, 80 cm height) whereas smaller tanks of about 72 2.3. Exploitation of the collected dataset
dm3 (41 cm diameter, 55 cm height) were used to capture runoff at the
BT plots. Three replicates were installed next to each other in case of 2.3.1. Expressing the effectiveness of the erosion control measures
terraced plots (PT and BT). NP plots were alternating with the BP in Runoff depth (RO) and soil loss (SL) were determined at event and
three replicates as well. seasonal basis. Runoff coefficients (RC), being the percentage ratio of
runoff to rainfall depth, were calculated.
2.2.3. Event-based monitoring of rainfall, runoff and sediment The effectiveness of the erosion control measures was determined by
Event-based rainfall amount, rainfall intensity, runoff and soil losses comparing runoff volumes and soil losses obtained from the terraced
were monitored during four consecutive agricultural rainy seasons: S1 plots with those on the non-protected land, using the following index:
(season A of 2015/16), S2 (season B of 2015/16), S3 (season A of 2016/
Yc − Ym
17) and S4 (season B of 2016/17). The sowing (start) and harvesting IE = x100 (1)
Yc
(end) dates of the four monitored seasons (Table 3) match with actually
observed rainfall conditions. As such, due to a late arrival of the rains, Whereby IE is the index of effectiveness (%), Yc is the seasonal runoff
season S1 started only mid-October in Murehe, but could count on volume (mm) or soil loss (kg) from the NP plot, and Ym is the seasonal
higher than normal rainfall amounts in the month of January. runoff volume (mm) or soil loss (kg) from the plot with erosion control
Two manual rain gauges were placed nearby the erosion plots for structures (PT and BT).
determining event-based rainfall records throughout the four agricul­
tural seasons. The average event intensity was calculated based on the 2.3.2. Determination of the conservation support practice factor
ratio of event rainfall amount and duration of the event. After each The experimental design allowed to provide local estimates of the
rainfall event, the runoff water volume was determined by multiplying conservation support practice (P) factor following the USLE model. The
the height level of water inside the graduated tank with its cross- P factor was calculated as the ratio of soil loss measured in the plots of
sectional area. A 1.5-L runoff water sample was taken after carefully PT or BT compared to the NP plots.
mixing runoff water with the sediment settled at the bottom of the tank
to produce a homogeneous water-sediment solution. The samples were
dried at the laboratory and weighted to determine the event-based 2.4. Physico-chemical soil characterization
sediment concentration. The total sediment lost was calculated by
multiplying the sediment concentration with runoff volume per each The source topsoils (20 cm depth) of the erosion plots were sampled
rainfall event. in August 2016 between the first and second land preparation and are
expected to represent the inherent fertility of the soil after several years

Table 3
Seasonal and average annual (n = 2) rainfall amounts, intensities, duration and number of (normal/extreme erosive) rainfall events at Tangata and Murehe during the
monitoring period (September 2015 to June 2017). Lower case letters (w to z) indicate significant (P < 0.05) differences between seasons.
Parameters Murehe Tangata

S1a S2a S3a S4a Annual S1a S2a S3a S4a Annual
Start date Oct 13 Mar24 Sep 6 Feb 11 Sep 2015 Sep 1 Mar 3 Aug 15 Feb 6 Sep 2015
End date Mar 19 Jun 22 Feb 3 May 12 June 2017 Feb 19 May 31 Jan 30 May 12 June 2017
Rainfall amount (mm)
Rainfall total 458.0 273.0 528.5 358.4 809.0 576.1 400.5 404.8 368 874.7
Average by event 9.5x 16.1xy 18.2xy 19.9 y 15.9 10.9xy 12.1xy 13.1xy 18.4 y 13.6
Max by event 46.3 46 65 67.5 56.9 34.5 25 30.5 40.5 37.5
Rainfall intensity (mm hr¡1)
Average by event 33.8bx 13.6yz 12.1 y 15.0yz 18.6 20.8xz 22.7x 23.6x 23.5x 22.7
Max by event 116.0b 28.6 36.4 41.2 78.6 60 50.5 44.3 50.8 55.4
Rain duration (min)
Duration’s total 1514 1759 3646 1563 4241 1824 1228 1107 1171 2665
Average by event 32x 104yw 126 y 87 y 87 34z 37zw 36zw 59yzw 42
Max by event 407 402 883 207 645 65 90 58 180 135
Number of events (¡)
Event’s total 48 17 29 18 56 53 33 31 20 69
N◦ erosive events NPc 23 13 26 15 39 53 33 31 20 69
N◦ erosive events BPc 24 13 26 15 39 53 33 31 20 69
N◦ extreme events NPd 1 2 2 1 3 1 1 1 1 2
N◦ extreme events BPd 1 2 2 2 4 2 1 2 3 4
a
S1 = season A of 2015/16, S2 = season B of 2015/16, S3 = season A of 2016/17 and S4 = season B of 2016/17 (see also Figure S7).
b
Reduces to 20.9 mm h− 1 average and 94.0 mm h− 1 maximum after removal of an outlier (116 mm h− 1) producing almost no runoff/soil loss and deemed erroneous.
c
Events that produced soil loss on non-protected (NP) and bare (BP) plots.
d
Events contributing to at least 50% of the total seasonal soil losses under traditional farming (NP) and bare plots (BP) when in descending order in cumulative soil
loss graphs.

4
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

of erosion (control). In each erosion plot, three composite soil samples Overall, a higher number of short, but more intense rainfall events
were collected, representing the upper, middle and lower part of the characterized Tangata, while at Murehe a lower number of longer-
plot, each composed of nine intimately mixed subsamples. The sub­ duration and low-intensity events have been reported. The minimum
samples were taken at least 1 m away from the upper and lower and maximum rainfall intensities at event basis varied in the similar
boundaries to avoid edge effects. In case of terraced plots, subsamples range as what Wagesho and Claire (2016) reported.
were taken in different terrace beds. The samples were air-dried, sieved At Tangata, all rainfall events caused soil losses in all treatments. At
(<2 mm) and sent to Ghent University for soil analysis. Particle-size Murehe, the percentage of rain events causing soil losses varied by
analysis was performed by the dry sieving and pipette method, season from a minimum of 48% (NP) or 50% (BP) to maximum 90%.
involving sample pre-treatment with H2O2 to remove the organic Nevertheless, at both sites, most soil losses were generated during a
matter. OC and Ntot were determined by dry combustion using the CNS- limited number of events. The characteristics of rainfall during obser­
2000 analyzer, available phosphorous by the Bray 2 method, with vation period vis-à-vis the long-term data are described in the supple­
sample absorbance measured in a spectrophotometer for extracted mentary data (Figure S7).
phosphorus (Van Ranst et al., 2004), and exchangeable cations by the
ICP-OES (iCAP 6300 series) analytical technique after extraction using 3.2. Characterization of the topsoil in erosion plots
NH4-acetate (Cottenie et al., 1982). pH in soil water suspension and 1 M
KCl solution of 1: 5 ratio was measured by potentiometer equipped with Table 4 summarises the topsoil properties of the erosion plots at both
a combined pH meter. sites. Average bulk density values range from 1.01 to 1.22 g cm− 3,
Also two undisturbed samples per plot, representing the upper and corresponding to typical values for clayey soils. Topsoils are moderately
lower part of the non-protected erosion plots and in different terrace acidic with moderate levels of organic carbon, total nitrogen and
beds for the terraced plots were collected using Kopecky rings for bulk amounts of exchangeable cations (Ca2+, Mg2+ and K+), and very low
density determination. Dry bulk density was determined using the core levels of available phosphorus according to Landon (1991).
method. Despite the generally low fertility level, significant differences have
been observed between treatments and sites as well as interaction effects
2.5. Statistical analysis of site on treatment, depending on specific topsoil properties considered.
Erosion and/or terracing did not lead to differences in topsoil texture
The collected rainfall, runoff, and soil losses were statistically between the treatments, but the clay and sand contents are respectively
analyzed using R software. Analysis of variance (ANOVA) (for normally significantly higher and lower at Murehe compared to Tangata. Both
distributed soil data) was used to test for significant differences between sites have similar silt contents ranging from 20 to 24% on average. These
sites, seasons and treatments at 95% confidence level. Comparison of site differences in texture lead to corresponding differences in bulk
means and assessment of interaction effects were done using the Tukey’s densities. With respect to this latter physical property, no differences
test method. The non-parametric Kruskal-Wallis test, followed by were observed between non-terraced and terraced land. While there are
Mann–Whitney’s test for multiple comparisons of means at 0.05 signif­ multiple site differences in chemical fertility of topsoil between Murehe
icance level, were adopted for non-normally distributed data (runoff, and Tangata non-protected farmland, terracing, especially BT, tends to
runoff coefficient, and soil loss). harmonize the fertility at both sites. From statistical point of view, the
PT at Murehe are characterized by a significantly lower chemical
3. Results fertility than both the BT as well as the NP lands. The chemical fertility
at the BT generally approaches the fertility of the NP farmland. At
3.1. Rainfall characteristics affecting runoff processes during Tangata, in contrast, the fertility of the BT generally outperforms the PT
experimental period and NP lands, which both share a similar fertility level.

Table 3 represents the distribution of rainfall amount, intensity, 3.3. Effects of terracing on water erosion
duration, and erosivity between the two study sites at annual, seasonal
and event scales in order to explore their effects on erosion. From 3.3.1. Runoff control
September 2015 to June 2017, 112 and 137 rainfall events, with rainfall At Murehe, seasonal RO values varied from 3.9 to 14.7 mm at the PT
amounts varying from 0.5 to 68.2 mm and 2.0–40.0 mm, were recorded plots, and from 0.9 to 2.5 mm with BT plots (Figs. 4 and 5). Seasonal RC
at the experimental sites in Murehe and Tangata, respectively values ranged from 0.5 to 4.3% (average = 2.3%) and 0.2–0.6%
(Figures S5 and S6). Event durations and average intensities ranged from (average = 0.4%), for PT and BT plots, respectively. The seasonal RO for
2.0 to 883.0 min and from 1.5 to 116.0 mm h− 1 at Murehe while they PT and BT plots at Tangata are 5.4–10.6 mm and 1.2–5.1 mm respec­
respectively ranged from 10.0 to 180.0 min and from 3.8 to 60.0 mm h− 1 tively, while the corresponding RC values ranged from 1.4 to 3% (1.9%
at Tangata. on average) and 0.3–0.9% (0.5% on average).
Rainfall in Tangata was characterized by a higher number of events At Tangata, PT and BT plots generated significantly lower RO and RC
of shorter duration and higher intensity compared to Murehe. Season A compared to NP plots. Specifically, BT had a significantly larger impact
recorded higher rainfall amounts than season B at both sites. Few on runoff reduction than PT. Significant reductions in RO and RC were
aggressive rainfall events in each season were responsible for most of the also reported for the BT plots at Murehe, while PT plots at this site in
soil losses. Statistical analysis of rainfall amount, duration and intensity contrast yielded significantly higher RO and RC values than NP plots
per event showed significant differences between seasons. (Fig. 2, Table 4).
Rainfall was generally higher at Tangata than Murehe, except for S3 Table 5 describes the impact of terracing on controlling the annual
within our monitoring period. During both years, season A (S1 and S3) RO and RC values by treatments and sites. At Murehe, annual RO ranged
recorded (slightly) higher rainfall amounts than season B (S2 and S4) at from 8.4 to 18.7 mm (13.6 on average) and from 2.0 to 4.7 mm (3.4 on
both sites. At Murehe, season A was also characterized by a higher average) respectively for PT and BT. The corresponding RC ranged from
number of events. Rainfall intensity in contrast did not vary much by 1.1 to 2.1% and 0.3–0.5%. During year 1, BT showed significantly lower
season, with averages being generally lower at Murehe than at Tangata. RO and RC values compared to NP and PT, the latter treatments not

5
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

Table 4
Means and standard deviation of the topsoil properties (n = 27 for disturbed samples and n = 18 for the core samples, each site) at the non-protected (NP), pro­
gressively terraced (PT) and radically terraced (BT) erosion plots located at Murehe (Eastern Plateau) and Tangata (Buberuka Highlands) experimental sites.
Site/Treatmenta Clay (%) Silt (%) Sand (%) BD (g cm-3) pH-H2O (¡) pH-KCl (¡)

Murehe
NP 57 ± 15 21 ± 12 22 ± 5 1.00 ± 0.03 5.83 ± 0.08 a 4.60 ± 0.07 b
PT 58 ± 14 23 ± 14 19 ± 4 1.02 ± 0.05 5.64 ± 0.06 a 4.40 ± 0.06 a
RT 54 ± 11 24 ± 10 22 ± 7 0.99 ± 0.02 5.99 ± 0.13 c 4.73 ± 0.14 c
Tangata
NP 36 ± 4 * 24 ± 4 40 ± 3 * 1.14 ± 0.16 * 5.58 ± 0.23 a* 4.27 ± 0.19 a*
PT 36 ± 4 * 24 ± 4 41 ± 2 * 1.19 ± 0.05 * 5.64 ± 0.09 a 4.34 ± 0.07 a
RT 37 ± 2 * 20 ± 5 43 ± 4 * 1.30 ± 0.11 * 5.95 ± 0.35 b 4.72 ± 0.39 b

Site/Treatmenta OC (%) Ntot (%) Pav (ppm) Caex (cmol(þ)/kg) Mgex (cmol(þ)/kg) Kex (cmol(þ)/kg)

Murehe
NP 2.35 ± 0.10 B 0.18 ± 0.01 b 1.75 ± 1.33 5.20 ± 0.34 b 2.10 ± 0.20 c 0.18 ± 0.02 A
PT 2.02 ± 0.08 A 0.16 ± 0.01 a 1.04 ± 1.21 3.97 ± 0.40 a 1.38 ± 0.11 a 0.20 ± 0.03 Ab
RT 2.29 ± 0.10 B 0.17 ± 0.01 b 1.06 ± 0.77 5.64 ± 0.66 b 1.62 ± 0.19 b 0.22 ± 0.03 B
Tangata
NP 2.08 ± 0.12 * 0.21 ± 0.01 * 4.79 ± 4.79 3.58 ± 0.85 a* 1.31 ± 0.35 ab* 0.36 ± 0.16 ab*
PT 1.66 ± 0.56 0.19 ± 0.01 * 5.87 ± 3.70 * 3.74 ± 0.34 a 1.18 ± 0.17 a* 0.25 ± 0.03 a*
RT 1.83 ± 0.67 0.20 ± 0.02 * 7.62 ± 6.87 * 5.98 ± 2.19 b 1.49 ± 0.22 b 0.37 ± 0.08 b*
a
Lowercase letters in the columns indicate significant differences (P < 0.05) in soil properties between treatments at the specific site, while asterisk * indicates
significant (P < 0.05) differences in soil properties between sites subjected to the same treatment.

impacts on runoff vary by site, season and terracing system, as


confirmed by the statistical analysis (Table 5). Figs. 3 and 4 illustrate
clear seasonal differences in performance of the progressive terraces at
Murehe. While in season B (S2, S4), similar runoff depths are recorded in
NP and PT plots, these latter turn out to generate higher runoff in season
A (S1, S3).
For NP plots as traditional slope farming practice, the seasonal RO
was mostly limited to 10 mm, except for season 3 at Murehe, while at
Tangata, RO ranged between 10 and 24 mm. RC values were very low.
Average seasonal RC values varied from 0.7 to 2.7% and from 2.5 to
6.8% at Murehe and Tangata, respectively. All rainfall events produced
runoff at Tangata while 69% of the rainfall events produced runoff at
Murehe. Annual runoff depth ranged from 13.18 to 34.17 mm yr− 1 with
corresponding runoff coefficient varying from 1.56 to 4.12% per year at
Murehe and Tangata, respectively.

3.3.2. Soil loss control


Average seasonal soil losses under PT and BT ranged from 0.09 to
4.30 t ha− 1 and from 0.01 to 0.10 t ha− 1, respectively at Murehe, and
from 1.04 to 2.10 and 0.01 to 0.06 t ha− 1, respectively at Tangata.
Average soil losses were 0.28 (excluding abnormal season 3, otherwise
2.57 t ha− 1 yr− 1) and 0.08 t ha− 1 yr− 1 for PT and BT, respectively at
Murehe. Tangata exhibited average soil losses of 3.4 and 0.06 t ha− 1 yr− 1
for PT and BT, respectively (Table 5).
Seasonal soil losses in NP plots ranged from 0.22 to 8.0 t ha− 1 and
from 17.9 to 28.8 t ha− 1 for Murehe and Tangata, respectively. Soil
Fig. 2. Boxplots of seasonal runoff depths and runoff coefficients (n = 12) re­ losses were thus significantly higher at Tangata than at Murehe. Event-
ported at non-protected (NP), progressively terraced (PT) and bench terraced based soil losses ranged from 0.0 to 3.3 t ha− 1 at Murehe and from 0.0 to
(BT) plots at Murehe and Tangata. Box plots with different letters are signifi­ 17.8 t ha− 1 at Tangata. The average annual soil losses were 4.7 and
cantly different at P < 0.05. 46.01 t ha− 1 yr− 1 at Murehe and Tangata, respectively (Table 5).
Annual soil losses at Murehe ranged from 0.76 to 4.38 t ha− 1 yr− 1
differing significantly from each other. In year 2, all treatments differed and from 0.04 to 0.11 t ha− 1 yr− 1 for PT and BT respectively. During the
in terms of RO and RC, the lowest values being reported in BT plots, two years of the experiment at Tangata, annual soil losses varied from
highest values on the non-protected plots. At Tangata, annual RO varied 2.86 to 3.94 t ha− 1 yr− 1 and from 0.03 to 0.09 t ha− 1 yr− 1 for PT and BT,
from 16.0 to 16.2 mm (16.1 on average) and 2.5–7.6 mm (5.1 on respectively. All treatments at both sites significantly differed in terms of
average) for PT and BT respectively, while the corresponding RC ranged soil losses, except for the PT and NP plots in year 1 (2015/2016) at
from 1.7 to 2.1% (1.9 on average) and 0.8 to 0.3% (0.6 on average). All Murehe.
three treatments significantly influenced RO and RC in both years of the Both terracing types (PT and BT) thus significantly reduced soil
experiment. The lowest annual RO and RC were reported for the BT losses compared to slope farming practice plots (NP) at Tangata. This
plots, while highest values were recorded for the NP plots. only applies to the BT in Murehe. The PT at this latter experimental site,
The event runoff took place with a delay of about three to eight despite the increased runoff values recorded, did show a trend towards
consecutive rainfall events (except in S4) in Murehe (Fig. 3) while in lower soil losses compared to the NP (Fig. 5).
Tangata, runoff started with the first rainfall event (Fig. 4). The terracing

6
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

Fig. 3. Cumulative event-based runoff (n = 3) by season during the monitoring period at the non-protected (NP), progressively terraced (PT) and bench terraced (BT)
for experimental site of Murehe.

Fig. 4. Cumulative event-based runoff (n = 3) by season during the monitoring period at the non-protected (NP), progressively terraced (PT) and bench terraced (BT)
for experimental site of Tangata.

3.3.3. Effectiveness of terracing of PT in runoff and soil loss control was observed in season B (S2 and S4)
The effectiveness of BT and PT plots on runoff and soil loss control than A (S1 and S3) at Murehe. At Tangata, runoff and soil loss controls
compared to NP plots in Murehe and Tangata during the four consecu­ under BT and PT varied seasonally without any specific pattern, simi­
tive seasons is presented in Table 6. BT effectively reduced runoff by 70 larly to BT at Murehe.
and 85% at Murehe and Tangata, respectively. PT plots proved to be
very ineffective in runoff control at Murehe and moderately effective 3.3.4. RUSLE conservation support practice (P) factor
(52%) at Tangata. The ineffectiveness of PT in Murehe is related to The erosion plot data were exploited to provide estimates of the
inappropriate establishment of hedgerows on the risers and their inad­ conservation support practice (P) factor (Table 2S). Seasonal P factors
equate maintenance by land owners (Table 2, Figures S2 & S3). for bench terracing varied little and averaged to annual values of 0.02
Despite the ineffectiveness of PT in runoff control at Murehe, soil loss and 0.001 at Murehe and Tangata, respectively. Especially at Tangata,
control was moderately effective up to 46%. BT however strongly BT almost stopped sediments leaving the terraced system downslope the
reduced soil losses up to 93%. At Tangata, both terrace types effectively risers. The P factors for PT showed a large seasonal as well as spatial
reduced soil losses up to 98 and 93%, respectively. Higher effectiveness variation. At Tangata, PT reduces soil losses to 93% of the NP values,

7
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

Table 5 Bekele-Tesemma, 2011). Several authors found that bench terraces


Annual runoff depth (RO, mm), runoff coefficient (RC, %) and soil loss (SL, t effectively reduce the runoff, soil and nutrient losses, even without
ha− 1 yr− 1) on non-protected (NP), progressive terracing (PT) and bench grasses or living hedges (Wickama et al., 2014; Widomski, 2011). The
terracing (BT) plots in Murehe and Tangata during the 2015/2016 (Year 1) and observed limited amount of runoff under bench terraces during this
2016/2017 (Year 2) monitoring seasons. study was primarily coming from the risers/embankments and not really
Plotsαα Var. Year 1 Year 2 Annual average from the flat terrace beds. Van Dijk and Bruijnzeel (2004) confirmed that
Murehe runoff and soil loss come from overland flow at terrace risers and some
limited soil transport by splash from the lower border of the terrace
NP RO 7.93 ± 0.24* 18.42 ± 0.06* 13.18 ± 7.42
RC 1.07 ± 0.0.03* 2.08 ± 0.01* 1.56 ± 0.69 beds.
SL 1.16 ± 0.03* 8.26 ± 0.11*$ 4.71 ± 5.02 Tenge (2005) evaluated the effectiveness of bench terraces using
PT RO 8.43 ± 0.19* 18.72 ± 0.41 13.58 ± 7.28 runoff plots and Gerlach troughs, finding however clearly lower effec­
RC 1.14 ± 0.03* 2.12 ± 0.05* 1.63 ± 0.69 tiveness compared to our results. BT soil losses measured by the Gerlach
SL 0.76 ± 0.02* 4.38 ± 0.17*$ 2.57 ± 2.56
BT RO 2.03 ± 0.24*$ 4.69 ± 0.33*$ 3.36 ± 1.88
troughs varied from 1.5 to 3.7 t ha− 1 depending on the slope ranging
RC 0.27 ± 0.03$ 0.53 ± 0.04$ 0.40 ± 0.18 from 6 to 15% to more than 32%, while soil losses reduced to 61 and
SL 0.04 ± 0.01$ 0.11 ± 0.01*$ 0.08 ± 0.05 46%. Data collected using runoff plots, reported soil losses that
Tangata seasonally vary between 2.1 and 4.5 t ha− 1 with respective effective­
NP RO 29.09 ± 3.21$ 39.24 ± 0.62$ 34.17 ± 7.18
nesses of 68 and 77%. As in this study, a lower effectiveness in runoff
RC 3.06 ± 0.34$ 5.18 ± 0.08$ 4.12 ± 1.50
SL 40.86 ± 5.38$ 51.15 ± 1.44$ 46.01 ± 7.28 reduction compared to soil loss reduction was observed (43–50%). In
PT RO 16.24 ± 0.17$ 15.93 ± 0.30$ 16.09 ± 0.22 contrast, fanya juu terraces, being hillside ditches made by throwing
RC 1.71 ± 0.02$ 2.10 ± 0.04$ 1.91 ± 0.28 excavated soil on the upslope part of the ditch along the contourlines,
SL 3.94 ± 0.31$ 2.88 ± 0.12$ 3.41 ± 0.75 were more effective in reducing soil loss than bench terraces in
BT RO 7.64 ± 0.02$ 2.66 ± 0.04$ 5.15 ± 3.52
RC 0.76 ± 0.00$ 0.34 ± 0.01$ 0.55 ± 0.30
Tanzania. The former technique exhibited similar effectiveness to the
SL 0.09 ± 0.02$ 0.03 ± 0.00$ 0.06 ± 0.04 bench terraces from our study.
Despite the slight seasonal variations observed in the performance of
An asterisk put after the Murehe parameter values indicates significant (P <
BT (Shimeles, 2012; Tenge, 2005), their annual average P factors of
0.05) difference with the corresponding parameter/treatment value at Tangata.
A dollar sign $ indicates significant (P < 0.05) within-site differences between
0.001–0.03 for Tangata and Murehe, respectively also reflect their large
treatments. effectiveness (Table 2S).
α
Each treatment (NP, PT, BT) has 3 replicates per site.

4.2. Progressive terraces


whereas in Murehe, these terraces fail to reduce soil losses by 50%.
Annual P f3actors for PT thus varied between 0.53 and 0.66 (0.55 on At both study sites, PT have generally demonstrated an effective
average) and 0.06–0.10 (0.07 on average) at Murehe and Tangata,
respectively. Seasonal differences in the P factor for PT reflect the
Table 6
variation in erosive rainfall and sediment concentration.
Effectiveness indices (%) of runoff and soil loss reduction under bench terraced
(BT) and progressively terraced (PT) plots compared to traditional slope farming
4. Discussion (NP) at Murehe and Tangata.
Site1 Plots S1 S2 S3 S4 Average
4.1. Bench terraces
Runoff
Murehe PT − 22 4 − 43 7
Annual runoff coefficients and soil losses under BT were less than 1%
¡21
BT 71 77 76 57 70
and 0.1 t ha− 1 yr− 1 respectively at both sites. As such, bench terraces Tangata PT 47 40 61 56 52
effectively reduced runoff by 70 and 85%, at Murehe and Tangata BT 72 76 91 95 85
respectively, while the corresponding soil loss reductions rise to 93 and Soil loss
Murehe PT 14 62 47 60 46
98%, respectively (Table 6).
BT 96 91 98 89 93
In agreement with previous studies, the agricultural BT in Rwanda Tangata PT 92 90 95 94 93
are very effective in protecting the soil against erosion, even in these BT 97 98 99 99 98
mountainous landscapes, as their almost flatbed fields store a maximum 1
The effectiveness was calculated using equation (1).
of water, reduce run-off, and avoid soil losses (Arnáez et al., 2015;

Fig. 5. Boxplots of seasonal soil loss (n = 12) reported at non-protected (NP), progressively terraced (PT) and bench terraced (BT) plots at Murehe and Tangata. Box
plots with similar letters are not significantly different at P < 0.05.

8
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

contribution to erosion control with low annual runoff coefficients rates at non-protected land at Tangata watershed. According to Van Dijk
varying between 3.9 (Tangata) and 4.6 (Murehe) % and strongly and Bruijnzeel (2004), well-managed grass strips reduced runoff co­
reduced soil losses ranging from 2.6 (Murehe) to 3.4 (Tangata) t ha− 1 efficients from 19 to 8% and soil losses by more than two-thirds.
yr− 1. When compared to NP plots, the PT at Tangata exhibited a mod­ The distance between consecutive hedges/risers of PT controls the
erate effectiveness (52%) for runoff reduction. However, at Murehe the erosion rate reduction by filtering the runoff and retaining the eroded
PT plots generated more runoff than the NP plots. Soil loss rates sediments (Subhatu et al., 2017). The stronger reduction in soil and
moderately reduced up to 50% at Murehe, and generally more than 80% nutrient losses than in runoff was especially observed under PT, illus­
at Tangata. By consequence, the derived P factor values varied by site trating the importance of the riser sediment filtering capacity.
and season. Average annual P factor values for PT ranged from 0.59 at Therefore, sustainability of these locally made terraces (PT) requires
Murehe to 0.07 at Tangata. Their seasonal variation indicates that their use of stable risers with living hedgerows or other agroforestry species
effectiveness varies with the rainfall characteristics and/or land man­ (Zuazo et al., 2011) to decrease soil losses and maintain soil fertility
agement (Table 2S). The latter includes the variation induced by rota­ (Baptista et al., 2015; Dass et al., 2011; Gabiri et al., 2015), as observed
tional cropping systems as crop cover changes causing within-season in Tangata.
variation in erosion rates and seasonal variation in terrace effectiveness.
The annual average P factor value observed at Murehe closely fits the
range of values (0.26–0.57) reported by other authors (Gebremichael 4.3. Comparative benefits of the two terracing types
et al., 2005). In Ethiopia, P factor values between 0.57 and 0.90 have
been reported for contouring at slope between 6 and 16% (Taye et al., 4.3.1. Erosion control
2018; Wischmeier and Smith, 1978) while stone bunds showed lower P PT can approach the effectiveness of BT in terms of soil erosion
factors, ranging between 0.31 and 0.39 (Nyssen, 2001). Grass strips control, once they are well established, managed and regularly main­
were found to reduce runoff up to 12–45% (Gebremichael et al., 2005; tained by land owners (farmers) as shown in the watershed of Tangata.
Tenge, 2005), showing a slightly poorer performance than the PT in As shown in our study, as well as by Maetens et al. (2012), both erosion
Tangata, but much better than PT in Murehe. In terms of soil loss control techniques are generally less effective in reducing RO than in
reduction, effectiveness values of grass strips ranged from 20 to 54% reducing SL. This implies that runoff water could still cause soil loss,
according to Tenge (2005), while it reached up to 79% according to particularly in steep landscapes.
Welle et al. (2006) in Ethiopia. The effectiveness of PT is thus strongly The mechanisms of bench and progressive terraces to reduce water
variable, reflecting the quality of their establishment, maintenance or erosion begin with their development up to exploitation. In general,
management (harvesting or regeneration of grasses on the risers), as radical terraces is mechanically constructed on sloping land with rela­
well as its interactions with the terrain, rainfall and soil type. This tively deep soils (>1.5 m depth) by cutting and filling to produce a series
seasonally variable effectiveness contrasts with the BT that generally of level steps namely as benches in order to allow water to slowly
displayed a more or less constant effectiveness in soil loss and runoff infiltrate into the soil rather than its accumulation on the long risers or
control. Also König (1992) reported the issue of PT non-uniformity in surface runoff. Agroforestry trees and grasses are used to stabilize the
erosion control following a four year observation period with plot embakments by spreadly covering them. In Rwanda, they are principally
measurements at Butare (Rwanda). Other studies reported soil losses on designed to reduce soil losses through enhanced retention and infiltra­
PT ranging from 5 to 111 t ha− 1 yr− 1 in Rwanda and in the larger region tion of runoff, to promote permanent agriculture on steep slopes and to
(Angima et al., 2001; Lenka et al., 2012; Lewis, 1988; Lewis and Nya­ promote land consolidation and intensive land use. Farmers cultivate
mulinda, 1996). Such low effectiveness can induce a high redistribution bench terraced lands across slope direction hence erosion is reduced.
of soil nutrients within individual PT plots (Tripathi et al., 2000), which The choice between levelled and reversed bench terraces relies on
actually develop as a consequence of tillage erosion. soil capacity to retain water as well as the rainfall amount present in the
The poor performance of PT plots in terms of reducing runoff, soil concerned region. Reversed terraces drain the excess of water through
and even nutrient losses at Murehe resulted from inappropriate estab­ waterways, particularly in the humid highlands of Rwanda (Tangata)
lishment of hedgerows on the risers or embankments and their inade­ (Bekele-Tesemma, 2011).
quate maintenance by farmers. At this study site, the distance between On the other hand, progressive terraces are slowly formed by
two consecutive risers with grass strips/hedgerows was about 22 m, establishing contour bunds with soil in combination with ditches and
therefore matching the length of non-protected plots (Table 2). Unstable agroforestry along contourlines (Roose and Ndayizigiye, 1997; Kagabo
hedgerows could not sufficiently slow down and reduce the accumulated et al., 2013). By time of natural process of erosion and sedimentation,
runoff volumes reaching the risers, resulting in the destruction of risers these terraces can be as more similar landform and effective as bench
and more eroded materials. Such vegetative barriers with long distance terrace depending on their management. Ploughing across slope direc­
(22 m) in between farm-made terraces (PT) are therefore sometimes tion definitely create a more alike bench terrace, otherwise they become
reported to be even less effective than non-protected plots (Inamdar and worse in terms of erosion control.
Dillaha, 2000; Yongmei et al., 2011). Especially during aggressive
rainfall events, overland flow detached large amounts of sediments from 4.3.2. Other benefits
the unstable risers. Both terracing types provide several (other) benefits to farmers.
At Tangata, despite the high slope gradient of the PT beds and In our study, BT had a higher chemical fertility compared to the
cultivation along the slope direction (upward tillage), PT was moder­ progressive terraces and the NP land, particularly in the steep land of
ately effective in runoff reduction and even very effective in soil loss Tangata. Although bench terraces minimize the topsoil fertility gradient
control, because of short distances between terrace risers (<8 m), stable (Table 4), terracing alone does not maintain soil fertility. Further studies
and well-managed napier grass hedgerows (Table 2). The stability of the are needed to understand the effect on crop productivity (Bizoza, 2011;
risers relies on short terrace intervals, as well as high density and height Lewis, 1992; Mupenzi et al., 2012).
of the grasses or other vegetation constituting the risers (Ng et al., 2008; The hedgerows on the risers provide fodder for animals, and trees for
Van Dijk et al., 1996). Kagabo et al. (2013) reported that well-managed home consumption and stakes. This compensates for the loss of culti­
progressive terraces (grass strips) reduced erosion up to 43–57% of the vated land area being occupied by the risers (Gebremichael et al., 2005;
Kagabo, 2013; Nyssen et al., 2008; Subhatu et al., 2017; Teshome et al.,

9
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

2014). Moreover, several studies recommend the use of agronomic guidelines for establishing and managing progressive terraces, as it has
measures on terraced land (Adimassu et al., 2017; Zhang et al., 2016). been the case for bench terraces, would support improvement of
Integration of such physical measures with agronomic practices, climate-smart agricultural production systems in Rwanda. Overall, this
particularly the supply of inorganic fertilizers, lime and manure, is research could be used to provide evidences of the appropriate erosion
recommended to control soil and associated nutrient losses (Adimassu control measures for sustainable land management and mitigate or
et al., 2017) while improving soil fertility. This latter aspect is essential adapt the effects of climate changes in diverse hilly landscapes.
to sustainable use of the bench-terraced land.
Declaration of competing interest
4.4. Importance of qualitative construction and maintenance
The authors declare that they have no known competing financial
In different parts of the country, bench-terraced land has been interests or personal relationships that could have appeared to influence
abandoned shortly after their installation due to soil fertility and acidity the work reported in this paper.
problems as well as inappropriate establishment, among others. Terrace
abandonment increases erosion rates, while simultaneously being un­ Acknowledgement
productive and economically non-viable. Arnáez et al. (2015) quoted
that the overland flow from abandoned terraces can connect to mass Financial support for this research was provided by VLIR-UOS [Grant
movements, pipes or gullies and cause soil losses exceeding 100 t ha− 1 no ST01007644] in partnership with Ghent University (UGent) and
yr− 1. Terrace abandonment was not a problem at our study sites given Rwanda Agriculture and Animal Resources Development Board (RAB).
the integrated approach used in their development and management. Many thanks go to farmers and field technicians who greatly contributed
The investment in terracing requires much attention from government, to the management of field experiments and data collection. We are
partners and beneficiaries (farmers) because a well-established terrace is grateful to the Belgian government-Flemish region that unceasingly
based on firstly removing a portion of terrace berm or topsoil and bring supports studies for innovative responses to local challenges in Southern
it back after terrace construction in order to maintain an adequate soil Countries.
fertility status (Lewis, 1992). Newly established bench terraces require
at least three years of cultivation and input application after their
Appendix A. Supplementary data
installation for ensuring their stability including fertility replenishment.
As the government promotes the two terracing types, it should may
Supplementary data to this article can be found online at https://doi.
attention to promote better installation and management techniques.
org/10.1016/j.jenvman.2020.111369. These supplementary materials
Sustainable use of the farm-made progressive terraces requires farmers’
include Figures S1 to S7, and Tables 1S–2S described in this article.
training into riser stability, harvesting or pruning of the hedges, and
tillage along the contour lines. Otherwise, PT risks becoming ineffective
References
in erosion control.
An integrated approach is to be recommended for development and Adimassu, Z., Langan, S., Johnston, R., Mekuria, W., Amede, T., 2017. Impacts of soil and
sustainable management of erosion control measures, including bench water conservation practices on crop yield, run-off, soil loss and nutrient loss in
and progressive terracings (Rutebuka et al., 2019). This requires Ethiopia: review and synthesis. Environ. Manage. 59, 87–101. https://doi.org/
10.1007/s00267-016-0776-1.
establishment and enforcement of a policy instrument and detailed Angima, S.D., O’Neill, M.K., Stott, D.E., 2001. On-farm Assessment of Contour Hedges for
technical guidelines for development and management of progressive Soil and Water Conservation in Central Kenya. International Centre for Research in
terraces, as has been the case for bench terracing. Agroforestry, Nairobi, Kenya.
Arnáez, J., Lana-Renault, N., Lasanta, T., Ruiz-Flaño, P., Castroviejo, J., 2015. Effects of
farming terraces on hydrological and geomorphological processes. A review. Catena
5. Conclusion 128, 122–134. https://doi.org/10.1016/j.catena.2015.01.021.
Baptista, I., Ritsema, C., Geissen, V., 2015. Effect of integrated water-nutrient
management strategies on soil erosion mediated nutrient loss and crop productivity
Results from this research provide field-based evidence of widely in cabo verde drylands. PloS One 10 (7), e0134244. https://doi.org/10.1371/
used terracing techniques in Rwandan rural farming systems from journal.pone.0134244.
contrasting agro-ecological zones. Runoff and soil loss measurements in Bekele-Tesemma, A., 2011. Illustrated Supervision Checklist for Assessment of the
Quality of Comprehensive Land-Husbandry Works at LWH Project Sites in Rwanda.
farmers’ fields illustrated large variation as related to natural conditions
Kigali (Unpublished report).
including soil types, rainfall, and topography, subjected to maize-bean Bidogeza, J.C., Berentsen, P.B.M., De Graaff, J., Oude Lansink, A.G.J.M., 2009.
rotation farming systems. Both terracing techniques effectively A typology of farm households for the Umutara Province in Rwanda. Food Sec 1,
reduced runoff and soil loss. Bench terrace proved to be the most 321–335. https://doi.org/10.1007/s12571-009-0029-8.
Bizoza, A.R., 2011. Farmers, Institutions and Land Conservation. Institutional Economic
effective, especially at Tangata. The P-factor at this site varied less be­ Analysis of Bench Terraces in the Highlands of Rwanda. Doctoral dissertation.
tween seasons compared to P-factor of progressive terracing. However, Wageningen University. https://library.wur.nl/WebQuery/wurpubs/fullte
when inadequately established, in terms of unstable risers placed at long xt/169360.
Bizoza, A.R., de Graaff, J., 2012. Financial cost-benefit analysis of bench terraces in
distances, and poorly or irregularly maintained, farmer-made progres­ Rwanda. Land Degrad. Dev. 23 (2), 103–115. https://doi.org/10.1002/ldr.1051.
sive terraces (PT) can lead to high runoff and soil loss, even higher than Bocco, G., Napoletano, B.M., 2017. The prospects of terrace agriculture as an adaptation
traditional slope farming plots as is the case in Murehe. In terms of to climate change in Latin America. Geogr. Compass 11 (10), 1–13. https://doi.org/
10.1111/gec3.12330 e12330.
chemical fertility, both erosion control systems reduced within-field Cerdan, O., Govers, G., Le Bissonnais, Y., Van Oost, K., Poesen, J., Saby, N., Gobin, A.,
topsoil fertility variation, but a highest soil fertility was obtained at Vacca, A., Quinton, J., Auerswald, K., Klik, A., Kwaad, F.J.P.M., Raclot, D., Ionita, I.,
the bench terraced lands of Tangata. At Murehe, the topsoil chemical Rejman, J., Rousseva, S., Muxart, T., Roxo, M.J., Dostal, T., 2010. Rates and spatial
variations of soil erosion in Europe: a study based on erosion plot data.
fertility of the progressive terraces was significantly lower than at the Geomorphology 122 (1–2), 167–177. https://doi.org/10.1016/j.
non-protected fields. geomorph.2010.06.011.
Further promotion of progressive terracing thus requires strong ef­ Clay, D., Dejaegher, Y., 1987. Agro-ecological zones: the development of a regional
classification scheme for Rwanda. Tropicultura 5 (4), 153–159.
forts to build capacity of farmers, with particular focus on the sustain­
Clay, D.C., Lewis, L.A., 1996. Land use, soil loss, and sustainable agriculture in Rwanda.
able use of these erosion control techniques. Attention must be paid to In: Bates, D.G., Lees, S.H. (Eds.), Case Studies in Human Ecology. Springer, Boston,
adopting appropriate distances between risers, stabilizing the risers MA, pp. 271–287.
using adapted tree or grass species, sustainable harvesting or pruning of Cottenie, A., Verloo, M., Kiekens, L., Velghe, G., Camerlynck, R., 1982. Chemical
Analysis of Plants and Soils. Lab. Agroch. State Univ. Gent, Belgium.
the hedges, and tillage practice along contour lines. A policy interven­ Dass, A., Sudhishri, S., Lenka, N.K., Patnaik, U.S., 2011. Runoff capture through
tion on integrated watershed management that involves clear technical vegetative barriers and planting methodologies to reduce erosion, and improve soil

10
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

moisture, fertility and crop productivity in southern Orissa, India. Nutrient Cycl. Nahayo, L., Lanhai, L., Alphonse, K., Fidele, K., Christophe, M., Felix, N., Enan, M.N.,
Agroecosyst. 89 (1), 45–57. https://doi.org/10.1007/s10705-010-9375-3. 2016. Agricultural impact on environment and counter measures in Rwanda. Afr. J.
De Graaff, J., 1996. The Price of Soil Erosion: an Economic Evaluation of Soil Agric. Res. 11 (25), 2205–2212. https://doi.org/10.5897/AJAR2016.10899.
Conservation and Watershed Development. PhD dissertation. Wageningen Ng, S.L., Cai, Q.G., Ding, S.W., Chau, K.C., Qin, J., 2008. Effects of contour hedgerows on
University. https://library.wur.nl/WebQuery/wurpubs/36110. water and soil conservation, crop productivity and nutrient budget for slope
Fleskens, L., Stroosnijder, L., 2007. Is soil erosion in olive groves as bad as often claimed? farmland in the Three Gorges Region (TGR) of China. Agrofor. Syst. 74 (3), 279–291.
Geoderma 141 (3–4), 260–271. https://doi.org/10.1016/j.geoderma.2007.06.009. https://doi.org/10.1007/s10457-008-9158-x.
Gabiri, G., Obando, J.A., Tenywa, M.M., Majaliwa, J.G., Kizza, C.L., 2015. Soil and Nyssen, J., 2001. Erosion Processes and Soil Conservation in a Tropical Mountain
nutrient losses under cultivated bush and climbing beans on terraced humid Catchments under Threat of Anthropogenic Desertification— A Case Study from
highland slopes of southwestern Uganda. J. Sci. Res. Reports 8, 1–16. https://doi. Northern Ethiopia. Ph.D. Thesis. KU Leuven, Belgium.
org/10.9734/JSRR/2015/18113. Nyssen, J., Poesen, J., Moeyersons, J., Haile, M., Deckers, J., 2008. Dynamics of soil
Gebremichael, D., Nyssen, J., Poesen, J., Deckers, J., Haile, M., Govers, G., erosion rates and controlling factors in the Northern Ethiopian Highlands – towards
Moeyersons, J., 2005. Effectiveness of stone bunds in controlling soil erosion on a sediment budget. Earth Surf. Process. Landforms 33, 695–711. https://doi.org/
cropland in the Tigray Highlands, northern Ethiopia. Soil Use Manag. 21 (3), 10.1002/esp.1569.
287–297. https://doi.org/10.1111/j.1475-2743.2005.tb00401.x. Ovuka, M., 2000. More people, more erosion? Land use, soil erosion and soil productivity
Government of Rwanda, 2009. Strategic plan for the transformation of agriculture in in Murang’a district, Kenya. Land Degrad. Dev. 11 (2), 111–124. https://doi.org/
Rwanda – phase II (PSTA II). In: Final Report February 2009 Ministry of Agriculture 10.1002/(SICI)1099-145X(200003/04)11:23.0.CO;2-I.
and Animal Resources. Kigali, Rwanda. REMA, 2008. Land Use and Agriculture, Republic of Rwanda. Rwanda Environment
Henao, J., Baanante, C., 1999. Nutrient Depletion in the Agricultural Soils of Africa. 2020 Management Authority.
Vision Briefs 62. International Food Policy Research Institute (IFPRI). REMA, 2010. Practical Tools on Soil and Water Conservation Measures. Rwanda
Hurni, H., Herweg, K., Portner, B., Liniger, H., 2008. Soil erosion and conservation in Environment Management Authority, Republic of Rwanda, p. 5. Tool and Guideline
global agriculture. In: Braimoh, A.K., Vlek, P.L.G. (Eds.), Land Use and Soil #.
Resources, 41–71. Springer, Dordrecht. Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., Yoder, D.C., 1997. Predicting
Ighodaro, I.D., Lategan, F.S., Yusuf, S.F.G., 2013. The impact of soil erosion on Soil Erosion by Water: A Guide to Conservation Planning with the Revised Universal
agricultural potential and performance of Sheshegu community farmers in the Soil Loss Equation (RUSLE). USA Department of Agriculture. Agriculture Handbook
Eastern Cape of South Africa. J. Agric. Sci. 5 (5), 140–147. https://doi.org/10.5539/ No. 703.
jas.v5n5p140. Roose, E., Ndayizigiye, F., 1997. Agroforestry, water and soil fertility management to
Inamdar, S.P., Dillaha, T.A., 2000. Relationships between drainage area, slope length, fight erosion in tropical mountains of Rwanda. Soil Technol. 11 (1), 109–119.
and slope gradient for riparian slopes in Virginia. Trans. ASAE (Am. Soc. Agric. Eng.) https://doi.org/10.1016/S0933-3630(96)00119-5.
43 (4), 861–866. https://doi.org/10.13031/2013.2981. Rutebuka, J., Kagabo, D.M., Verdoodt, A., 2019. Farmers’ diagnosis of current soil
Kagabo, D., 2013. Participatory Integrated Watershed Management in the North-Western erosion status and control within two contrasting agro-ecological zones of Rwanda.
Highlands of Rwanda. Doctoral Thesis Wageningen University. https://library.wur. Agric. Ecosyst. Environ. 278, 81–95. https://doi.org/10.1016/j.agee.2019.03.016.
nl/WebQuery/wurpubs/438852. Shimeles, D., 2012. Effectiveness of Soil and Water Conservation Measures for Land
Kagabo, D.M., Stroosnijder, L., Visser, S.M., Moore, D., 2013. Soil erosion, soil fertility Restoration in the Wello Area, Northern Ethiopian Highlands. PhD diss..
and crop yield on slow-forming terraces in the highlands of Buberuka, Rwanda. Soil Universitäts-und Landesbibliothek, Bonn.
Tillage Res. 128, 23–29. https://doi.org/10.1016/j.still.2012.11.002. Steiner, K.G., 1998. Using farmers’ knowledge of soils in making research results more
Karamage, F., Zhang, C., Ndayisaba, F., Shao, H., Kayiranga, A., Fang, X., Nahayo, L., relevant to field practice: experiences from Rwanda. Agric. Ecosyst. Environ. 69 (3),
Nyesheja, E.M., Tian, G., 2016. Extent of cropland and related soil erosion risk in 191–200. https://doi.org/10.1016/S0167-8809(98)00107-8.
Rwanda. Sustain. Times 8 (7), 1–19. https://doi.org/10.3390/su8070609. Subhatu, A., Lemann, T., Hurni, K., Portner, B., Kassawmar, T., Zeleke, G., Hurni, H.,
König, D., 1992. The potential of agroforestry methods for erosion control in Rwanda. 2017. Deposition of eroded soil on terraced croplands in Minchet catchment,
Soil Technol. 5 (2), 167–176. https://doi.org/10.1016/0933-3630(92)90017-U. Ethiopian Highlands. Int. Soil Water Conserv. Res. 5 (3), 212–220. https://doi.org/
König, D., 1993. Degradation et erosion des sols au Rwanda. In: Neuvième réunion du 10.1016/j.iswcr.2017.05.008.
Réseau Erosion. Poitiers (FRA), vol. 13, pp. 31–40, 1992/11/05-07. Réseau Eros. Taye, G., Vanmaercke, M., Poesen, J., Van Wesemael, B., Tesfaye, S., Teka, D., Nyssen, J.,
Bull. http://www.documentation.ird.fr/hor/fdi:39484. Deckers, J., Haregeweyn, N., 2018. Determining RUSLE P- and C-factors for stone
Lal, R., 1995. Erosion-crop productivity relationships for soils of Africa. Soil Sci. Soc. Am. bunds and trenches in rangeland and cropland, North Ethiopia. Land Degrad. Dev.
J. 59 (3), 661. https://doi.org/10.2136/sssaj1995.03615995005900030004x. 29 (3), 812–824. https://doi.org/10.1002/ldr.2814.
Landon, J.R., 1991. Booker tropical soil manual : a handbook for soil survey and Tenge, A.J.M., 2005. Participatory Appraisal for Farm-Level Soil and Water Conservation
agricultural land evaluation in the tropics and subtropics. In: Pbk (Ed.), Booker Planning in West Usambara Highlands, Tanzania. PhD Diss. Wageningen Univerisity.
Agriculture International Limited. Harlow, Essex, England : Longman Scientific & Teshome, A., De Graaff, J., Stroosnijder, L., Fleskens, L., Graaff, J. De, Stroosnijder, L., De
Technical ; 1991, London. Graaff, J., Stroosnijder, L., Fleskens, L., 2014. Evaluation of soil and water
Lenka, N.K., Dass, A., Sudhishri, S., Patnaik, U.S., 2012. Soil carbon sequestration and conservation practices in the north-western Ethiopian highlands using multi-criteria
erosion control potential of hedgerows and grass filter strips in sloping agricultural analysis. Front. Environ. Sci. 2, 60. https://doi.org/10.3389/fenvs.2014.00060.
lands of eastern India. Agric. Ecosyst. Environ. 158, 31–40. https://doi.org/ Tripathi, B.P., Shrestha, S.P., Acharya, G.P., 2000. Summary and Updating with Season
10.1016/j.agee.2012.05.017. Soil and Nutrient Losses from Bari Land Terraces in the Western Hills of Nepal.
Lewis, L., 1988. Measurement and assessment of soil loss in Rwanda. Catena Suppl. 12, Lumle, Kaski, Nepal technical paper No 2000/3.
151–165. https://eurekamag.com/research/001/882/001882094.php. Van Dijk, A.I.J.M., Bruijnzeel, L.A., 2004. Runoff and soil loss from bench terraces. 2. An
Lewis, L.A., 1992. Terracing and accelerated soil loss on Rwandian steeplands: a event- based erosion process model. Eur. J. Soil Sci. 55 (2), 317–334. https://doi.
preliminary investigation of the implications of human activities affecting soil org/10.1111/j.1365-0754.2004.00605.x.
movement. Land Degrad. Dev. 3 (4), 241–246. https://doi.org/10.1002/ Van Dijk, P., Kwaad, F.J.P.M., Klapwijk, M., 1996. Retention of water and sediment by
ldr.3400030405. grass strips. Hydrol. Process. 10 (8), 1069–1080. https://doi.org/10.1002/(SICI)
Lewis, L.A., Nyamulinda, V., 1996. The critical role of human activities in land 1099-1085(199608)10:8<1069::AID-HYP412>3.0.CO;2-4.
degradation in Rwanda. Land Degrad. Dev. 7 (1), 47–55. https://doi.org/10.1002/ Van Ranst, E., Delvaux, B., Baert, G., Imerzoukène, S., Jamagne, P., Ndayiragije, S.,
(SICI)1099-145X(199603)7:1<47::AID-LDR213>3.0.CO;2-M. Pineros Garcet, J., 2000a. Carte pédologique du Rwanda, feuille 09: gakenke. Gent:
Li, L., Du, S., Wu, L., Liu, G., 2009. An overview of soil loss tolerance. Catena 78 (2), universiteit Gent. Vakgroep Geologie en Bodemkunde. Laboratoire de Pédologie,
93–99. https://doi.org/10.1016/j.catena.2009.03.007. Gand. ISBN 90-76769-17-6.
MINAGRI, 2013. Strategic Plan for the Transformation of Agriculture in Rwanda Phase Van Ranst, E., Delvaux, B., Baert, G., Imerzoukène, S., Jamagne, P., Ndayiragije, S.,
III (2013-2017) (Kigali, Rwanda). Pineros Garcet, J., 2000b. Carte pédologique du Rwanda, feuille 25: zaza. Gent:
MINAGRI, 2014. National Fertilizer Policy in Rwanda. Ministry of Agriculture and universiteit Gent. Vakgroep Geologie en Bodemkunde.
Animal resources, Kigali, Rwanda. Van Ranst, E., Verloo, M., Demeyer, A., Pauwels, M.J., 2004. Manual for the Soil
Moges, D.M., Taye, A.A., 2017. Determinants of farmers’ perception to invest in soil and Chemistry and Fertility Laboratory-Analytical Methods for Soils and Plants,
water conservation technologies in the North-Western Highlands of Ethiopia. Int. Equipment, and Management of Consumables, vol. 835. NUGI, Ghent, Belgium. ISBN
Soil Water Conserv. Res. 5 (1), 56–61. https://doi.org/10.1016/j.iswcr.2017.02.003. 90- 76603-01-4.
Moradi, A., Teh, C., Sung, B., Joo, K., Husni, A., Hanif, M., Fauziah, C., 2015. Effect of Wagesho, N., Claire, M., 2016. Analysis of rainfall intensity-duration-frequency
four soil and water conservation practices on soil physical processes in a non- relationship for Rwanda. J. Water Resour. Protect. 8 (7), 706–723. https://doi.org/
terraced oil palm plantation. Soil Tillage Res. 145, 62–71. https://doi.org/10.1016/ 10.4236/jwarp.2016.87058.
j.still.2014.08.005. Wang, Y., Tan, S., Liu, B., Yang, Y., 2017. Estimating rainfall erosivity by incorporating
Morgan, R., 1995. Soil Erosion and Conservation, second ed. Longman publishing Group: seasonal variations in parameters into the Richardson model. J. Geogr. Sci. 27 (3),
Harlow and J. Wiley &. Sons, New York. 275–296. https://doi.org/10.1007/s11442-017-1376-6.
Mupenzi, J., Li, L., Jiwen, G., Habumugisha, J., Habiyaremye, G., Ngamije, J., Welle, S., Chantawarangul, K., Nontananandh, S., Jantawat, S., 2006. Effectiveness of
Baragahoranye, I., 2012. Radical terraces in Rwanda. East African J. Sci. Technol. 1 grass strips as barrier against runoff and soil loss in Jijiga area, northern part of
(1), 53–58. http://www.eajscience.com. ISSN 2227-1902. Somali Region, Ethiopia. Kasetsart J./Nat. Sci. 40, 549–558.

11
J. Rutebuka et al. Journal of Environmental Management 277 (2021) 111369

Wickama, J., Okoba, B., Sterk, G., 2014. Effectiveness of sustainable land management Electronics, Communications and Networks. CECNet), XianNing, pp. 2943–2946.
measures in West Usambara highlands, Tanzania. Catena 118, 91–102. https://doi. https://doi.org/10.1109/CECNET.2011.5769391.
org/10.1016/j.catena.2014.01.013. Zhang, J.H., Wang, Y., Li, F.C., 2016. Soil organic carbon and nitrogen losses due to soil
Widomski, M., 2011. Terracing as a measure of soil erosion control and its effect on erosion and cropping in a sloping terrace landscape. Soil Res. 53 (1), 87–96. https://
improvement of infiltration in eroded environment. In: Godone, D., Stanchi, S. doi.org/10.1071/SR14151.
(Eds.), Soil Erosion Issues in Agriculture, ISBN 978-953-307-435-1, pp. 314–335. Zuazo, V.H.D., Pleguezuelo, C.R.R., Peinado, F.J.M., Graaff, J. De, Martínez, J.R.F.,
Wischmeier, W.H., Smith, D.D., 1978. Predicting Rainfall Erosion Losses: a Guide to Flanagan, D.C., 2011. Environmental impact of introducing plant covers in the
Conservation Planning. U.S. Department of Agriculture. https://doi.org/10.1029/ taluses of terraces : implications for mitigating agricultural soil erosion and runoff.
TR039i002p00285. Handbook No. 537, 1 69. Catena 84 (1–2), 79–88. https://doi.org/10.1016/j.catena.2010.10.004.
Yongmei, D., Xihuan, S., Xianghong, G., Shijun, N., Juanjuan, M., 2011. Analysis of slope
length on water soil erosion. In: 2011 International Conference on Consumer

12

You might also like