Agriculture, Ecosystems and Environment 203 (2015) 127–139

Contents lists available at ScienceDirect

Agriculture, Ecosystems and Environment

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

Soil erosion in the humid tropics: A systematic quantitative review

Nicolas Labrière a,b, * , Bruno Locatelli a,c , Yves Laumonier a,c, Vincent Freycon a ,
Martial Bernoux d
a
UPR BSEF, CIRAD (Centre de Coopération Internationale en Recherche Agronomique pour le Développement), Avenue Agropolis, 34398 Montpellier Cedex 5,
France
b
AgroParisTech, Doctoral School ABIES, 19 Avenue du Maine, 75732 Paris Cedex 15, France
c
Center for International Forestry Research, Jalan CIFOR, Situ Gede, Sindang Barang, Bogor (Barat) 16115, Indonesia
d
UMR Eco&Sols, IRD, 2 place Viala, 34060 Montpellier Cedex 2, France

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

Article history: Healthy soils provide a wide range of ecosystem services. But soil erosion (one component of land
Received 22 October 2014 degradation) jeopardizes the sustainable delivery of these services worldwide, and particularly in the
Received in revised form 26 January 2015 humid tropics where erosion potential is high due to heavy rainfall. The Millennium Ecosystem
Accepted 30 January 2015
Assessment pointed out the role of poor land-use and management choices in increasing land
Available online xxx
degradation. We hypothesized that land use has a limited influence on soil erosion provided vegetation
cover is developed enough or good management practices are implemented. We systematically reviewed
Keywords:
the literature to study how soil and vegetation management influence soil erosion control in the humid
Ecosystem services
Systematic review
tropics. More than 3600 measurements of soil loss from 55 references covering 21 countries were
Quantitative analysis compiled. Quantitative analysis of the collected data revealed that soil erosion in the humid tropics is
Landscape dramatically concentrated in space (over landscape elements of bare soil) and time (e.g. during crop
Land use rotation). No land use is erosion-prone per se, but creation of bare soil elements in the landscape through
Land-use type particular land uses and other human activities (e.g. skid trails and logging roads) should be avoided as
Management practices much as possible. Implementation of sound practices of soil and vegetation management (e.g. contour
planting, no-till farming and use of vegetative buffer strips) can reduce erosion by up to 99%. With limited
financial and technical means, natural resource managers and policy makers can therefore help decrease
soil loss at a large scale by promoting wise management of highly erosion-prone landscape elements and
enhancing the use of low-erosion-inducing practices.
ã 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction disturbed hydrological regimes such as increased flood risk due

to riverbed filling and stream plugging (Chomitz and Kumari, 1998;
The ecosystem service of soil erosion control, for the delivery of Lal, 2003; Millennium Ecosystem Assessment, 2005; Morgan,
which vegetation cover plays an important role, has been 2005; Locatelli et al., 2011).
degrading worldwide (Millennium Ecosystem Assessment, Research on factors influencing soil loss has resulted in widely
2005). As this regulating service is lost, soil formation can no used models, such as the RUSLE (revised universal soil loss
longer compensate for soil loss due to an increase in erosion, which equation). This model was built from plot data of experiments
depletes soil resources and the ecosystem services they support carried out in the United States and predicts soil loss from climatic
(Lal, 2003; Morgan, 2005). The Millennium Ecosystem Assessment (rainfall erosivity), edaphic (soil erodibility) and topographic
(2005) identified unwise land-use choices and harmful crop or soil (slope length and slope steepness) factors, as well as soil and
management practices as the major drivers of increasing soil vegetation management practices (Wischmeier and Smith, 1978;
erosion. Soil erosion has multiple on- and off-site consequences Renard et al., 1997). Management of soil and vegetation has long
such as decreasing crop yields, increasing atmospheric CO2 been recognized as the most efficient and effective way to
concentration, decreasing water quality (turbidity and influence the extent of soil loss, and therefore soil erosion control
particle-born pollutants), sedimentation of reservoirs, and (Goujon, 1968).
The humid tropics are rich in carbon and biodiversity
and attract major attention because of the rapid loss of
* Corresponding author. Tel.: +33 467 593 726; fax: +33 467 593 909. rainforests (Strassburg et al., 2010; Saatchi et al., 2011; Tropek
E-mail address: [email protected] (N. Labrière). et al., 2014). Because of the large amount and high intensity of

http://dx.doi.org/10.1016/j.agee.2015.01.027
0167-8809/ã 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
128 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139

rainfall in the humid tropics, soil erosion can potentially reach (2005) showed that erosion risk prediction was poorly achieved by
dramatic levels in this region (El-Swaify et al., 1982; Lal, 1990). the USLE (universal soil loss equation) in a watershed of western
Tropical ecosystems with healthy soils can support multiple Kenya, and called for ground surveys to properly calibrate the USLE
ecosystem services (e.g. water regulation, climate regulation and similar empirical models.
through carbon storage and biodiversity support) and support In the face of this lack of agreement, studies that directly
local livelihoods. A better understanding of soil erosion control in measure soil loss are of great interest as they can help shed light on
the humid tropics is therefore vital (Locatelli et al., 2014). the influence of vegetation and soil management on soil erosion
Theoretically, empirical models of erosion prediction should control. Synthesizing and analyzing available data from multiple
only be applied under conditions and for purposes similar to those sources is necessary given the diversity of study contexts and the
of their development (e.g. predicting erosion from croplands in the impossibility of drawing general conclusions from a single study.
United States for the RUSLE). Adapting an empirical model to Such syntheses are available for some regions of the world.
out-of-range conditions would require parameter calibration, Focussing on Europe and the Mediterranean, Maetens et al. (2012)
which can consume both time and resources (Nearing et al., reviewed data from 227 stations and 1056 soil erosion plots to
1994). While some studies have adapted temperate model factors analyze the effect of land use on erosion and runoff. They found that
to their own geographical contexts (e.g. Streck and Cogo, 2003 for (semi-) natural vegetation produced lower erosion (<1 Mg/ha/yr)
surface soil consolidation and Diodato et al., 2013 for rainfall than vegetation directly influenced by human activities
erosivity), others have directly applied models developed for a (e.g. croplands and vineyards; 6–20 Mg/ha/yr). Montgomery
temperate context to predict soil erosion in the humid tropics (2007) also compiled erosion data from globally distributed studies
(e.g. Angima et al., 2003; Hoyos, 2005). (some in the humid tropics) and showed that conventional
Yet there is little consensus about the direct applicability of agriculture, i.e. with tillage, produced 10–100 times more soil loss
models such as RUSLE (and its predecessors) to a tropical context. than conservation agriculture, i.e. with no-tillage, but conditions
Despite over- and under-estimation of soil loss depending on the were highly variable. For example, plots under conventional
cropping phase, Almas and Jamal (2000) found the RUSLE model to agriculture were more erosion-prone (with maximum slope of 37
correctly predict the overall soil loss from a banana–pineapple and maximum annual precipitation of 5600 mm/yr) than those of
intercropping system in Malaysia. On the other hand, Cohen et al. plots under conservation agriculture (17 and 2000 mm/yr).

Table 1
Land-use types and subtypes.

Land-use type Land-use subtype Definitions

Bare Land has been opened and kept bare for various reasons (includes pre-sowing and post-harvesting cropland
and skid trails).
Tilled High-disturbance soil management techniques (e.g. ploughing and raking) are used.
Untilled Low-disturbance soil management techniques (e.g. slash and burn and weeding with a knife) are used.

Cropland Crops are sown and harvested within a single agricultural year, sometimes more than once (excludes
perennial crops).
Crop, non-established, without Crop was recently planted and crop cover is not developed; no conservation techniques are practiced.
conservation practices
Crop, established, without Crop cover is developed; no conservation techniques are practiced.
conservation practices
Crop with vegetation-related Crop cover may or may not be fully developed. Vegetation-related conservation techniques (e.g. hedgerows,
conservation practices intercropping and mulching) are practiced.
Crop with vegetation- and soil-related Crop cover may or may not be fully developed. Both vegetation-related (e.g. hedgerows, intercropping and
conservation practices mulching) and soil-related (e.g. no-till farming and contour planting) conservation techniques are practiced.

Grassland Vegetation is dominated by grasses (includes open grasslands and pastures).

Pasture Land is used for grazing and managed through agricultural practices such as seeding, irrigation and use of
fertilizer.
Open grassland Land is unmanaged and has no trees or shrubs.

Shrubland Vegetation is dominated by shrubs but can also include grasses, herbs and geophytes.
Open shrubland A transitional plant community occurs temporarily as the result of a disturbance such as logging or fire.

Tree-dominated Planted vegetation is dominated by trees, including perennial tree crops such as rubber, fruit and nut trees.
agrosystem
Tree plantation A group of planted trees is grown in the form of an agricultural crop, usually with the aim of harvesting
wood.
Tree crop without contact cover A permanent crop has been planted; it has no contact cover (such as grass or cover crops) underneath.
Tree crop with contact cover A permanent crop has been planted and has contact cover (such as grass or cover crops) underneath.
Simple agroforest One woody perennial species is planted with one annual crop.
Complex agroforest Multiple species of woody perennials, often with natural vegetation regrowth, are planted (usually
intercropped) with annual crops.

Forest Ground is covered with natural vegetation dominated by trees (excludes tree plantations).
Secondary forest Forest has regenerated naturally after clear-cutting, burning or other land-clearing activities and contains
vegetation in early successional stages.
Old-growth forest Forest is ecologically mature, containing trees of various sizes and species (the last stage in forest
succession).
Logged-over forest Forest has been logged-over.
Degraded forest Forest has been degraded by human activities other than logging or by a naturally occurring event such as a
fire or severe storm.
 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139 129

Selecting erosion measurements available for the two agriculture erosion measurement method, the database shrank to 114 refer-
types under the same conditions substantially reduced the sample. ences. Finally, after screening the full texts of those references, we
No synthesis (to our knowledge) has been done so far for the kept 55 of them (more details are available in Appendix B). For each
humid tropics. The purpose of this study was therefore to reference, we retrieved data on soil loss (expressed as quantity of
quantitatively analyze available data (collected via systematic soil mass per unit of area) in one or more cases. A case was defined
review of the literature) on soil erosion in the humid tropics to as one erosion measurement, characterized by an associated
study how soil and vegetation management influence soil erosion measurement method (profile meter, root exposition, sediment
control in this region. Effects of the measurement protocol trap, unbounded plot or runoff plot, all with natural rainfall, and
(method, duration and area) and context (rainfall, slope length, runoff plot with simulated rainfall), area and duration, topograph-
slope steepness and soil erodibility) were controlled for to keep a ical features (slope length and steepness), rainfall, and land-use
consistent dataset and focus on the influence of soil and vegetation type and subtype (see definitions in Table 1). For each case,
management on soil erosion. building on the classification proposed by Moench (1991),
The underlying hypothesis is that land use has a limited influence vegetation cover was also described by the presence or absence
on soil erosion provided vegetation cover is developed enough or of four layers: high (4 m), intermediate (at least 1 m but <4 m),
good management practices are implemented. This hypothesis was low (at least 0.1 m but <1 m) and ground (<0.1 m).
previously conclusively tested in a few single studies on ecosystems The final data set consisted of 3649 measurements from
such as rangelands (e.g. Snelder and Bryan, 1995; Chartier and 55 references covering 21 countries in the humid tropics (Fig. 1,
Rostagno, 2006), but never systematically nor for the humid tropics. Table 2). Most references originated from peer-reviewed journals
This study aims to contribute to the scientific understanding of the (n = 44) and used runoff plots to quantify soil loss (n = 48).
relationship between soil erosion and vegetation/soil management Publication years ranged from 1973 to 2012, with half of the
in the humid tropics, to help clarify the applicability of widely used references published before 1997 (Fig. 2a). The number of cases per
models such as the RUSLE, and to provide to stakeholders involved in study was highly variable, and the six references with the most
natural resource management and protection a synthesis on soil cases contributed half the total number of cases in the final data set
erosion control and its sound management. (Fig. 2b, Table 2). Study length ranged from two days (studies under
simulated rainfall) to 17 years (Fig. 2c). References generally
2. Materials and methods reported erosion values per rainfall event, per year or for the
duration of the study (Fig. 2d). Most references assessed one to
2.1. Materials three land-use types (Fig. 2e), of which bare soils and croplands
were the most studied (Fig. 2f).
We searched for studies of erosion in the humid tropics, defined Rainfall erosivity and soil erodibility were assessed for each
for the purpose of this review as the “Af” (tropical rainforest case. An indicator of rainfall erosivity sensu Renard et al. (1997)
climate) and “Am” (tropical monsoon climate) Köppen climatic could not be obtained or computed for most cases because
classes (Köppen, 1936; Peel et al., 2007). Queries were built on the monthly data were not available or because measurement duration
conjunction of elements from three thematic clusters: “scope” and was too short to apply an annual erosivity index. We thus used total
“outcome” and “measurement”. The “scope” cluster corresponded rainfall as an indicator of rainfall erosivity based on the finding by
to: tropic* or region (list of broadly defined relevant regions, e.g. Maetens et al. (2012) that soil loss does not correlate better with
Africa) or specific country (all countries under either Af or Am erosivity indices than with total rainfall.
climate were considered, e.g. Brazil). The “outcome” cluster For soil erodibility, we combined different indices because of
encompassed the following terms: soil erosion, water erosion, the diverse ways soils were described in the studies. For each case,
soil loss, soil depletion, land degradation, sedimentation, sediment we calculated three soil erodibility indices from soil texture and
production and siltation. The “measurement” cluster included organic matter data with an empirical table and two different
keywords defining methodological approaches and measurement equations (Stewart et al., 1975; Sharpley and Williams, 1990; Torri
methods such as “runoff plot” and “sediment trap”. In order to et al., 1997). If soil data were not available in a study, we extracted
select studies with homogeneous land use; we excluded measures them from the ISRIC global soil dataset (resolution of 1 km) using
at the catchment scale. Additionally, to avoid bias in the analysis of measurement coordinates (ISRIC-World Soil Information, 2013).
reported measurements, indirect measures and estimates (e.g. the For each index, soils were split into low-, medium- and high-
use of 137Cs as a tracer—see Sidle et al., 2006) were not considered. erodibility classes of equal sizes. A soil was then classified as highly
As suggested by the Collaboration for Environmental Evidence erodible if it was considered highly erodible by at least two of the
(2013), a variety of peer-reviewed and grey literature sources were three indices, low if it was considered low by at least two indices
searched. Details about queries and sources are available in and medium otherwise (more details are available in Appendix C).
Appendix A. Queries were carried out during the second half of
April 2013 in English, French and Spanish. 2.2. Data analysis
Searches led to 5183 references after removing duplicates. After
irrelevant references were removed, based on information in All data transformation and statistical analysis were done using
article titles and abstracts about topic, geographical scope and R (R Core Team, 2013). Due to highly skewed distributions, all

Fig. 1. Location of study sites (n = 61). Some dots represent several references, and some references contribute more than one dot. Red dots show locations provided by the six
references with the most cases. Af (tropical rainforest) climate ranges are displayed in dark blue and Am (tropical monsoon) climate ranges in light blue. (For interpretation of
the references to color in this figure legend, the reader is referred to the web version of this article.)
130 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139

Table 2
Contributing references by geographical location. References from Southeast Asia and Northeast Australia (n = 29) made up more than half of all references (n = 55). The
30 references with the fewest cases provided about 10% of all cases (n = 3649). The 6 references with the most cases are printed in bold.

Reference Country Source type Method Rainfall type Soil Land-use Cases Case time frame(s) Study length
dataa type(s)b
Africa (n = 11)
Ambassa-Kiki and Nill Cameroon Journal article Runoff plot Natural ST + OM 3 (B, C, T) 3 Study 2 years
(1999)
Boye and Albrecht (2004) Kenya Project report Runoff plot Simulation ST + OM 1 (B) 10 Rainfall event 2 days
Collinet (1983) Côte d’Ivoire Project report Runoff plot Natural None 2 (C, F) 24 Year, study 3 years
Collinet (1988) Côte d’Ivoire PhD thesis Runoff plot Simulation None 2 (B, C) 189 Rainfall event 2 months
Defersha and Melesse Kenya Journal article Runoff plot Natural ST + OM 3 (B, C, G) 87 Rainfall event, month 1 month
(2012)
Kamara (1986) Sierra Leone Journal article Runoff plot Natural ST + OM 2 (B, C) 14 Month 2 years
Lundgren (1980) Tanzania Journal article Runoff plot Natural ST + OM 2 (F, T) 33 Year, study 2 years
Ngatunga et al. (1984) Tanzania Journal article Runoff plot Natural ST + OM 3 (B, C, G) 36 Season, year 1 year
Odemerho and Nigeria Journal article Runoff plot Natural ST 2 (C, G) 126 Rainfall event, study 5 months
Avwunudiogba (1993)
Roose (1973) Côte d’Ivoire PhD thesis Runoff plot Natural None 5 (B, C, F, G, 431 Rainfall event, day, month, 17 years
T) season, year
Våje et al. (2005) Tanzania Journal article Runoff plot Natural ST + OM 2 (B, C) 10 Rainfall event, season 2 years

America & North Pacific Ocean (n = 10)

Alegre and Cassel (1996) Peru Journal article Runoff plot Natural OM 3 (B, C, F) 4 Study 52 months
Alegre and Rao (1996) Peru Journal article Runoff plot Natural OM 3 (B, C, F) 50 Season, year, study 5 years
Bellanger et al. (2004) Venezuela Journal article Runoff plot Natural ST + OM 3 (B, C, T) 41 Rainfall event, week, season 5 months
Dangler and El-Swaify USA (Hawaii) Journal article Runoff plot Simulation None 1 (B) 16 Rainfall event 1.75 years
(1976)
Francisco-Nicolas et al. Mexico Journal article Runoff plot Natural OM 1 (C) 18 Year, study 8 years
(2006)
Fritsch and Sarrailh France (French Journal article Runoff plot Natural None 2 (B, F) 38 Month, season, year, study 32 months
(1986) Guiana)
McGregor (1980) Colombia Journal article Runoff plot natural ST 3 (C, F, G) 7 Study 8 week
Ruppenthal et al. (1997) Colombia Journal article Runoff plot Natural None 2 (B, C) 32 Season 2 years
Sarrailh (1981) France (French Project report Runoff plot Natural None 2 (F, G) 50 Month, season, year, study 20 months
Guiana)
Wan and El-Swaify USA (Hawaii) Journal article Runoff plot Simulation ST + OM 2 (B, C) 6 Rainfall event 2 days
(1999)

SE Asia & NE Australia (n = 29)

Afandi et al. (2002a) Indonesia Journal article Runoff plot Natural ST + OM 1 (T) 54 Month 3.5 years
Afandi et al. (2002b) Indonesia Journal article Sediment Natural ST + OM 4 (C, F, G, T) 77 Month, study 11 months
trap
Almas and Jamal (2000) Malaysia Journal article Runoff plot Natural None 3 (B, C, T) 52 Season 9 months
Baharuddin et al. (1995) Malaysia Journal article Runoff plot Natural None 3 (B, F, G) 90 Month, year 2 years
Bons (1990) Indonesia Conference Runoff plot Natural None 2 (S, T) 2 Year, study 26 months
proceedings
Chatterjea (1998) Singapore Journal article Runoff plot Natural None 2 (B, G) 30 Rainfall event 1.3 years
Comia et al. (1994) Philippines Journal article Runoff plot Natural ST + OM 1 (C) 16 Year, study 3 years
Daño and Siapno (1992) Philippines Conference Runoff plot Natural None 1 (T) 22 Year, study 2 years
proceedings
Hartanto et al. (2003) Indonesia Journal article Runoff plot Natural None 2 (B, F) 135 Rainfall event, season 2.5 months
Hashim et al. (1995) Malaysia Journal article Runoff plot Natural ST + OM 2 (B, T) 152 Rainfall event, season, study 1.5 years
Jaafar et al. (2011) Malaysia Journal article Runoff plot Natural ST + OM 1 (F) 6 Year 1 year
Leigh (1982) Malaysia Journal article Sediment Natural ST 1 (F) 11 Year 1 year
trap
Malmer (1996) Malaysia Journal article Unbounded Natural None 2 (B, F) 3 Year, study 1 year
plot
Moehansyah et al. (2004) Indonesia Journal article Runoff plot Natural ST 3 (C, G, T) 156 Rainfall event, season, study 8 months
Moench (1991) India Journal article Runoff plot Natural OM 1 (T) 21 Study 9 months
Pandey and Chaudhari India Journal article Runoff plot Natural ST 3 (C, F, T) 44 Year, study 3 years
(2010)
Paningbatan et al. (1995) Philippines Journal article Runoff plot Natural ST + OM 1 (C) 168 Rainfall event, season 3 years
Poudel et al. (1999) Philippines Journal article Runoff plot Natural ST + OM 1 (C) 35 Season, study 2.5 years
Poudel et al. (2000) Philippines Journal article Runoff plot Natural OM 1 (C) 12 Year 2.5 years
Presbitero (2003) Philippines PhD thesis Runoff plot Natural OM 2 (B, C) 433 Rainfall event 2.5 years
Prove et al. (1995) Australia Journal article Profile meter Natural None 1 (C) 14 Year 6 years
Ross and Dykes (1996) Brunei Book chapter Runoff plot Natural ST 1 (F) 24 Month 8 months
Shimokawa (1988) Indonesia Book chapter Root Natural None 1 (F) 21 Year 1 year
exposition
Siebert and Belsky (1990) Indonesia Journal article Runoff plot Natural ST + OM 1 (C) 3 Season 9 months
Sinun et al. (1992) Malaysia Journal article Runoff plot Natural None 3 (B, F, G) 78 Month, year 1 year
Sudarmadji (2001) Indonesia Conference Runoff plot Natural ST 1 (F) 3 Study 4 months
proceedings
Syed Abdullah and Al- Malaysia Journal article Sediment Natural ST + OM 1 (F) 12 Year 1 year
Toum (2000) trap
van der Linden (1980) Indonesia Journal article Runoff plot Natural ST + OM 3 (B, C, G) 88 Rainfall event, study 3 months
Verbist et al. (2010) Indonesia Journal article Runoff plot Natural ST + OM 2 (F, T) 18 Year 4 years
 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139 131

Table 2 (Continued)
Reference Country Source type Method Rainfall type Soil Land-use Cases Case time frame(s) Study length
dataa type(s)b
Caribbean islands (n = 5)
Khamsouk (2001) France PhD thesis Runoff plot Natural, ST + OM 3 (B, C, T) 429 Rainfall event 1.5 years
(Martinique) simulation
Larsen et al. (1999) USA (Puerto Journal article Unbounded Natural ST 3 (B, G, S) 177 Month, season, year 3.75 years
Rico) plot
McDonald et al. (2002) Jamaica Journal article Runoff plot Natural ST + OM 3 (B, C, F) 24 Year, study 5 years
Mohammed and Gumbs Trinidad and Journal article Runoff plot Natural ST + OM 2 (B, C) 6 Rainfall event, season 3 months
(1982) Tobago
Ramos Santana et al. USA (Puerto Journal article Runoff plot Natural None 3 (B, G, T) 8 Month 1 month
(2003) Rico)
a
ST: soil texture; OM: organic matter.
b
B: bare; C: cropland; G: grassland; F: forest; S: shrubland; T: tree-dominated agrosystem.

continuous variables (erosion, duration, area, rainfall, slope length number of cases, we tested whether they had a dominant effect on
and slope steepness) were log10-transformed to normalize their the overall results. To do so, we reanalyzed the data after removing
distribution. If not specified, further mention of values of these these references one by one, but no significant changes in the
variables will refer to their log10-transformed values. Because null results and no changes in the findings were observed.
values cannot be log10-transformed, each null value of measured
soil loss (664 values, expressed in g after transforming values 3. Results
reported in other units in the papers) was replaced by a random
value taken from a uniform distribution in the range of 0.001–1 g, Soil loss was maximum on bare soils and strikingly exceeded
an interval arbitrarily chosen in which 1 g represents a measure- that of all other land-use types (Fig. 3). Minimum soil loss was
ment detection threshold (Chiappetta et al., 2004). After substi- found in forests. Croplands had the second highest soil loss value
tuting the null values, measured soil loss (g) was converted into soil among land-use types. Mean soil loss values for grasslands and
loss per unit of area and per year (g/m2/yr). Replicating the shrublands were about half that of croplands. The ratio (of
substitution process 10 times, we checked that the randomness of geometric means in the natural scale) shrank to 1:3 for mean soil
the data replacement did not affect the subsequent results. loss between tree-dominated agrosystems and croplands. The
In order to analyze the effect of soil or vegetation management erosion rate in forests was ca. one-tenth and one-150th than that of
on soil erosion, we controlled first for the effect of the croplands and bare soils, respectively. The ratio of soil loss values
measurement protocol (method, duration and area) (Hair et al., between two consecutive land uses (sorted by decreasing mean
2006). Annual soil loss values obtained from extrapolation of soil loss) was much higher between bare soils and croplands
measures taken over a single rain event are likely to be larger than (ca. 20:1) than between other land-use types (ratios below 3:1).
values from measures over one year, and soil loss values per unit of Soil loss differed significantly between subtypes of land uses
area are probably higher in small plots than in larger areas because within the same type. Soil loss was minimum for tree crops with
of sediment deposition (Boix-Fayos et al., 2006). We used only the contact cover (e.g. grass or cover crop) and maximum on tilled bare
two quantitative descriptors of measurement protocol (area and soils, with a ratio of 1:1,200 between the two values (Fig. 4). Among
duration), as they were good proxies for method (60% correct bare soils, soil loss was 40% higher with tillage than without (the
determination, jackknifed classification following discriminant latter still had a high absolute value of soil loss). Among croplands,
function analysis). We transformed the log10 values of soil loss and recently planted crops without vegetation-related conservation
context variables (rainfall, soil erodibility, slope length and slope practices (e.g. hedgerows, mulching or intercropping) had erosion
steepness) into the residuals resulting from a linear regression rates similar to those of bare soils (either tilled or not), whereas
against duration, area and the interaction between the two well-established crops on similar lands reduced soil loss by 89% on
variables (all three significant at p < 0.001; Table D1). Residuals average. Vegetation-related conservation practices reduced soil
were further adjusted to correspond to a reference protocol of loss by 93% in recently planted cropland but did not reduce soil loss
measurements over one year and 100 m2 (this value corresponding significantly in land with established crops. Simultaneous soil- and
to the order of magnitude of the median area). vegetation-related conservation practices (e.g. no-till farming and
We then controlled for the effect of context on soil loss by hedgerows) decreased soil loss in croplands (up to 99% compared
calculating the residuals of a general linear model relating soil loss to to no conservation practices in land with recently planted crops).
context (values of rainfall, slope length and slope steepness, after Among tree-dominated agrosystems, tree crops with contact
factoring out the effects of protocol, as well as soil erodibility classes). cover faced 99% less soil loss on average than tree crops without
All the context variables had a significant effect on soil loss (p < 0.05; contact cover. Simple agroforests had greater soil loss than complex
Table D2). The residuals were adjusted to a “reference scenario” ones (3:1 ratio); however, the difference was not significant. Among
with the median values for annual rainfall (exclusively from cases the five least erosion-prone land-use subtypes, three were of forest
where rainfall was measured for one year or more), slope length, type (old-growth, secondary, and logged-over forests).
slope steepness (back-transformed values being 2444 mm, 16.4 m The number of layers constituting the vegetation cover had a
and 16.5%, respectively), and a soil erodibility of class “medium”. significant impact on soil loss. Soil loss was maximal without any
All subsequent statistical analyses (ANOVA and Tukey’s HSD) layer and minimal with four layers. Soil loss was one-tenth as much
used these log10-transformed soil loss values, corrected for the with one layer as without, and one-70th as much with two layers as
effect of the measurement protocol and context and scaled to without (Fig. 5). The 90% reduction in soil loss between one and two
correspond to a reference scenario. We tested for differences (at layers was also significant. Conversely, no significant difference in
p < 0.001) in soil loss depending on (1) land-use type, (2) land-use mean soil loss was found between two and four layers.
subtype and (3) the number and (4) nature of layers constituting The type of layers constituting the vegetation cover had a
the vegetation cover. As six references provided half the total significant impact on soil loss. The presence of high, intermediary,
132 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139

(a) (b)
5

15
4

3
Frequency

Frequency
10

5
1

0 0
1980 1990 2000 2010 0 100 200 300 400
Year of publication Number of cases per reference

(c) (d)
25

20
10

15
Frequency

Frequency

5 10

0 0
0 5 10 15 r. event day week month season study year
Length of the study (years) Case time frames

(e) (f)

30

15

20
Frequency

Frequency

10

10
5

0 0
1 2 3 4 5 6 bare cropland forest grassland shrubland tree-dom.
Number of land-use types per reference Land-use type

Fig. 2. Frequency distribution of (a) year of publication of the contributing references (n = 55), (b) number of cases per reference (total cases = 3649), (c) length of the study, (d)
case time frames, (e) number of land-use types investigated per reference, (f) land-use types investigated. Total for (d) >55 because some references provide data on more than
one time frame; total for (f) >55 because most references reported on more than one land use. R. event: rainfall event; tree-dom.: tree-dominated agrosystem.
 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139 133

Fig. 3. Impact of land-use type on soil loss under reference scenario (significant difference at p < 0.001). Geometric means along with 95% confidence intervals on the natural
scale are plotted on a log10 scale for the sake of readability (bottom panel). Log10-transformed mean soil loss values with the same letter are not significantly different (Tukey’s
HSD, p < 0.01). Geometric means are also plotted on the natural scale to highlight the loss of soil erosion control from cropland to bare land (top panel). Tree-dom.:
tree-dominated agrosystem.

low and ground layers influenced soil loss significantly and to be ca. 20:1 in the humid tropics, the ratio ranged from 2:1 to
differently (Table 3): soil loss under a unique layer of high 10:1 in Europe and the Mediterranean (Cerdan et al., 2010;
vegetation (4 m) was twice that occurring on bare soils, whereas Maetens et al., 2012). This suggests that soil erosion control is still
other layers decreased soil loss compared to bare soils by a factor of provided in the humid tropics to a certain extent for crop-
5, 8 and 5 for intermediary, low and ground layers respectively, and and grass-dominated land uses but is alarmingly depleted in
a factor of 200 for a combination of the three layers. bare soils, with dramatic consequences on soil loss. The 2-order-
of-magnitude difference in soil loss between one and zero
4. Discussion vegetation layer also suggests that some vegetation cover is
necessary for soil erosion control to be provided. Consequently,
4.1. Soil erosion is concentrated in space and time bare soils should be avoided at all times.
The abrupt loss of soil erosion control depicted in Figs. 3–5
Soil erosion control can abruptly be lost when vegetation cover suggests that, in most land uses, erosion is concentrated spatially
is not developed enough and/or when poor soil and vegetation (over bare soil, e.g. logging roads or non-protected crop fields
management practices are implemented (Figs. 3–5). While we between rotations) and temporally (e.g. before vegetation is fully
found the ratio of soil loss values between bare soils and croplands established). Soil loss was lowest in plots under tree crops with
134 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139

Fig. 4. Impact of land-use subtype on soil loss under reference scenario (significant difference at p < 0.001). Geometric means along with 95% confidence intervals on the
natural scale are plotted on a log10 scale for the sake of readability (bottom panel). Log10-transformed mean soil loss values with the same letter are not significantly different
(Tukey’s HSD, p < 0.01). Geometric means are also plotted on the natural scale to highlight the loss of soil erosion control from tree crops with contact cover to tilled bare soils
(top panel). B: bare; C: cropland; G: grassland; F: forest; S: shrubland; T: tree-dominated agrosystem; estab.: established; VCP: vegetation-related conservation practice(s);
V&SCP: vegetation- and soil-related conservation practice(s).

contact cover, but such crops might not be totally erosion-neutral. Although this finding has been reported before (Sarrailh, 1981;
Similarly, the fact that soil loss in logged-over forests is not Baharuddin et al., 1995; Anderson and Macdonald, 1998; Bruijnzeel
different from that in old-growth forests should not lead to the et al., 1998; Rijsdijk, 2005; Defersha and Melesse, 2012), our study
delusive conclusion that logging does not increase soil erosion. brings a strong quantitative endorsement to it because of the
Bare soil elements exclusively related to logging and farming (e.g. number of studies and cases taken into consideration.
roads and trails) contribute to disproportionately increase the Studies investigating the consequences of land-use changes for
overall erosion rate of such activities (e.g. Rijsdijk, 2005; Gómez- soil erosion often used a synchronic approach (comparing different
Delgado, 2010). Much attention should therefore be given to land uses in different plots to infer the consequences of a conversion,
managing these elements (e.g. through water diversion, use of in a single plot, from one land use to the other). Unlike a diachronic
vegetative buffer strips and trail consolidation) so as to reduce the approach measuring soil loss before, during and after land use
overall impact of such activities. change (e.g. Fritsch and Sarrailh, 1986; Malmer, 1996), a synchronic
Attention must also be given to temporal transitions between approach does not record the transition (e.g. through clear-cutting or
land uses, for example when establishing crops or plantations. tillage) from one land use to the other. This transition appears to be
 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139 135

Fig. 5. Impact of the number of vegetation layers on soil loss under reference scenario (significant difference at p < 0.001). Geometric means along with 95% confidence
intervals on the natural scale are plotted on a log10 scale for the sake of readability (bottom panel). Log10-transformed mean soil loss values with the same letter are not
significantly different (Tukey’s HSD, p < 0.01). Geometric means are also plotted on the natural scale to highlight the loss of soil erosion control from one layer of vegetation to
none (top panel).

critical for understanding the consequences of land-use changes for In this respect, the sequence of land uses—bare untilled,
soil loss in the humid tropics, where vegetation regrowth is rapid but cropland, open grassland, open shrubland, secondary forest and
most of the annual soil loss is potentially caused by a limited number old-growth forest—can be interpreted as snapshots of different
of extreme rainfall events (e.g. Poudel et al., 1999; Defersha and successional stages following shifting cultivation (after clearing,
Melesse, 2012). Comparing synchronic and diachronic approaches cultivation, and subsequent natural regeneration). This review
for soil carbon sequestration assessment, Costa Junior et al. (2013) showed that soil erosion decreased along the sequence, attesting to
found that results depended on the selected approach, and the recovery of soil erosion control. Martin et al. (2013) highlighted
recommended use of the diachronic approach whenever possible. a similar increasing trend for carbon storage and plant diversity
Because of intrinsic variations in soil characteristics (e.g. texture) during post-disturbance forest recovery. This suggests a synergy
between sites under the same land use or management practice, a (or a joint increase in multiple ecosystem services following
diachronic approach should always be preferred. On the other hand, implementation of a practice—forest regeneration in this case)
a synchronic approach using multiple replicates makes it possible to between soil erosion control, carbon storage and plant diversity.
highlight trends in the consequences of land use change or But the evaluation of a wider range of ecosystem services
management. (including e.g. water regulation) is advised so as to avoid
136 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139

Table 3 soil loss in the humid tropics (although intuitively one would
Coefficients of the generalized linear model regression of annual soil loss (log10-
expect lower soil erosion under natural vegetation). For example,
transformed values) against presence/absence of high (4 m), intermediate
(1 m  height < 4 m), low (0.1 m  height <1 m) and ground (<0.1 m) vegetation we found that soil loss in old-growth forest is higher than in tree
layers. crops with contact cover. Soil erosion is a natural phenomenon that
also occurs in old-growth forest despite its complex vegetation
Estimate Standard error p
structure and high ground cover (mostly leaf litter or wood debris).
***
Intercept (bare) 2.97 0.044
**
In Tanzania, Lundgren (1980) suggested that good land manage-
High 0.22 0.071
Intermediary 0.66 0.054 *** ment practices (e.g. mulching and no burning) accounted for lower
Low 0.91 0.058 *** erosion rates in agrosystems than in natural forest, even though
***
Ground 0.71 0.068 this observation was made during normal rainfall conditions and it
Adjusted R2: 0.204 was impossible to predict how the human-managed system would
Number of observations: 3649
have reacted to extreme rainfall events. In South Andaman island,
**
p < 0.01. Pandey and Chaudhari (2010) showed that coconut plantations
***
p < 0.001. with a contact cover of Pueraria phaseoloides had similar soil loss as
nearby native evergreen forest and therefore recommended the
promoting measures (e.g. afforestation) that would be detrimental use of contact cover in plantations for soil erosion control on
for the delivery of other services. the island.
Our quantitative analysis strongly supports the idea that no
4.2. What matters in soil erosion control by vegetation? land use (except bare soils) is erosion-prone per se and that sound
management of soil and vegetation can reduce soil erosion in
The change of slope in Fig. 4 highlights four land uses in which managed areas to levels even lower than in areas under natural
soil erosion control is depleted. In addition to two situations of bare vegetation.
soils, recently planted croplands without vegetation-related
conservation practices also provide a low level of soil erosion 4.4. Differences in soil erosion control between tropical vs. temperate
control. This highlights the importance of good management of regions
croplands: vegetation-related conservation practices (such as
hedgerows) can ensure that, even during inter- or early-rotation Comparing the effect of land use on soil erosion in the humid
periods when crop cover is not yet developed, erosion can be tropics (this review) and in temperate regions (Renard et al., 1997;
prevented or minimized. Burke and Sugg, 2006), we found that changes in soil erosion
Tree crops without contact cover also provide critically low levels control along a gradient of land uses had similar shape in both
of soil erosion control, which is confirmed by the analysis of the effect temperate and tropical areas (Fig. 6). A difference between these
of vegetation layers: the presence of a sole high layer increases climatic zones is observed in grasslands and croplands, where soil
erosion compared to bare soil. This is consistent with other studies erosion control is higher in the humid tropics than suggested by
that pointed out the role of tree canopy in modifying rainfall kinetic the RUSLE. Our analysis shows a much more pronounced threshold
energy (e.g. Wiersum, 1985; Brandt, 1988; Calder, 2001). Leaves of effect in the relation between vegetation and soil erosion control
the canopy layer help break the kinetic energy of raindrops, but than given by the RUSLE, which suggests that soil erosion is more
secondary drops falling from the canopy (particularly from large concentrated in space and time in the humid tropics than
leaves) are often larger than the raindrops and reach the ground with elsewhere. The difference can be explained by the more rapid
a higher kinetic energy than in areas without a canopy layer development of dense vegetation protecting soil in croplands and
(Wiersum, 1985; Brandt, 1988). This results in increased soil erosion, grasslands of the humid tropics. Because of the “universal” nature
particularly when the canopy is high and there is no understorey of the mechanism of soil erosion, the RUSLE, an empirically-based
vegetation. Teak (Tectonia grandis L.f.) plantations, for example, have model that integrates all the factors known to influence soil
often been associated with high erosion rates because of lack of erosion (e.g. soil erodibility, rainfall erosivity), could potentially be
understorey and large tree leaves (Calder, 2001). But a recent study
showed that poor vegetation and soil management rather than
intrinsic teak leaf morphology was responsible for those high
erosion rates (Fernández-Moya et al., 2014).
Litter and understorey both help break the kinetic energy of
raindrops and therefore decrease splash erosion (Brandt, 1988).
Multiple layers of vegetation are necessary in plantations to
minimize soil erosion, and non-compliance with sound manage-
ment rules (e.g. the repeated use of fire to clear ground cover and
understorey) directly and dramatically increases soil loss (Wier-
sum, 1984). Overall, whatever the land use, we found low and
ground layers of vegetation to be essential in decreasing soil loss
(Table 3). This is consistent with plot-derived results from
northern Vietnam, which identified a critical value of understorey
biomass (130 g/m2) above which soil loss was negligible (Anh et al.,
2014). Therefore, low and ground covers should be restored and/or
maintained whatever the land use.

4.3. Soil erosion under human-impacted or managed vs. natural

vegetation
Fig. 6. Ratio of cover-management factors for the RUSLE for 5 different land uses
This study also showed that the difference between “human- (reference being erosion on bare soils), and ratio of soil loss per land use to soil loss
impacted or managed” and “natural” vegetation does not explain on bare soils from our systematic review (SR).
 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139 137

used to predict soil erosion for any geographical context. But should be conducted to calibrate model parameters. As a
factors’ parameters were computed from data collected exclusively preliminary step to answer the UN call for action to reverse land
in temperate regions and the direct application of the RUSLE to a degradation (UN, 2012), we stress the need to establish standard
tropical context would lead to soil loss misestimation especially for measurement procedures for soil erosion and influencing factors,
croplands and grasslands. Properly calibrating all RUSLE factors’ to mirror what was achieved for terrestrial carbon measurement
parameters (especially those related to soil and vegetation (Walker et al., 2012). For improving soil and vegetation
management) using data acquired in a tropical context is therefore management, uncovered or unprotected soils should be avoided
critical to achieve accurate prediction of soil erosion in the humid at all times, and low and ground layers of vegetation should be
tropics. restored and/or maintained whatever the land use.
No land use (except bare soils) is erosion-prone per se and
4.5. Limitations of the study natural resource managers and policy makers need to promote
sound management of soil and vegetation (e.g. contour planting,
This analysis faced challenges related to data availability. As no-till farming, intercropping and use of cover crops) to reduce soil
soils were sometimes poorly described, we had to use a global loss from erosion-prone landscape elements. Because of the
database to estimate texture and carbon content, which probably relative affordability and simplicity of such management practices,
influenced the accuracy of our soil erodibility indices. The structure substantial decrease in soil loss can be attained at the catchment or
of the vegetation cover (e.g. number and height of layers, planting regional scale with limited financial and technical means. Since soil
density and presence or absence of ground cover) was not always erosion appears to decrease during the different phases of forest
well described. For example, Sinun et al. (1992) studied an regeneration, soil ecosystem services (e.g. nutrient cycling, flood
abandoned logging track where a sharp decrease in soil loss was regulation, water purification), the delivery of which is greater in
recorded over time; but while soil loss was measured on a monthly healthier soils, might be good candidates for ecosystem services
basis over one year, vegetation was not described over time. Two bundling with biodiversity protection and carbon storage.
noticeable exceptions were Khamsouk (2001) and Presbitero
(2003), in which vegetation cover was regularly and systematically Acknowledgements
estimated, but with different approaches (e.g. crown cover and
contact cover). We thank the library staffs of the Center for International
The aim of this study was to quantitatively analyze soil erosion Forestry Research and Centre de Coopération Internationale en
control in the whole humid tropics, but references only covered Recherche Agronomique pour le Développement (CIRAD) for their
21 countries and some sub-regions were critically under- help in retrieving some studies used in this systematic review. We
represented, e.g. the Brazilian part of the Amazon and the Congo also thank Ervan Rutishauser and Ghislain Vieilledent for their
basin (Fig. 1, Table 2). Yet Köppen climatic classes “Af” and “Am” are recommendations about statistical analysis. We are grateful to the
homogeneous in term of temperature, rainfall pattern and two anonymous reviewers for their constructive comments that
vegetation type (Köppen, 1936), which supports the applicability helped to improve the quality of the manuscript. This research
of this study’s findings to under-represented sub-regions. Research received financial support from CIRAD, AusAID (Agreement 63650
should nevertheless be carried out in the Amazon and the with the Center for International Forestry Research), ABIES (Ecole
Congo basin to document the effect of local human activities Doctorale Agriculture, Alimentation, Biologie, Environnements et
(e.g. small- and large-scale agriculture, fuelwood collection and Santé) and CRP-FTA (Consortium Research Program on Forests,
industrial logging) on soil erosion. Trees, and Agroforestry).
Because six references (from four countries) represented half
the total number of cases, we tested for their dominant effect on Appendix A. Supplementary data
the overall results, but no such effect was found; this further
supports the relevance of this study to the whole humid tropics. Supplementary data associated with this article can be found, in
Mean annual soil loss values in this study appeared to be in the line the online version, at http://dx.doi.org/10.1016/j.agee.2015.01.027.
of benchmarks provided by other studies. For example, annual
erosion rates ranged from 0.1 to 90 and 3 to 750 Mg/ha in humid References
West Africa for croplands and bare soils, respectively (Morgan,
2005), compared to 1 and 16 Mg/ha on average in our analysis. Afandi, Manik, T.K., Rosadi, B., Utomo, M., Senge, M., Adachi, T., Oki, Y., 2002a. Soil
erosion under coffee trees with different weed management in humid tropical
Other benchmarks are 0.03 to 6.2, 0.1 to 5.6, and 1.2 to 183 Mg/ha
hilly area of Lampung, South Sumatra, Indonesia. J. Jpn. Soc. Soil Phys. 91, 13–14.
for old-growth forests and tree crops with and without Afandi, Rosadi, B., Maryanto, Nurarifani, Utomo, M., Senge, M., Adachi, T., 2002b.
contact cover, respectively (Wiersum, 1984), compared to 0.1, Sediment yield from various land use practices in a hilly tropical area of the
Lampung region, South Sumatra, Indonesia. J. Jpn. Soc. Soil Phys. 91, 25–38.
2 and 5 Mg/ha in our analysis.
Alegre, J.C., Cassel, D.K., 1996. Dynamics of soil physical properties under alternative
Since we used log10-transformed data to carry out statistical systems to slash-and-burn. Agric. Ecosyst. Environ. 58, 39–48.
analyses, back-transforming means led to geometric means in the Alegre, J.C., Rao, M.R., 1996. Soil and water conservation by contour hedging in the
natural scale that are intrinsically less sensitive to extreme values humid tropics of Peru. Agric. Ecosyst. Environ. 57, 17–25.
Almas, M., Jamal, T., 2000. Use of RUSLE for soil loss prediction during different
(Bland and Altman, 1996). This explains the fact that our values lie growth periods. Pak. J. Biol. Sci. 3, 118–121.
in the lower part of the range. Ambassa-Kiki, R., Nill, D., 1999. Effects of different land management techniques on
selected topsoil properties of a forest Ferralsol. Soil Tillage Res. 52, 259–264.
Anderson, D.M., Macdonald, L.H., 1998. Modelling road surface sediment production
5. Conclusion using a vector geographic information system. Earth Surf. Processes Landforms
23, 95–107.
Soil erosion in the humid tropics is dramatically concentrated 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. Agric. Ecosyst.
both spatially (over bare soil) and temporally (before vegetation Environ. 97, 295–308.
cover establishes), and low and ground layers of vegetation are Anh, P.T.Q., Gomi, T., MacDonald, L.H., Mizugaki, S., Khoa, P.V., Furuichi, T., 2014.
essential in mitigating soil erosion. Because soil erosion appears Linkages among land use, macronutrient levels, and soil erosion in northern
Vietnam: a plot-scale study. Geoderma 232, 352–362.
more concentrated in space and time in the humid tropics than
Baharuddin, K., Mokhtaruddin, A.M., Nik Muhamad, M., 1995. Surface runoff and soil
elsewhere, models developed in temperate regions should not be loss from a skid trail and a logging road in a tropical forest. J. Trop. For. Sci. 7,
directly applied in the humid tropics, and thorough research 558–569.
138 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139

Bellanger, B., Huon, S., Velasquez, F., Vallès, V., Girardin, C., Mariotti, A., 2004. Fritsch, J.M., Sarrailh, J.M., 1986. Les transports solides dans l’écosystème forestier
Monitoring soil organic carbon erosion with d13C and d15N on experimental tropical humide guyanais: effets du défrichement et de l’aménagement de
field plots in the Venezuelan Andes. Catena 58, 125–150. pâturages. Cah. ORSTOM, Sér Pédol. 22, 209–222.
Bland, J.M., Altman, D.G., 1996. Transformations, means, and confidence intervals. Gómez-Delgado, F., 2010. Hydrological, Ecophysiological and Sediment Processes in
Br. Med. J. 312, 1079. a Coffee Agroforestry Basin: Combining Experimental and Modelling Methods
Boix-Fayos, C., Martínez-Mena, M., Arnau-Rosalén, E., Calvo-Cases, A., Castillo, V., to Assess Hydrological Environmental Services. Montpellier, France.
Albaladejo, J., 2006. Measuring soil erosion by field plots: understanding the Goujon, P., 1968. Conservation des sols en Afrique et à Madagascar: 1ère partie: les
sources of variation. Earth Sci. Rev. 78, 267–285. facteurs de l’érosion et l’équation universelle de Wischmeier. Bois For. Trop. 118,
Bons, C.A., 1990. Accelerated erosion due to clearcutting of plantation forest and 3–17.
subsequent Taungya cultivation in upland West Java, Indonesia. In: Ziemer, R.R., Hair, J.F., Tatham, R.L., Anderson, R.E., Black, W., 2006. Multivariate Data Analysis,
O’Loughlin, C.L., Hamilton, L.S., International Union of Forestry Research 6th Ed Pearson Prentice Hall, Upper Saddle River, NJ.
Organizations (Eds.), Research Needs and Applications to Reduce Erosion and Hartanto, H., Prabhu, R., Widayat, A.S.E., Asdak, C., 2003. Factors affecting runoff and
Sedimentation In Tropical Steeplands. IAHS Publication No. 192. International soil erosion: plot-level soil loss monitoring for assessing sustainability of forest
Association of Hydrological Sciences, Wallingford, Oxfordshire, UK, pp. 279–288. management. For. Ecol. Manage. 180, 361–374.
Boye, A., Albrecht, A., 2004. Soil erodibility control and soil carbon losses under Hashim, G.M., Ciesiolka, C.A.A., Yusoff, W.A., Nafis, A.W., Mispan, M.R., Rose, C.W.,
short term tree fallows in western Kenya. Bull. Réseau Eros. 23, 123–143. Coughlan, K.J., 1995. Soil erosion processes in sloping land in the east coast of
Brandt, J., 1988. The transformation of rainfall energy by a tropical rain-forest Peninsular Malaysia. Soil Technol. 8, 215–233.
canopy in relation to soil-erosion. J. Biogeogr. 15, 41–48. Hoyos, N., 2005. Spatial modeling of soil erosion potential in a tropical watershed of
Bruijnzeel, L.A., Van Eijk, B., Purwanto, E., 1998. Runoff and soil loss from bench the Colombian Andes. Catena 63, 85–108.
terraces in upland West Java, Indonesia and implications for process modelling. ISRIC - World Soil Information, 2013. SoilGrids: an automated system for global soil
In: Summer, W., Klaghofer, E., Zhang, W. (Eds.), Modelling Soil Erosion, Sediment mapping. Available for download at http://soilgrids1km.isric.org.
Transport and Closely Related Hydrological Processes. IAHS Publication No. 249. Jaafar, O., Syed Abdullah, S.M., Al-Toum, S., 2011. The Tekala Forest Reserve: a study
International Association of Hydrological Sciences, Wallingford, Oxfordshire, on surface wash and runoff using close system erosion plots. Geogr. Malay. J.
UK, pp. 211–220. Soc. Space 7, 1–13.
Burke, L., Sugg, Z., 2006. Hydrologic modeling of watersheds discharging adjacent to Kamara, C.S., 1986. Mulch-tillage effects on soil loss and soil properties on an ultisol
the Mesoamerican Reef. World Resources Institute, Washington, DC. in the humid tropics. Soil Tillage Res. 8, 131–144.
Calder, I.R., 2001. Canopy processes: implications for transpiration, interception and Khamsouk, B., 2001. Impact de la culture bananière sur l’environnement: influence
splash induced erosion, ultimately for forest management and water resources. des systèmes de cultures bananières sur l’érosion, le bilan hydrique et les pertes
Plant Ecol. 153, 203–214. en nutriments sur un sol volcanique en Martinique (cas du sol brun rouillé à
Cerdan, O., Govers, G., Le Bissonnais, Y., Van Oost, K., Poesen, J., Saby, N., Gobin, A., halloysite). Montpellier, France.
Vacca, A., Quinton, J., Auerswald, K., Klik, A., Kwaad, F.J.P.M., Raclot, D., Ionita, I., Köppen,W.,1936.DasgeographiscaSystemderKlimate.In:Köppen,W.,Geiger,G.(Eds.),
Rejman, J., Rousseva, S., Muxart, T., Roxo, M.J., Dostal, T., 2010. Rates and spatial Handbuch der klimatologie. Gebrüder Borntraeger, Berlin, Germany, pp. 1–44.
variations of soil erosion in Europe: a study based on erosion plot data. Lal, R., 1990. Soil erosion in the tropics: principles and management. McGraw-Hill,
Geomorphology 122, 167–177. New York.
Chartier, M.P., Rostagno, C.M., 2006. Soil erosion thresholds and alternative states in Lal, R., 2003. Soil erosion and the global carbon budget. Environ. Int. 29, 437–450.
northeastern patagonian rangelands. Rangeland Ecol. Manage. 59, 616–624. Larsen, M.C., Torres-Sánchez, A.J., Concepción, I.M., 1999. Slopewash, surface runoff
Chatterjea, K., 1998. The impact of tropical rainstorms on sediment and runoff and fine-litter transport in forest and landslide scars in humid-tropical
generation from bare and grass-covered surfaces: a plot study from Singapore. steeplands luquillo experimental forest, Puerto Rico. Earth Surf. Processes
Land Degrad. Dev. 9, 143–157. Landforms 24, 481–502.
Chiappetta, P., Roubaud, M.C., Torrésani, B., 2004. Blind source separation and the Leigh, C., 1982. Sediment transport by surface wash and throughflow at the Pasoh
analysis of microarray data. J. Comput. Biol. 11, 1090–1109. Forest Reserve Negri Sembilan, Peninsular Malaysia. Geogr. Ann. Ser. A. Phys.
Chomitz, K.M., Kumari, K., 1998. The domestic benefits of tropical forests: a critical Geogr. 64, 171–180.
review. World Bank Res. Obser. 13, 13–35. Locatelli, B., Imbach, P., Vignola, R., Metzger, M.J., Hidalgo, E.J.L., 2011. Ecosystem
Cohen, M.J., Shepherd, K.D., Walsh, M.G., 2005. Empirical reformulation of the services and hydroelectricity in Central America: modelling service flows with
universal soil loss equation for erosion risk assessment in a tropical watershed. fuzzy logic and expert knowledge. Reg. Environ. Change 11, 393–404.
Geoderma 124, 235–252. Locatelli, B., Imbach, P., Wunder, S., 2014. Synergies and trade-offs between
Collaboration for Environmental Evidence, 2013. Guidelines for Systematic Review ecosystem services in Costa Rica. Environ. Conserv. 41, 27–36.
and Evidence Synthesis in Environmental Management. Version 4.2. Lundgren, L., 1980. Comparison of surface runoff and soil loss from runoff plots in
Environmental Evidence: www.environmentalevidence.org/Authors.htm. forest and small-scale agriculture in the Usambara Mts. Tanzania. Geogr. Ann.
Collinet, J., 1983. Hydrodynamique superficielle et érosion comparées des sols Ser. A. Phys. Geogr. 62, 113–148.
représentatifs des sites forestiers et cultivés de la station écologique de Taï (Sud- Maetens, W., Vanmaercke, M., Poesen, J., Jankauskas, B., Jankauskien, G., Ionita, I.,
Ouest ivoirien): premier bilan sur parcelles expérimentales recevant des pluies 2012. Effects of land use on annual runoff and soil loss in Europe and the
naturelles (campagnes 1978-1979-1980) et simulées (campagne de novembre Mediterranean: a meta-analysis of plot data. Prog. Phys. Geogr. 36, 597–651.
1978 et mars 1979). Cahiers ORSTOM. Malmer, A., 1996. Hydrological effects and nutrient losses of forest plantation
Collinet, J., 1988. Comportements hydrodynamiques et érosifs de sols de l’Afrique de establishmentontropicalrainforestlandinSabah,Malaysia.J.Hydrol.174,129–148.
l’Ouest: évolution des matériaux et des organisations sous simulations de Martin, P.A., Newton, A.C., Bullock, J.M., 2013. Carbon pools recover more quickly
pluies. Strasbourg, France. than plant biodiversity in tropical secondary forests. Proc. R. Soc. Lond., Ser. B:
Comia, R.A., Paningbatan, E.P., Håkansson, I., 1994. Erosion and crop yield response Biol. Sci. 280, 20132236.
to soil conditions under alley cropping systems in the Philippines. Soil Tillage McDonald, M.A., Healey, J.R., Stevens, P.A., 2002. The effects of secondary forest
Res. 31, 249–261. clearance and subsequent land-use on erosion losses and soil properties in the
Costa Junior, C., Corbeels, M., Bernoux, M., Piccolo, M.C., Neto, M.S., Feigl, B.J., Cerri, Blue Mountains of Jamaica. Agric. Ecosyst. Environ. 92, 1–19.
C.E.P., Cerri, C.C., Scopel, E., Lal, R., 2013. Assessing soil carbon storage rates McGregor, D.F.M., 1980. An investigation of soil erosion in the Colombian rainforest
under no-tillage: comparing the synchronic and diachronic approaches. Soil zone. Catena 7, 265–273.
Tillage Res. 134, 207–212. Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-Being:
Dangler, E.W., El-Swaify, S.A., 1976. Erosion of selected Hawaii soils by simulated Synthesis. Island Press, Washington, DC.
rainfall. Soil Sci. Soc. Am. J. 40, 769–773. Moehansyah, H., Maheshwari, B.L., Armstrong, J., 2004. Field Evaluation of Selected
Daño, A.M., Siapno, F.E., 1992. The effectiveness of soil conservation structures in Soil Erosion Models for Catchment Management in Indonesia. Biosys. Eng. 88,
steep cultivated mountain regions of the Philippines. In: Walling, D.E., Davies, T. 491–506.
R.H., Hasholt, B. (Eds.), Erosion, Debris flows and Environment in Mountain Moench, M., 1991. Soil erosion under a successional agroforestry sequence: a case
Regions. IAHS Publication No. 209. International Association of Hydrological study from Idukki District, Kerala, India. Agrofor. Syst. 15, 31–50.
Sciences, Wallingford, Oxfordshire, UK, pp. 399–405. Mohammed, A., Gumbs, F.A., 1982. The effect of plant spacing on water runoff: soil
Defersha, M.B., Melesse, A.M., 2012. Field-scale investigation of the effect of land use erosion and yield of maize (Zea mays L.) on a steep slope of an ultisol in Trinidad.
on sediment yield and runoff using runoff plot data and models in the Mara J. Agric. Eng. Res. 27, 481–488.
River basin, Kenya. Catena 89, 54–64. Montgomery, D.R., 2007. Soil erosion and agricultural sustainability. Proc. Natl.
Diodato, N., Knight, J., Bellocchi, G., 2013. Reduced complexity model for assessing Acad. Sci. USA 104, 13268–13272.
patterns of rainfall erosivity in Africa. Global Planet. Change 100, 183–193. Morgan, R.P.C., 2005. Soil Erosion and Conservation. Wiley-Blackwell, Oxford, UK.
El-Swaify, S.A., Dangler, E.W., Armstrong, C.L., 1982. Soil erosion by water in the Nearing, M.A., Lane, L.J., Lopes, V.L., 1994. Modeling soil erosion. In: Lal, R. (Ed.), Soil
tropics. College of Tropical Agriculture and Human Resources. University of Erosion: Research Methods. Soil and Water Conservation Society and St. Lucie
Hawaii, Honolulu. Press, Ankeny, Iowa, pp. 127–156.
Fernández-Moya, J., Alvarado, A., Forsythe, W., Ramírez, L., Algeet-Abarquero, N., Ngatunga, E.L.N., Lal, R., Uriyo, A.P., 1984. Effects of surface management on runoff
Marchamalo-Sacristán, M., 2014. Soil erosion under teak (Tectona grandis L.f.) and soil erosion from some plots at Mlingano, Tanzania. Geoderma 33, 1–12.
plantations: general patterns, assumptions and controversies. Catena 123, Odemerho, F.O., Avwunudiogba, A., 1993. The effects of changing cassava
236–242. management practices on soil loss: a Nigerian example. Geogr. J. 159, 63–69.
Francisco-Nicolas, N., Turrent-Fernandez, A., Oropeza-Mota, J.L., Martinez-Menes, Pandey, C.B., Chaudhari, S.K., 2010. Soil and nutrient losses from different land uses
M.R., Cortes-Flores, J.I., 2006. Soil loss and erosion-productivity relationships in and vegetative methods for their control on hilly terrain of South Andaman.
four soil management systems. Terra Latinoamericana 24, 253–260. Indian J. Agric. Sci. 80, 399–404.
 N. Labrière et al. / Agriculture, Ecosystems and Environment 203 (2015) 127–139 139

Paningbatan, E.P., Ciesiolka, C.A.A., Coughlan, K.J., Rose, C.W., 1995. Alley cropping for Siebert, S.F., Belsky, J.M., 1990. Bench terracing in the Kerinci uplands of Sumatra,
managing soil erosion of hilly lands in the Philippines. Soil Technol. 8, 193–204. Indonesia. J. Soil Water Conserv. 45, 559–562.
Peel, M.C., Finlayson, B.L., McMahon, T.A., 2007. Updated world map of the Sinun, W., Meng, W.W., Douglas, I., Spencer, T., 1992. Throughfall, stemflow,
Köppen-Geiger climate classification. Hydrol. Earth Syst. Sci. 11, 1633–1644. overland flow and throughflow in the Ulu Segama Rain Forest, Sabah, Malaysia.
Poudel, D.D., Midmore, D.J., West, L.T., 1999. Erosion and productivity of vegetable Philos. Trans. R. Soc. Lond., Ser. B: Biol. Sci. 335, 389–395.
systems on sloping volcanic ash-derived Philippine soils. Soil Sci. Soc. Am. J. 63, Snelder, D.J., Bryan, R.B., 1995. The use of rainfall simulation tests to assess the
1366–1376. influence of vegetation density on soil loss on degraded rangelands in the
Poudel, D.D., Midmore, D.J., West, L.T., 2000. Farmer participatory research to Baringo District, Kenya. Catena 25, 105–116.
minimize soil erosion on steepland vegetable systems in the Philippines. Agric. Stewart, B.A., Woolhiser, D.A., Wischmeier, W.H., Caro, J.H., Frere, M.H., 1975. Control
Ecosyst. Environ. 79, 113–127. of Water Pollution from Cropland. U.S. Environmental Protection Agency,
Presbitero, A.L., 2003. Soil Erosion Studies on Steep Slopes of Humid-tropic Washington, DC.
Philippines. Griffith, Australia. Strassburg, B.B.N., Kelly, A., Balmford, A., Davies, R.G., Gibbs, H.K., Lovett, A., Miles, L.,
Prove, B.G., Doogan, V.J., Truong, P.N.V., 1995. Nature and magnitude of soil erosion Orme, C.D.L., Price, J., Turner, R.K., Rodrigues, A.S.L., 2010. Global congruence of
in sugarcane land on the wet tropical coast of north-eastern Queensland. Aust. J. carbon storage and biodiversity in terrestrial ecosystems. Conserv. Lett. 3,
Exp. Agric. 35, 641–649. 98–105.
R Core Team, 2013. R: A language and environment for statistical computing. R Streck, E.V., Cogo, N.P., 2003. Reconsolidation of the soil surface after tillage
Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, discontinuity, with and without cultivation, related to erosion and its prediction
URL http://www.R-project.org/. with RUSLE. Rev. Bras. Cienc. Solo 27, 141–151.
Ramos Santana, R., Martínez, G., Macchiavelli, R., Rodríguez, J.E., Guzmán, J.L., 2003. Sudarmadji, T., 2001. Impact of logging and forest fires on soil erosion in tropical
Potential of trees grasses, and turf legumes for restoring eroded soils. Commun. humid forest in Kalimantan. Rehabilitation of Degraded Tropical Forest
Soil Sci. Plant Anal. 34, 2149–2162. Ecosystems: Workshop Proceedings, 2–4 November 1999, Center for
Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D., Yoder, D., 1997. Predicting soil International Forestry Research, Bogor, Indonesia, pp. 35–44.
erosion by water: a guide to conservation planning with the revised universal Syed Abdullah, S.M., Al-Toum, S., 2000. A study on surface wash and runoff using
soil loss equation (RUSLE). Agriculture Handbook No. 703. U.S. Department of open system erosion plots. Pertanika J. Trop. Agric. Sci. 23, 43–53.
Agriculture, Washington DC. Torri, D., Poesen, J., Borselli, L., 1997. Predictability and uncertainty of the soil
Rijsdijk, A., 2005. Evaluating sediment sources and delivery in a tropical volcanic erodibility factor using a global dataset. Catena 31, 1–22.
watershed. In: Horowitz, A.J., Walling, D.E. (Eds.), Sediment Budgets 1. IAHS Tropek, R., Sedlacek, O., Beck, J., Keil, P., Musilova, Z., Simova, I., Storch, D., 2014.
Publication No. 291. International Association of Hydrological Sciences, Comment on High-resolution global maps of 21st-century forest cover change.
Wallingford, Oxfordshire, UK, pp. 1–9. Science 344, 981.
Roose, E., 1973. Dix-sept années de mesures expérimentales de l’érosion et du UN, 2012. Conference on sustainable development, Outcome of the conference, A/
ruissellement sur un sol ferrallitique sableux de basse Côte d’Ivoire: CONF.216/L.1. Rio de Janeiro, Brazil.
contribution à l’étude de l’érosion hydrique en milieu intertropical. Abidjan, Våje, P.I., Singh, B.R., Lal, R., 2005. Soil erosion and nutrient losses from a
Côte d’Ivoire. volcanic ash soil in Kilimanjaro Region, Tanzania. J. Sustainable Agric. 26,
Ross, S.M., Dykes, A., 1996. Soil conditions: erosion and nutrient loss on steep slopes 95–117.
under mixed dipterocarp forest in Brunei Darussalam. Monogr. Biol. 74, van der Linden, P., 1980. The application of a parametric grey box type approach to
259–270. investigate surface runoff and erosion during rainstorms. An erosion plot case
Ruppenthal, M., Leihner, D.E., Steinmüller, N., El-Sharkawy, M.A., 1997. Losses of study in central Java, Indonesia. Indones. J. Geogr. 10, 23–42.
organic matter and nutrients by water erosion in cassava-based cropping Verbist, B., Poesen, J., van Noordwijk, M., Widianto Suprayogo, D., Agus, F., Deckers,
systems. Exp. Agric. 33, 487–498. J., 2010. Factors affecting soil loss at plot scale and sediment yield at catchment
Saatchi, S.S., Harris, N.L., Brown, S., Lefsky, M., Mitchard, E.T.A., Salas, W., Zutta, B.R., scale in a tropical volcanic agroforestry landscape. Catena 80, 34–46.
Buermann, W., Lewis, S.L., Hagen, S., Petrova, S., White, L., Silman, M., Morel, A., Walker, S.M., Pearson, T.R.H., Casarim, F.M., Harris, N., Petrova, S., Grais, A., Swails, E.,
2011. Benchmark map of forest carbon stocks in tropical regions across three Netzer, M., Goslee, K.M., Brown, S., 2012. Standard Operating Procedures for
continents. Proc. Natl. Acad. Sci. USA 108, 9899–9904. Terrestrial Carbon Measurement: Version 2012. Winrock International,
Sarrailh, J.M., 1981. Parcelles élémentaires d’étude du ruissellement et de l’érosion: Arlington, Virginia.
analyse des résultats obtenus durant les deux premières campagnes de mesure. Wan, Y., El-Swaify, S.A., 1999. Runoff and soil erosion as affected by plastic mulch in a
Ecosyst. For. Guyan. Bull. Liaison Groupe Trav. ECEREX 4, 45–51. Hawaiian pineapple field. Soil Tillage Res. 52, 29–35.
Sharpley, A.N., Williams, J.R., 1990. EPIC-erosion/productivity impact calculator: 1. Wiersum, K.F., 1984. Surface erosion under various tropical agroforestry systems. In:
Model documentation. USDA Technical Bulletin No. 1768. U.S. Department of O’Loughlin, C.L., Pearce, A.J. (Eds.), Effects of Forest Land Use on Erosion and
Agriculture, Washington, DC. Slope Stability. International Union of Forestry Research Organizations, Austria,
Shimokawa, E., 1988. Effect of a fire of tropical rainforest on soil erosion. In: Tagawa, pp. 231–239.
H., Nirawan, N. (Eds.), A Research on the Process of Earlier Recovery of Tropical Wiersum, K.F., 1985. Effects of various vegetation layers in an Acacia auriculiformis
Rainforest after a Large Scale Fire in Kalimantan Timur, Indonesia. Occasional forest plantation on surface erosion in Java, Indonesia. In: El-Swaify, S.A.,
paper N 14, Kagoshima University, Research Center for the South Pacific, Japan, Moldenhauer, W.C., Lo, A. (Eds.), Soil Erosion and Conservation. Soil
pp. 2–11. Conservation Society of America, Ankeny, Iowa, pp. 79–89.
Sidle, R.C., Ziegler, A.D., Negishi, J.N., Nik, A.R., Siew, R., Turkelboom, F., 2006. Erosion Wischmeier, W.H., Smith, D.D., 1978. Predicting rainfall erosion losses: a guide to
processes in steep terrain – truths myths, and uncertainties related to forest conservation planning. Agriculture Handbook No. 537. U.S. Department of
management in Southeast Asia. For. Ecol. Manage. 224, 199–225. Agriculture, Washington, DC.