Ecological Engineering 85 (2015) 132–143
Contents lists available at ScienceDirect
Ecological Engineering
journal homepage: www.elsevier.com/locate/ecoleng
Biophysical and anthropogenous determinants of landscape patterns
and degradation of plant communities in Mo hilly basin (Togo)
Badabate Diwediga a,b,∗ , Kperkouma Wala b , Fousseni Folega b , Marra Dourma b ,
Yao A. Woegan b , Koffi Akpagana b , Quang Bao Le c
a
West African Science Service Centre on Climate Change and Adapted Land Use (WASCAL) Graduate Research Programme, Department of Civil Engineering,
Kwame Nkrumah University of Science and Technology, Kumasi, Ghana
b
Laboratory of Botany and Plant Ecology, University of Lomé, 01 BP, 1515 Lomé 01, Togo
c
CGIAR Research Program in Dryland Systems, International Centre for Agricultural Research in Dry Areas (ICARDA), Amman 11195, Jordan
a r t i c l e
i n f o
Article history:
Received 14 April 2015
Received in revised form 8 September 2015
Accepted 13 September 2015
Keywords:
Landscape patterns
DCA
Eco-edaphical factors
Human disturbances
Ecological conservation
Mo basin
Togo
West Africa
a b s t r a c t
In mountainous areas, biophysical settings and human disturbances strongly influence landscape patterns and dynamics requiring a permanent understanding of their combined influence. In this study, we
investigated the diversity and patterns of wild landscapes in relation to ecological factors, human disturbances and land protection regime in the Mo river basin (Central Togo). First, we used geographical
information systems (GIS) and remote sensing techniques to quantify and spatially explicit the major
land cover types occurring in the area. We performed a supervised classification of Landsat 8 image from
2014. Next, we used vegetation ordination and classification methods to detect vegetation group types
and their similarity level from forest inventory data collected at 75 sites. Data from soil samples at the
75 sites and DEM-based topographical indices were used as biophysical variables to analyze factors of
current landscape structure. Both satellite image classification and the outputs from ordination methods
indicated that three major vegetation types (forestlands, woodlands and savannahs/shrubs) occurred
in protected (PA) and unprotected (UPA) areas. Image classification showed that savannahs/shrubs are
the most widespread vegetation types (54.4%) while forestlands and woodlands cover 10.4% and 26.4%
of the total area, respectively. UPA showed high rates of human disturbances that shaped the occurrence of a fourth vegetation type made up mainly by degraded savannahs and woodlands. Along the
land protection gradient, the landscapes are driven by soil nutrients and moisture, in combination with
the influence of topography and human disturbances. In both PA and UPA, and along protection gradient, majority of features exhibited significant differences among plant communities. The spatial analyses
combined with the field data providing information on vegetation cover, species richness, and human
footprint indices suggested that some areas outside protection still exhibit high potentials for land conservation. In multifunctional landscapes of Mo basin, conservation strategies could also be encouraged
in the wild landscapes of community common lands to promote both biodiversity conservation and
sustainable provision of ecosystem services.
© 2015 Elsevier B.V. All rights reserved.
1. Introduction
Diverse landscapes are host of biodiversity and provide a wide
range of ecosystem services, requiring knowledge on the socialecological interactions. Understanding the interactions between
human imprints and biophysical components defining the landscape heterogeneity has been considered as a fundamental for
∗ Corresponding author at: 01 P.O. Box, 1515 Lomé 01, Togo.
E-mail addresses: diwedigaba@gmail.com, diwedigaba@yahoo.fr (B. Diwediga),
kpwala75@yahoo.fr (K. Wala), q.le@cgiar.org, q.le@alumini.ethz.ch (Q.B. Le).
http://dx.doi.org/10.1016/j.ecoleng.2015.09.059
0925-8574/© 2015 Elsevier B.V. All rights reserved.
landscape management and biological conservation (Ali et al.,
2014). In tropical regions, natural landscapes, especially woodlands
and forests provide many functions and services, such as biodiversity conservation, climate regulation and livelihood support to
millions of people (Zeleke and Hurni, 2001; Shackleton et al., 2007;
Tindan, 2015). In majority, population depend on these natural
resources that they manage and conserve according to their usage
and resources (Appiah et al., 2009; Pare et al., 2010; Pouliot et al.,
2012; Ouedraogo et al., 2013; Steele et al., 2015). This situation led
to management options through protected areas and community
forest zones in order to serve as guards for landscape conservation. Unfortunately, increasing human pressures on land resources
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
affect these protected areas, aggravating the failure of their biological conservation role (Folega et al., 2010a; Wala et al., 2012;
Damnyag et al., 2013; Dimobe et al., 2014; Folega et al., 2014b).
It hence appears that human disturbances are taking over natural
ecological factors in shaping and changing the functions, structure
and aesthetics of the landscapes. This broad-scale human-related
disturbance and destruction of native vegetation is considered as
landscape fragmentation and degradation (Bennett and Saunders,
2010).
Globally, landscape fragmentation and heterogeneity changes
are primarily induced by increasing human disturbances, especially
agriculture, wood extraction of timber and charcoal production
(Pare et al., 2009b; Norris et al., 2010; Pare et al., 2010; Onojeghuo
and Blackburn, 2011; Wale et al., 2012; Vu et al., 2014; Wampembe
et al., 2014). Landscape heterogeneity is most often driven by
complex mechanisms being the result of a natural dynamic phenomenon influenced by human imprints, policy response, climate
change, market, and poverty (Appiah et al., 2009; Zhang and Zang,
2011; Appiah, 2013) at a given time and location. In Togo, especially in the Central region covered by mountainous ecosystems,
numerous human pressures are shaping and disturbing the landscape patterns, even in protected areas. In these areas, intensive
wood extraction of timber, firewood and charcoal production, and
small scale farming system strongly affect the landscape structure
and induce degradation of natural ecosystems (Dourma et al., 2009;
Wala et al., 2012). In this context, land resource availability and
location are changing following these human footprint gradients,
calling for the spatial mapping of the landscape patterns in order
to provide an understanding of how certain factors influence these
changes at landscape level.
While earth observation technology provides insights to assess
the spatial patterns of the landscapes, field surveys are required to
provide stand characteristics of each landscape component, especially vegetation types and soils. Interestingly, a combination of
both methods could provide more understanding of the factors governing landscape structure for targeting options of sustainable use
and management of resources (N’da et al., 2008; Hoffmann et al.,
2012). Such information on intrinsic and exogenous factors can
contribute to managing a more sustainable the landscapes considered as commons that are life-support systems on which people
undeniably depend (Shackleton et al., 2007; Thondhlana et al.,
2012; Tieguhong and Nkamgnia, 2012). Landscape planning needs
not only knowledge of the nature of factors causing the dynamic of
land cover, but also necessitate spatial information to target areas
likely or not to undergo changes.
Mo river basin, embedded in one of the richest landscapes
in Togo covering three protected areas, is undergoing continuous transformation. Despite its importance, the ecological status
and appropriate conservation management remain poorly understood. Furthermore, as in the whole Togo, there is no master plan
aiming at promoting the sustainable allocation of land resources.
Though protected areas are erected in the region to ensure biological conservation, the public policies failed due to weaknesses in law
enforcement and illegal incursions (Wala et al., 2012). Attempts to
propose pathways for sustainable management of landscapes were
undertaken fundamentally on the characterization of vegetation
structure and floristic composition in relation to environmental
variables and human disturbances (Woegan, 2007; Dourma, 2008).
Acute attention has only been paid to the use of species composition, and vegetation stand structure as evaluation indicators of land
performance and landscape heterogeneity. No insight is provided
on the spatial patterns of the evaluated land resources in order
to inform about the potential resource availability. Furthermore,
potentials of the soils in terms of chemical contents are not well
understood, although correlations between soil conditions and vegetation influence the landscape patterns (Galal and Fahmy, 2012).
133
In the context of insufficient information to tackle landscape
fragmentation issues, further researches that integrate spatial
dimensions still need to be undertaken at local and national levels. Therefore, in this study, we used an approach that integrates
geographical information systems and remote sensing combined
with field measurements in order to spatially explicit landscape
patterns and assist the sustainable management of the multifunctional landscapes of the Mo river basin. The specific objectives of
the study were to (i) provide a spatially explicit map of the current
landscape heterogeneity in natural and semi-natural landscapes of
both protected and unprotected areas of Mo basin; (ii); analyze biophysical and human disturbances prevailing at the landscape level,
and (iii) determine the stand characteristics as well as soil chemical
conditions in each vegetation type. By hypothesizing that protected
areas exhibit better indicators of land conservation performance
than unprotected areas, we investigated on the structure and stand
characteristics as well as edaphic-ecological variables according to
land protection regimes. The outputs of the study are suggested as
potential indicators of landscape configuration, threats to land conservation and land characteristics that could help in re-addressing
poor land management issues and landscape planning in the Central Togo.
2. Methodology
2.1. Study area
Mo watershed is a sub-unit of the Volta basin (West Africa)
located in the Central Region of Togo (Fig. 1). With a total area of
148,592 ha, the basin is particularly sensitive as it contains great
parts of the Fazao-Malfakassa National Park (FMNP), the Aledjo
Wildlife reserve, and Kemeni Forest reserve. Due to the increasing demand for land resources, these protected areas are likely
to undergo more human pressures (Woegan, 2007; Wala et al.,
2012). The Mo basin is embedded in the Ecological Zone 2 of
Togo, characterized by a mosaic of dry and riparian forests, woodlands, guinea-soudanian savannahs. Dominant land uses within
the human-dominated landscapes (non-protected zones) are small
scale agro-systems (Woegan, 2007). The predominant plant species
in the area are Isoberlinia doka Craib & Stapf, Isoberlinia tomentosa (Harms) Craib & Stapf, Monotes kerstingii Gilg, Detarium
microcarpum Juss. and Uapaca togoensis Pax., etc. (Woegan, 2007;
Dourma, 2008; Dourma et al., 2009). The climate is tropical subhumid characterized by a rainy season from April to October (Petit,
1981). Mean annual rainfall is between 1200 and 1300 mm with an
irregular spatial-temporal distribution. Mean minimal and maximal temperatures reach respectively 19 ◦ C in January with the
Harmattan winds and 30 ◦ C in April. Evapotranspiration is generally high, especially during dry season and can reach 1600 mm
per annum. Some parts of the hilly lands have elevation above sea
level higher than 800 m, especially in Aledjo Mounts. Other mounts
are of variable heights, comprising the massifs of Mazela (704 m),
mount Akitili (861 m), mounts Kouzé (625 m) and Kpeya (652 m).
Mounts Malfakassa composed of Ouassi (568 m), Zandebou, Tchakouya, Timbou et Balankan (Woegan, 2007). The rivers/streams
network is heavily developed in accordance with the mountainous
relief. Mo, Loukoulou, Kamasse, and Bouzalo are the most important streams of the basin. On morpho-structural angle, the Mo
basin is dominated by sericite and muscovite dominant quartzites.
Lithosoils and ferruginous tropical soils are the dominant soil
types with some patches of ferralitic soils (Lamouroux, 1969).
Foremost of the land uses in the area is small-scale subsistence
farming, pasture lands, protected areas and built-up areas. The
prominent environmental issues are land degradation due to overgrazing, unsustainable agricultural land use, fuel wood harvesting
134
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
Fig. 1. Location of the study area.
and charcoal production (Fontodji et al., 2009; Fontodji et al., 2011;
Aboudou, 2012). Illicit incursions for hunting and tree logging in
protected areas are also concerns that cause conflicts between
land users and state agencies protecting lands. In addition, protection enforcement regarding the protected areas is weak and
non-inclusive. The catchment is relatively important for tourism
and one of the breadbaskets (crop production) of the country. The
population is mainly composed of rural farm households and cattle
herders living in villages and hamlets interlinked by a poor network of rural tracks serving in transportation of goods and other
services.
2.2. Deriving Mo landscape patterns based on land use/cover
mapping
In order to provide a spatial glimpse of the different vegetation types in both protected and unprotected landscapes of
Mo basin, this study relied on the use of Landsat 8 satellite image of March 2014. This data was downloaded from
https://earthexplorer.usgs.gov. Land cover classification followed
the Yangambi’s vegetation nomenclature system defined based
on tree heights, canopy cover density, and grassy layer coverage
(Aubreville, 1957). We defined six land use cover types (hereafter, LUC) based on landscape features that were depicted during
reconnaissance field works and previous studies in the same area
(Woegan, 2007; Dourma, 2008). Namely, these LUC are (1) forested
lands (riparian and dry forests), (2) woodlands (open canopy
forests and woody savannahs), (3) savannahs (tree savannahs
and shrubs, and scattered grasslands), (4) cultivated lands (farms,
young fallows, and parklands), (5) residential areas (built up areas
including urban constructions and rural settlements), and (6) water
bodies (rivers and reservoirs). Paved surfaces and bare rocks are
mostly confused among settlements and agricultural lands since
they reflect in the quite similar range. We used ENVI 4.7 and ArcGIS
10.0 software to perform the maximum likelihood classification.
Since the focus of the study is vegetation-related analyses, Normalized Difference Vegetation Index (NDVI) is selected among many
other vegetation indices for it is widely used as a powerful indicator
of vegetation greenness, and less sensitive to topographic factors
in mountainous areas (Matsushita et al., 2007). Original bands and
NDVI layer were stacked in a multilayer image. Analyses were
pixel-based using on-screen digitizing procedure to produce training areas for each LUC on an RGB colour composite image. Though
topography is a common source of biases in land cover classification in mountainous areas (Vanonckelen et al., 2013), elevation data
were not integrated during the classification process. The main reason is that the maximal elevation above sea level in the study area
is around 850 m and does not really provide significant hill shade
effect at the sensor passing time (Diallo et al., 2010). Image classification relied on 177 ground truth points collected during forest
inventory, based on random sampling within the different vegetation cover types. Since the number of field-registered points was
insufficient to serve as training and validation data, we generated
training sites for each class based on Google Earth images of the
same date combined with a topographical map of Togo and other
field knowledge. As the acquired image corresponds to the dry season, there was sharp distinctions between the different cover types,
especially vegetation and lands under crops. At least 60 training
pixels were selected for each LUC from homogeneous large areas
(Zhai et al., 2013). Finally, all the 177 geo-located points were used
to assess the accuracy of the classified image through a confusion
matrix.
2.3. Collection of vegetation data and ecological variables
Since topography was the main constraint during the sampling,
we set 75 forest inventory plots along a topographical gradient
from valleys to top-hill or summits without any predefined plot
number for each location. Thus, we randomly set these plots according to the accessibility, the representativeness of flora biodiversity
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
and the vegetation homogeneity in such a way to represent the
different vegetation types in the landscape. In total, the number of plots varied according to protection regime (39 and 36
plots in protected and unprotected areas, respectively) and vegetation types (9 in dry forests, 19 in riparian forests, 10 in shrubs,
19 in tree savannahs, and 18 in woodlands). Squared plots of
20 m × 20 m and 30 m × 30 m were set in dry forests and savannahs,
respectively. Meanwhile, plot dimensions were about 50 × 10 m in
riparian forests in order to match the linear shape and the width
of these ecosystems in savannah-dominated landscapes (Folega
et al., 2010a; Wala et al., 2012; Dimobe et al., 2014). In each plot,
we focused on woody plant species and recorded the following
attributes for each tree: species name, total height, diameter at
breast height (DBH), crown diameters (North-South and West-East
directions), crown height, and trunk height. In addition, the geographical variables (longitude, Latitude, Altitude above sea level)
were collected using a GPS Garmin 62S. Human disturbances,
i.e., fire occurrence, grazing, selective tree logging and charcoal
production, were recorded as the presence/absence (1 = presence,
0 = absence). Other ecological attributes such as vegetation type,
soil type, soil submersion (0 = No, 1 = Yes), canopy cover density
(coded as 0 = very low, 1 = low, 2 = medium, and 3 = high) and protection status (1 = protected, 0 = free-access) were collected as well.
For data processing purpose, six major topographical positions have
been defined (i.e., flat terrain, hill-foot, mid-hills, hill summits,
riverbank, and inland valleys) and coded from 1 to 6, respectively.
In addition to the aforementioned ecological characteristics, soil
samples were collected at two depths (0–10 cm and 10–30 cm)
to provide soil chemical properties of each forest inventory plot.
For each depth at each site, five sub-samples were mixed thoroughly before collecting one representative composite sample.
For both depth, 150 samples were collected and air-dried for
chemical analyses. Three basic chemical properties were determined: soil organic carbon (SOC, in %), total nitrogen (TN, in
%), and pH water. SOC was determined using Walkley-Black
method, which consists of titration of excess potassium dichromate used with sulphuric acid to react with 3 g of dried soil.
SOM was derived from the SOC using the Van Bemmelen conversion factor (1.724) commonly used for the estimation of
organic matter content (Sebastia et al., 2008; Fontodji et al.,
2009; Agboadoh, 2011). The total nitrogen was analyzed using
Total Kjeldahl Nitrogen method. The potential of hydrogen (pH),
which measures the acidity or alkalinity of a solution, was determined using 1/2.5 soil/water ratio. The pH of the solution is
electronically and directly measured using a glass electrode pHmetre.
Furthermore, to assess the potential influence of topography
on vegetation types in Mo basin, other biophysical indices were
derived from the DEM-SRTM 1 arc-second (approx. 30 m) downloaded from earthexplorer.usgs.gov. These indices included terrain
attributes such as slope, Topographic Wetness Index (TWI), and
mean altitude above the channel level (Alt a.c.l.). The TWI is a topographic variable indicating the spatial distribution of soil moisture
conditions, a potential indicator of species preference to moisture.
Alt a.c.l. is used to indicate the potential depth to groundwater/free
water for plant species, and therefore to moisture content in the
substratum. Maps of slope, TWI and Alt a.c.l were derived using
SAGA GIS 2.0.8 platform that embeds the algorithms for computing these variables. Soil texture was obtained from HWSD but
was not considered in detailed analyses for no difference existed
in soil properties according to soil types. The various maps were
then exported to ArcGIS 10 that helped in extracting values of
the variables to geographical coordinates of the 75 plots. IBM
SPSS 20 and Microsoft Excel served in data preparation for further
analyses.
135
2.4. Vegetation and ecological data analyses
The ordination method was used to identify environmental
gradients defining species distribution and landscape patterns.
Among the various multivariate methods used for this purpose in
plant community analyses, the indirect gradient algorithms (DCA),
cluster analyses (CA) and two ways indicator of species analysis
(TWINSPAN) were used to analyze the variation of plant communities and their relationships with environmental variables. The
first reason is that they are commonly used for plant community
analyses (Dourma et al., 2009; Tavili and Jafari, 2009; Folega et al.,
2010b; Zhang and Zhang, 2010; Folega et al., 2011; Wala et al.,
2012; Wale et al., 2012; Kebede et al., 2013; Dimobe et al., 2014;
Folega et al., 2014a). The second reason is that the dataset was too
heterogeneous and too many species deviated from the assumed
model of linear response (Leps and Smilauer, 2003), for the length of
gradients (which measures the beta diversity in community composition, i.e., the extent of species turnover along the individual
independent ordination axes) obtained during trials exhibited values between 3 and 4. In such cases, it was compelling to select
unimodal method using the indirect detrended DCA method by
segments down-weighting rare species without any transformation of the initial information. DCA methods summarize variation
in the relative frequencies of the response variables (species). An
important implication is that these methods cannot work with
‘empty’ samples, i.e. records in which no species is present (Leps
and Smilauer, 2003). The DCA were performed in CANOCO 4.5 and
CanoDraw 4.1 for Windows. TWINSPAN in CAP 2.15 (Community
Analysis Package) clustered plant communities according to their
level of similarity. First, a matrix of 75 plots × 142 plant species
subjected to a DCA ordination did not help in depicting plant communities along any gradient. Therefore, the data were split into
36 unprotected relevés and 39 relevés in PA. The two matrices,
39 relevés × 121 species and 36 relevés × 100 species were subject
separately to partial ordination, respectively for PA and UPA.
Three measures of species diversity, i.e. species richness (S),
Shannon-Wiener’s species diversity index (H ), and Pielou equitability index (E), were computed to characterize each plant
community. Species richness (S) was computed for each plant community as the total number of woody species recorded in the relevés
of the community. Because of different plot sizes, mean species
richness at plot level was calculated as the number of species in
a plot divided by the log of the area sampled (White et al., 2014).
Shannon-Wiener’s species diversity index (H ) is calculated using
the formula of the form:
S
H = −
(Pi)log 2(Pi)
i=1
where Pi = Ni/N with Ni is the number of individuals of species i and
N is the total number of individual of all recorded species.
Pielou equitability index (E) is calculated based on H and S as:
E=−
H
log 2(S)
Plot basal area at breast height (G in m2 ha−1 , sum of crosssectional area of all trees with DBH ≥10 cm in a given plot) and
mean values of several other features were calculated for each plant
community according to protection status:
0.0001sqr(di)
4A
n
G=
i=1
where n is the number of trees recorded in the plot, A is the sampled
plot area (plot size, in ha), and di is the DBH (in cm) of the ith tree.
136
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
5
Protection status
Fire occurrence
Soil submersion
Cattle grazing
Axis 2
Canopy density
Topography
Tree logging
-1
-2
Axis 1
5
Fig. 3. DCA of 75 relevés × 142 woody species in protected (green diamonds) and
free-access areas (black circles). (For interpretation of the references to color in
figure legend, the reader is referred to the web version of the article.)
Fig. 2. Current spatial patterns of land cover types in Mo landscape.
In addition, vegetation stands were characterized by computing
their tree (DBH ≥ 10 cm) density. Regeneration of all mixed species
was also calculated to overview landscape dynamics. Mean soil
chemical properties (pH, SOC, and TN), and stand conditions (mean
indices of soil submersion, fire occurrence, canopy cover, selective
tree logging, and Alt a.c.l.) were calculated for the various plant
communities in both free-access lands and protected areas.
Similarity between plant communities was assessed using Jaccards’s index (Sij) and plotted using TWINSPAN in CAP 2.15. The
performance of Jaccards’s Similarity was used to compare the plant
communities under the different protection status:
Sij =
C
A+B−C
where A is the number of species belonging to the plant community
i, B is the number of species belonging to the plant community j,
and C the number of species belonging to both plant communities
i and j. If Sij ≥ 50%, communities exhibit similarity; otherwise (i.e.
Sij ≤ 50%), there is no similarity (Woegan, 2007).
Supplementary statistical analyses were performed through
analysis of variance to compare characteristics of the different
plant communities, assuming equality of variance and normality
in data distribution. Outputs from analyses of variance were contrasted using post hoc multiple comparison of Fisher and Tukey
at p < 0.05. Pairwise correlation was performed to depict the relationship between biophysical and ecological variables within the
vegetation types.
northeast parts of the basin (Fig. 2). The spatial analysis showed that
Mo basin is a heterogeneous landscape where small-scale farming,
pasture, wood extraction and charcoal production dominate unprotected areas. Areas under protected status are relatively greener
though human incursions are noticeable, especially at the edges.
Overall, woodlands and savannahs/shrubs are the two dominant
land cover types at the landscape level. Built up areas made up
essentially by rural settlements covered about 0.3% of the areas
outside PA. Water bodies showed the lowest coverage due to roughness that do not favour the occurrence of water bodies though the
river network is highly developed. Overall, the southern and southwestern parts of Mo basin present wilder landscapes attributable to
protection status and the increasing distance from settlements as
well. However, these factors in combination with biophysical and
soil conditions could provide explanations on the spatial patterns
of the landscape.
3.2. Discrimination of plant communities and biophysical
conditions in natural landscapes
3.1. Current landscape patterns from satellite image
The outputs from the ordination of 75 relevés × 142 woody
species (see species list in Table S8) in relation to human disturbances, protection status and other biophysical factors are shown
in Fig. 3. Axis 1 of the DCA plot correlated positively with protection status (0.58) and topography (0.63) but negatively correlated
to human footprints (−0.53 for tree logging and −0.63 for fire occurrence) (Tables S2 and S3). Meanwhile, axis 2 showed a positive
correlation with protection status (0.64). These results indicate that
the most important ecological factors determining landscape patterns are protection status and topography, in opposition to human
disturbances. However, the ordination plot does not allow a net
discrimination of the different vegetation types in the overall landscape. Hence, the 75 relevés were split into free access landscapes
(36 relevés, black circles) and protected areas (39 relevés, green
diamonds) for further partial ordination.
The overall accuracy of the land cover classification was 83.1%
with a kappa index of agreement of 0.79 (Table S1). Statistics
on the areal distribution of the classification showed that the
mosaic savannahs/shrubs are the most widespread vegetation
types (80,836 ha) while forestlands covered 15,459 ha (10.4% of
the total area) (Table S1). Human-made landscapes (croplands and
fallows) occupying about 8.5%, dominated the central and the
3.2.1. Partial ordination of relevés of free-access landscapes
The partial ordination of the 36 relevés × 100 species in unprotected areas (Fig. 4a) showed four vegetation types (U1, U2, U3, U4,
U stands for unprotected areas; see supplemental file). Tables S4
and S5 showed that the first two axes of Fig. 4b defined gradients of
soil moisture conditions and topographical positions. Axis 1 defines
plant communities of moist and finer soils in lowlands (U1) versus
3. Results
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
2.0
a
3.5
R66
R96
U4
(a)
R67
R54
R4
P3
Axis 2 (5.4 % of variance)
Axis 2 (6.3 % of variance)
R89
R28
R46
R37
R12
R14
R30
R10
R59
R25
U1
U2
R13
R48
R65
R63
R7 R19
R5
R57
R21
R11
R9
R97
U3
R8
R87
R95
R72
R71
R33
R27
R84
R91
R90
R78
R81 R92
R62
R47
R68
R70
R50
R55
R34
R85
R76
R69
R39
R3
R56
137
R80
R75
R52
R58
R45
R74
R16
R77
P1
R88
R43
R1
R73
R86
R60
R18
R83
R15
R6
R64
R20
R79
R82
P2
-0.5
-0.5
b
Axis 2 (7.4 % of variance)
-0.5
-1.0
0.2
5.0
Axis 1 (8.9 % of variance)
(b)
Grazing
Fire occurence
pH10
Soil texture
SOC20
SOC10
Soil submersion
Canopy cover density
TN10
pH20
Soil texture
Alt.a.c.l
Slope
Axis 2 (9.1 % of variance)
0.20
3.5
Axis 1 (13.6 % of variance)
Alt_a.c.l.
TN20
TWI
Topography
Fire occurrence
Grazing
SOC10
Soil submersion
pH20
Slope
pH10
Wetness index
SOC20
Tree logging
-0.20
-0.40
Axis 1 (18.9 % of variance)
TN10
TN20
Topography
Tree logging
0.30
Canopy cover density
-0.3
Fig. 4. (a) DCA ordination of 36 relevés and 100 woody species taken from freeaccess landscapes showing the four plant communities. (b) DCA plots showing
biophysical and soil properties under the 36 samples in free access landscape areas.
those of dry and rocky soils located on relatively high lands (U2, U3
and U4). The DCA plot of environmental variables indicated a gradient of canopy cover density related to increasing soil moisture
content and topographical position (axis 1) while axis 2 denoted
mostly the effects of anthropogenic factors.
From the combined analysis of the above-mentioned DCA plots
in UPA, it is highlighted that U1 occurs more on richer soils in
organic matter in topsoil. High density of canopy coverage, low
slope and high soil moisture (TWI) characterize the community U1
though human pressures tend to modify the natural patterns (both
variables are located at antipode sides of the DCA plot). The horizontal axis denotes a gradient of human pressures increasing from
the left to the right side. Groups U1 and U4 are antipodes, with U1 in
lowlands and U4 at hill summits. The relative high number of plots
occurring in the right part of the DCA indicates that the majority of
the vegetation types experienced human pressures such as illegal
logging and grazing, and bush fires.
3.2.2. Partial ordination of relevés of protected areas
The DCA ordination biplot (Fig. 5a) showed three groups of
relevés clustered by axis 1 (8.9%), and axis 2 (5.4%) (Tables S6 and
S7). The DCA plot (Fig. 5; see supplemental file).
Combined analyses of above DCA plots in PA showed the coexistence of the biophysical and human threats that shape landscape
composition. Indeed, P1 regroups sites in which species prefer
richer soils in organic matter and TN in topsoils, and richer subsoils in TN. Releves of P1 are often in low topographic with high
soil moisture (high TWI). Meanwhile, P2 occurs on soil with high
pH and rich sublayers. They often experience more tree logging
because of their relative low slope and their canopy cover density.
-0.4
0.4
Axis 1 (12.3 % of variance)
Fig. 5. (a) DCA ordination of 39 relevés and 121 woody species taken from protected
areas exhibiting 3 communities (P1, P2, and P3). (b) DCA plot showing biophysical
and soil properties under the 39 samples in protected areas.
However, because of the relative low canopy cover density in P3,
fire and grazing do occur often in these plots because of their richness in grasses and fodder herbaceous. In sum, the horizontal axis
(axis 1) is comparable to an increasing gradient of topsoil nutrients and soil moisture from the right to the left side (summits and
mid-hills versus lowlands and river banks). A combined effect of
increasing gradient of human pressures from right to left is also
interpreted from this plot (highly disturbed vegetation versus less
disturbed vegetation). However, the second axis indicates mostly
a decreasing gradient of vegetation physiognomy, which is denser
from negative to positive canonical axes. Though groups exhibited
low similarity, P2 and P3 are more similar (0.46) than P1.
3.3. Structural and dendrometric characteristics of plant
communities of Mo basin
Tables 1–3 show the stand characteristics and diversity indices
of the various plant communities in free-access lands, protected
areas, and merged analyses for PA versus UPA, respectively.
In free-access landscapes, features exhibited significant differences among groups (ANOVA, p < 0.05), except for mean
species richness (ANOVA, p = 0.362), tree density with saplings
(ANOVA, p = 0.580), and sapling density (ANOVA, p = 0.509). Forested
stands clustered in group U1 showed the highest mean values of tree height (10.5 ± 6.3 m), diameter (28.0 ± 17.4 cm), and
density (947 ± 328 stems ha−1 ). With a mean basal area of
58.8 ± 14.1 m2 ha−1 , U1 exhibited the highest record, compared
to U2, U3, and U4 with 18.6 ± 6.2 m2 ha−1 , 15.9 ± 7.7 m2 ha−1 , and
138
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
Table 1
Stand characteristics in the four plant communities within none-protected areas.
Characteristics
Cluster U1 (8 relevés)
Cluster U2 (7 relevés)
Cluster U3 (11 relevés)
Cluster U4 (10 relevés)
ANOVA (˛ = 0.05)
Species richness
Mean species richness (per plot)
Mean height (m)
Mean DBH (cm)
Mean diameter with saplings (cm)
Basal area (m2 ha−1 )
Tree density (trees ha−1 )
Density without saplings (trees ha−1 )
Sapling density (saplings ha−1 )
Shannon index (unitless)
Pielou evenness index (unitless)
58
19 ± 5aA
10.5 ± 6.3aA
28.0 ± 17.4aA
21.8 ± 17.6aA
58.80 ± 14.07aA
947 ± 328aA
678 ± 217aA
269 ± 219aA
5.04 ± 0.08
0.9
51
20 ± 6aA
6.7 ± 4.2bB
19.3 ± 8.7bB
13.2 ± 9.1bB
18.58 ± 6.19bB
921 ± 490aA
487 ± 167abB
433 ± 445aA
4.56 ± 0.10
0.8
66
19 ± 5aA
5.4 ± 3.7cC
18.3 ± 9.3bB
10.6 ± 8.1cC
15.93 ± 7.69bB
846 ± 432aA
311 ± 142bcC
534 ± 484aA
5.07 ± 0.07
0.8
49
16 ± 4aA
5.0 ± 3.5cC
19.4 ± 11.2bB
8.8 ± 7.9dD
7.66 ± 3.33bC
692 ± 428aA
156 ± 101cD
537 ± 444aA
4.65 ± 0.09
0.8
na
p = 0.362 ns
p = 0.000 ss
p = 0.000 ss
p = 0.000 ss
p = 0.000 ss
p = 0.580 ns
p = 0.000 ss
p = 0.509 ns
na
na
Note: Outputs of one-way ANOVA: na, not available; ns, not statistically significant; ss, statistically significant (p < 0.05). Values that do not share a letter are significantly
different at 95% confidence internal (CI) using ANOVA with post hoc test (p < 0.05). Capitalized letters are the outputs from Tukey comparison test whereas small letters are
the results of Fisher comparison method. Group U1, Mosaic of riparian and dry forests; Cluster U2, Open forests/woodlands; Cluster U3, Tree savannahs/degraded woodlands;
Cluster U4, Shrubs.
Table 2
Stand characteristics in the three plant communities within protected areas.
Characteristics
Cluster P1 (10 relevés)
Cluster P2 (10 relevés)
Cluster P3 (19 relevés)
ANOVA (at ˛ = 0.05)
Species richness
Mean species richness (per plot)
Mean height (m)
Mean DBH (cm)
Mean diameter with saplings (cm)
Basal area (m2 ha−1 )
Tree density (trees ha−1 )
Density without saplings (trees ha−1 )
Sapling density (saplings ha−1 )
Shannon index (unitless)
Pielou evenness index (unitless)
68
14 ± 4aA
15.5 ± 6.5aA
28.0 ± 16.6aA
26.3 ± 16.9aA
46.26 ± 31.77aA
604 ± 301aAB
560 ± 280aA
50 ± 72aA
4.38 ± 0.10
0.78
70
16 ± 3aA
10.6 ± 5.4bB
24.0 ± 14.7bB
20.8 ± 14.9aB
32.55 ± 18.97abA
712 ± 533aA
553 ± 297aA
163 ± 282aA
4.61 ± 0.09
0.79
70
15 ± 4aA
9.5 ± 4.1cC
20.5 ± 10.1cC
18.7 ± 10.6cC
15.09 ± 7.99bB
407 ± 201aB
356 ± 156aB
51 ± 97aA
5.02 ± 0.08
0.86
na
p = 0.245 ns
p = 0.000 ss
p = 0.000 ss
p = 0.000 ss
p = 0.001 ss
p = 0.065 ns
p = 0.036 ss
p = 0.176 ns
na
na
Note: Outputs of one-way ANOVA: na, not available; ns, not statistically significant; ss, statistically significant. Values that do not share a letter are significantly different at
95% CI using ANOVA with post hoc test (p < 0.05). Capitalized letters are the outputs from Tukey comparison test whereas small letters are the results of Fisher comparison
method. Group P1, Mosaic of riparian and dry forests; Cluster P2, Open forests/woodlands; Cluster P3, Tree savannahs/shrubs.
7.7 ± 3.3 m2 ha−1 , respectively. Meanwhile, U3 is the species-richer
group (66 woody species), exhibiting then high values of diversity features (5.07 and 0.84, for H and E, respectively) close to the
values obtained for U1. The high values of H and E are indicators
of stability and homogeneity in studied landscapes. U4 which is
the record of low values was dominated by saplings which density
is the highest (537 ± 444 saplings ha−1 ). In this group U4, species
richness is about 49 species with mean diameter and height of
19.4 ± 11.2 cm and 5.0 ± 3.5 m, respectively. Saplings had a strong
effect on mean diameters in all clusters resulting in mean basal
areas that decreased from U1 to U4.
In the other hand, plant communities in PA are characterized
by relative high values of the stand and diversity features compared to those in free-access lands. An exception was observed
for sapling density and diversity indices H and E. Apart from the
average species richness per plot and the sapling density, all other
features varied significantly according to plant communities. Mean
DBH was about 28.0 ± 16.6 cm, 24.0 ± 14.7 cm, and 20.5 ± 10.1 cm
in P1, P2 and P3, respectively. This trend in tree diameters
resulted in similar trend in the basal areas with highest value
in P1 (46.3 ± 31.8 m2 ha−1 ) and lowest in P3 (15.1 ± 8.0 m2 ha−1 ).
Though the basal area and the tree density were the lowest
in P3, this plant community exhibited high diversity of species
(5.02 and 0.86 for H and E, respectively). With a density of
712 ± 533 trees ha−1 , P2 is the most dense plant communities compared to P1 (604 ± 301 trees ha−1 ) and P3 (407 ± 201 trees ha−1 ).
Basal areas in these protected areas are more shaped by the large
contribution of mature individuals than saplings which exhibited low densities, especially in P1 (50 saplings ha−1 ) and P3
(51 saplings ha−1 ).
A broad analysis showed that most stand features and diversity differed significantly between PA and UPA. PA showed higher
Table 3
Stand characteristics according to land protection status.
Characteristics
Protected landscapes (39 relevés)
Free-access landscapes (36 relevés)
ANOVA (at ˛ = 0.05)
Species richness
Mean species richness (per plot)
Mean height (m)
Mean DBH (cm)
Mean diameter with saplings (cm)
Basal area (m2 ha−1 )
Tree density (trees ha−1 )
Density without saplings (trees ha−1 )
Sapling density (saplings ha−1 )
121
15 ± 4aA
11.2 ± 5.6aA
23.2 ± 13.5aA
21.0 ± 13.8aA
27.56 ± 23.04aA
536 ± 354bB
459 ± 247aA
79 ± 164aA
100
18 ± 5bB
6.4 ± 4.6bB
21.3 ± 12.6bB
12.5 ± 11.1bB
23.67 ± 21.10aA
840 ± 416aA
384 ± 247aA
456 ± 416bB
na
p = 0.000 ss
p = 0.000 ss
p = 0.001 ss
p = 0.000 ss
p = 0.450 ns
p = 0.001 ss
p = 0.192 ns
p = 0.000 ss
Note: Outputs of one-way ANOVA: na, not available; ns, not statistically significant; ss, statistically significant. Values that do not share a letter are significantly different at
95% CI using ANOVA with post hoc test (p < 0.05). Capitalized letters are the outputs from Tukey comparison test whereas small letters are the results of Fisher comparison
method.
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
139
Table 4
Woody species-based similarity between discriminated vegetation stands.
Groups of vegetation stands
P1
P2
P3
U1
U2
U3
P2
P3
U1
U2
U3
U4
0.366
0.289
0.326
0.202
0.186
0.206
0.458
0.438
0.441
0.447
0.368
0.333
0.407
0.432
0.384
0.473
0.333
0.372
0.539
0.493
0.456
values of stand features, except tree density and sapling density.
Only basal area and tree density without saplings did not display
statistical significant differences between the two supra-groups.
There were 122 woody species counted in PA while 100 were
recorded in free-access landscapes. Mean sapling density is very
low in PA (79 ± 164) compared to those in UPA (456 ± 416). This is
probably due to low potential of vegetative multiplication through
natural process (suckering and seedlings) in PA. On average, trees
in PA are taller (11.2 ± 5.6 m) and bigger (23.2 ± 13.5 cm) than those
in free-access lands (6.4 ± 4.6 m and 21.3 ± 12.6 cm, respectively for
mean height and diameter).
Based on stand characteristics, no significant similarity existed
between the seven groups, except between U2 and U3 (Sij = 0.539)
(Table 4 and Fig. 6). Closest groups in similarity were P2–P3
(Sij= 0.458), U1–U2 (Sij = 0.473), U2–U4 (Sij = 0.493), and U3–U4
(Sij = 0.456). However, the obtained similarity values indicated substantial common characteristics among groups (Sij ≥ 0.333 for most
of them).
3.4. Ecology and human disturbances within plant communities
The analyses of ecological features and human impacts in the
different plant communities (Table 5) indicated differences related
to in situ conditions. In general, soils of the seven plant communities are acidic (pH <7) with relative high chemical properties
in topsoil (0–10 cm). Though pH and SOC in the lower 20 cm did
not vary significantly among groups (p > 0.05), other ecological
variables characterized the plant communities. Based on ecology
and human footprints, similarity dendrogramme (Fig. 7) indicates four clusters of plant communities: (1) U1 and P1 are sites of
riparian/dry forests developed on nutrient-rich soils (high contents
of SOC and TN) located on riverbanks and inland valleys where
floods often occurred (soil submersion higher than 30%). For both
U1 and P1, there is negligible traces of grazing with low tree
logging rate. On average, U1 is closer to the riverbed than P1
(mean altitude above the channel level of 7.2 m versus 11.2 m,
respectively) resulting in high density of canopy in U1 (coefficient
of 3.00 versus 2.10). (2) Relevés of U2 are particular stands as
despite the high human disturbances, they exhibited nutrient-rich
soils with less moisture content (33.2 m above channel). (3) U3
and P2 develop on soils with medium nutrient contents associated
with high rate of illegal tree logging (0.70 in P2) and fire occurrence (0.70 and 0.91 in P2 and U3, respectively). This group is the
timber-rich stands, explaining the high rate of tree logging, even in
protected areas. However, cattle herders do not make incursions
in P2. Their soils have low potential of submersion related to their
relatively high locations above channel levels (24.4 m for U3 and
25.1 m for P2). (4) U4 and P3 are relevés with less nutrient contents
experiencing moderate human disturbances, especially in U4. On
average, they occur at the same altitude above the channel (19 m).
Soils are less submersible with low canopy density favourable to
grazing (coefficients of 0.70 and 0.79) but not to tree logging. The
number of plant communities in UPA is due to the vegetation is less
disturbed and allow an easy demarcation of the three vegetation
types. Meanwhile, in UPA, human effects changed the physiognomy of some stands increasing the patterns defining vegetation
types.
In sum, though each relevés or each plant community denotes
its intrinsic ecological and biophysical features, some levels of
similarity are observable related to certain undeniable common
points (topography and protection status). Further, broad analyses
of soil conditions in protected areas versus free-access landscapes
showed significant differences for TN and SOC at both depths,
and cattle grazing. Plant communities under land protection status exhibited high chemical values and low coefficients of human
disturbances. Though there were no statistical differences, fire
occurrence has a coefficient of 0.78 in free-access landscapes versus
0.62 in protected areas. Coefficients of grazing was quite high
Fig. 6. Dendrogramme of similarity between the seven plant communities based on stand characteristics.
Fig. 7. Dendrogramme of similarity between the seven plant communities based on ecology of and human disturbances.
Mean altitude
above channel
level (Alt .a.c.l)
(in m)
Site level
7.17
33.15
24.36
19.52
11.24
25.07
19.36
0.543 ns
20.90
18.74
0.733 ns
Site level
0.00
0.14
0.36
0.60
0.00
0.00
0.16
0.001 ss
0.31
0.08
0.011 ss
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
Grazing
140
in UPA compared to PA (0.30 versus 0.08). The same comparison was observed for tree logging coefficients (0.42 in UPA versus
0.26 in PA). These human footprint coefficients in PA indicate the
weakness of protection laws, especially at the edges of protected
lands.
4. Discussion
0.13
0.57
0.64
0.30
0.10
0.70
0.11
0.001 ss
0.42
0.26
0.145 ns
3.00
1.86
1.45
1.30
2.10
2.40
2.00
0.004 ss
1.83
2.13
0.207 ns
6.19 ± 0.58abA
6.03 ± 0.13abAB
6.16 ± 0.57abA
6.03 ± 0.45abAB
5.70 ± 0.28bB
6.29 ± 0.50aA
6.01 ± 0.34abAB
0.086 ns
6.10 ± 0.47
6.00 ± 0.42
0.325 ns
6.41 ± 0.39abA
6.31 ± 0.26abAB
6.23 ± 0.23abAB
6.48 ± 0.49aA
5.95 ± 0.45bB
6.35 ± 0.17abA
6.36 ± 0.37abA
0.047 ss
6.35 ± 0.37
6.25 ± 0.39
0.247 ns
36
39
Note: Outputs of one-way ANOVA: na, not computed; ns, not statistically different; ss, statistically significant difference at the indicated ˛ value.
0.75aA
0.14abB
0.09bB
0.10bB
0.30abB
0.20abB
0.37abB
0.024 ss
0.25
0.31
0.584 ns
2.31 ± 0.77aA
2.05 ± 0.53aAB
1.93 ± 0.57aAB
2.11 ± 0.99aAB
1.72 ± 0.43aB
1.84 ± 0.46aAB
1.72 ± 0.39aB
0.242 ns
2.09 ± 0.73
1.75 ± 0.42
0.015 ss
2.44 ± 1.29abBC
3.09 ± 0.96abAB
2.21 ± 0.69bC
2.40 ± 0.71abBC
3.63 ± 1.46aA
3.01 ± 0.68bAB
2.55 ± 0.56abBC
0.010 ss
2.48 ± 0.92
2.94 ± 0.98
0.041 ss
0.071 ± 0.04abBC
0.074 ± 0.01abBC
0.057 ± 0.02bBC
0.052 ± 0.02bC
0.101 ± 0.04aA
0.077 ± 0.04abAB
0.062 ± 0.02bBC
0.002 ss
0.062 ± 0.02
0.076 ± 0.03
0.038 ss
8
7
11
10
10
10
19
U1
U2
U3
U4
P1
P2
P3
Anova at p = 0.05
Overall U
Overall P
Anova at p = 0.05
0–10 cm
0.133 ± 0.08bB
0.119 ± 0.02bBC
0.081 ± 0.03bC
0.077 ± 0.04bC
0.215 ± 0.09aA
0.129 ± 0.05bB
0.104 ± 0.04bBC
0.000 ss
0.099 ± 0.05
0.139 ± 0.07
0.009 ss
10–30 cm
0–10 cm
10–30 cm
Site level
10–30 cm
0.50abBC
1.00aA
0.91aA
0.70abAB
0.20bC
0.70abAB
0.79aAB
0.002 ss
0.78
0.62
0.131 ns
Mean ± standard deviation
Site level
Mean ± standard deviation
0–10 cm
Soil organic carbon (SOC in %)
Total nitrogen (TN in %)
Potential of hydrogen (pH)
Nb plots
Plant
communities
Table 5
Soil conditions and human disturbance rates in the seven plant communities of both free-access lands and protected areas.
Soil
submersion
Fire
occurrence
Canopy
cover
Site level
Tree
logging
Site level
4.1. Vegetation patterns, structure and dynamics
At landscape level, Mo basin composes of wide types of natural
and human-influenced ecosystems. Indeed, apart from productive
managed landscapes (farms, fallows, and pastures), dry forests,
gallery forests, woodlands, tree/shrub savannahs are the common cover types, defined according to soil conditions in relation
to topography as well as species composition. These biophysical variables determine the physiognomy of the vegetation types
which stand characteristics and soil conditions significantly differed (Dourma, 2008; Wala et al., 2012). Along the topographical
gradient denoting soil moisture conditions, canopy cover is denser
in lowlands (inland valleys, hill foots and river banks) than midhills and hill summits, indicating that the wetness index plays
an important role in species composition and vegetation growth
(Aynekulu, 2011). Though the stand charactersitics do not really
help in defining the vegetation types, it was evident that larger
trees barely occurred in top-hills dominated by tree savannah and
shrubs, as a consequence of the coarser and rocky soil conditions
which do not favour soil moisture for plant growth. These consequently resulted in the stand basal area values which showed
a decreasing trend from lowlands to upper lands (top-hills dominated by shrubs/tree savannahs). Along the protection status
gradient, mean basal area (27.6 m2 ha−1 ) is higher than that of
the unprotected areas (23.7 m2 ha−1 ) but included in the range
of values recorded by studies in other similar landscapes made
up of subtropical dry forests and woodlands (23.8–78.8 m2 ha−1 )
(Dourma et al., 2009; Folega et al., 2012; Wala et al., 2012). These
results are the evidence of an over-exploitation of the landscape
resources, especially of tree species.
A broad analysis showed that most stand features and diversity differed significantly between PA and UPA. PA showed higher
values of stand features, except sapling density, which is in contrast with the findings of Dourma et al. (2009) who found high
stand characteristics in protected dry forests compared to UPA. PA
showed a higher woody species-diversity with very low sapling
density probably due to low potential of vegetative multiplication
through natural process (suckering and seedlings). In general, the
poor representation of saplings could be due to the rocky nature
of the soils that do not favour quick germination of seeds, and to
landscape roughness inducing seeds transportation downward in
lowland and their carrying away by water. This could explain the
high density of saplings in unprotected areas, which lie, mostly on
less rough landscapes. On the other hand, the prevalence of tall and
big trees in protected lands compared to UPA suggested the effects
of intensive and selective tree logging, charcoal production and
mortar making on vegetation structure. These mentioned human
impacts often target specifically high and big trees, inducing the
loss of trees of big size (Appiah et al., 2009; Dourma et al., 2009;
Appiah, 2013). Subsequently, human pressures convert tree-rich
stands into tree-less and induce svananisation process. The protected lands showed a more stable landscape with the bell-shaped
distribution of tree densities. This can be resulted from the fact that
trees of these sizes are subject to tree logging for wood and charcoal production. In non-protected areas, the high rate of human
threats stimulate multiplication of trees through clonal propagation, suckers and seedlings (Pare et al., 2009a). These stressing
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
factors induce an increase in stem density in unprotected areas
compared to protected lands undergoing less threat, especially
tree cutting. This study assumed that disturbances in unprotected
areas induced local vanishing of plant species due to their replacement by direct economic plant species such as Tectona grandis,
Eucalptus spp., Anacardium occidentale, Eleais guinensis, etc. However, several studies (Zhai et al., 2013; Zhai et al., 2015) found
that disturbed natural forests are more vulnerable to invasion by
exotic plant species compared to undisturbed forests, which tend
to exhibit more homogeneous and monospecific stands, and therefore less species diversity. Pare et al. (2009b) reported in Burkina
Faso that unprotected forests were most diverse compared to protected areas, due to the low population density in unprotected
lands.
4.2. Relationships between vegetation types and environment
Ecology and biophysical factors usually define vegetation types
which often contribute to the maintenance of these factors, especially edapho-ecological variables. As shown by previous studies,
the distribution of plant community or vegetation types is defined
along environmental gradients such as topography, soil conditions,
microclimate, and human disturbances. In this study, whatever the
protection status, vegetation types were mostly defined according to topographical gradient inducing different soil conditions
(moisture and nutrient contents). As reported by several authors,
topography is a factor that defines vegetation patterns, especially in mountainous landscapes. Riparian and dry forests and
woodlands occurred on moist soils with finer particles and high
nutrient contents. These findings aligned with those former authors
(Woegan, 2007; Dourma, 2008; Wala et al., 2012) who worked in
the same zones. Regarding soil nutrients and particles, the flow
gradient from top-hills to hill-foots and inland valleys induces a
typical gradient of soil conditions. From top to bottom, the soils
are finer and nutrient-richer downwards (Ofori et al., 2013). Naturally, each land cover occurs on particular soil physico-chemical
properties which highly vary spatially, even within the same LUC
type (Wiesmeier et al., 2014). Accordingly, landscape positions and
elevation induced the spatial variability of soil conditions (Solon
et al., 2007; Jabeen and Ahmad, 2009; Zhang and Zhang, 2010).
On the second hand, the general trend of soil chemicals is sharply
shaped by the landscape positions and the hydrological processes
occurring at the various sites. Based on previous findings in similar landscapes of the same study area, topography in relation
with soil conditions is the most determinant factor of vegetation
physiognomy and patterns in Mo basin. Topography and somehow elevation have significant effects on the vegetation patterns
through their determinant effects on soil conditions (Aynekulu,
2011). Soils in lowlands and hill-foots do have deeper profiles
with high moisture and nutrient content favourable the growth
of plant species. In contrast, on hill summits and hillsides with
steep slopes, the soil conditions are drier, rockier, and less deep.
These ecological indicators explain the occurrence of forests along
riversides and lowland soils, even in savannah-dominated landscapes.
Besides, environmental factors, anthropogenic disturbances
affect species diversity, size classes and land cover types through
habitat loss and fragmentation. Studies evidenced that human
disturbances usually have substantial effects on ecosystem functioning, and therefore their structure, physiognomy and species
composition (Paré, 2006; Dourma et al., 2009; Folega et al., 2010a;
Ouedraogo, 2010; Appiah, 2013). Human footprints caused by grazing, tree logging, wildfire, and cropping have been reported as
negative drivers changing habitat fragmentation with less tree
species diversity and loss of landscape functions and aesthetics,
even in protected areas (Appiah, 2013; Tchabsala and Mbolo, 2013).
141
Fires in forest and wild lands are cited as a threat that affect forest structure and biodiversity (Bowman and Murphy, 2010), and
contribute significantly to atmospheric CO2 and nitrogen emissions as well (Slik et al., 2008; Kugbe, 2012). The induced land
degradation also has subsequent effects on soil conditions, especially nutrient flow, carbon and nitrogen cycles (Traoré et al.,
2015).
4.3. Land management, human disturbances and implications for
sustainable management
Along land protection status, three and four vegetation types
were identified in PA and UPA respectively. In both landscapes,
topography and soil conditions were the most prominent variables
defining the landscape patterns. This indicates that topographical conditions (i.e. elevation, slope, Alt a.c.l, and soil submersion)
in combination with human disturbances (i.e. fire, logging, grazing) determine species composition and vegetation structure and
dynamics. It often appears that landscape fragmentation and deforestation largely occur in accessible areas such as lowlands. In
these easy accessible areas, the transformation of wild landscapes
(forests and woodlands) into other land-use types (farms and pastures, plantations) is a result of people livelihood support (Appiah
et al., 2009; Pare et al., 2010). It is suggested that land conservation
measures should target these vulnerable areas of the landscapes
in order to ensure effectiveness and efficiency of implementation. As the PA showed somehow their weakness in land resource
conservation due to inefficiency of management regime and law
enforcement (Folega et al., 2010a; Damnyag et al., 2013; Folega
et al., 2014b), it is important to investigate on alternatives for better management (Pare et al., 2009a; Pare et al., 2009b; Wala et al.,
2012; Appiah, 2013; Damnyag et al., 2013).
Besides, wild unprotected lands, especially in rural areas with
low population density, could be of great importance in biological
conservation, even if an explicit role of conservation is not devoted
to them. The lack of properties/rights over lands, the weakness of
policy/laws regarding resource conservation, and the inefficiency
in PA managements are among the sources of unsustainable land
management (Wala et al., 2012; Folega et al., 2014b). In this regard,
land management in the basin should evolve adapted strategies
that define clearly property rights, reinforce laws and policies, and
involve all stakeholders. These strategies involving all stackholders compel to a rethinking of collective management of common
resource pools, especially forests and lands, to avoid “tragedy of
commons”.
The main issues in the innovative forms of adapted land use
and planning in multifunctional landscapes are the participatory
resource allocation and incentives for perennial ecosystem conservation. Governance and management systems combined with
socioeconomic conditions of people being of the underlying factors
in decision-making regarding land use (Ellis and Porter-Bolland,
2008; Kaye-Zwiebel and King, 2014; Specht et al., 2015), law
enforcement for protected areas without any implication of local
stakeholders will guarantee failure in sustainable conservation
strategies. The role of landscapes and other social-ecological systems in ensuring people livelihood while mitigating climate change
is still a crucial issue to be solved. In such conditions, agroforestry
appears to be of great potential for sustainable development and
climate mitigation (Mbow et al., 2014), especially using native
tree species of local economic importance such as V. paradoxa
and Isoberlinia spp. (Dourma et al., 2009). The quasi-stability of
human-affected landscapes and ecosystems is an indicator that
lands outside PA can gain more attention to ensure adapted
land use/conservation in production landscapes (Ellis and PorterBolland, 2008; Ellis, 2013; Gu and Subramanian, 2014).
142
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
5. Conclusions
Diversity and status of plant species as biodiversity-related indicator of landscape stability is studied in Mo landscapes. This helped
in the identification of the linkage between social-ecological systems, and the configuration of landscape patterns under different
land protection status. In total, 142 woody species were identified within four plant communities in unprotected areas (UPA)
and three vegetation types in protected areas (PA). The analyses of
ecological features and human impacts in the different plant communities indicated differences related to in situ conditions and land
protection status. In general, the soils of the seven plant communities are acidic (pH <7) with high contents of chemical properties in
topsoil (0–10 cm). Land management/protection status is an important factor impacting and shaping the vegetation physiognomy in
Mo basin. Accordingly, the common environmental threats were
the high level of wood extraction (firewood, charcoal production
and tree logging), bush fires, and cattle grazing. Though the natural
biophysical factors shape landscape physiognomy, human disturbances affect the structure and composition of plant communities
through the provided multiple and valuable ecosystem services to
local people. As a limitation of the study, the sampling approach
could have drawn an equal number of plots in PA and UPA in
order to draw conclusions based on more comparable results. However, the study showed that biological conservation should not
only target landscapes in protected areas but also on some wild
landscapes located in inaccessible and low populated areas. This
could promote both biological conservation and livelihood support to riparian population. In connivance with local resource users,
their capacity for effective management of common resource pools
should be strengthened by increasing people awareness of the land
degradation phenomenon.
Acknowledgements
The authors are grateful to the German Federal Ministry of
Education and Research (BMBF), which supported this research
through WASCAL research programme. Special thanks to WASCAL
Graduate Research Programme in Climate Change and Land Use
(Kwame Nkrumah University of Science and Technology – KNUST,
Ghana). We also thank the Ministry of Environment and Forest
Resources of Togo, and the Central Region office of Environment
in Togo. We gratefully acknowledge the technical support from the
Foundation Franz Weber of Fazao-Malfakassa National Park during
our field-works. The CGIAR Research Program in Dryland Systems
(CRP-DS), Overarching Cluster of Integrated Systems Analysis and
Modelling (budget code 910101) funded for Quang Bao Le’s contribution. Thanks to the anonymous reviewers for their valuable
comments to the quality of the work.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.ecoleng.2015.09.
059.
References
Aboudou, M., 2012. Etat des lieux de l’occupation du sol dans et autour du parc
national Fazao-Malfakassa. MERF-Togo/Fondation Franz Weber, 19 pp.
Agboadoh, M.Y.D., 2011. Estimation and Mapping of Soil Organic Carbon in Croplands of the Bechem Forest District, Ghana. ITC, University of Twente, pp. 67.
Ali, A., de Bie, C.A.J.M., Skidmore, A.K.R.G.S., Lymberakis, P., 2014. Mapping the heterogeneity of natural and semi-natural landscapes. Int. J. Appl. Earth Obs. Geoinf.
26, 176–183.
Appiah, M., 2013. Tree population inventory, diversity and degradation analysis of
a tropical dry deciduous forest in Afram Plains, Ghana. For. Ecol. Manag. 295,
145–154.
Appiah, M., Blay, D., Damnyag, L., Dwomoh, F.K., Pappinen, A., Luukkanen, O., 2009.
Dependence on forest resources and tropical deforestation in Ghana. Environ.
Dev. Sustain. 11, 471–487.
Aubreville, A., 1957. Accord a yangambi sur la nomenclature des types africains de
vegetation. Bois et Forets des Tropiques 51, 22–27.
Aynekulu, B.E., 2011. Forest Diversity in Fragmented Landscapes of Northern
Ethiopia and Implications for Conservation. Bonn.
Bennett, A.F., Saunders, D.A., 2010. Habitat fragmentation and landscape change.
In: Ehrlich, S.A. (Ed.), Conservation Biology for All. Oxford University Press, pp.
88–106.
Bowman, D.M.J.S., Murphy, B.P., 2010. Fire and biodiversity. In: Ehrlich, S.A. (Ed.),
Conservation Biology for All. Oxford University Press, pp. 163–180.
Damnyag, L., Saastamoinen, O., Blay, D., Dwomoh, F.K., Anglaaere, L.C.N., Pappinen,
A., 2013. Sustaining protected areas: identifying and controlling deforestation
and forest degradation drivers in the Ankasa Conservation Area, Ghana. Biol.
Conserv. 165, 86–94.
Diallo, Y., Hu, G., Wen, X., 2010. Assessment of land use cover changes using NDVI
and DEM in Puer and Simao Counties, Yunnan Province, China. Rep. Opin. 2,
7–16.
Dimobe, K., Wala, K., Dourma, M., Kiki, M., Woegan, Y.A., Folega, F., Batawila, K.,
Akpagana, K., 2014. Disturbance and population structure of plant communities
in the wildlife reserve of Oti-Mandouri in Togo (West Africa). Annu. Res. Rev.
Biol. 4, 2501–2516.
Dourma, M., 2008. Les forets claires a Isoberlinia sp. dans la zone soudanienne
du Togo: Ecology, Regeneration naturelle et impacts humains. Department of
Botany, University of Lome, pp. 185.
Dourma, M., Wala, K., Bellefontaine, R., Batawila, K., Atsu, G.K., Akpagana, K., 2009.
Comparaison de l’utilisation des ressources forestières et de la régénération
entre deux types de forêts claires à Isoberlinia au Togo. Bois et Forets des
Tropiques 302, 5–19.
Ellis, E.A., Porter-Bolland, L., 2008. Is community-based forest management more
effective than protected areas?: a comparison of land use/land cover change in
two neighboring study areas of the Central Yucatan Peninsula, Mexico. For. Ecol.
Manag. 256, 1971–1983.
Ellis, E.C., 2013. Sustaining biodiversity and people in the world’s anthropogenic
biomes. Curr. Opin. Environ. Sustain. 5, 368–372.
Folega, F., Wala, K., Zhang, C., Zhao, X., Akpagana, K., 2011. Woody vegetation of protected areas in northern Togo. Cases of Barkoissi, Galangashi and Oti-Keran:
ecological and structure analyses of plant communities. For. Stud. China 13,
23–35.
Folega, F., Zhan, C.Y., Zhao, X.H., Wala, K., Akpagana, K., 2010a. Floristic
diversity in most dry and environmentally disturbed areas of Northern
Togo. In: International Conference on Biology Environment and Chemistry,
pp. 241–243.
Folega, F., Zhang, C., Woegan, Y.A., Wala, K., Dourma, M., Batawila, K., Seburanga, J.L.,
Zhao, X., Akpagana, K., 2014a. Structure and ecology of forest plant community
in Togo. J. Trop. For. Sci. 26, 225–239.
Folega, F., Zhang, C., Zhao, X., Wala, K., Batawila, K., Huang, H., Dourma, M., Akpagana,
K., 2014b. Satellite monitoring of land-use and land-cover changes in northern
Togo protected areas. J. For. Res. 25, 385–392.
Folega, F., Zhao, X., Batawila, K., Zhang, C., Huang, H., Dimobe, K., Perky, H., Bawl, A.,
Wala, K., Akpagana, K., 2012. Quick numerical assessment of plant communities
and land use change of Otis prefecture protected areas (North Togo). Afr. J. Agric.
Res. 7, 1011–1022.
Folega, F., Zhao, X., Zhang, C., Wala, K., Akpagana, K., 2010b. Ecological and numerical
analyses of plant communities of the most conserved protected area in NorthTogo. Int. J. Biodivers. Conserv. 2, 359–369.
Fontodji, K.J., Atsri, H., Adjonou, K., Radji, A.R., Kokutse, A.D., Nuto, Y., Kokou, K., 2011.
Impact of charcoal production on biodiversity in Togo (West Africa). In: Pujol,
J.L. (Ed.), The Importance of Biological Interactions in the Study of Biodiversity.
InTech, pp. 215–230.
Fontodji, K.J., Mawussi, G., Nuto, Y., Kokou, K., 2009. Effects of charcoal production on
soil biodiversity and soil physical and chemical properties in Togo, West Africa.
Int. J. Biol. Chem. Sci. 3, 870–879.
Galal, T.M., Fahmy, A.G., 2012. Plant diversity and community structure of Wadi
Gimal protected area, Red Sea Coast of Egypt. Afr. J. Ecol. 50, 266–276.
Gu, H., Subramanian, S.M., 2014. Drivers of change in socio-ecological production
landscapes: implications for better management. Ecol. Soc. 19.
Hoffmann, W.A., Geiger, E.L., Gotsch, S.G., Rossatto, R., Silva, C.R.L., Lau, O.L.,
Haridasan, M., Franco, A., 2012. Ecological thresholds at the Savanna-forest
boundary: how plant traits, resources and fire govern the distribution of tropical
biomes. Ecol. Lett. 15, 759–768.
Jabeen, T., Ahmad, S.S., 2009. Multivariate analysis of environmental and vegetation
data of Ayub National Park Rawalpindi. Soil Environ. 28, 106–112.
Kaye-Zwiebel, E., King, E., 2014. Kenyan pastoralist societies in transition: varying
perceptions of the value of ecosystem services. Ecol. Soc. 19.
Kebede, M., Yirdaw, E., Luukkanen, O., Lemenih, M., 2013. Plant community analysis
and effect of environmental factors on the diversity of woody species in the
moist Afromontane forest of Wondo Genet, South Central Ethiopia. Biodivers.:
Res. Conserv. 29.
Kugbe, J.X., 2012. Spatio-temporal Dynamics of Bush-fire Nutrient Losses and Atmospheric Depositional Gains Across the Northern Savanna Region of Ghana.
Lamouroux, M., 1969. Notice explicative de la carte pédologique du Togo. ORSTOM,
Paris.
Leps, J., Smilauer, P., 2003. Multivariate Analysis of Ecological Data using CANOCO.
New York, USA.
B. Diwediga et al. / Ecological Engineering 85 (2015) 132–143
Matsushita, B., Yang, W., Chen, J., Onda, Y., Qiu, G., 2007. Sensitivity of the enhanced
vegetation index (EVI) and normalized difference vegetation index (NDVI) to
topographic effects: a case study in high-density cypress forest. Sensors 7,
2636–2651.
Mbow, C., Smith, P., Skole, D., Duguma, L., Bustamante, M., 2014. Achieving mitigation and adaptation to climate change through sustainable agroforestry
practices in Africa. Curr. Opin. Environ. Sustain. 6, 8–14.
N’da, D.H., Adou, Y.C.Y., N’guessan, K., Edouard, Kone, M., Sagne, Y.C., 2008. Analyse
de la diversité floristique du parc national de la Marahoué, Centre-Ouest de la
Côte d’Ivoire. Afr. Sci. 4, 552–579.
Norris, K., Asase, A., Collen, B., Gockowksi, J., Mason, J., Phalan, B., Wade, A., 2010.
Biodiversity in a forest-agriculture mosaic – the changing face of West African
rainforests. Biol. Conserv. 143, 2341–2350.
Ofori, E., Atakora, E.T., Kyei-Baffour, N., Antwi, B.O., 2013. Relationship between landscape positions and selected soil properties at a Sawah site in Ghana. Far. J. Agric.
Res. 8, 3646–3652.
Onojeghuo, A.O., Blackburn, G.A., 2011. Forest transition in an ecologically important
region: patterns and causes for landscape dynamics in the Niger Delta. Ecol.
Indic. 11, 1437–1446.
Ouedraogo, A., Kakai, R.G., Thiombiano, A., 2013. Population structure of the
widespread species, Anogeissus leiocarpa (DC.) Guill. & Perr. across the climatic
gradient in West Africa semi-arid area South African. J. Bot. 88, 286–295.
Ouedraogo, I., 2010. Land use dynamics and demographic change in southern Burkina Faso. Acta Universitatis Agriculturae Sueciae.
Paré, S., 2006. Land Use Dynamics, Tree Diversity and Local Perception of Dry Forest
Decline in Southern Burkina Faso, West Africa. Swedish University of Agricultural Sciences, pp. 79.
Pare, S., Savadogo, P., Tigabu, M., Oden, P.C., Ouadba, J., 2009a. Regeneration and
spatial distribution of seedling populations in Sudanian dry forests in relation
to conservation status and human pressure. Trop. Ecol. 50, 339–353.
Pare, S., Savadogo, P., Tigabu, M., Ouadba, J., Oden, P.C., 2010. Consumptive values
and local perception of dry forest decline in Burkina Faso, West Africa. Environ.
Dev. Sustain. 12, 277–295.
Pare, S., Tigabu, M., Savadogo, P., Oden, P.C., Ouadba, J., 2009b. Does designation of
protected areas ensure conservation of tree diversity in the Sudanian dry forest
of Burkina Faso? Far. J. Ecol. 48, 347–360.
Petit, M., 1981. Géomorphologie. Atlas du Togo, Paris.
Pouliot, M., Treue, T., Obiri, B., Ouedraogo, B., 2012. Deforestation and the limited
contribution of forests to rural livelihoods in West Africa: evidence from Burkina
Faso and Ghana. Ambio 41, 738–750.
Sebastia, M.-T., Marks, E., Poch, R.M., 2008. Soil carbon and plant diversity distribution at the farm level in the Savannah region of Northern Togo (West Africa).
Biogeosci. Discuss. 5, 4107–4127.
Shackleton, C.M., Shackleton, S.E., Buiten, E., Bird, N., 2007. The importance of dry
woodlands and forests in rural livelihoods and poverty alleviation in South
Africa. For. Policy Econ. 9, 558–577.
Slik, J.W., Bernard, C.S., Van Beek, M., Breman, F.C., Eichhorn, K.A., 2008. Tree diversity, composition, forest structure and aboveground biomass dynamics after
single and repeated fire in a Bornean rain forest. Oecologia 158, 579–588.
Solon, J., Marek, D., Ewa, R., 2007. Vegetation response to a topographical-soil gradient. CATENA 71, 309–320.
Specht, M.J., Pinto, S.R.R., Albuqueque, U.P., Tabarelli, M., Melo, F.P.L., 2015.
Burning biodiversity: fuelwood harvesting causes forest degradation in humandominated tropical landscapes. Glob. Ecol. Conserv. 3, 200–209.
Steele, M.Z., Shackleton, C.M., Uma Shaanker, R., Ganeshaiah, K.N., Radloff, S., 2015.
The influence of livelihood dependency, local ecological knowledge and market
proximity on the ecological impacts of harvesting non-timber forest products.
For. Policy Econ. 50, 285–291.
143
Tavili, A., Jafari, M., 2009. Interrelations between plants and environmental variables.
Int. J. Environ. Res. 3, 239–246.
Tchabsala, A., Mbolo, M., 2013. Characterization and impact of wood logging on
plant formations in Ngaoundéré District, Adamawa Region, Cameroon. J. Ecol.
Nat. Environ. 5, 265–277.
Thondhlana, G., Vedeld, P., Shackleton, S., 2012. Natural resource use, income and
dependence among San and Mier communities bordering Kgalagadi Transfrontier Park, southern Kalahari, South Africa. Int. J. Sustain. Dev. World Ecol. 19,
460–470.
Tieguhong, J.C., Nkamgnia, E.M., 2012. Household dependence on forests around
Lobeke National Park, Cameroon. Int. For. Rev. 14, 196–212.
Tindan, P.D., 2015. Savanna primary livelihoods at the edge of land degradation:
linkages and impacts in Ghana. Int. J. Innov. Appl. Stud. 10, 119–131.
Traoré, S., Ouattara, K., Ilstedt, U., Schmidt, M., Thiombiano, A., Malmer, A., Nyberg,
G., 2015. Effect of land degradation on carbon and nitrogen pools in two soil
types of a semi-arid landscape in West Africa. Geoderma 241–242, 330–338.
Vanonckelen, S., Lhermitte, S., Rompaey, A.V., 2013. The effect of atmospheric and
topographic correction methods on land cover classification accuracy. Int. J.
Appl. Earth Observ. Geoinf. 24, 9–21.
Vu, Q.M., Le, B.Q., Frossard, E., Vlek, P.L.G., 2014. Socio-economic and biophysical
determinants of land degradation in Vietnam: an integrated causal analysis at
the national level. Land Use Policy 36, 605–617.
Wala, K., Woegan, Y.A., Borozi, W., Dourma, M., Atato, A., Batawila, K., Akpagana, K.,
2012. Assessment of vegetation structure and human impacts in the protected
area of Aledjo (Togo). Far. J. Ecol. 50, 355–366.
Wale, H.A., Bekele, T., Dalle, G., 2012. Plant community and ecological analysis of
woodland vegetation in Metema Area, Amhara National Regional State, Northwestern Ethiopia. J. For. Res. 23, 599–607.
Wampembe, U., Liwa, E., Martine, H., 2014. Assessment of vegetation cover degradation and regeneration after removal of pastoralists in Usangu Wetlands – Mbeya,
Tanzania. Int. J. Ecosyst. 4, 12–18.
White, R.J., Carreiro, M.M., Zipperer, W.C., 2014. Woody plant communities along
urban, suburban, and rural streams in Louisville, Kentucky, USA. Urban Ecosyst.
17, 1061–1094.
Wiesmeier, M., Barthold, F., Spörlein, P., Geuß, U., Hangen, E., Reischl, A., Schilling, B.,
Angst, G., von Lützowa, M., Kögel-Knabner, I., 2014. Estimation of total organic
carbon storage and its driving factors in soils of Bavaria (southeast Germany).
Geoderma Reg. 1, 67–78.
Woegan, Y.A., 2007. Diversite des formations végétales de deux aires protégées
de l’Atakora Nord: la réserve de faune d’Alédjo et Malfakassa. Department of
Botany, University of Lome, pp. 173.
Zeleke, G., Hurni, H., 2001. Implications of land use and land cover dynamics
for mountain resource degradation in the northwestern Ethiopian Highlands.
BioOne 21, 184–191.
Zhai, D.-L., Xu, J.-C., Dai, Z.-C., Cannon, C.H., Grumbine, R.E., 2013. Increasing tree
cover while losing diverse natural forests in tropical Hainan, China. Reg. Environ.
Change 14, 611–621.
Zhai, D.L., Cannon, C.H., Dai, Z.C., Zhang, C.P., Xu, J.C., 2015. Deforestation and
fragmentation of natural forests in the upper Changhua watershed, Hainan,
China: implications for biodiversity conservation. Environ. Monit. Assess. 187,
4137.
Zhang, J.-T., Zhang, F., 2010. Ecological relations between forest communities and
environmental variables in the Lishan Mountain Nature Reserve, China. Far. J.
Agric. Res. 6, 248–259.
Zhang, Z.D., Zang, R.G., 2011. Relationship between species richness of plant functional groups and landscape patterns in a tropical forest of Hainan Island, China.
J. Trop. For. Sci. 23, 289–298.