Content uploaded by Nobert Tafadzwa Mukomberanwa
Author content
All content in this area was uploaded by Nobert Tafadzwa Mukomberanwa on Mar 06, 2025
Content may be subject to copyright.
1 of 17
African Journal of Ecology, 2025; 63:e70034
https://doi.org/10.1111/aje.70034
African Journal of Ecology
RESEARCH ARTICLE
Landscape Connectivity Modelling for the African
Savannah Elephant With Spatial Absorbing Markov Chain
and Predicting the Regenerative Power of the Range in
a Mesic Protected Area
NobertTafadzwaMukomberanwa1 | PhillipTaru1 | BeavenUtete2 | PatmoreNgorima3 |
HonestKomboreroMadamombe1
1Chinhoyi University of Technology, Department of Geoinformatics and Environmental Conser vation, P. Bag, Chinhoyi, Zimbabwe | 2Chinhoyi University
of Technology, Department of Freshwater and Fishery Sciences. P. Bag, Chinhoyi, Zimbabwe | 3Zimbabwe Parks and Wildlife Management Authority,
Scientific Services Unit- Marongora Field Station, Karoi,Zimbabwe
Correspondence: Nobert Tafadzwa Mukomberanwa (nobertmukomberanwa@gmail.com)
Received: 16 January 2025 | Revised: 18 February 2025 | Accepted: 19 Februar y 2025
Funding: The authors received no specific funding for this work.
Keywords: connectivity| corridors| Markov chain| resource selection functions
ABSTRACT
Landscape connectivity is a critical factor influencing the survival and ecological roles of large terrestrial herbivores within
dynamic ecosystems. Yet, the increasing fragmentation of habitats due to human activities, such as agricultural expansion and
infrastructure development, disrupts natural movement patterns and limits access to essential resources. T his is particularly con-
cerning in mesic protected areas, where moderate rainfall supports diverse vegetation but is often bordered by human- dominated
landscapes. To address this challenge, the use of Spatial Absorbing Markov Chain (SAMC) provides a robust framework to sim-
ulate the African savannah elephant (Loxodonta africana) dispersal and identify critical connectivity nodes within fragmented
landscapes. Additionally, assessing and understanding the regenerative potential of these landscapes is vital for evaluating their
capacity to sustain wildlife populations and maintain ecological balance. The objectives of this study were to (i) model the eco-
logical connectivity of Mana Pools National Park (MPNP) by assessing spatial and functional linkages among African savannah
elephant herds and (ii) predict the regenerative potential of the park's range. We used multi- temporal satellite data (2003, 2013,
and 2023), GPS collar data, road transects, and plot- based surveys. The study employed a cellular automata artificial neural net-
work (CA- ANN) to forecast the regenerative potential of the range. Connectivity maps illuminated vital pathways that sustain
the elephants' migratory and foraging behaviours, underscoring the holistic interplay of land cover, slope, and terrain in shaping
movement patterns. The study identified core micro- corridors and broader sub- landscape linkages essential for maintaining the
park's ecological vitality. This interconnectedness serves as a testament to the resilience and regenerative power of the semi-
arid savannah. CA- ANN projections predicted a high landscape regenerative capacity by the year 2083. Highlighting diverse
geographical priorities for connectivity conservation, the research advocates for integrated, multi- scale actions to preserve these
vital linkages. Such insights are pivotal in nurturing the relational integrity of MPNP, ensuring its long- term viability as a sanctu-
ary for elephants and other coexisting life forms. By integrating connectivity modelling and habitat regeneration predictions, this
study advances conservation strategies. It highlights the importance of maintaining functional landscapes to preserve ecosystem
resilience, enhance biodiversity, and mitigate human- wildlife conflicts in increasingly fragmented ecosystems.
© 2025 Joh n Wiley & Sons Ltd.
2 of 17 African Journal of Ecology, 2025
1 | Introduction
Effective conservation and land management require a
thorough understanding of how landscape elements across
both spatial and temporal scales affect population distribu-
tion, abundance, and connectedness (Battisti 2024; Naidoo
etal.2024; Roever, etal.2013). Thus, the integrity and func-
tionality of an ecosystem, along with its biodiversity and eco-
system services maintenance, are made possible by the flow
of species, materials, energy, and information across the land-
scape (Suksavate etal.2019). This understand ing has generated
substantial interest in conservation biology, notably landscape
connectivity investigations (Cushman etal.2010). Landscape
connection between habitats and protected areas is vital for
the protection of wildlife species, especially vulnerable um-
brella species that require broad home ranges covering a vari-
ety of environments (Huang etal.2019). Such species are often
prioritised by managers and planners when making decisions
about landscape design (Osipova et al.2019). As such, land-
scape connectedness functions at multiple scales, contingent
upon the area's geography and the specific species or ecolog-
ical processes being examined (Mukomberanwa etal.2024a;
Mukomberanwa etal.2024b, 2024c; Naidoo etal.2024). Most
connection studies, however, generally concentrate on a sin-
gular scale, which, in the context of resistance- based connec-
tivity modelling, frequently encompass the entire landscape
or protected area (PA) network (Naidoo et al. 2024; Turner
etal.2022). This extensive, singular- scale emphasis may over-
look regions crucial for connectivity at finer scales, which can
be recorded by observed animal movements without the need
for broad landscape statistical modelling and extrapolation
methods (Battisti 2024; Kaszta et al. 2021). The establish-
ment of transfrontier conservation areas (TFCA) and corri-
dors integrating several protected areas, such as the Kavango
–Zambezi TFCA in southern Africa and the Yellowstone to
Yukon Corridor in North America, has emerged as a promis-
ing trend (Naidoo etal.2024). Large mammals are of special
importance for conservation networks because these species
function at vast spatial scales and thus their populations are
more susceptible to fragmentation (Struhsaker et al. 1996;
Tiller et al. 2024). Large mammals often face direct rivalry
with humans; predators compete with human- hunters for
game, ungulates and ruminants harbour diseases which can
transmit to domestic animals, and herbivores may damage
gardens and crops (Hema etal.2017; Pittiglio etal.2012; Sheil
and Salim2004).
Understanding the African savannah elephant (Loxodonta
africana) space usage and connection among their habitats
can counter impediments to biological flow among scattered
populations (Mukomberanwa et al. 2024a; Mukomberanwa
etal.2024b, 2024c). With growing threats such as habitat loss,
fragmentation, and global environmental changes, maintain-
ing ecological connectivity between wildlife populations has
become a top conser vation priority. This connectivity enables
species dispersal, reduces biodiversity loss, and preserves eco-
logical integrity at a landscape scale (Tripathy et al. 2021).
Habitat connection has be en investigated extensively in e cology,
referring to the degree of functional connectedness between
patches of ideal habitats for a certain species (Correa Ayram
et al. 2016; Puyravaud et al. 2017). This sort of connectivity
is based on the notion that there is a complex interaction be-
tween species mobility (set by physiology and behaviour) and
landscape structure (set by landscape composition and config-
uration) which determines the ability of the species to move
through the landscape (Zacarias and Loyola2018). Land cover
modifications have enormous repercussions on species popula-
tions and habitats and are the fundamental causes of the cur-
rent biodiversity catastrophe (Roever et al. 2013). Landscape
connectedness has possible consequences on survival, fitness,
gene flow, diversity, and colonisation of various small popu-
lations (Chibeya et al.2021). Habitat fragmentation and loss
usually harm species in tiny and occasionally isolated popu-
lations (Battisti2024). This raises their danger of extinction as
they are subject to demographic and environmental stochastic
events, together with a lack of gene flow or inbreeding depres-
sion (Mills et al. 2018). The survival of a majority of species
amid growing human- dominated landscapes is strongly de-
pendent upon the ecological connection among their geograph-
ically fragmented populations and habitats (Rood etal.2010).
This is particularly true of wide- ranging mammals who are
declining worldwide, as the connectivity among their habitat
patches is being deteriorated due to the conversion of the nat-
ural landscape into various land use patterns (e.g. transporta-
tion networks, patches of cropland, or other land use factors
that hinder species movement) (Tiller etal.2024). Thus, find-
ing habitat patches and the linkages that establish connectivity
among them is vital to permit efficient spatial planning to pro-
tect these habitat networks.
Savannah dry forests are the most vulnerable tropical terrestrial
ecosystems (Clegg and O'Connor2 017). However, despite their
ecological significant, little research have been undertaken on
the natural regeneration processes essential for restoring these
forests (Brokaw1985; Bussmann2001; Clegg and O'Connor2017;
Hardwick etal. 1997; Moe etal.2009; Omeja etal. 2016; Sheil
and Salim 2004; Skarpe et al. 2004; Vieira and Scariot 2006).
Top- down management of ecosystems by large herbivores is a
matter of current debate between scientists and managers, and
a prominent example is the interaction between elephants and
trees in African savannahs. A prevalent belief among wildlife
managers is that a local reduction in elephant numbers will ul-
timately allow woodland to self- restore to a desired former state
(Moe etal. 2009). Such regeneration is, however, dependent on
the survival of seedlings of damaged tree species (Vieira and
Scariot2006). Savannah systems are characterised by a mixture
of trees and gra sses where the relative dominance of woody cover
is governed by abiotic elements like water, nutrients, and fire,
and biotic factors like herbivory (Tiller et al. 2024). However,
when high numbers of ungulates consume huge quantities of
biomass, fire becomes less crucial in determining the balance
between woodland and grassland (Brokaw 1985). African sa-
vannah elephants are widely regarded as the key element deter-
mining savannah woody cover because of their very enormous
body size permitting (Chibeya etal.2021). One of Africa's big-
gest conservation accomplishments is the recovery of elephant
populations within protected areas (Aleper and Moe2006) such
as those in northern Botswana. This recuperation provides var-
ious problems, however. First, habitat within protected areas is
increasingly destroyed from high- intensity elephant browsing.
Second, the expanding elephant and human populations in the
region have led to huge increases in human–elephant conf lict
3 of 17
around the boundary of protected areas. Tropical landscapes
are changing rapidly as a result of human alterations; nonethe-
less, despite rising deforestation, human population increase,
and the demand for more agricultural land, deforestation rates
have overtaken the rate at which land is converted to agricul-
ture or pasture. For deforested lands to have conservation value
requires an understanding of regeneration rates of vegetation,
the rates at which animals colonise and flourish in regenerating
areas, and the nature of interactions between plants and ani-
mals in the given region.
GPS collar data from wide- ranging animals, such as wolves and
elephants, have supplied the empirical basis for locating and
safeguarding migration corridors, particularly between large,
protected wilderness regions (Mills etal.2018). African savan-
nah elephants exploit corridors to acquire restricted resources
that is forage and water patchily distributed across various envi-
ronments they wander (Gara etal.2021). The existence of small
elephant metapopulations depends on the intactness of these
corridors to access the restricted supplies (Gara etal.2021). By
facilitating movement between core habitats, migration corri-
dors enhance habitat connectivity and support the distribution
of elephants across their landscapes (Gara etal.2017a, 2017b).
Migration via the usage of corridors is proven to ease ecologi-
cal pressure on ecosystems and allow habitat restoration (Gara
etal.2017a , 2017b). However, human land uses, including urban
development and agriculture, have imposed great pressure on
corridors, resulting in the narrowing of corridors and hence
loss of genetic flow among elephant populations (Chaiyarat
etal.2022). The spatial absorbing Markov chain was utilised to
model landscape connectivity (Marx etal.2020). For organisms
dispersing across complex landscapes, for each time step an in-
dividual can either survive and stay at the same location (i.e. site
fidelity), survive and move to a nearby site, or suffer mortality.
The SAMC framework honours this simple idea by considering
‘transient’ states of fidelity and movement, and an ‘absorbing’,
permanent state of mortality (Marx etal.2020). The step selec-
tion functions (SSFs) were employed to analyse the influence of
impediments on elephant locomotion. Connectivity maps were
constructed based on the estimated resistance surfaces by util-
ising the least- cost path and spatial absorbing Markov chain
through a proposed analytical method. The predictors consist
of landcover, elevation, digital elevation model, roads, and dis-
tance to the next cell of specific landcover. The objectives of this
study were to (i) model the ecological connectivity of Mana Pools
National Park (MPNP) by assessing spatial and functional link-
ages among African savannah elephant herds and to (ii) predict
the regenerative capacity of the range. The research questions
for this study were: (1) how do spatial and functional linkages
among African savannah elephant herds influence ecological
connectivity within Mana Pools National Park?, what are the
key landscape features and environmental factors that facilitate
or hinder elephant movement and connectivity in MPNP?, how
can a spatial absorbing Markov chain model be applied to assess
habitat connectivity and predict movement patterns of elephant
herds in MPNP?, what is the regenerative potential of different
habitat patches within MPNP based on elephant movement pat-
terns and vegetation dynamics?. We hypothesise that Landscape
connectivity for African savannah elephants, as modelled using
a spatial absorbing Markov chain, significantly influences the
elephants' ability to traverse the landscape and correlates with
the regenerative capacity of vegetation in mesic protected areas.
The connectivity maps highlighted many potential connecting
pathways, bottlenecks, and varying essential linkages between
core sites in the region. This provides an alternate possibility in
gathering information on landscape connectivity for offering aid
in conservation planning at the landscape level.
2 | Materials and Methods
2.1 | Study Area
This study was carried out in Mana Pools National Park (MPNP)
which is ~2124 km2 in size. MPNP is located within 29.15° E-
29.74° E and 15.67° S- 16.29° S in the northern part of Zimbabwe
(Figure 1). The area experiences a single rainy season from
November to April. Mean annual rainfall is 724 mm, and the
elevation ranges from 329 to 1180 m above sea level (Dunham
and du Toit2013). July is the coldest month, and November is
the hottest. Although the park occupies 2196 km2, it is a part of
nearly 15646km2 of wildlife area (Mid Zambezi), the majority
of which is employed for sport hunting within the surrounding
Safari Areas. A great diversity of vegetation kinds is found in
MPNP, which supports a diversified ecology (Ndoro etal.2 016).
Mixed Miombo and Brachystegia forests dominate the Chewore
area and the Zambezi escarpment, creating a particular envi-
ronment for a diversity of animals. Huge mopane trees fill the
floodplains of MPNP, with some pockets of ‘jesse bush’, which
form dense thickets that ser ve as food and shelter for the African
savannah elephant (Matsa etal.2022). The alluvial soils of the
floodplains of Mana Pools National Park are home to open for-
ests of Faidherbia albida trees. During the dry season, these
trees grow fruit, attracting large terrestrial mammals like the
African savannah elephants and other herbivores. Diverse
woodland specie, such as Kigelia africana, Trichilia emetica, and
superb Ficus bussei specimens, are among the other noteworthy
tree species found in Mana Pools National Park (Dunham and
du Toit2013). These trees help to ensure the general diversity
of habitats inside the park. During the dry season, large mam-
mals concentrate on the Zambezi alluvium, with populations
increasing as the season goes on. The Zambezi River's drinking
water and a few alluvial flood plains, F. alb ida fruits that fall
to the ground and verdant grass near to the river all drew large
animals like the African savannah elephant (Mukomberanwa
et al. 2024a; Mukomberanwa et al. 2024b, 2024c). Elephants
in Mana Pools National Park demonstrate a remarkable prev-
alence across multiple vegetation types, reflecting their re-
silience and ecological importance within the environment
(Mukomberanwa et al. 2024a; Mukomberanwa et al. 2024b,
2024c). The Mid- Zambezi Valley region is recognised for its
healthy elephant population, which is one of the four elephants
range areas in Zimbabwe. The park's diverse ecosystems, rang-
ing from riverine forests to open savannahs, provide abundant
food supplies for the African savannah elephants throughout
the year. Anthropogenic activities around Mana Pools National
Park include agricultural expansion, illegal hunting, fishing,
and tourism- related disturbances. Human settlements near the
park, particularly in areas like Mahuwe and Kanyemba, con-
tribute to habitat fragmentation. Road networks and increasing
human- wildlife conflicts further impact elephant movement
and ecological connectivity in the region.
4 of 17 African Journal of Ecology, 2025
2.2 | Data Preparation for the SAMC Object
This investigation utilises a comprehensive, high- resolution GPS
tracking database containing roughly eleven thousand GPS co-
ordinates from nearly seven monitored elephants and their re-
spective herds. Data regarding elephant locations were gathered
over the preceding three years (2021–2024). The methodology
includes road transects, plot- based surveys, and GPS collar data.
Satellite data for the years 2003, 2013, and 2023 were used to
analyse Land Use Land Cover (LULC) and existing maps were
utilised to identify prospective transects along roadways that cut
through different habitat categories. We marked 10 road tran-
sects that were 5–10 km in length and were spaced 10–15 km
apart. To pick plot areas, we used eleven thousand GPS points
to identify crucial locations. We established 26 plots that were
positioned in diverse habitat types (Figure2). The marked plots
were 50 × 50 m and were randomly selected within critical areas,
but we established a minimum distance of 10 km between plots
to prevent spatial autocorrelation. Plots and road transects were
marked in March, July, and November 2024.
FIGUR E | Location of Mana Pools National Park, Zimbabwe.
FIGUR E | Plot spacing and their basal area in MPNP.
5 of 17
The package SAMC necessitates, at a minimum, data on ab-
sorption (or mortality) and terrain resistance about movement
behaviour (Mar x etal.2020). Moreover, fidelity and occupancy
(or abundance) statistics may also be utilised. Similar to other
applications that define ‘resistance’ for landscape connectiv-
ity, the parameterisation of the transition matrix utilised to
analyse movement behaviour in the SAMC (Q) can be derived
via behavioural tests, telemetry, habitat utilisation, or expert
assessment (Fletcher and Fortin2018; Kautz etal.2006; Zeller
et al. 2012). Parameterising Q of the SAMC is analogous to
parameterising least- cost and circuit theory methodologies.
We parameterise resistance for African savannah elephants
in MPNP according to the analysis by Kautz etal.(2006) util-
ising land- cover data and resistance coding as demonstrated
in Fletcher and Fortin (2018). This study utilised data from
GPS- collared African savannah elephants to assess habitat
preferences through resource selection functions, which were
subsequently transformed into a rank- based measure of move-
ment resistance. We consider a raster- based, land- cover map
(Figure3), aggregated to a 100 m resolution (representing the
approximate grain of telemetry error; post- aggregation, 54,933
cells on the map). We set the GPS collar data as source loca-
tions (as in Kautz etal.2006) for individual dispersers by cre-
ating a raster map where the centroids of these two locations
are set to one and all other locations on the raster are set to
0 (Figure A1). Fidelity rates, R, can be approximated using
numerous ways, including employing field trials (Fletcher Jr.
etal.2019) and habitat- specific survival estimates (Low and
Merry2010). For example, if annual survival rates are accessi-
ble, these rates could be adjusted to represent the anticipated
fidelity rate per time step (where a ‘time step’ denotes the av-
erage duration expected for movement between neighbour-
ing cells). Furthermore, if the average dispersal distance for
a species is established, R could be adjusted to provide dis-
persal distance estimates from the SAMC that correspond to
known dispersal lengths (Marx et al. 2020). The SAMC im-
plies non- zero fidelity risk because if fidelity risk is 0, it would
predict that all places are eventually reached by mobile spe-
cies. To parameterise the fidelity risk of spatial variation for
African savannah elephants, we concentrate on telemetered
data from highways, a significant factor disrupting elephant
migrations. Using these sites, we evaluated the fraction of
movement places within 100 m (the map resolution) of major
roadways relative to what would be expected based on avail-
able habitat (landcover map). Movement locations (n = 11,000)
were 3.7× more likely to be within 100 m of roadways than
> 100 m based on these data. We utilise this information to
alter relative predicted movement rates near roads. We then
FIGUR E | Landcover data used for modelling landscape connectivity in MPNP (0 = water, 1 = forest, 2 = bareland, 3 = shrubland). Coordinates
are in Africa Albers Equal A rea Conic (ESRI 102022).
6 of 17 African Journal of Ecology, 2025
calibrate absolute rates of movement based on known disper-
sal lengths in African savannah elephants, where the mean
dispersal distance has been estimated as 10.3 km. To do so, we
first investigate different baseline rates of movement across a
homogeneous 100 × 100 cell (80 K) grid with the cell resolution
considered in mapping (100 m). The software can be used to
calculate the likelihood of visiting sites as a function of dis-
tance with the function (D) to build probable dispersal kernels
based on known mortality risk (Figure6). We measured prob-
abilities from the source location to locations as a function of
distance to determine expected mean movement lengths based
on the parameters of an exponential dispersal kernel. In this
example, baseline fidelity risk would be 6.5 × 10–6 per cell to
create mean movement distances of around 10 km (Figure6).
We then enhanced this basic mortality probability 3.7× near
roadways, given movement locations to construct a fidelity
risk map for the SAMC. Information on species distribution
and abundance can also be included as a map, which is neces-
sary for some (but not all) metrics. In most situations, we ex-
pect that such information will be supplied into the model to
capture source- destination locations (Cushman et al. 2009),
local populations (Larkin et al. 2004), or individual dispers-
ers (Fletcher Jr. etal.2018). This information can either re-
flect the probability that an individual occurs at a location,
in which output generally reflects probabilities of individual
movement or mortality, or it can reflect the number of indi-
viduals, in which output generally reflects the number of indi-
viduals moving or suffering mortality. Including information
on distribution and abundance increases computational effi-
ciency. If such information is not supplied, users are implicitly
assuming that individuals start from all locations, similarly
to some graph theory applications to resistance modelling
(Carroll etal.2020) and roughly related to factorial least- cost
analysis (Elliot etal.2014). All sets of data must be saved either
as matrices or raster maps. Mixing matrices and rasters is not
supported due to underlying differences in the way they store
data and potential difficulties that can develop when users are
not aware of these differences. Every collection of data must
have the same dimensions (number of rows and columns), and
data are supported as long as they occur in the same positions
in every set of data. If the datasets are saved as rasters, then
they must have the same geographic coordinate extents and
coordinate reference system (CRS). The coordinate reference
system supported in the SAMC package are in Africa Albers
Equal Area Conic (ESRI 102022). The function is provided to
simplify the process of checking that all these conditions are
met. The next step is to establish a huge SAMC object to cal-
culate the various metrics stated in Fletcher Jr. etal.(2019).
2.3 | Modelling Landscape Connectivity Using
the Spatial Absorbing Markov Chain
The Spatial Absorbing Markov Chain (SAMC) method is a pow-
erful tool for modelling landscape connectivity, particularly
for species like the African savannah elephant in Mana Pools
National Park. This method provides a probabilistic framework
to evaluate how elephants traverse the landscape, taking into ac-
count habitat heterogeneity, movement barriers, and preferred
pathways, which are crucial for understanding their spatial ecol-
ogy and informing conservation strategies.
In this approach, the landscape is conceptualised as a grid of
nodes, where each node represents a spatial location such as
a patch of habitat or a waterhole. Movement between nodes is
governed by transition probabilities, which are calculated based
on ecological factors influencing elephant movement. For the
savannah elephants in Mana Pools, these factors might include
proximity to water sources, vegetation density, terrain rugged-
ness, human settlements, and the availability of food. Nodes
with favourable attributes, such as access to water and forage,
have higher transition probabilities, while areas with high re-
sistance, such as steep terrain or human activity zones, have
lower probabilities. The SAMC method incorporates absorbing
states, which are specific nodes in the landscape where the pro-
cess terminates. In the context of Mana Pools, these absorbing
states could represent key destinations for elephants, such as
critical waterholes, grazing areas, or safe habitats. Once an el-
ephant reaches one of these absorbing states, it is assumed to
stay there, reflecting the goal of the movement (e.g. accessing
water or forage). The first step in applying the SAMC method
is constructing a transition matrix that defines the probabilities
of movement between all nodes in the landscape. These proba-
bilities are influenced by a combination of ecological data, such
as habitat quality and movement resistance, often derived from
Geographic Information System (GIS) layers. The next step is
designating the absorbing states, which are typically based on
known locations of critical resources or habitats. By analysing
this transition matrix, several key metrics can be derived. One
primary output is the probability of an elephant reaching an
absorbing state from any given starting point in the landscape.
This provides insights into connectivity, highlighting which
areas are well connected and which may act as barriers. Another
useful metric is the expected number of steps or movements re-
quired to reach an absorbing state, offering an understanding of
how accessible critical areas are within the landscape.
In the case of Mana Pools National Park, the SAMC method
can help identify vital movement corridors that elephants use
to access waterholes during the dry season or migrate between
different parts of the park. It can also pinpoint areas where
connectivity is compromised due to natural barriers like es-
carpments or anthropogenic pressures such as agricultural ex-
pansion or poaching risk. Conservation managers can use these
findings to prioritise interventions, such as mitigating human-
elephant conflict or restoring degraded corridors to ensure safe
passage for elephants.
The SAMC method's probabilistic framework makes it espe-
cially valuable for modelling elephant movement in dynamic
and heterogeneous environments like Mana Pools. By provid-
ing detailed insights into landscape connectivity, this approach
supports evidence- based decision- making to balance elephant
conservation with sustainable land use, ensuring the long- term
survival of these iconic animals in the park.
2.4 | Predicting the Regenerative Power
of the Range Using LULC Data
Landsat 8 images for the years 2003, 2013, and 2023 were
classified using Google Earth Engine (GEE) (FigureA4). The
image classification for the study period was performed by
7 of 17
supervised classification using a maximum likelihood clas-
sifier. The classified images were assigned to the respective
classes, which are forest, water, grassland, and bare land. The
satellite images for the years 2003, 2013, and 2023 were used
to identify the LULC in MPNP, and ground truthing was done
to verify the data obtained in the images. GEE was preferred
due to its computation power. Google Earth Engine (GEE)
JavaScript was used for the classification (Mukomberanwa
etal.2024a; Mukomberanwa etal.2024b, 2024c). To perform
a supervised classification of Landsat 8 images in GEE, the fol-
lowing steps were employed: (1) collect training data: this was
done by drawing polygons on the map to represent different
land cover classes or by importing predefined training data
from an Earth Engine table asset. Training data are instrumen-
tal in supervised image classification. The training dataset is a
labelled set of data that are used to inform or ‘train’ a classifier.
The trained classifier can then be applied to new data to create
a classification (Mukomberanwa etal.2024a; Mukomberanwa
etal.2024b, 2024c). In this study, land cover training data con-
tained examples of each class in the study's legend. Based on
the labels, the classifier can predict the most likely land cover
class for each pixel in an image. The categorical classification
and the training labels are therefore categorical. (2) Assemble
features: We selected the bands that we wanted to use as pre-
dictors and created a feature collection that includes the class
labels and predictor variables. (3) Instantiate a classifier: We
then choose a classifier from the ee.Classifier package, such
as ee.Classifier.cart or a Classification and Regression Trees
(CART) classifier. Image classification using CART in GEE in-
volves leveraging its powerful computational capabilities and
vast remote sensing datasets (Mukomberanwa et al. 2024a;
Mukomberanwa etal.2024b, 2024c). CART is a decision tree-
based classification method that partitions data into subsets
based on recursive binary splits, making it effective for clas-
sifying satellite imagery. (4) Train the classifier: The next step
was to use the train method to train the classifier on the train-
ing data. (5) Classify the image: Finally, we used the classify
method to classify the image or feature collection. To extract
the area of each class in hectares, we employed the ee.image.
pixel Area method to calculate the area of each pixel in square
meters, then multiply by the number of pixels in each class and
divide by 10,000 to convert to hectares. This methodology was
used to obtain land use land cover changes in MPNP as illus-
trated. The user guide of Climate Change Initiatives (CCI Land
Cover (LC) team, n.d.) product gives accuracy ratings of the
CCI- LC map year 2018 using GlobCover 2009 validation data-
set (Mukomberanwa etal.2024a; Mukomberanwa etal.2024b,
2024c). Overall accuracies turned out to be 82.45% employing
‘certain’ whether ‘homogeneous’ or ‘heterogeneous’ points
what's more, 76.4% was identified utilising only ‘homogeneous’
and ‘certain’ locations (ESA Climate Change Initiative—Land
Cover project, 2017).
Projection of future changes in LULC for the MPNP for the year
2083 was done using Cellular automation (CA- ANN). For the
purpose of producing future simulation maps, the classified
maps of 2003, 2013, and 2023 (Figure A4) were used as inputs
and integrated w ith spatial var iables (DEM and slope) (Figure8).
Digital Elevation Models (DEM) do not normally change every
year. Very often, they may change depending on weather pro-
cesses like sedimentation and erosion processes. Slope and DEM
are important inputs for LULC simulation because they provide
information about the topography of the land and influence
the distribution of woodlands, grasslands, and water resources
across the landscape (Mashapa et al.2021). Validation checks
were done to ensure whether the model can accurately predict
known outcomes and ensure its reliability for future projections
(e.g. the year 2083) (Figure 9). Metrics like overall accuracy,
Kappa coefficient, or spatial similarity indices are often used to
measure the degree of agreement between the predicted LULC
map changes, which were validated using the Kappa statistics.
The neural network learning curve shows the learning process
the model underwent, approaching zero, signifying model reli-
ability (Figure A5). The Neural Network Learning Curve in a
cellular automata (CA) model is a graphical representation of
the performance of a neural network (NN) that may be used to
learn or approximate the rules or dynamics of the CA. It tracks
the progress of the network's training and helps in understand-
ing how well it learns the underlying patterns and relationships
in the data. The training curve shows decreasing loss, indicating
that the NN is improving its prediction of cell state transitions
based on the training dataset. For the validation curve, if the
validation loss decreases and stabilises near the training loss,
the NN has generalised well. Low training and validation errors
indicate a well- trained model. Persistent high error suggests
problems with the dataset (e.g. noise, insufficient data) or the
network architecture.
3 | Results
3.1 | Modelling Landscape Connectivity Using
the Spatial Absorbing Markov Chain
Vector findings obtained by analysis match to cells in the land-
scape data and can be examined appropriately (cell size = 100).
The function is supplied to assist this procedure; it guarantees
that the values in the vector results are mapped to the land-
scape in the correct order and properly handles results based
on datasets with values. The result is a raster that may be plot-
ted instantly using without any additional processing. For the
African savannah elephant, we mapped both D and B with the
function. Mapping dispersal from two source places provides a
tool to understand the possibility of dispersers visiting any lo-
cation on the terrain (Figure A2), indicating that dispersal is
likely to decrease with distance from the source locations. We
remark that utilising the algorithm for each source location in-
dependently helps isolate the roles of source habitats on possible
connections. We also map the estimated chance of mortality of
individuals spreading from these two source areas (Figure 6).
Expected mortality locations are usually comparable to move-
ment locations, except that mortality risk jumps near highways
that are within the dispersal range from the source locations.
Finally, we observed that life expectancy (number of expected
time steps before dying, z) was estimated to be 2% greater for
persons dispersing from the Zambezi River (z = 2746 steps) than
from the Artificial water pans (z = 2461 steps). Other types of in-
terpretation and visualisation are possible, such as demarcation
of potential corridors, using models to parameterise network (or
patch- based graph) links in connectiv ity models, and identif ying
the extent to which protected areas may contribute dispersers to
larger metapopulations for discussion. We highlight that at least
8 of 17 African Journal of Ecology, 2025
three sources of uncertainty emerge in predictions and interpre-
tation of this framework: (1) uncertainty in the model frame-
work (e.g. assumption of biased random walks); (2) uncertainty
in inputs (e.g. resistance, mortality uncertainty as a function of
land use); and (3) uncertainty in the metrics themselves (e.g. the
variance of D). The first source can be evaluated by comparing
the SAMC to other modelling frameworks, the second can be
addressed when resistance and mortality inputs have estimates
of precision (e.g. SEs), whereas the third requires new model
derivations (e.g. see Ross 2010 for an example for estimating the
variance in life expectancy with an absorbing Markov chain).
The results represents the land use and land cover (LULC) clas-
sification of Mana Pools National Park (MPNP) for the year 2023
(Figure 3), utilised in habitat connectivity modelling using the
SAMC approach. The classification categories in the legend are
as follows: areas in dark blue represent water bodies, including
rivers and wetlands. These regions are crucial for wildlife, par-
ticularly as water sources for species like the African savannah
elephants and other fauna in the park. Green areas correspond
to forested regions, which provide essential habitat for many
species, offering shelter, foraging opportunities, and critical
ecological functions. Forests are key contributors to habitat con-
nectivity within MPNP. Cyan- coloured regions represent bare
or sparsely vegetated land. These areas may have limited habi-
tat value but could serve as transitional zones between different
habitats. They may also reflect human- disturbed zones or natu-
ral clearings. Yellow areas indicate shrubland, which likely con-
stitutes the dominant land cover type in the park. Shrublands
are significant for many species as foraging habitats and corri-
dors connecting forest patches. The spatial distribution of these
land cover types highlights the heterogeneity of habitats within
the park. Shrublands dominate the landscape, while forests are
scattered and concentrated in specific zones. Water bodies are
primarily located in the northern part of the park, potentially
serving as critical attractors for wildlife during dry periods
(Figure3). The integration of this LULC map into SAMC mod-
elling allows for the identification of key habitats and movement
corridors. It helps assess how different land cover types influ-
ence connectivity and movement patterns. This information is
vital for conservation planning, focusing on maintaining eco-
logical connectivity and protecting critical habitats in MPNP.
The findings (Figure4) also depict habitat connectivity in Mana
Pools National Park, with values ranging from high (green, up
to 3.0) to low (yellow and beige). The predominantly green areas
suggest that most of the park maintains strong habitat connec-
tivity, particularly in the northern regions, which likely serve as
critical corridors for elephant movement and ecological inter-
actions. Patches of yellow and beige indicate zones with lower
connectivity, possibly due to habitat fragmentation or natural
barriers (Figure4). These areas may hinder species movement
and reduce ecological functionality. The overall connectivity
highlights the park's ecological resilience, but targeted conser-
vation efforts are necessary to address less connected zones.
The results also illustrate disconnected regions in Mana Pools
National Park identified using the Spatial Absorbing Markov
Chain (SAMC) method (Figure5). Areas shaded in green rep-
resent higher values, indicating regions with greater disconnec-
tion, while lighter pink areas correspond to more connected
regions (Figure5). The green zones highlight potential barri-
ers to movement, which could be caused by natural features
such as rivers or cliffs, or human influences such as roads or
settlements. Identifying these disconnected regions is essential
for conservation efforts, as they can hinder wildlife movement
and ecosystem functionality. Addressing these barriers through
FIGUR E | Habitat connectivity in Mana Pools National Park, ranging from high (green, up to 3.0) to low (yellow and beige). Northern green
areas indicate strong connectivity, while yellow/beige suggest fragmentation. Coordinates are in Africa Albers Equal Area Conic (ESRI 102022).
9 of 17
targeted interventions can improve habitat connectivity and
support ecological resilience.
The results also illustrate a homogeneous grid within Mana
Pools National Park, derived from a SAMC object to assess land-
scape connectivity (Figure A3). The left map (binary connec-
tivity) shows potential movement paths, with purple indicating
high connectivity and yellow highlighting low connectivity. The
right map represents predicted movement distances, ref lecting
the expected dispersal capability in the landscape. The resis-
tance maps illustrate the spatial distribution of movement resis-
tance for African savannah elephants in Mana Pools National
Park, based on telemetry data and resource selection functions
(Figure A3). The gradient represents resistance values, with
higher values (red) indicating areas with greater difficulty for
elephant movement, likely due to factors such as proximity to
major roads or unsuitable habitat. Lower resistance values
(green) correspond to areas more conducive to movement. The
concentrated red zone suggests a localised barrier or high-
resistance feature, possibly near roads or human activity. The
map underscores the importance of mitigating anthropogenic
barriers to support elephant connectivity and conservation. The
findings also illustrate the expected probability of mortality for
individuals dispersing from source areas within Mana Pools
National Park, based on a SAMC object analysis (Figure A4).
Higher probabilities of mortality are indicated by warmer co-
lours (yellow to red), while lower probabilities are represented
by cooler colours (green). Areas with no dispersal inf luence ap-
pear in grey, highlighting regions of limited connectivity or un-
suitable habitat. This visualisation helps assess the risks faced
by dispersing individuals and informs conservation strategies to
mitigate mortality. The findings display a resistance surface for
Mana Pools National Park, derived from resource selection func-
tions based on telemetry data from African savannah elephants.
Green areas represent low resistance, indicating preferred or
easily traversable habitats, while warmer colours (yellow to
red) indicate higher resistance, representing less favourable or
challenging areas. Major roads are clearly delineated and act as
barriers to movement, influencing the connectivity of the land-
scape. This resistance map is vital for understanding elephant
movement patterns and identifying areas where roads or other
features may hinder connectivity.
3.2 | Predicting the Regenerative Power
of the Range Using LULC Data
The changes in land use and land cover in Mana Pools National
Park between 2003 and 2083 reveal insights into the regenera-
tive capacity of the range (Table1). Over the study period, the
area covered by water showed significant fluctuations, decreas-
ing sharply by 58.7% between 2003 and 2013 but recovering by
50.6% between 2013 and 2023. However, a subsequent decline of
8.0% between 2023 and 2083 suggests challenges in maintain-
ing consistent water levels, possibly due to climatic or anthropo-
genic influences. Forested areas displayed a steady and robust
increase, growing by 0.396% from 2003 to 2013, 10.280% from
2013 to 2023, and 10.676% from 2023 to 2083. This sustained ex-
pansion points to a strong regenerative potential, likely driven by
effective conservation measures and favourable ecological con-
ditions. Shrubland exhibited the most dynamic growth, expand-
ing by 20.813% from 2003 to 2013, 4.154% from 2013 to 2023,
and a remarkable 24.967% from 2023 to 2083. This suggests that
shrubland is adapting well to environmental changes, serving as
a critical buffer for biodiversity. Conversely, bareland decreased
significantly throughout the period, dropping by 20.622% from
2003 to 2013, 14.941% from 2013 to 2023, and 35.56% from 2023
to 2083. This reduction highlights successful land restoration ef-
forts, indicating the park's strong regenerative power. The neu-
ral network learning network curve shows the learning rate and
the Root Mean Square Error (FigureA5) to depict how well the
model learns from the data. The final outcome is a predicted
landscape for the year 2083 (Figure9).
FIGUR E | Disconnected regions in Mana Pools National Park
identified using the Spatial Absorbing Markov Chain (SAMC). Areas
shaded in green represent higher values, indicat ing regions with greater
disconnection, while lighter pink areas correspond to more connected
regions.
FIGUR E | Predicted movement distances and relating predictions
to known dispersal distances. (Mapping dispersal from two source ar-
eas provides a means to underst and the probability of dispersers visiti ng
any location on the landscape, illustrating that dispersal is expected to
decrease with distance from the source locations). Coordinates are in
Africa Albers Equal Area Conic (ESR I 102022).
10 of 17 African Journal of Ecology, 2025
4 | Discussion
4.1 | Modelling Landscape Connectivity Using
the Spatial Absorbing Markov Chain
The findings represent landscape connectivity and predicted
movement of African savannah elephants in Mana Pools
National Park (MPNP) using a spatial absorbing Markov chain
(SAMC) approach. It provides a detailed visualisation of hab-
itat suitability and connectivity, critical for understanding
elephant movement patterns and identifying barriers or facil-
itators of dispersal. The results highlight predicted movement
pathways, reflecting areas where elephants are likely to traverse
based on resource selection functions and landscape resistance
(Figure 4). The colour gradient indicates movement potential:
green areas denote higher connectivity and easier movement
paths, while orange to brown regions signify lower connectivity
or resistance. The variation in connectivity aligns with factors
such as habitat quality, resource availability, and anthropogenic
barriers, including roads. The SAMC model incorporates prob-
abilistic movement patterns (Marx etal.2020), where elephants
are assumed to transition between landscape nodes based on
habitat preferences and resistance. This approach captures the
cumulative effects of connectivity and provides insights into po-
tential corridors and pinch points. Areas with high connectivity
suggest critical zones for conservation, facilitating movement
between habitats and reducing the risk of population fragmen-
tation (Battisti2024). Additionally, the findings integrate telem-
etry data from elephants to refine predictions. The inclusion of
major roads, water sources, and vegetation density enhances
the model's realism, offering a practical tool for conservation
planning. Roads, for instance, may act as barriers or mortality
hotspots, reducing connectivity in adjacent regions. This analy-
sis is vital for MPNP, a critical conservation area in the Zambezi
Valley, as it identifies priority zones for interventions, such as
mitigating road impacts or restoring degraded habitats. By link-
ing predicted elephant movement with connectivity, the SAMC
approach supports evidence- based strategies to maintain eco-
logical processes and safeguard elephant populations in MPNP.
The results depict a homogeneous grid of constant abundance,
derived from the spatial absorbing Markov chain (SAMC) anal-
ysis, to predict movement distances within Mana Pools National
Park (MPNP). This modelling approach is designed to assess
landscape connectivity and relate predictions of movement po-
tential to known dispersal distances, particularly for key species
like African savannah elephants (Figure7). The left map shows
binary connectivity, where purple areas represent accessible or
connected regions (value = 1), and yellow indicates inaccessible
or disconnected zones (value = 0). The visualisation highlights
the road network and other linear features influencing move-
ment patterns across the park. These features act as structural
constraints, defining regions of high connectivity along path-
ways and reduced connectivity in isolated areas. The right map
represents a uniform dispersal scenario, where the entire land-
scape is treated as having a constant abundance, facilitating the
estimation of predicted movement distances (Figure 6). This
approach provides a baseline for evaluating the dispersal ca-
pacity of species, allowing comparisons to observed movement
behaviours. Uniformly yellow regions emphasise the underly-
ing assumption of homogeneity in dispersal potential across the
landscape, offering a theoretical framework for understanding
how elephants or other animals might traverse the terrain with-
out specific resource constraints. Ecologically, these maps are
essential for understanding the inf luence of landscape structure
on movement and connectivity. The presence of roads and frag-
mented areas underscores the potential barriers to dispersal,
which can result in population isolation and genetic bottlenecks.
For species like elephants, which rely on wide- ranging move-
ments for foraging, reproduction, and maintaining ecosystem
dynamics, identifying and mitigating barriers is crucial. This
analysis informs conservation efforts by pinpointing critical
areas for intervention, such as mitigating road impacts, restor-
ing degraded habitats, and prioritising connectivity corridors.
Ultimately, the SAMC- based predictions provide a robust frame-
work for ensuring ecological resilience and promoting sustain-
able biodiversity in MPNP.
FIGUR E | The expected probability of mortality of individuals
dispersing from these two source areas. Expected mortality locations
are generally similar to movement locations, except that mortality risk
spikes near roads that are within the dispersal range from the source
locations. Coordinates are in Africa Albers Equal Area Conic (ESRI
102022).
TABLE | Change trends of each land cover class.
2003 2013 % change 2003–2013 2023 % change 2013–2023 2083 % change 2023–2083
Water 301.54 156.87 −0.587 281.77 0.506 261 −0.080
Forest 5114.4 5212 0.396 7 747.69 10.280 8023 10.676
Shrubland 7911.15 13044.6 20.813 14069.28 4.154 14,702 24.967
Bareland 11337.56 6251.17 −20.622 2565.9 −14.941 1677 −35.56
11 of 17
FIGUR E | Spatial variables digital elevation model (DEM) and slope used to predict the future landscape scenario.
FIGUR E | Predicted landscape for the year 2083.
12 of 17 African Journal of Ecology, 2025
The findings display the expected probability of mortality for
African savannah elephants dispersing from two source areas
within Mana Pools National Park, as modelled using the Spatial
Absorbing Markov Chain (SAMC) approach. The probabilities,
ranging from 0.0 (low risk) to 0.4 (high risk), are represented
by a gradient of colours. This spatial depiction highlights the
landscape's inf luence on dispersal success and potential mor-
tality risks faced by elephants during their movement. The
highest probabilities of mortality are concentrated around the
source areas, with the core regions displaying prominent yellow
or orange hues. These areas may represent challenging condi-
tions for dispersal, possibly due to anthropogenic factors such as
proximity to roads, settlements, or other human disturbances.
Alternatively, these zones may lack adequate resources, such as
water or food, or exhibit dense vegetation or terrain that impedes
movement, increasing vulnerability to mortality. Identifying
these zones is critical for targeted conservation efforts, such as
mitigating human- wildlife conf lict or restoring habitat quality.
As dispersing individuals move further from the source areas,
the probability of mortality decreases, transitioning to green and
eventually grey areas. These regions are indicative of lower- risk
zones, likely offering more suitable habitat conditions or fewer
obstacles to movement. Such areas could serve as important cor-
ridors for safe dispersal, facilitating connectivity between pop-
ulations and supporting the long- term viability of the species.
The findings of this study highlight the ecological connectivity
of Mana Pools National Park (MPNP) by assessing the spatial
and functional linkages among African savannah elephant
herds. The spatial absorbing Markov chain (SAMC) approach ef-
fectively mapped dispersal patterns, identifying movement cor-
ridors and connectiv ity zones. The results indicate that dispersal
probability decreases with distance from source locations, with
notable differences between elephants dispersing from the
Zambezi River and artificial water pans. Mortality risk was
highest near roads, suggesting that anthropogenic features sig-
nificantly impact elephant movement and survival. The connec-
tivity analysis revealed that northern areas of the park exhibit
strong habitat connectivity, serving as critical corridors for el-
ephant movement, while fragmented zones in yellow and beige
indicate reduced connectivity. The resistance surface analysis
further identified major roads as barriers to movement, reinforc-
ing the importance of mitigating human- induced obstacles to
improve ecological connectivity. The predictive capacity of the
SAMC framework underscores its potential for parameterising
network- based connectivity models, assessing conservation pri-
orities, and guiding landscape management decisions. Given the
uncertainties in model predictions—stemming from assump-
tions, resistance values, and mortality estimates—comparisons
with alternative modelling approaches and enhanced data preci-
sion remain necessary for refining conservation strategies.
4.2 | Predicting the Regenerative Power
of the Range Using LULC Data
The changes in land use and land cover in Mana Pools National
Park from 2003 to predicted 2083 map provide valuable insights
into the park's regenerative power and ecological dynamics.
The data reveal distinct trends in water, forest, shrubland,
and bareland, each with significant ecological implications
for understanding the park's resilience and long- term sustain-
ability. Water cover in the park showed significant fluctua-
tions over the 80- year period. From 2003 to 2013, water cover
declined sharply by 58.7%, suggesting potential impacts of
droughts, climate variability, or anthropogenic pressures such
as water extraction. However, between 2013 and 2023, water
cover rebounded with a 50.6% increase, indicating possible im-
provements in water management or hydrological conditions.
The subsequent 8.0% decline from 2023 to predicted 2083 may
signal renewed challenges, such as reduced rainfall or in-
creased evaporation rates due to climate change. These fluc-
tuations underscore the vulnerability of aquatic ecosystems in
the park and their dependence on consistent hydrological cycles
(Johnson and Host 2010). Declining water availability could
stress aquatic and riparian species, reducing biodiversity and
the overall health of the ecosystem (Crook etal.2015). Forested
areas displayed steady and significant growth throughout the
study period, increasing by 0.396% between 2003 and 2013,
10.280% between 2013 and 2023, and 10.676% between 2023
and 2083. This consistent growth reflects the park's ability to
regenerate forested areas, likely due to effective conservation
policies, reduced human disturbances, and favourable climatic
conditions. Forests are critical for ecological stability, as they
act as carbon sinks, regulate local climates, and provide habitat
for a wide range of species. The expansion of forest cover sug-
gests a strong regenerative capacity and resilience of the park's
ecosystem, enhancing its role in mitigating climate change and
supporting biodiversity. However, increased forest cover also
raises questions about the balance between forest and other
land covers, particularly shrubland, which might be compet-
ing for space. Shrubland exhibited the most dynamic changes,
with a substantial 20.813% increase from 2003 to 2013, a mod-
erate 4.154% rise from 2013 to 2023, and a remarkable 24.967%
expansion from 2023 to predicted 2083. This growth suggests
that shrubland is highly adaptable to changing environmen-
tal conditions and may benefit from reduced disturbances
and improved management practices. Shrublands often act as
transitional ecosystems, playing a vital role in soil stabilisa-
tion and supporting a diverse range of fauna. The expansion of
shrubland could indicate a natural succession process, where
degraded areas are gradually restored, further reinforcing the
park's regenerative potential (Mukomberanwa et al. 2024a;
Mukomberanwa et al. 2024b, 2024c). Conversely, bareland
decreased consistently throughout the period, shrinking by
20.622% from 2003 to 2013, 14.941% from 2013 to 2023, and
a significant 35.56% from 2023 to 2083. This reduction high-
lights successful restoration efforts, including reforestation,
soil conservation, and the natural regrowth of vegetation. The
decline in bareland indicates an overall improvement in soil
health and reduced erosion, which are critical for sustaining
the park's productivity and biodiversity. However, the complete
loss of bareland could reduce habitat heterogeneity, potentially
impacting species that rely on open habitats.
The neural network learning cur ve from cellular automata mod-
elling of land cover change in Mana Pools National Park shows
the Root Mean Square d Error (RMSE) for trai ning and validation
over 1000 iterations. Initially, RMSE values are high, indicating
poor model performance, but they decrease over iterations as the
model learns. The training RMSE is consistently lower than the
validation RMSE, suggesting possible overfitting. Fluctuations
13 of 17
in validation RMSE reflect variations in data complexity. By the
final iterations, RMSE stabilises, indicating model convergence.
While the model effectively predicts land cover changes, further
optimisation could address overfitting and improve its generali-
sation to unseen data.
The regenerative potential of the MPNP landscape demonstrates
strong resilience over time, as indicated by land cover changes
between 2003 and 2083. While water bodies fluctuated, forest
cover expanded steadily, suggesting successful conservation
measures and favourable ecological conditions. Shrubland, the
dominant land cover, exhibited significant growth, reinforcing
its role as a crucial ecological buffer. The steady decline in bare
land further supports the park's regenerative capacity, highlight-
ing ongoing land restoration efforts. The long- term predictive
modelling of habitat transformation suggests that Mana Pools
maintains the potential for sustaining biodiversity and ecologi-
cal functions. These findings have broader implications beyond
the local scale, offering ecologists valuable insights into habitat
connectivity modelling, landscape regeneration, and conserva-
tion planning across protected savannah ecosystems.
Generally, the integration of spatial absorbing Markov chain
(SAMC) modelling for landscape connectivity and land use/
land cover (LULC) predictions for regenerative power provides
a comprehensive understanding of Mana Pools National Park's
ecological dynamics. SAMC effectively maps elephant disper-
sal corridors and identifies movement constraints, while LULC
projections reveal long- term habitat transformations. Their com-
plementarity lies in linking connectivity patterns with habitat
sustainability, ensuring conservation interventions target both
movement pathways and regenerative zones. However, SAMC
assumes fixed resistance values, potentially oversimplifying
dynamic environmental changes, while LULC models may not
fully capture short- term disruptions in connectivity caused by
anthropogenic influences, requiring continuous data ref inement.
Acknowledgements
The authors acknowledge Zimbabwe Parks and Wildlife Management
Authority (ZPWMA) for granting us permission to conduct this re-
search. Data use for this research was approved under Permit Number:
23(1) (C) (II) 33/202 4 issued by the ZPWMA. T he authors would like to
acknowledge the role of anonymous reviewers who helped to enhance
the quality of the work.
Ethics Statement
Handling of African savannah elephants for GPS collaring and data
collection for this research was done by the research team, the Africa
Wildlife Tracking (AWT) in compliance with the Zimbabwe Parks
and Wildlife Management Authority (ZPWMA) and CITES ethical
requirements.
Conflicts of Interest
The authors declare no conflicts of interest.
Data Availability Statement
The data that support the findings of this study are available on request
from the corresponding author. The data are not publicly available due
to privacy or ethical restrictions.
References
Aleper, D., and S. R. Moe. 2006. “The African Savannah Elephant
Population in Kidepo Valley National Park, Uganda: Changes in Size
and Structure From 1967 to 2000.” African Journal of Ecolog y 44, no.
2: 157–164.
Battisti, C. 2024. “Ecological Networks as Planning Tools for African
Fragmented Landscapes: Overcoming Weaknesses for an Effective
Connectivity Conservation.” African Journal of Ecology 62, no. 1:
e13186.
Brokaw, N. V. 1985. “Gap- Phase Regeneration in a Tropical Forest.”
Ecology 66, no. 3: 682–687.
Bussmann, R. W. 2001. “Succession and Regeneration Patterns of East
African Mountain Forests. A Review.” Systematics and Geography of
Plants 71, no. 2: 959–974. https:// doi. org/ 10. 2307/ 3668731.
Carroll, K. A., A. J. Hansen, R. M. Inman, R. L. Lawrence, and A. B.
Hoegh. 2020. “Testing Landscape Resistance Layers and Modeling
Connectivity for Wolverines in the Western United States.” Global
Ecology and Conservation 23: e01125.
Chaiyar at, R., M. Wettasin, N. Youngpo y, a nd N. Cheachean. 2022 . “Use
of Human Dominated Landscape as Connectivity Corridors Among
Fragmented Habitats for Wild Asian Elephants (Elephas Maximus) in
the Eastern Part of Thailand.” Diversity 15, no. 1: 6.
Chibeya, D., H. Wood, S. Cousins, K. Carter, M. A. Nyirenda, and
H. Maseka. 2021. “How Do African Elephants Utilize the Landscape
During Wet Season? A Habitat Connectivity Analysis for Sioma
Ngwezi Landscape in Zambia.” Ecology and Evolution 11, no. 21:
14916–14931.
Clegg, B. W., and T. G. O'Connor. 2017. “Determinants of Seasonal
Changes in Availability of Food Patches for Elephants (Loxodonta afri-
cana) in a Semi- Arid African Savanna.” PeerJ 5: e3453.
Correa Ayram, C. A., M. E. Mendoza, A. Etter, and D. R. P. Salicrup.
2016. “Habitat Connectivity in Biodiversity Conservation: A Review of
Recent Studies and Applications.” Progress in Physical Geography 40,
no. 1: 7–37.
Crook, D. A., W. H. Lowe, F. W. Allendorf, et al. 2015. “Human
Effects on Ecological Connectivity in Aquatic Ecosystems: Integrating
Scientific Approaches to Support Management and Mitigation.” Science
of the Total Environment 534: 52–64.
Cushman, S. A., K. S. McKELVEY, and M. K. Schwartz. 2009. “Use
of Empirically Derived Source- Destination Models to Map Regional
Conservation Corridors.” Conservation Biology 23, no. 2: 368–376.
Cushman, S. A., M. Chase, and C. Griffin. 2010. “Mapping Landscape
Resistance to Identify Corridors and Barriers for Elephant Movement
in Southern Africa.” In Spatial Complexity, Informatics, and Wildlife
Conservation, 349–367. Springer.
Dunham, K. M., and A. J. du Toit. 2013. “Using Citizen- Based Survey
Data to Determine Densities of Large Mammals: A Case Study From
Mana Pools National Park, Zimbabwe.” African Journal of Ecology 51,
no. 3: 431–440.
Elliot, N. B., S. A. Cushman, D. W. Macdonald, and A. J. Loveridge.
2014. “The Devil Is in the Dispersers: Predictions of Landscape
Connectivity Change With Demography.” Journal of Applied Ecology
51, no. 5: 1169–1178.
Fletcher, R., and M. Fortin. 2018. Spatial Ecology and Conservation
Modeling. Springer.
Fletcher, R. J., Jr., R. K. Didham, C. Banks- Leite, etal. 2018. “Is Habitat
Fragmentation Good for Biodiversity?” Biological Conservation 226:
9–15.
Fletcher, R. J., Jr., T. J. Hefley, E. P. Robertson, B. Zuckerberg, R. A.
McCleery, and R. M. Dorazio. 2019. “A Practical Guide for Combining
Data to Model Species Distributions.” Ecology 100, no. 6: e02710.
14 of 17 African Journal of Ecology, 2025
Gara, T. W., T. Wang, A. K. Skidmore, S. M. Ngene, T. Dube, and M.
Sibanda. 2017a. “Elephants Move Faster in Small Fragments of Low
Productivity in Amboseli Ecosystems: Kenya.” Geocarto International
32, no. 11: 1243–1253.
Gara, T. W., T. Wang, A. K. Skidmore, etal. 2017b. “Understanding the
Effect of Landscape Fragmentation and Vegetation Productivity on
Elephant Habitat Utilization in Amboseli Ecosystem, Kenya.” African
Journal of Ecology 55, no. 3: 259–269.
Gara, T. W., T. Wang, T. Dube, S. M. Ngene, and K. S. Mpakairi. 2021.
“African Elephant (Loxodonta africana) Select Less Fragmented
Landscapes to Connect Core Habitats in Human- Dominated
Landscapes.” African Journal of Ecology 59, no. 2: 370–377.
Hardwick , K., J. Healey, S. Elliot t, N. Garwo od, and V. Anusarnsunthor n.
1997. “Understanding and Assisting Natural Regeneration Processes in
Degraded Seasonal Evergreen Forests in Northern Thailand.” Forest
Ecology and Management 99, no. 1–2: 203–214.
Hema, E. M., R. F. Barnes, M. di Vittorio, L. Luiselli, and W. Guenda.
2017. “Selective Disturbance by Elephants (Loxodonta africana) on
Eight Tree Species in a West African Savannah.” Ecological Research
32: 205–214.
Huang, C., X. Li, L. Khanal, and X. Jiang. 2019. “Habitat Suitability
and Connectivity Inform a Co- Management Policy of Protected Area
Network for Asian Elephants in China.” PeerJ 7: e6791.
Johnson, L. B., and G. E. Host. 2010. “Recent Developments in
Landscape Approaches for the Study of Aquatic Ecosystems.” Journal of
the North American Benthological Societ y 29, no. 1: 41–66.
Kaszta, Ż., S. A. Cushman, and R. Slotow. 2021. “Temporal Non-
stationarity of Path- Selection Movement Models and Connectivity: An
Example of African Elephants in Kruger National Park.” Frontiers in
Ecology and Evolution 9: 553263.
Kautz, T., C. López- Fando, and F. Ellmer. 2006. “Abundance and
Biodiversity of Soil Microarthropods as Inf luenced by Different Types
of Organic Manure in a Long- Term Field Experiment in Central Spain.”
Applied Soil Ecology 33, no. 3: 278–285.
Larkin, I. A., M. I. Stockman, M. Achermann, and V. I. Klimov. 2004.
“Dipolar Emitters at Nanoscale Proximity of Metal Surfaces: Giant
Enhancement of Relaxation in Microscopic Theory.” Physical Review B
69, no. 12: 121403.
Low, S. M., and S. E . Merry. 2010. “Engage d Anthropology: D iversity and
Dilemmas: A n Introduction to Supplement 2.” Current Anthropology 51,
no. S2: S203–S226.
Marx, A. J., C. Wang, J. A. Sefair, M. A. Acevedo, and R. J. Fletcher
Jr. 2020. “Samc: An R Package for Connectivity Modeling With Spatial
Absorbing Markov Chains.” Ecography 43, no. 4: 518–527.
Mashapa, C., E. Gandiwa, M. Never, and P. Mhuriro- Mashapa. 2021.
“Land Use and Land Cover Changes in a Human- Wildlife Mediated
Landscape of Save Valley Conser vancy, South- Eastern Low veld of
Zimbabwe.” Journal of Animal and Plant Sciences 31: 583–595.
Matsa, M., C. Shuche, T. Musasa, and R. Defe. 2022. “Surface Water
Distribution Challenges and Elephant Impacts on Woody Species in Mana
Pools National Park, Zimbabwe.” Tropical Ecology 63, no. 4: 604–614.
Mills, E. C., J. R. Poulsen, J. M. Fay, et al. 2018. “Forest Elephant
Movement and Habitat Use in a Tropical Forest- Grassland Mosaic in
Gab on.” PLoS One 13, no. 7: e0199387.
Moe, S. R., L . P. Rutina, H. Hytteborn, and J. T. du Toit. 2009. “What
Controls Woodland Regeneration After Elephants Have Killed the Big
Trees?” Journal of Applied Ecology 46, no. 1: 223–230.
Mukomberanwa, N. T., P. Taru, and B. Utete. 2024a. “Space Use by
African Savannah Elephants in an Arid Protected A rea Dominated by
a Mosaic of Different Vegetation.” African Journal of Ecology 62, no. 1:
e13240.
Mukomberanwa, N. T., P. Taru, B. Utete, and H. K. Madamombe.
2024b. “Predicting the Dry Season Habitat Occupancy of African
Savann ah Elephant Using Vegetation Indices a nd Modelling Lands cape
Variability in a Mesic Protected Area.” African Journal of Ecolog y 62,
no. 3: e13318.
Mukomberanwa, N. T., P. Taru, B. Utete, and P. Ngorima. 2024c.
“Modelling Tree Allometries to Understand the Impact of African
Savannah Elephant Herbivory Dynamics on the Vegetation Structure
and Tree Cover Change in a Protected Area.” African Journal of Ec ology
62, no. 4: e13336.
Naidoo, R., P. Beytell, A. Brennan, etal. 202 4. “Landscape Connectivity
for Afr ican Elephants in the world's Largest Transfrontier Conser vation
Area: A Collaborative, Multi- Scalar Assessment.” Journal of Applied
Ecology 61, no. 10: 2483–2496.
Ndoro, O., C. Mashapa, S. Kativu, and E. Gandiwa. 2016. “Impact
of African Elephant on Baobab Along a Surface Water Availability
Gradient in Mana Pools National Park, Zimbabwe.” Tropical Ecolog y 57,
no. 2: 333–341.
Omeja, P. A., M. J. Lawes, A. Corriveau, etal. 2016. “Recover y of Tree
and Mammal Communities During Large- Scale Forest Regeneration in
Kibale National Park, Uganda.” Biotropica 48, no. 6: 770–779.
Osipova, L., M. Okello, S. Njumbi, et al. 2019. “Using Step- Selection
Functions to Model Landscape Connectivity for African Elephants:
Accounting for Variability Across Individuals and Seasons.” Animal
Conservation 22, no. 1: 35–48.
Pittiglio, C., A. K. Skidmore, H. A. van Gils, and H. H. Prins. 2012.
“Identifying Transit Corridors for Elephant Using a Long Time- Series.”
International Journal of Applied Earth Observation and Geoinformation
14, no. 1: 61–72.
Puyravaud, J. P., S. Cushman, P. Davidar, and D. Madappa. 2017.
“Predicting Landscape Connectivity for the A sian Elephant in Its Largest
Remaining Subpopulation.” Animal Conservation 20, no. 3: 225–234.
Roever, C. L., R. J. van Aarde, and K. Leggett. 2013. “Functional
Connectivity Within Conservation Networks: Delineating Corridors for
African Elephants.” Biological Conservation 157: 128–135.
Rood, E., A. A. Ganie, and V. Nijman. 2010. “Using Presence- Only
Modelling to Predict Asian Elephant Habitat Use in a Tropical Forest
Landscape: Implications for Conservation.” Diversity and Distributions
16, no. 6: 975–984.
Sheil, D., and A. Salim. 2004. “Forest Tree Persistence, Elephants, and
Stem Scars.” Biotropica 36, no. 4: 505–521.
Skarpe, C., P. A. Aarrestad, H. P. Andreassen, etal. 2004. “The Return
of the Giants: Ecological Effects of an Increasing Elephant Population.”
Ambio: A Journal of the Human Environment 33, no. 6: 276–282.
Struhsaker, T. T., J. S. Lwanga, and J. M. Kasenene. 1996. “Elephants,
Selective Logging and Forest Regeneration in the Kibale Forest,
Uga nda.” Journal of Tropical Ecology 12, no. 1: 45–64.
Suksavate, W., P. Duengkae, and A. Chaiyes. 2019. “Quantifying
Landscape Connectivity for Wild Asian Elephant Populations Among
Fragmented Habitats in T hailand.” Global Ecology and Conservation 19:
e00685.
Tiller, L., T. Humle, R. Amin, etal. 2024. “Elephant Pathway Use in a
Human- Dominated Landscape.” Wildlife Biolog y 2025: e01204.
Tripathy, B. R., X. Liu, M. Songer, B. Zahoor, W. Wickramasinghe, and
K. K. Mahanta. 2021. “Analysis of Landscape Connectivity Among the
Habitats of As ian Elephants in Keonjhar Forest Di vision, India.” Remote
Sensing 13, no. 22: 4661.
Turner, J., N. Owen- Smith, L. Kruger, C. Wigley- Coetsee, and T. G.
O'Connor. 2022. “Shifting Patch Dynamics of Floodplain Woodlands
in an Arid Savanna Region: Effects of Soil Form, Flooding and Large
Herbivores.” South African Journal of Botany 151: 866–875.
15 of 17
Vieira, D. L., and A. Scariot. 20 06. “Principles of Natural Regeneration of
Tropical Dr y Forests for Restoration.” Restoration Ecology 14, no. 1: 11–20.
Zacarias, D., and R. Loyola. 2018. “Distribution Modelling and Multi-
Scale Landscape Connectivity Highlight Important Areas for the
Conservation of Savannah Elephants.” Biological Conservation 224: 1–8.
Zeller, K. A., K. McGarigal, and A. R. Whiteley. 2012. “Estimating
Landscape Resistance to Movement: A Review.” Landscape Ecolog y 27:
777–797.
Appendix
FIGUR E A | Centroids.
FIGUR E A | Resistance map based on resource selection functions
created by telemetered African savannah elephants (including w ith re-
spect to major roads in MPNP). Coordinates are in Africa Albers Equal
Area Conic (ESRI 102022).
16 of 17 African Journal of Ecology, 2025
FIGUR E A | The left map (bina ry connectivit y) s hows potential movement paths, with pu rple indicating high connectivity and yellow hig hlight-
ing low connectivity. The right map represents predicted movement distances, ref lecting the expected dispersal capability in the landscape.
FIGUR E A | Land cover maps derived from Landsat images to show vegetation regeneration over time.
17 of 17
FIGUR E A | Neural network learning cur ve.
