European Journal of Experimental Biology, 2015, 5(11):12-19
Prediction of suitable habitats for Syzygium caryophyllatum, an
Keywords: Modelling algorithms, S. caryophyllatum, AUC, Bioclim, MaxEnt, GARP.
Species distribution models (SDMs) or ecological niche models (ENM) that use environmental factors based on
historical collections are increasingly being used to analyze species distributions and also to predict the presence or
absence of species in unrecorded areas [1-3]. These models establish relationships between occurrences of species
and biophysical and environmental conditions in the study area. SDMs have been used to predict potentially suitable
areas for the conservation of endangered and rare species [4-8] for the identification of suitable sites for
reintroduction or restoration [9,10] and for assessing potential effects of future climate change on species
distributions as well as on local species diversity [11,12]. It is also used to enable the analysis of the impacts of
climate change on species, it is essential to quantify the relative importance of climate relative to other descriptors of
the environment [13,14].
The ecological niche models have also been used in a wide range of applications such as in locating rare and
to global climate change . Recent works in this field deals with methodological challenges specific to best
ENM-based predictions of suitable areas and identification of conservation priorities[19,20].
The Western Ghats comprises the major portion of the Western Ghats and Sri Lanka Hotspot, one of 34 global
in biodiversity. Although the total area is less than 6 percent of the land area of India, the Western Ghats contains
more than 30 percent of fauna and flora found in India. Like other hotspots, the Western Ghats has a high proportion
of endemic species. The Western Ghats contains numerous medicinal plants and important genetic resources such as
the wild relatives of grains, fruits and spices. In addition to rich biodiversity, the Western Ghats is home to diverse
social, religious, and linguistic groups. The high cultural diversity of rituals, customs, and lifestyles has led to the
establishment of several religious institutions that strongly influence public opinion and the political decision-
making process. Conservation challenges lie in engaging these heterogeneous social groups and involving them in
community efforts aimed at biodiversity conservation and consolidation of fragmented habitats in the hotspot.
Syzygium caryophyllatum (L.) Alston., commonly known as Wild black plum, a medium sized tropical evergreen
tree belongs to the family Myrtaceae. The vernacular name of this plant is Jangli jamun in Hindi, Kattunjara,
Kanipazham or Jnarapazham in Malayalam, Kunta nerale in Kannada. S.caryophyllatum is native to India and Sri
Lanka; in India the distribution mainly occur in the forests of Western Ghats. Tree grows along margin
of evergreen forests or in open formations from low to higher elevations. Fruits are edible, sweet and astringent in
taste and they are useful in stomatitis and intestinal disorder. The seeds and bark were dried and its decoction was
used for the treatment of Diabetes mellitus  . The leaf and bark extracts of this plant are well known for its
antibacterial and antioxidant efficacy  . Tribal peoples were considering this plant as a boon of nature and its
fruits and seeds were consumed by Paniya tribal community of Waynad district, Kerala .
Based on the previous reports the extended distribution of this plant species is reported in Kalakad Mundanthurai
South India  and Annamalai hills that form of the Western Ghats-Sri Lanka biodiversity hotspots  . Based on
the threat perception due to its habitat loss and human activities natural populations of this species are on the decline
mode. Due to these pressures this species has been listed under the endangered category of IUCN Red List. The
ecological conditions necessary for the survival of this species can greatly help in conservation scenario. The aim of
this study was to predict suitable habitat distribution for threatened tree species Syzygium caryophyllatum using
known presence observations with three different modelling algorithms (BIOCLIM, GARP and MaxEnt). To
identify the environmental factors associated with S.caryophyllatum habitat distribution and to predict suitable
habitat for reintroduction and future conservation of this species.
MATERIALS AND METHODS
Species occurrence data for Ecological niche modelling
The occurrence points of S. caryophyllatum were identified based on the field surveys in the Western Ghats region
of Tamilnadu, Kerala, Maharastra and Karnataka states of India and also from the secondary data collected from the
literature survey. Thirty two occurrence points of S. caryophyllatum were used in the present study. The primary
presence only data was used for modelling the distribution of this endangered species.
The environmental variables were nineteen bioclimatic variables used for all the three modelling algorithms (Table
1). These bioclimatic variables were derived from the monthly temperature and rainfall values in order to generate
more biologically significant variables. The bioclimatic variables represent, annual trends (e.g. mean annual
temperature, annual precipitation), seasonality (e.g. annual range in temperature and precipitation) and extreme or
limiting environmental factors (e.g. temperature of the coldest and warmest month, precipitation of wet and dry
quarters). These variables were obtained from globally interpolated datasets (source: http://www.worldclim.org)
which are presumed to be relevant to plant existence [12, 26, 27]. Analyses were conducted at the 1 x 1 km pixels
spatial resolution of the environmental data sets.
Three different modelling algorithms were used in the present study following the Open Modeller Desktop version
1.1.0 (downloaded from http://openmodeller.sourceforge.net) software. BIOCLIM, GARP (GARP with the best
subsets – new open modeller implementation) and MaxEnt algorithms were run using the above software [28-31].
BIOCLIM is one of the earlier modelling techniques, based on climatic envelop theory. For each given
environmental variable the algorithm finds the mean and standard deviation (assuming normal distribution)
associated with the occurrence points. Each variable has its own envelope represented by the interval [M - CO*
StDev, M + C* StDev], where 'M' is the mean; 'CO' is the cut off input parameter; and ’StDev’ is the standard
deviation. Besides the envelope, each environmental variable has additional upper and lower limits taken from the
maximum and minimum values related to the set of occurrence points .
Table 1 Bioclimatic variables used in the model development
Maximum temperature of warmest month (º C)
Minimum temperature of coldest month(º C)
GARP (genetic algorithm for rule set prediction) is an ecological niche modelling method based on a genetic
algorithm. This modelling approach predicts the suitable environmental conditions under which the species should
be able to maintain populations. For input, GARP uses a set of point localities where the species is known to occur
and a set of geographical layers that might limit the specie’s capabilities to survive. This model is a random set of
mathematical rules which can be read as limiting environmental conditions .
The maximum entropy (MaxEnt) approach estimates a target probability distribution of the species by finding the
set of environmental variables). Default values of different parameters, maximum iterations = 500, convergence
threshold = 0.00001 and 50% of data points were used as a random test percentage in the present study [29,30].
A receiver operating characteristics (ROC) plot was generated by incorporating the sensitivity values, the true
positive fraction against the false positive fraction for all available probability thresholds to measure prediction
accuracy of the models output [32-34]. The sensitivity values were calculated using confusion matrix. A curve
which maximizes sensitivity against low false positive fraction values is considered as good model which was
evaluated by using the area under the curve (AUC). Cross-validated AUC values were summarized to present
overall model performance by taking mean AUC values of all model accuracies. The range of AUC is from 0.0
to1.0. A model providing excellent prediction has an AUC higher than 0.9, a fair model has an AUC between 0.7
and 0.9 and a model with AUC below 0.7 is considered poor (Swets 1988). The Jackknife procedure was used to
assess the importance and percentage of contribution of bioclimatic variables. The final potential species distribution
maps had a range of values from 0 to 1 which were regrouped into three classes of potential habitats viz., ‘high
potential’ (>0.6), ‘medium potential’ (0.2–0.4) and ‘low potential’ (<0.2).
RESULTS AND DISCUSSION
The prediction of suitable habitats for conservation of the endangered tree species S. caryophyllatum was
with the modelling techniques used in the present study (Figure 1 a, b & c). SDMs outputs revealed that, MaxEnt
predicted largest area (75.95%) under potential distribution compared to BIOCLIM (1.73%) and GARP (5.26%).
Both GARP and MaxEnt showed a wide range of distribution from low to high probability area in India whereas,
BIOCLIM output is restricted to Western ghats regions of India. The AUC values for the current potential
distribution of S. caryophyllatum were high indicating good predictive model performance. BIOCLIM and GARP
models showed a good performance with AUC values of 0.99 and 0.97 compared with MaxEnt AUC value of 0.91
(Table 2 ).
Figure 1 Predicted suitable habitats for S. caryophyllatum a) BioClim b) GARP c) MaxEnt
Table 2 Comparative account of various model outputs
The jacknife test of variable importance in MaxEnt has identified the precipitation of coldest quarter (bioclimatic
variable 19) as the most important environmental variable contributed to the model development (Figure 2). Other
variables like Isothermality (bioclimatic variable 3), Temperature seasonality (bioclimatic variable 4), Temperature
annual range (bioclimatic variable 7), Annual precipitation (bioclimatic variable 12), Precipitation of driest quarter
(bioclimatic variable 17), Annual mean temperature (bioclimatic variable 1) and Mean diurnal temperature range
(bioclimatic variable 2) also have considerable predictive values with regard to distribution of S.caryophyllatum
(Figure. 3). Considering the permutation importance, only 9 variables out of 19 contributed to the model output.
followed by annual precipitation (bio 12) (3.5%) and precipitation of driest quarter (bio 17) (1.8 %). All three
variables together contributed to 99.3% of the variation (Table. 3). Potential distribution maps show various
possibilities for conservation and management of this endangered tree species. Both MaxEnt and GARP have shown
a wider range of distribution from medium to high probability in South-west and north-east states of India and also
showed high probability in some areas of Sri Lanka, it is an another native region of S. caryophyllatum whereas,
Bioclim distribution is restricted as high probability areas in and around the occurrence points of its existing natural
The major role of SDM is to estimate the probability of occurrence of a given species based on observed presence
and (or absence locations) as well as environmental and climatic covariates. A common application of this method is
to predict species ranges with climate data as predictors. Several studies were done successfully and predict suitable
distribution habitats for many threatened plant species using different modelling algorithms like Artemisia sieberi
and Artemisia aucheri, Justicia adhatoda ,Coscinium fenestratum, Tapirus pinchaque and Monotropa uniflora [35-
39]. Ray et al. (2011)  reported the predictive distribution modelling of a rare Himalayan medicinal plant
of suitable habitat for conservation of Nepeta septemcrenata  and an endangered tree Canacomyricca moniticola
 using MaxEnt algorithm with low omission error was predicted.
Distribution data on threatened species often have few records and are geographically close together, making it
data provide limited information for determining the relationships between the species and their environments .
Maximum entropy (Maxent) models present good results even for small sample size . Maxent is a multivariate
approach to study the geographic distribution of species on a large scale using only presence data of the species .
Figure 3 Response curves of the variables that most contributed to explain the potential distribution of S.caryophyllatum (Isothermality
The presence points of the species were considered as appropriate places for their occurrence. The essential data
habitat of distribution pattern of threatened and endangered plant species. Maxent has several strong characteristics,
it needs only species occurrence data and environmental factors; it can examine factor importance by way of a
jackknife procedure and also facilitates model interpretation [30,43].
Geographical data of threatened species provide the degree of species threats, their distributions, and habitat
threatened species is the one with narrow habitat range, low climate tolerance; specialised adaptation requiring an
outside agency for pollination, poor dispersal strategies, few seeds per fruit and poor viability of seeds . The
conservation status of a species can best be developed by synthesising information on each of its known populations,
viewed together with any information on changes in historical range and evidence of vulnerability of its
characteristic habitat. Mapping by such intrinsic features of the land as natural regions and physiographic areas is
the best way of presenting plant distributions data. Species and habitat relationship modelling with precise locality
data on microclimate, topography and soil in association with site-specific location data of concerned taxa helps in
understanding the interrelationships and controls of biotic and abiotic factors on species distribution pattern .
Natural populations of S. caryophyllatum have always been small, but our modelling distribution prediction results
showed potential habitat greater than the area of the actual distribution. These results give an insight into the
availability of areas suitable for the species’ regeneration, possibly through ex vitro conservation planning.
Our results of potential habitat distribution maps for S. caryophyllatum may help to discover new populations,
identify top-priority study sites or set priorities to restore its natural habitat for more effective conservation.
Moreover the effective conservation planning is necessary for this tree species for its further existence in the natural
First author thank University Grant Commission for the award of meritorious fellowship. Corresponding author
thank partial financial support through UGC for the funding of UGC-CAS, DST-IPLS and DST-Purse programme.
 Araujo MB, Pearson RG, Thuiller W, Erhard M, Glob Change Biol,2005,11, 1504.
 Guisan A, Hofer U, J Biogeogr, 2003, 30, 1233.
 Gallagher RV, Hughe L, Leishman MR, Wilson PO, Biol Invasions, 2010, 12, 4049.
 Papes M, Gaubert P, Divers Distrib, 2007, 13, 890.
 Rebelo H, Jones G, J Appl Ecol, 2010, 47, 410.
 Solano E, Feria TP, Biodiv and Conser, 2007, 16, 188.
 Arun KD, Rameshprabu N, Swamy PS, Paper Proceedings of International Conference on Biodiversity (ICBD),
Colombo, Sri Lanka, 2013, pp 5.
 Klar N, Fernandez N, Kramer-Schadt S, Herrmann M, Trinzen M, Buttner I, Niemitz C, Biol Conser, 2008, 141,
 Hole DG, Willis SG, Pain DJ, Fishpool LD, Butchart SHM, Collingham YC, Rahbek C, Huntley B, Ecol Lett,
 Pearson RG, Dawson TP, Global Ecol Biogeogr, 2003, 12, 361.
 Morueta-Holme N, Flojgaard C, Svenning JC, PLos One, 2010, 5(4), 10.
 Newbold T, Progress in Physical Geography, 2010, 34(1), 3.
 Jackson CR, Robertson MP, J Nat Conserv, 2011, 19, 87.
 Peterson AT, Martínez-Campos C, Nakazawa Y, Martínez-Meyer E, Transactions of the Royal Society of
 Ganeshaiah KN, Barve N, Nath N, Chandrashekara K, Swamy M, Cur Sci, 2003, 85, 1526.
 Barve N, Bonilla AJ, Brandes J, Brown JC, Brunsell N, Rev Mex Biodivers, 2012, 83, 817.
 Loiselle BA, Howell CA, Graham CH, Goerck JM, Brooks T, Smith KG, Williams PH, Conserv Biol, 2003, 17,
 Peterson AT, Kluza DA, Anim Conserv, 2003, 6, 47.
 Shilpa KJ, Krishnakumar G, Int J of Pharm Pharm Sci, 2012, 4, 198.
 Ratheesh NMK, Anilkumar N, Balakrishnan V, Sivadasan M, Ahmed Alfarhan H, Alatar AA, J Med Plants
Res, 2011, 5(15), 3520.
 Parthasarathy N, Biodiv and Conser, 2009, 8, 1365.
 Muthuramkumar S, Ayyappan N, Parthasarathy N, Mudappa D, Shankar Raman TR, Arthur Selwyn M, Arul
Pragasan L, Biotropica, 2006, 38(2), 143.
 Hijmans RJ, Cameron S, Parra J, Jones P, Jarvis A, Int J Climatol, 2005, 25, 1965.
 Irfan-Ullah M, Giriraj A, Murthy MSR, Peterson AT, Biodiv Conser, 2007, 16(6), 1917.
Nature Conservation: Cost Effective Biological Surveys and Data Analysis. Canberra: CSIRO, 1991, pp 64.
 Peterson AT, The Condor, 2001, 103(3), 599.
 Phillips SJ, Dudik M, Schapire RE, In proceedings of the 21
AMC Press, New York, 2004, pp 655.
 Stockwell DRB, Peters D, Int J Geogr Inf Sci, 1999, 13 (2), 143.
 Boubli JP, de Lima MG, Int J Primatol, 2009, 30.
 Lobo JM, Jime´nez-Valverde A, Real R, Global Ecol Biogeogr, 2008, 17, 145.
 VanDerWal J, Shoo LP, Graham C, Williams SE, Ecol Model, 2009, 220, 589.
 Hosseini SZ, Kappas M, ZareChahouki MA, Ecol Inform, 2013, 18, 61.
 Yang XQ, Kushwaha SPS, Saran S, Ecol Eng, 2013, 51, 83.
 Thriveni HN, Srikanth V, Gunaga HN, Babu R, Vasudeva R, Trop Ecol,2015, 56(1), 101.
 Ortega-Andrade HM, Prieto-Torres DA, Gómez-Lora I, Lizcano DJ, PLoS ONE, 2015, 10(3), e0121137.
 Pradhan P, Biodiversitas, 2015, 16( 2), 109.
 Ray R, Gururaja KV., Ramachandra TV, J Env Biol, 2011, 32(6), 725.
 Khafaga O, Hatab EE, Omar K, Academia Arena, 2011, 3(7), 45.
 Phillips SJ, Anderson RP, Schapired RE, Ecol Model, 2006, 190, 231.
 Scheldeman X, van Zonneveld M, Biodiversity International, 2010, 139.
 Nayar MP, Hot spots of endemic plants of India, Nepal and Bhutan, TBGRI, Trivandrum, 1996, pp 252.