Ecological Indicators 154 (2023) 110631

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

Species distribution models to predict the potential niche shift and priority
conservation areas for mangroves (Rhizophora apiculata, R. mucronata) in
response to climate and sea level fluctuations along coastal India
Pujarini Samal a, c, Jyoti Srivastava a, b, *, Bipin Charles d, S.R. Singarasubramanian c
a
Birbal Sahni Institute of Palaeosciences, 53 University Road, Lucknow 226 007, India
b
Academy of Scientific and Innovative Research (AcSIR), Ghaziabad 201002, India
c
Department of Earth Sciences, Annamalai University, Tamil Nadu 608002, India
d
Institute for Biodiversity Conservation and Training, 7thMain Road, Shankar Nagar, Bangalore, Karnataka 560096, India

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

Keywords: Mangroves are more than just a tree as they are home to thousands of species, carbon absorbers and a natural
Climate change coastal fortress against floods and storm surges. Hence, conservation planning and decision making for this
Species-specific restoration valuable ecosystem must involve identification of priority conservation areas (PCAs) at species-specific di­
Core distributional shift
mensions. In the present study, we adopted an ensemble modelling approach for the distribution of two
Ensemble model
R. mucronata
mangrove species (Rhizophora apiculata, Rhizophora mucronata) using high-resolution environmental and edaphic
R. apiculata datasets, to identify the PCAs for future conservation. We also identified the key environmental variables shaping
their distribution and precisely estimated the core distributional shift along the Indian coastline under changing
climate scenario. The findings revealed that about 5844 km2 and 7846 km2 areas were identified as extremely
suitable areas, which were distributed along Maharashtra coast and Kerala coast for R. mucronata and
R. apiculata, respectively, in the current climate scenario. The distribution of R. apiculata was found to be mostly
shaped by mean diurnal range and annual mean temperature, whereas the distribution of R. mucronata was
primarily shaped by annual mean temperature and altitude. The highest range expansion of mangrove species
occurred during middle Holocene due to high precipitation and sea-level rise and this finding is further supported
by fossil pollen evidence. The suitable habitat range for R. apiculata is predicted to increase along Kerala coast
under RCP2.6 scenario by 6.90% and 6.93% and under RCP8.5 scenario by 9.33% and 9.90% in the year 2050
and 2070, respectively, whereas the range for R. mucronata is getting reduced in the future climate scenario.
Overall, our predictions reveal a steady migration of conducive mangrove habitat towards land or higher ele­
vations due to relative sea level rise in future. These results would aid in planning a long-term species-specific
conservation and management strategy for mangrove ecosystem along coastal India.

1. Introduction typically store thrice as much carbon as comparable terrestrial forests

do. Yet, mangroves are today severely threatened by anthropogenic
Mangroves are a unique ecosystem comprising of the intertidal ma­ activity and climate change, resulting in a sharp decline in their range
rine plant community that predominates the margins of tropical coast­ and extensive degradation (Zhang et al., 2021; Samal et al., 2022). It is
lines all over the world. They offer a variety of essential services to projected that coastal vegetation will be significantly impacted by
human communities and the biosphere, including the flood control, climate change through direct biophysical regime impacts such as
filtration of water, preservation of wildlife habitat, reducing the effects drought, changing rainfall patterns and overheating, and also indirect
of climate change and preserving cultural values (Cao and Fox, 2009; effects such as wildfires, changes in land cover and sea level rise (Dwire
Zhang et al., 2021). They also make up 20– 30% of the global carbon et al., 2018; Middleton, 2012). Meanwhile, because of the agricultural
pool, making them essential for carbon sequestration and controlling activities, >50% coastal wetlands are getting reduced worldwide (Ver­
impacts of changing climatic condition (Kayranli et al., 2010). They hoeven and Setter, 2009). Especially in Asia, about 30–50% of coastal

* Corresponding author at: Birbal Sahni Institute of Palaeosciences, 53 University Road, Lucknow 226 007, India.
E-mail address: jyoti.srivastava@bsip.res.in (J. Srivastava).

https://doi.org/10.1016/j.ecolind.2023.110631
Received 20 April 2023; Received in revised form 2 July 2023; Accepted 6 July 2023
Available online 14 July 2023
1470-160X/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
P. Samal et al. Ecological Indicators 154 (2023) 110631

wetlands have degraded over the last three decades (Alongi, 2015). the association between species and its environment and also predicting
One of the most significant mangrove genera in the tropical region is species response to such environmental changes (Elith and Leathwick,
Rhizophora, commonly occurring in the tidal creeks. In terms of benefits 2009; Elith and Franklin, 2013; Sahana et al., 2022). The power of SDM
of this genera, it serves as coastal guard by protecting the coastal eco­ to forecast the potential geographical distribution of the species in
systems from tsunamis, storms and hurricanes (Alongi, 2008). It also future time periods facilitate numerous ecological applications, such as
prevents coastal erosion due to its roots’ ability to capture suspended predicting the highly suitable habitat under different climate change
solids, wherein the upland runoff is filtered, preserving sea grass beds scenarios (Pearson and Dawson, 2003; Hijmans and Graham, 2006;
and coral reefs from the damaging effects of the suspended solids Kaky et al., 2020), helping with conservation and management plans,
(Alongi et al., 2005; Alongi, 2008). Additionally, because of its extensive and also providing information about the ecological preferences,
prop root systems, aquatic organisms such as fish and prawns can paleobiology (Svenning et al., 2011) and biogeography (Guisan et al.,
reproduce and the juveniles can find shelter (Negelkerken et al., 2000). 2006). For the sake of forecasting distributions of the species based on
Apart from its ecological and economic benefits, it also has medicinal environmental factors, an extensive variety of SDM techniques have
attributes for its traditional usage to treat ulcers, hepatitis, haematoma, been designed (Thuiller, 2003). Frequently utilised algorithms are;
and elephantiasis (Ravikumar et al., 2005). maximum entropy modelling (MaxEnt) (Phillips et al., 2006), general­
These stilt mangroves occur widely throughout the Indo-west pacific ized additive models (GAM) and generalized linear models (Guisan
region along the tropical and subtropical intertidal wetlands. The target et al., 2002; Hastie and Tibshirani, 2004), boosted regression trees (brt)
species i.e. R. apiculata and R. mucronata are widely distributed along the (De’Ath, 2007), random forests (rf) (Svetnik et al., 2003) and other
West Bengal, Odisha, Andhra Pradesh, Tamil Nadu, Kerala, Karnataka machine learning algorithms such as Bayesian hierarchical modelling
and Gujarat coast. Precisely, these species are distributed in Bali Island (Latimer et al., 2006) or artificial neural networks (ANN) (Zurada,
of Sundarban Delta along West Bengal coast, Bhitarkanika, Kansaridia, 1992). Many of these techniques are simple to apply concomitantly in
Khola and Kharnasi region of Odisha coast, Ramannapalem, Chinnava­ statistical programmes and packages, and generally, the frequently used
lasala and Gadimoga area of Coringa mangrove region of Andhra Pra­ algorithms have the more user-friendly and simple interfaces (Norberg
desh coast and Pichavaram, Muthupet and Gulf of Mannar along Tamil et al., 2019). The sample size, spatial distribution, choice of predictor
Nadu coast on the east coast of India. Similarly, the current occurrence variables, and the SDM method, are possible causes of uncertainty in
of these species is widely distributed in Kannur and Ernakulam districts predictions, each of which can affect predictive power and accuracy of
of Kerala coast, Tarkarli, Achara and Kolamb area of Konkan region and the model (Buisson et al., 2010). To overcome these uncertainties, based
also along Gorai creek of Mumbai of Maharashtra coast along the west on evaluation metric e.g. AUC, ensemble model has been used in the
coast of India. However, along the Gujarat coast only R. mucronata is present study that integrate prediction of single SDM model outcomes
found along Gulf of Kachchh, Gulf of Khambhat, Kori creek and Purna into one average binary map (Thuiller et al., 2004).
estuary region. Although they appear to have continuous distributions, Despite its socio-economic value, mangrove ecosystems are at a risk
different species within the same genus have distinct preferred zonation of extinction, particularly in developing countries (Friess et al., 2020).
throughout coastal habitats, where they are dominant. Rhizophora Loss of suitable habitat of mangroves in countries like Malaysia and
mucronata grows in regions experiencing regular freshwater flows, Myanmar exceeds the global average, where proper conservation mea­
Rhizophora stylosa is mainly found in marine conditions and prefers sures should be planned (Friess et al., 2020). Several studies have been
exposed offshore areas, R. apiculata is encountered in embayments and conducted to map the potential distribution of mangroves in future
mid-lower sites of large riverine estuaries while the hybrid species R. × climatic scenarios globally as well as regionally (Dang et al., 2021;
lamarckii has been located from mid-high intertidal regions of estuaries Banerjee et al., 2021; Sahana et al., 2022; Samal et al., 2022). Banerjee
(Duke, 2006). These stilt mangroves flourish in subtropical and tropical et al. (2021) have suggested prioritizing sites for both true mangroves
climates with somewhat high and evenly distributed rainfall. Their and mangrove associates in the Indo-West Pacific region. However, the
distributional range, form and biomass are a reflection of their response species-specific response to the future climate change in a developing
to variations in the temperature and precipitation pattern through time country like India is still missing. Hence in the present study, we tried to
and space. As these mangroves are limited to elevations between the assess the niche distribution and environmental factors responsible for
highest tides and mean sea level, their communities will need to relocate shaping the distribution, for which we adopted ensemble modelling
inland during the time of sea level rise. As a result of the responses of approach to predict the habitat suitability of two Rhizophora species
these mangrove habitats and their distributional shift, these crucial (R. apiculata and R. mucronata) in Indian subcontinent using a wide­
climate change markers may be recognized and mapped. Accurate un­ spread collection of occurrence records. The key objectives of the pre­
derstanding of these shifts and their root causes enables more accurate sent study are: (1) to identify the significant bioclimatic and edaphic
forecasting of upcoming changes. factors influencing the niche range shift of Rhizophora species; (2) to
The IPCC’s Fifth Assessment Report (AR5) predicts that global evaluate the core distributional shift of these species in past (Mid-Ho­
warming will continue, with an average increase in temperature of 0.3 locene), present and future (2050, 2070 under RCP 2.6 and RCP 8.5 i.e.
to 4.5 ◦ C by 2100 compared to 1986 to 2005. Later, the Intergovern­ the two greenhouse gas emission scenarios and (3) to measure the
mental Panel on Climate Change (IPCC)’s Working Group One (WG1) spatial extent of the habitat suitability for both Rhizophora species in
issued the Sixth Assessment Report (AR6), which states that the rise in future climatic condition so that it will help to plan a more precise
global surface temperature during the period of 1850–1900 was 1.09 ◦ C species specific conservation strategy in different sectors of the Indian
(0.95◦ –1.20 ◦ C) for the decade 2011–2022 (Arias et al., 2021). The coastline.
report also states that, the rise is expected to increase by 1.0 ◦ C- 1.8 ◦ C in
the lowest CO2 emission scenario and 3.3 ◦ C- 5.7 ◦ C in the highest CO2 2. Materials and methods
emission scenario for the period of 2081–2100 in comparison to
1850–1900. Climate change remarkably affect the species diversity and The methods employed in this study include:
their distributional changes will consequently show how change in cli­
matic condition has impacted the landscape (Lawler et al., 2009; Zhang 1. Compiling species occurrence record and the environmental
et al., 2021). Hence, species-specific knowledge of responses of datasets.
mangrove to climate variability and sea level changes would aid in 2. Predicting the habitat suitability of two Rhizophora species i.e.
formulating more accurate management and conservation strategies in R. apiculata and R. mucronata based on ensemble modelling by using
the high priority suitable habitats. sdm package available in R software.
Species distribution models (SDMs) are the key tool to understand

2
P. Samal et al. Ecological Indicators 154 (2023) 110631

3. Identifying the pattern of the centroid shift of the suitable habitat From the general circulation model (GCM), coupled model inter-
area using SDM toolbox version 2.5. comparison project (CMIP5) were downscaled and verified using the
baseline climate from Worldclim 1.4, which has been often used in
A detailed workflow of the analyses used in the present study has earlier research, for the projection of past, current and future climate
been outlined in (Fig. 1). (Hu et al., 2020; Banerjee et al., 2022; Samal et al., 2022). Further, from
the GCMs, Hadley Global Environment Model 2-Earth System
(HadGEM2-ES) was used to model the future climate scenarios. Two
2.1. Species occurrence record
representative concentration paths (RCPs), i.e., RCP2.6 (lowest) and
RCP8.5 (highest) greenhouse gas concentrations reported by IPCC
Species occurrence points are the geographic points (latitude and
Assessment Report 5 (AR5), were used for past (~ 6000 years ago), and
longitude) where a species has occurred. In an ensemble model, these
future (for the year 2050 and 2070) prediction. Following the assess­
records provide crucial information for predicting habitat suitability.
ments of SDM projection under past and future climatic condition, the 7
The geo-referenced occurrence points for R. mucronata and R. apiculata
soil parameters (pH, OCD, SOCS, BD, CC, CEC and TN) remained un­
have been compiled after a thorough literature survey conducted using
changed. The overall spatial resolution of these environmental factors
databases i.e. Global Biodiversity Information Facility (GBIF, 2023)
was compiled to 30 arc-seconds spatial resolution (c. 1 km2 area).
database (10.15468/dl.aq8r9con January 16, 2023) and published lit­
It is advised to eliminate the highly correlated variables before
eratures. Globally 6298 and 2478 occurrence points have been compiled
adding the environmental variables to model the species distribution
from GBIF for R. apiculata and R. mucronata and processed for the Indian
(Graham, 2003). We executed collinearity test of these environmental
subcontinent using ArcGIS 10.5. The occurrence records were spatially
variables through usdm package in R software (Naimi et al., 2014), to
rarefied and filtered within a grid cell of 10 km × 10 km by using SDM
remove the highly correlated variables. Variables having the variance
toolbox version 2.5 available in ArcGIS 10.5. At the end of the process, a
inflammation factor VIF > 10 along with the Pearson correlation coef­
total 67 (Fig. 2) and 101 (Fig. 3) presence records were used in con­
ficient ≥ 0.7 were removed (Dormann et al., 2013). A list of the resulting
structing the models.
variables were provided in Table 1.

2.2. Environmental variable selection

2.3. Development and validation of model
Based on the environmental factors that can potentially affect the
distribution of R. apiculata and R. mucronata, total 27 environmental An ensemble modelling approach were adopted in this study to
variables were selected to model the current distribution pattern of these predict the ecological niche for R. apiculata and R. mucronata. We used
species. Amongst these, a total of 19 bioclimatic variables were eight modelling algorithms from the sdm package available in R software
compiled from Worldclim database, version 1.4 (https://www.world (Naimi and Arajo, 2016). A number of statistical methods with various
clim.org/) (Hijmans et al., 2005) with 30 arc-seconds spatial resolu­ characteristics and levels of complexity offer a range of potential pro­
tion (c. 1 km2 area). Seven soil parameters such as soil pH, soil organic jections, and it has frequently been shown that an ensemble of methods
carbon stock (SOCS), bulk density (BD) of fine earth fraction, organic chosen in accordance with the evaluation criteria gradually enhances
carbon density (OCD), proportion of clay content (CC), cation exchange the transferability of models (Guisan et al., 2017). These model includes:
capacity (CEC) and total nitrogen (TN) were extracted from the Soil­ two classification algorithms, classification and regression trees
Grids™ database version 2.0 (de Sousa et al., 2020). Lastly, one topo­ (RPART) (Therneau and Atkinson, 1997), and flexible discriminant
graphical variable i.e. elevation has been added to project the niche analysis (FDA) (Hastie et al., 1994); a maximum entropy algorithm-
range of these species. MAXENT (Phillips et al., 2006); a climatic envelope approach

Fig. 1. Methodology in flow diagram.

3
P. Samal et al. Ecological Indicators 154 (2023) 110631

Fig. 2. Current occurrence records of Rhizophora apiculata along (A) West Bengal (B) Odisha (C) Andhra Pradesh (D) Tamil Nadu (E) Kerala (F) Karnataka (G)
Maharashtra (H) Gujarat coast.

(BIOCLIM) (Busby, 1991); two modern machine learning approach, 2004). Similarly, based on TSS value (ranges between + 1 and − 1),
random forests (RF) algorithm (Breiman, 2001), and support vector model performance is classified as good to excellent (>0.8), useful
machine (SVM) (Vapnik, 1995); and two regression algorithms, gener­ (0.6–0.8) and poor (0.2–0.6) (Coetzee et al., 2009; Shabani et al., 2016).
alized linear model (GLM) (McCullagh and Nelder, 1989) and multi­ In the present study, we constructed an ensemble model using the
variate adaptive regression splines (MARS) (Friedman, 1991). weighted averaging approach based on TSS statistics and the optimal
For each species, the occurrence points were arbitrarily splited into threshold criterion 2 (i.e. Max (Sens + Spec)). By utilising the ’ensemble’
80% for training and remaining 20% for validating. The background function of the sdm package, the models that were successfully fitted
points (30 points) were chosen in R software depending on the envi­ were taken into consideration to produce ensemble maps. The final
ronmental variables to enhance the model’s accuracy (Senay et al., predicted potential distribution for past, current and future (2050 and
2013). The presence data for the species with a background mask was 2070) of these species had a range value from 0 to 1 which were
used to generate the background or pseudo-absence data. To strengthen reclassified into five groups showing extreme suitability (>0.8), high
the evaluation of model performance, each algorithm was set to develop suitability (0.6–0.8), moderate suitability (0.4–0.6), low suitability
6 models (3 × 2 replication method), thus generating a total of 36 (0.2–0.4) and no suitability (0–0.2) to climatic and edaphic conditions
models. To evaluate the importance of each predictor variable and its by using ArcGIS 10.5.
contribution to the niche shift of these Rhizophora species, permutation
constructed on two metrics, such as the Pearson Correlation and AUC 2.4. Core distribution shift
test, has been used in the sdm package.
The model’s evaluation robustness was calibrated and validated The pattern of change in niche area was also estimated, and the
using threshold-dependent true skill statistics (TSS) (Allouche et al., centroid of the shift for past, present and future suitable habitats were
2006) and threshold-independent receiver operating characteristics of compared using a Python-based GIS software i.e. the SDM tool-box
area under the curve (ROC/AUC) (Fielding and Bell, 1997) available in (Brown, 2014). This study also demarcates the centre of niche shift of
sdm package. Based on the AUC value (ranges between 0 and 1), per­ R. apiculata and R. mucronata. To perform this, the distribution of each
formance of the model is categorised as excellent (>0.9), good (0.8–09), species was narrowed down to a centroid (centre) point, and a vector file
fair (0.7–0.8), poor (0.6–0.7) and fail (0.5–0.6) (Swets, 1988; Vanagas, was produced to show the direction and magnitude of the forecasted

4
P. Samal et al. Ecological Indicators 154 (2023) 110631

Fig. 3. Current occurrence records of Rhizophora mucronata. along (A) West Bengal (B) Odisha (C) Andhra Pradesh (D) Tamil Nadu (E) Kerala (F) Karnataka (G)
Maharashtra (H) Gujarat coast.

change over different time slices. Finally, we investigated the changes in extremely suitable habitats for R. apiculata and R. mucronata encompass
the niche distribution by monitoring the shifts in the centre points with about 7981 km2 and 6022 km2, respectively. Projected past (6000 years
various SDMs. BP) distribution depicted that the areas of extreme to high suitable
habitat occurred along West Bengal (Sundarban delta), Odisha (Chilika
3. Results Lagoon), Andhra Pradesh (Coringa mangroves), Tamil Nadu (Pich­
varam, Muthupet), Kerala (Vembanad) and Maharashtra (Vikroli) coast
3.1. Species distribution and evaluation for R. apiculata (Fig. 6a,b). Similarly, for R. mucronata the projected
extreme to high suitable habitat included Sundarban delta, Coringa
The model performance of R. apiculata and R. mucronata were better mangrove, Pichvaram and Muthupet (Cauvery delta) and Vembanad of
than random, with the provided training and testing data set. Based on Indian peninsula (Fig. 7a,b). The areas of extremely suitable habitat of
the AUC and TSS values, the best model performance was produced by past distribution for R. apiculata and R. mucronata encompass about
maxent model for both R. apiculata (AUC-1, TSS-1) and R. mucronata 21,267 km2 and 8626 km2, respectively.
(AUC-1, TSS-1), however the poor performance was generated by bio­ Under RCP2.6–2050 climate change scenario, the ensemble model
clim model for both R. apiculata (AUC- 0.67, TSS- 0.33) and R. mucronata predicts the increase in R. apiculata extremely suitable habitat area in
(AUC-0.78, TSS-0.56). The evaluation of ensemble model also depicted Sundarban delta (3.95%), Mahanadi delta and Chilika Lagoon (1.5%),
that the predicted models for both the species generated excellent model Coringa mangrove (0.77%), Pichavaram and Muthupet (1.50%), Kerala
performance with AUC value 0.99 and TSS value 0.967 for R. apiculata coast (6.90%) and Udupi region (Karnataka coast, 0.22%) (Fig. 6a,b),
and with AUC and TSS value 1 for R. mucronata (Table 2). accounting to ~ 21,563 km2. Under RCP2.6–2070 climatic scenario, the
In the current scenario, the evaluated areas of extreme to high suit­ increased extremely suitable habitat area remain unchanged in com­
able habitat for R. apiculata included West Bengal, Tamil Nadu (Cauvery parison with 2050. Based on the current climate scenario, the overall
Delta), Kerala and Maharashtra (Vasai, Vikroli) coast of the Indian extremely suitable habitat area are expected to increase by 1.70% and
peninsula (Fig. 4a,b). Similarly, the extreme to high suitable habitat for 1.56% under RCP2.6 2050 and 2070 scenario, respectively in the Indian
R. mucronata occur in Tamil Nadu (Cauvery delta), Andhra Pradesh subcontinent. However, in R. mucronata, model a loss in extremely
(Coringa mangrove), Maharashtra (Kundalika, Vikroli) and Gujurat suitable habitat area in Coringa mangroves (0.39%), Pichavaram and
(Dwarka) coast of the Indian peninsula (Fig. 5a,b). The current areas of Muthupet (0.82%), Kerala coast (2.77%), Maharashtra coast (0.33%)

5
P. Samal et al. Ecological Indicators 154 (2023) 110631

Table 1 influence the distribution of both R. apiculata and R. mucronata in the

Percentage contribution of the environmental variables included to model Rhi­ Indian subcontinent. Six bioclimatic variables namely annual mean
zophora apiculata and Rhizophora mucronata. (- indicates removed variables temperature (bio 1), mean diurnal range (bio 2), mean temperature of
because of their high collinearity). wettest quarter (bio 8), mean temperature of coldest quarter (bio 11),
Environmental Sources Abbreviations Percentage contribution precipitation of warmest quarter (bio 18) and precipitation of coldest
variables
Rhizophora Rhizophora quarter (bio 19) are common for both the species. Precipitation of driest
apiculata mucronata month (bio 14) and precipitation for driest quarter (bio 17) can also
Annual Mean WorldClim- bio 1 20.8 10.2
influence the distribution of R. apiculata and R. mucronata, respectively.
Temperature Global climate In addition to climate, two soil parameters such as BD (bulk density of
Mean Diurnal Data htt bio 2 62.4 43.8 the fine earth fraction) and TN (total nitrogen) impacted the distribution
Range ps://www. of R. apiculata whereas for R. mucronata. Total Nitrogen in sediment and
Mean worldclim. bio 8 6.9 4.4
one topographic parameter i.e. altitude contributed the most to the
Temperature org/
of Wettest distribution. Based on both correlation and AUC metric, assessment of
Quarter the average variable importance (average of eight algorithms) revealed
Mean bio 11 6.9 17.5 that bio 2 and bio 1 contributed maximum to the distribution for
Temperature R. apiculata. Similarly, bio 2 and altitude are the major factors control­
of Coldest
Quarter
ling the distribution of R. mucronata.
Precipitation of bio 14 3.6 –
Driest Month 3.3. Core distributional shifts
Precipitation of bio 17 – 2.7
Driest Quarter
Along the West Bengal coastline, the centroid for current habitat for
Precipitation of bio 18 3.1 2.3
Warmest R. apiculata is located at Kakdwip region. Later, the centroid of suitable
Quarter habitat would shift towards Diamond Harbour region under future
Precipitation of bio 19 2 2 climate scenario (Table 3). The centroid of past (Mid-Holocene) distri­
Coldest
bution for R. apiculata was also located at Diamond Harbour region
Quarter
Bulk density BD 8.2 –
(Fig. 4a,b). Similarly, the centroid for R. mucronata distribution in cur­
Total nitrogen TN 8.4 2.9 rent and future climatic conditions is located towards the south of
Altitude Alt – 17.9 Diamond Harbour region. Under past climate scenario the centroid was
positioned at Diamond Harbour region (Fig. 5a,b).
Along Odisha coastline, the centroid for current habitat R. apiculata
Table 2 was positioned at Jagatsinghpur region. The centroid of future suitable
Ensemble model performance of the targeted species. habitat under RCP2.6–2050 and 2070 would be located at the same
Species AUC TSS Sensitivity (model Specificity (model
region as in current scenario. Under RCP8.5–2050 and 2070 scenario,
correctly predicting correctly predicting the centroid will shift to Kendrapara region. The core of R. apiculata
the presence of the absence of distribution during middle Holocene was located at Jagatsinghpur re­
species) species) gion (Fig. 4 a,b). Likewise, for R. mucronata the centroid was located at
R. apiculata 0.99 0.967 1.0 0.967 Puri region under past and current climate scenario. It will remain at the
R. mucronata 1.0 1.0 1.0 1.0 same position under RCP2.6–2050 and 2070 scenario but would shift
towards Chilika under RCP8.5–2050 and 2070 (Fig. 5 a,b).
In Andhra Pradesh coastal regions, the centroid of current suitable
and Karnataka coast (0.49%) under RCP2.6–2050 and 2070 scenario
habitat area for R. apiculata is in Kottapeta region whereas for future
accounting to ~ 868 km2 and ~ 565 km2, respectively. In contrast,
suitable habitat it will move towards Narsapur region under
based on the current climatic scenario, the predicted area for
RCP2.6–2050, 2070 and RCP8.5–2050, 2070 scenario. The centroid for
R. mucronata remains unchanged under RCP2.6–2050 and 2070 sce­
past climate scenario for R. apiculata was situated in the Bay of Bengal
nario. The overall decrease in extremely habitat suitability is estimated
(Fig. 4a,b). For R. mucronata, the current and future suitable habitat
to be 0.41% and 0.43% under RCP2.6–2050 and 2070, respectively
centre under RCP2.6–2050 and 2070 scenario, would be positioned at
(Fig. 7a,b).
East Godavari region. Later, the centroid for R. mucronata shifted to
Under the RCP8.5–2050 climatic scenario, the ensemble model
Visakhapatnam coast under RCP-8.5–2.50 and then to Srikakulam coast
predicted gains in suitable habitat area for R. apiculata by about
under RCP8.5–2070. The centroid for past suitable habitat area was
33,815km2in Sundarban delta (11.69%), Mahanadi delta and Chilika
situated at West Godavari region (Fig. 5a,b).
Lagoon (2.41%), Coringa mangrove (1.48%), Pichavaram and Muthupet
Along Tamil Nadu coast the centroid for past and current suitable
(2.02%), Kerala coast (9.33%) and Udupi region (Karnataka coast,
habitat of R. apiculata was positioned at Manargudi region. However, the
0.40%) by 2050 (Fig. 6a,b). The increase in the habitat suitability pre­
future suitable habitat area centre will shift to Pattukkottai region under
dicted to be almost same under 2070 scenario except Sundarban delta
RCP2.6–2050 and 2070 scenario. Under RCP8.5–2050 and 2070 sce­
(14.33%) and Mahanadi delta and Chilika Lagoon (3.70%). The model
nario, the centroid would again shift back to Manargudi region (Fig. 4a,
forecasts an overall increase in the habitat suitability for R. apiculata
b). Similarly, the distribution core of past suitable habitat for
along the Indian coastline. However, in R. mucronata, the loss of suitable
R. mucronata was located at Thanjavur region. Later, it shifted to
habitat is predicted in regions namely Pichavaram and Muthupet
Nagapattinum coast in current climatic conditions. Then, the future
(0.82%), Kerala coast (2.77%), Mahanadi delta and Chilika Lagoon
suitable habitat would shift towards Kancheepuram region under
(0.08%) and Karnataka coast (0.56%) under RCP8.5–2050 and 2070.
RCP2.6–2050 and RCP2.6–2070 scenario. Under RCP8.5–2050 and
The overall suitable habitat is forecasted to decrease for R. mucronata
2070 scenario, the model predicts a complete loss in the suitable habitat
under RCP8.5 scenario in the Indian subcontinent (Fig. 7a,b).
for R. mucronata (Fig. 5a,b).
Along the Kerala coast, the centroid for current suitable habitat of
3.2. Contribution of environmental variables R. apiculata is located at Kodungallur region. Further, the centroid for
future suitable habitat will shift towards Trichur region under both
The VIF analysis suggested that 9 variables each can potentially RCP2.6 and 8.5, 2050 and 2070 scenario. The distribution core for

6
P. Samal et al. Ecological Indicators 154 (2023) 110631

Fig. 4. Predicted current distribution and core distributional shift for Rhizophora apiculata (a) along the east coast and (b) along the west coast of the Indian
subcontinent.

R. apiculata for past climatic scenario was positioned at Irinjalakuda shifted broadly in the Karnataka state. During the past and current cli­
region (Fig. 4a,b), which shifted to Malappuram coast as the current matic conditions, habitat suitable area for R. mucronata has been located
suitable area for R. mucronata. Then, the centroid would shift towards at Udupi region. The centroid for future habitat suitable area will move
Kasargod region under RCP2.6–2050 and 2070 scenario while under to Uttar Kannada region under RCP2.6–2050 and 2070 scenario and
extreme global warming condition it will shift towards Kannur region would further move to Kodagu region under RCP8.5–2050 and 2070
during 2050 s then towards Wayanad region in 2070 s (Fig. 5a, b). scenario (Fig. 5a,b).
The centroid shift for R. apiculata from current to future habitat In Maharashtra state, the centre for R. apiculata distribution under
suitable area in the state of Karnataka was not so prominent. The core of past climate scenario was located at Mandangarh region whereas during
the distribution for past, current and future habitat suitable area situated present condition the core habitat suitable area is located in Mangaon
at Kundapura region (Fig. 4a,b). However, the centroid for R. mucronata region. Later, the centroid for future habitat suitable area will again shift

7
P. Samal et al. Ecological Indicators 154 (2023) 110631

Fig. 5. Predicted current distribution and core distributional shift for Rhizophora mucronata (a) along the east coast and (b) along the west coast of the Indian
subcontinent.

towards Mandangarh region under both RCP2.6 and-2050 and 2070 habitat suitable area was positioned at Junagadh region under both
scenario (Fig. 4a,b). The core of R. mucronata distribution for the past RCP2.6 and 8.5–2050 and 2070 scenario (Fig. 4a,b). Similarly, the
and current climatic conditions was located at Ratnagiri region. How­ centroid for R. mucronata for the past, current and future (under
ever, it will shift towards Raigarh region under RCP2.6–2050 and RCP2.6–2050 and 2070 scenario) habitat suitable area will remain at
RCP2.6–2070 scenario. On the other hand, the centroid for future Ranavav region. However, under future extreme global warming con­
habitat suitable area would move to Greater Bombay under highest ditions in 2050 (RCP8.5–2050) it will move towards Porbandar and
global warming scenario in 2050 (RCP8.5–2050) and after that will shift would further shift to Kalyanpur region under RCP8.5–2070 scenario
to Satara under RCP8.5–2070 scenario (Fig. 5a,b). (Fig. 5a, b).
In the state of Gujarat, the centroid for R. apiculata for past habitat
suitable area was located at Porbandar which shifted to current suitable
habitat was shifted to Junagadh region. Later, the centroids of future

8
P. Samal et al. Ecological Indicators 154 (2023) 110631

Fig. 6. Predicted past, present and future distribution for Rhizophora apiculata (a) along the east coast and (b) along the west coast of the Indian subcontinent.

4. Discussion 4.1. Model prediction

Each mangrove species’ potential to survive in the changing climate Selecting a best suited single method to model every species is
is determined by its distinct physiological traits. Each mangrove species absolutely critical and unpredictable (Elith and Graham, 2009). Even
is predicted to react swiftly and decisively to climate changes like though, the prediction of single model algorithm differs in accordance
temperature, rainfall, and sea level rise/fall due to its specific tolerance with the selected species, in this study, maxent is the best performed
level for each environmental component. Thus, we mapped the niche model based on evaluation metrics (AUC and TSS score). The maxent
shift of two stilt mangrove species (R. apiculata and R. mucronata) along model is one of the most popular and effective models for analysing how
the Indian coastline across spatial and temporal scales. species respond to environmental factors and how their distribution
changes (Guisan and Thuiller, 2005), because it can render good

9
P. Samal et al. Ecological Indicators 154 (2023) 110631

Fig. 7. Predicted past, present and future distribution for Rhizophora mucronata (a) along the east coast and (b) along the west coast of the Indian subcontinent.

performance even in small size in comparison to other single models 4.2. Climate and edaphic factors affecting the current distribution of
(Elithet al., 2006; Phillips and Dudík, 2008). Several previous studies Rhizophora species
also reported the maxent as the best performed algorithm to model the
potential distribution of various species (Hu et al., 2020; Dang et al., The present study revealed that factors like precipitation, tempera­
2021; Radha and Khwarahm, 2022). The results of this study further ture and elevation strongly influenced the current distribution of both
support the validity of the ensemble modelling strategy over the use of a the targeted species, which was consistent with the pattern seen at both
single model to project species distribution using the AUC and TSS the worldwide (Simard et al., 2019) and regional scales (Banerjee et al.,
evaluation methods (Thuiller et al., 2009; Shabani et al., 2016; Dong 2022). However, comparing the contribution of these environmental
et al., 2020; Dang et al., 2021; Samal et al., 2022). variables to the habitat distribution of the targeted species suggested
that separate sets of environmental variables potentially affected the

10
 P. Samal et al.
Table 3
Potential niche shift locations for targeted mangrove species in response to climatic changes N.B: Distance of past and future centroids has been calculated with respect to current centroid point.
Species Area Centroid shifts

Past Present Future-2050 Future-2070

Mid-Holocene Current RCP2.6 RCP8.5 RCP2.6 RCP8.5

Coordinates Distance Coordinates Distance Coordinates Distance Coordinates Distance Coordinates Distance
(km) (km) (km) (km) (km)

R. apiculata West Bengal 22.21◦ N, 50.80 22.65◦ N, 22.09◦ N, 62.80 22.22◦ N, 49.74 22.09◦ N, 62.79 22.07◦ N, 65.11
88.29◦ E 88.4◦ E 88.31◦ E 88.26◦ E 88.31◦ E 88.32◦ E
Odisha 20.30◦ N, 25.20 20.08◦ N, 20.30◦ N, 24.75 20.42◦ N, 39.46 20.28◦ N, 22.52 20.40◦ N, 35.74
86.17◦ E 86.16◦ E 86.19◦ E 86.27◦ E 86.19◦ E 86.20◦ E
Andhra 16.25◦ N, 61.24 16.59◦ N, 16.59◦ N, 16.30 16.54◦ N, 13.86 16.52◦ N, 19.76 16.62◦ N, 7.02
Pradesh 81.47◦ E 81.92◦ E 81.77◦ E 81.80◦ E 81.75◦ E 81.86◦ E
Tamil Nadu 10.58◦ N, 16.48 10.48◦ N, 10.52◦ N, 8.83 10.60◦ N, 14.82 10.48◦ N, 5.62 10.65◦ N, 21.16
79.36◦ E 79.47◦ E 79.40◦ E 79.41◦ E 79.42◦ E 79.38◦ E
Kerala 10.38◦ N, 3.32 10.38◦ N, 10.42◦ N, 7.00 10.46◦ N, 10.44 10.42◦ N, 7.89 10.47◦ N, 11.38
76.14◦ E 76.11◦ E 76.16◦ E 76.16◦ E 76.17◦ E 76.16◦ E
Karnataka 13.64◦ N, 9.27 13.70◦ N, 13.65◦ N, 7.01 13.64◦ N, 10.07 13.64◦ N, 8.55 13.61◦ N, 13.24
74.71◦ E 74.65◦ E 74.69◦ E 74.72◦ E 74.70◦ E 74.73◦ E
11

Maharashtra 18.04◦ N, 25.24 18.26◦ N, 18.04◦ N, 25.24 18.03◦ N, 26.47 18.04◦ N, 25.36 18.04◦ N, 25.36
73.19◦ E 73.13◦ E 73.19◦ E 73.20◦ E 73.20◦ E 73.20◦ E
Gujarat 21.68◦ N, 154.52 20.74◦ N, 21.33◦ N, 100.16 21.34◦ N, 97.34 21.31◦ N, 95.95 21.35◦ N, 96.54
69.65◦ E 70.75◦ E 70.02◦ E 70.07◦ E 70.04◦ E 70.08◦ E

R. mucronata West Bengal 22.06◦ N, 40.58 21.70◦ N, 21.88◦ N, 20.43 21.79◦ N, 15.08 21.84◦ N, 15.24 21.72◦ N, 15.89
88.35◦ E 88.41◦ E 88.37◦ E 88.52◦ E 88.41◦ E 88.26◦ E
Odisha 19.97◦ N, 21.02 19.83◦ N, 19.84◦ N, 1.57 19.59◦ N, 53.45 19.84◦ N, 1.56 19.65◦ N, 46.58
85.88◦ E 85.74◦ E 85.75◦ E 85.30◦ E 85.73◦ E 85.34◦ E
Andhra 16.57◦ N, 33.81 16.72◦ N, 16.72◦ N, 0 17.26◦ N, 107.12 16.76◦ N, 13.12 18.54◦ N, 320.82
Pradesh 81.72◦ E 81.97◦ E 81.97◦ E 82.81◦ E 82.07◦ E 84.29◦ E
Tamil Nadu 10.49◦ N, 75.94 11.09◦ N, 11.53◦ N, 50.05 12.31◦ N, 138.39 Not suitable - Not suitable -
79.37◦ E 79.70◦ E 79.85◦ E 80.00◦ E
Kerala 10.51◦ N, 61.27 11.0◦ N, 12.36◦ N, 162.31 12.14◦ N, 130.12 12.38◦ N, 164.21 11.76◦ N, 83.73
76.14◦ E 75.83◦ E 75.28◦ E 75.52◦ E 75.28◦ E 75.85◦ E
Karnataka 13.53◦ N, 6.91 13.49◦ N, 14.31◦ N, 97.37 12.42◦ N, 136.37 14.41◦ N, 113.04 12.43◦ N, 140.10
74.73◦ E 74.79◦ E 74.45◦ E 75.41◦ E 74.40◦ E 75.48◦ E
Maharashtra 17.99◦ N, 1.13 18.00◦ N, 18.54◦ N, 60.79 18.99◦ N, 112.26 18.99◦ N, 111.94 17.92◦ N, 54.81

Ecological Indicators 154 (2023) 110631

73.16◦ E 73.16◦ E 73.04◦ E 72.93◦ E 72.94◦ E 73.67◦ E
Gujarat 21.61◦ N, 6.12 21.65◦ N, 21.62◦ N, 6.22 21.69◦ N, 6.80 21.62◦ N, 5.32 21.95◦ N, 44.19
69.71◦ E 69.67◦ E 69.72◦ E 69.62◦ E 69.71◦ E 69.39◦ E
P. Samal et al. Ecological Indicators 154 (2023) 110631

habitat distribution of these two Rhizophora species along the Indian coastal regions (except Sundarban Delta, Mahanadi delta and Karnataka
coastal regions. coast) but the suitable habitat for R. mucronata is predicted to decrease
Temperature was identified as the key factor followed by precipita­ under future climatic conditions (Samal et al., 2022). As the sea level
tion for driving the mangrove distribution. One of the temperature rises, the mangrove community relocates inland but are restricted to
related variables i.e. bio 2 contributed the most to the habitat distri­ heights between mean sea level and peak tides (Duke, 2006). Mangroves
bution of these two mangrove species. Our finding is confirming to the have a limited range of ideal temperature conditions, therefore when
conclusion drawn from the study on physiology of mangrove ecosystems temperature rises, their potential niche will shift north or south to lo­
(Lin et al., 1994). Additionally, growth and reproduction of mangroves cations with a conducive climate condition for their survival, and they
are highly controlled by temperature. Stomatal conductance and will recede from locations that are unsuitable for their growth and
evapotranspiration are influenced by the temperature of the air and its survival. The suitable habitat loss in case of R. mucronata under future
impact on relative humidity, whereas sea surface temperature (SST) climate scenarios is mainly due to rise in temperature all over the Indian
controls the metabolism and root growth of the aboveground plant subcontinent. For the period of 2081–2100 compared to 1850–1900, the
through tidal inundations (Robertson and Alongi, 1992; Quisthoudt anticipated rise in temperature is by 1.0 ◦ C to 1.8 ◦ C in the low CO2
et al., 2012). Hong et al., (2020) reported that, the fruit of each tree of emission scenario and 3.3 ◦ C to 5.7 ◦ C in the high CO2 emission scenario
some of the mangrove species increases with latitudinal decrease, and a (Arias et al., 2021). Based on the IPCC, AIR5 report, the future rise in
rise in temperature increases the survival rate of seedlings (Gillis et al., global mean sea level is projected as 0.43 m under RCP2.6 and up to
2019). A study on the biodiversity of mangroves found that species 0.84 m under RCP8.5 scenario by 21st century (Stocker et al., 2013). In
richness and the composition of mangrove spatial assemblages are accordance with this, the mangrove community will also shift as each
constrained by the winter temperature (Wu et al., 2018). For the Indo- mangrove species need specific saline water condition for its growth.
West Pacific region, the recorded mean annual temperature was Subsequently, R. apiculata commonly occurs in the mid-lower estuary
20◦ –30 ◦ C, mean maximum temperature of hottest month was and embayments where they can get sufficient tidal inundation to grow,
23◦ –38 ◦ C, average minimum temperature of the coldest month was whereas, R. mucronata preferably grows in areas subjected to regular
13◦ –18 ◦ C and minimum temperature tolerated was 10 ◦ C (Duke, 2006). freshwater inputs (Duke, 2006). Thus, increase in the area of highly
Thus, the models’ predictions of the best ranges for environmental suitable habitat for R. apiculata, and significant reduction for
variables mainly agreed with the results of the field experiments. Our R. mucronata in the present study confirms the impact of climate change
findings are consistent with numerous previous studies that reported and sea level rise on the coastal ecosystem along the Indian coastline.
temperature as a key driving force for the potential distribution of During Middle Holocene, there was moderate temperature and pre­
mangroves (Wu et al., 2018; Hu et al., 2020; Dang et al., 2021). cipitation conditions which encouraged the mangrove species to expand
Additionally, the present study also revealed altitude as a significant their ranges to their largest extent. Based on past distribution pro­
factor for the distribution of R. mucronata. The elevation has frequently jections, the highly suitable habitat for both R. apiculata and
been identified as the significant variable for the distribution of R. mucronata was found to occur in Sundarbans delta, Pichavaram
mangrove species because it is well-associated with the tidal fluctuations mangroves and along the Kerala coastal regions of the Indian subcon­
and its amplitude, one of the important environmental parameters for tinent during 6000 yrs BP. The location of centroid points of niche range
the zonation of mangrove species along the coastline (Leong et al., 2018; for both the species shifted mostly towards the north and landward from
Banerjee et al., 2021). Moreover, R. mucronata usually occurs at a reg­ mid-Holocene to the current time suggesting a higher sea-level specif­
ular freshwater input zone fringing along the estuarine channels towards ically in these sectors of the Indian coastline. These results agree with
the landward edges where the slope is higher. Our findings further Cannon et al., (2009), which noted that during the deglaciation, coastal
revealed that bulk density of soil can also be a potential driver for the mangroves and swamp forests experienced total biogeographic
distribution of these mangrove species because higher bulk density of replacement. The predicted maps for both the species under the past
soil can help to increase the root diameter and biomass which in turn (Middle Holocene) are complementing many previous studies based on
increases mangrove productivity (Ola et al., 2018, 2019). Our result is fossil pollen records and reveals wide distribution of both the Rhizophora
consistent with the previous study on restoration of mangrove in the species along the east coast during the Middle Holocene time period
Indo-Pacific region (Banerjee et al., 2021). The growth of these (Figs. 8 and 9) (Farooqui and Vaz, 2000; Bhattacharyya et al., 2013; Hait
mangrove species have been found to be most likely restricted by the and Behling, 2019; Srivastava and Farooqui, 2017) and west coast of
nutrients such as nitrogen. Due in part to the anoxic soil properties, India (Limaye and Kumaran, 2012; Limaye et al., 2014; Alappat et al.,
ammonium is the main compound of nitrogen in mangrove soils, and 2021).
ammonium uptake primarily supports tree development (Reef et al., Subsequently, the relative sea level changes may also have a po­
2010; Ray et al., 2014; Reis et al., 2017). In the present study, total tential impact on the distribution of these species because salinity acts as
nitrogen of soil is also detected as an influencing factor that can affect a primary factor for the zonation and growth of these species. Studies on
the distribution of these mangrove species in the coastal ecosystem. Middle Holocene Sea level changes show a + 5 m rise for 6000 years BP
(Nageswara Rao and Sadakata, 1993), +3 m rise around 7300 years BP
4.3. Niche shift in response to future climate change scenarios (Banerjee, 2000) along the east coast of India and + 1.5 m rise around
8000–7000 years BP (Alappat et al., 2021) along the west coast of India.
Coastal vegetation has been identified as being more affected by Consequently, the salinity gradient is also optimum because of moderate
climate change, which might indicate either a niche range contraction i. precipitation and temperature conditions throughout the year along
e. loss of suitable habitat or a range expansion/shift i.e. migration of with arise in sea level, which encourages the growth of stilt mangroves
species towards inland. We precisely analysed the species-specific re­ like Rhizophora species.
sponses to projected climatic changes and the pattern of their niche
range shifts. The hypothesis of niche contraction under future climate 4.4. Limitations of the study
scenario postulated by Ellison (1994) was found to be valid for
R. mucronata whereas, we found a landward shift of the suitable habitat The future projections in this study are only based on climate data for
under future climate scenarios for R. apiculata. It has been observed in future scenarios although there are several other physicochemical and
the present study that the centroid of the future suitable habitat for environmental factors which influence the distribution and growth of
R. apiculata is showing a northward shift along both east and west coast coastal vegetation. Hence, it is crucial to add such variables also while
of India (except Karnataka and Maharashtra coast). The centroid of the projecting the future distribution of species. Presently, such data are
future habitats for R. mucronata also showed a northward shift along the available only for current climatic conditions and that also in a broad

12
P. Samal et al. Ecological Indicators 154 (2023) 110631

Fig. 8. Mid-Holocene fossil pollen records of Rhizophora apiculata.

spatiotemporal scale. Therefore, it is crucial to develop finer resolution suited for the selected species. The areas that are presently suitable for
environmental and edaphic variables for future climatic scenarios to these species and will continue to be favourable even in future climatic
provide more robust future projections for species niche shift and conditions should receive the highest priority for restoration and con­
habitat suitability. Additionally, other potential driving forces which servation plans. Other studies have also investigated the potential dis­
influence the distribution of the mangrove species includes the coastal tribution of endemic species for the implication of conservation
geomorphological features which might have been a limiting factor strategies globally (Radha and Khwarahm, 2022; Hama and Khwarahm,
causing mass loss or restricting mangroves in narrow fringes as they shift 2023). The coastal regions are one of the densely populated areas with
landward due to rising sea levels (Asbridge et al., 2015).Genetic ana­ numerous developmental activities causing to freshwater restrictions,
lyses, controlled experiments, and the data availability on these species’ environmental pollution and encroachment along the coasts. Conse­
basic habitats from fossil records may further increase the accuracy of quently, to achieve efficient coastal management and restoration plan­
the prediction models. Since environmental conditions can fluctuate at a ning we should integrate the anthropogenic aspect in the framework of
smaller scale, particularly in estuarine and coastal ecosystem, modelling the conservation policies. Most prominently, the highly suitable habitats
at various spatial and temporal scales and including climate proxy data i.e., Sundarbans delta, Chilika lagoon, Mahanadi delta and Kerala
from direct observations can further improve the predictive accuracy. coastline must be given high priority in addition to other sectors of the
However, by giving these errors equal weight, the threshold used in this coastline for protection and conservation. Loss in species richness along
study TSS minimizes the total commission and omission errors. An the coastline will also increase the vulnerability of mangroves in
evaluation criterion that prioritises sensitivity over specificity may also response to future climatic conditions. It is important to note that by
be preferred but this would also produce significant commission errors summing the binary rasters for each species, we were able to assess
(Das et al., 2019). species richness as well as identify the priority conservation areas.
Although this method has been employed frequently in multispecies
modelling research (McKerrow et al., 2018), this does not take into ac­
4.5. Future priority conservation areas (PCAs) along the Indian coastline count the biotic interactions (like competition and facilitation), which
could cause an underestimation or overestimation of species richness.
Future loss of habitats with significant loss in species richness Thus, the biotic interconnections between species may be incorporated
highlights the necessity to preserve habitats that are environmentally

13
P. Samal et al. Ecological Indicators 154 (2023) 110631

Fig. 9. Mid-Holocene fossil pollen records of Rhizophora mucronata.

into the more complicated community-level modelling method to study suggests the highly suitable habitats along West Bengal, Odisha
further strengthen the conservation priority target areas determined by and Kerala coast as the priority conservation areas (PCAs) in addition to
our simplified species-level modelling framework (Dormann et al., other sectors of the coastline for mangrove protection and conservation.
2018). Emphasis should also be given to the loss in species richness which will
also increase the vulnerability of mangroves in response to future cli­
5. Conclusions matic conditions. Hence, the priority conservation targets established by
this streamlined species-level modelling method can be strengthened if
Understanding mangroves’ responses to changes in temperature, the complex community-level modelling framework incorporates the
precipitation and sea level patterns over time and space requires an biotic relationships between species, coastal geomorphological features
estimation of species-specific distributional ranges and shift along their as well as by generating finer resolution environmental and edaphic
habitats. In this study we mapped the shift in the spatial and temporal variables for future climatic scenarios.
distribution of two stilt mangrove species (R. apiculata, R. mucronata)
along the Indian coastline. It showed that temperature, precipitation, CRediT authorship contribution statement
and elevation had an impact on both the targeted species’ present dis­
tribution, which was consistent with the trend seen at both the global Pujarini Samal: Investigation, Data curation, Formal analysis,
and regional scales. Mean diurnal range of temperature is identified as Writing – original draft, Validation. Jyoti Srivastava: Supervision,
the key factor followed by precipitation, bulk density of soil and total Resources, Conceptualization, Writing – review & editing. Bipin
nitrogen in mangrove sediments which are driving the species’ distri­ Charles: Resources, Methodology, Software, Writing – review & editing.
bution. However, surface elevation also played a significant role in the S.R. Singarasubramanian: Data curation, Formal analysis.
distributional shift of R. mucronata. The centre for suitable habitat for
R. apiculata along east and west coast of India under future climate
change scenarios would be shifting northwards except along Karnataka Declaration of Competing Interest
and Maharashtra coast while for R. mucronata the core distributional
shift would also be towards north except along West Bengal, Odisha and The authors declare the following financial interests/personal re­
Karnataka coast. Habitat suitability for both the species shifted towards lationships which may be considered as potential competing interests:
north and landward from mid-Holocene to the current time suggesting a Pujarini Samal reports financial support was provided by India Ministry
rise in the sea-level along different sectors of the coastline. Finally, this of Science & Technology Department of Science and Technology.

14
P. Samal et al. Ecological Indicators 154 (2023) 110631

Data availability distribution modelling: 10 questions to guide interpretation and avoid false
conclusions. Glob. Ecol. Biogeogr. 27 (9), 1004–1016.
Duke, N.C., 2006. Rhizophora apiculata, R. mucronata, R. stylosa, R. × annamalai, R. ×
Data will be made available on request. lamarckii (Indo-West Pacific Stilt Mangroves). Species Profiles for Pacific Island
Agroforestry. Permanent Agriculture Resources, Hōlualoa, Hawai‘i, USA.
Acknowledgement Dwire, K.A., Mellmann-Brown, S., Gurrieri, J.T., 2018. Potential effects of climate change
on riparian areas, wetlands, and groundwater-dependent ecosystems in the Blue
Mountains, Oregon. USA. Clim. Serv. 10, 44–52.
The authors thank the Director, Birbal Sahni Institute of Palae­ Elith*, J., H. Graham*, C., P. Anderson, R., Dudík, M., Ferrier, S., Guisan, A., J. Hijmans,
osciences for providing necessary facilities to accomplish this work. PS is R., Huettmann, F., R. Leathwick, J., Lehmann, A., Li, J., 2006. Novel methods
improve prediction of species’ distributions from occurrence data. Ecography, 29
thankful to DST INSPIRE (DST/INSPIRE FELLOWSHIP/2017/IF170769) (2).129-151.
for providing financial support as a SRF. Elith, J., Franklin, J., 2013. Species distribution modeling. In: Encyclopedia of
Biodiversity, Second Edition. Elsevier Inc., pp. 692–705
Elith, J., Graham, C.H., 2009. Do they? How do they? Why do they differ? On finding
References reasons for differing performances of species distribution models. Ecography 32 (1),
66–77.
Alappat, L., Frechen, M., Tsukamoto, S., Anupama, K., Prasad, S., Gopakumar, P.G., Elith, J., Leathwick, J.R., 2009. Species distribution models: ecological explanation and
Kumar, S.S., 2021. Evidences of early to mid-Holocene land–sea interactions and prediction across space and time. Annu. Rev. Ecol. Evol. Syst. 40 (1), 677–697.
formation of wetlands of Central Kerala in the south west coast of India. Reg. Stud. Ellison, J., 1994. Climate Change and Sea Level Rise Impacts on Mangrove Ecosystems.
Mar. Sci. 48, 102009. Impacts of Climate Change on Ecosystems and Species, A Marine Conservation and
Allouche, O., Tsoar, A., Kadmon, R., 2006. Assessing the accuracy of species distribution Development Report.
models: prevalence, kappa and the true skill statistic (TSS). J. Appl. Ecol. 43 (6), Farooqui, A., Vaz, G.G., 2000. Holocene sea-level and climatic fluctuations: Pulicat
1223–1232. Lagoon-A case study. Curr. Sci. 79 (10), 1484–1488.
Alongi, D.M., 2008. Mangrove forests: resilience, protection from tsunamis, and Fielding, A.H., Bell, J.F., 1997. A review of methods for the assessment of prediction
responses to global climate change. Estuar. Coast. Shelf Sci. 76 (1), 1–13. errors in conservation presence/absence models. Environ. Conserv. 24 (1), 38–49.
Alongi, D.M., 2015. The impact of climate change on mangrove forests. Curr. Clim. Friedman, J.H., 1991. Multivariate adaptive regression splines. The Annals of Statistics
Change Rep. 1 (1), 30–39. 19 (1), 1–67.
Alongi, D.M., Pfitzner, J., Trott, L.A., Tirendi, F., Dixon, P., Klumpp, D.W., 2005. Rapid Friess, D.A., Yando, E.S., Abuchahla, G.M.O., Adams, J.B., Cannicci, S., Canty, S.W.J.,
sediment accumulation and microbial mineralization in forests of the mangrove Cavanaugh, K.C., Connolly, R.M., Cormier, N., Dahdouh-Guebas, F., Diele, K.,
Kandeliacandel in the Jiulongjiang Estuary. China. Estuar. Coast. Shelf Sci. 63 (4), Feller, I.C., Fratini, S., Jennerjahn, T.C., Lee, S.Y., Ogurcak, D.E., Ouyang, X.,
605–618. Rogers, K., Rowntree, J.K., Sharma, S., Sloey, T.M., Wee, A.K.S., 2020. Mangroves
Arias, P.A., Bellouin, N., Coppola, E., Jones, R.G., Krinner, G., Marotzke, J., Zickfeld, K., give cause for conservation optimism, for now. Curr. Biol. 30 (4), R153–R154.
2021. Technical summary. Clim. Change 33–144. GBIF, 2023. GBIF.org (16 January 2023) GBIF Occurrence Download 10.15468/dl.
Asbridge, E., Lucas, R., Accad, A., Dowling, R., 2015. Mangrove response to aq8r9c.
environmental changes predicted under varying climates: case studies from Gillis, L.G., Hortua, D.A.S., Zimmer, M., Jennerjahn, T.C., Herbeck, L.S., 2019.
Australia. Curr. For. Rep. 1 (3), 178–194. Interactive effects of temperature and nutrients on mangrove seedling growth and
Banerjee, P.K., 2000. Holocene and late Pleistocene relative sea level fluctuations along implications for establishment. Mar. Environ. Res. 151, 104750.
the east coast of India. Mar. Geol. 167 (3–4), 243–260. Graham, M.H., 2003. Confronting multicollinearity in ecological multiple regression.
Banerjee, A.K., Feng, H., Lin, Y., Liang, X., Wang, J., Huang, Y., 2022. Setting the Ecology 84 (11), 2809–2815.
priorities straight-Species distribution models assist to prioritize conservation targets Guisan, A., Edwards Jr, T.C., Hastie, T., 2002. Generalized linear and generalized
for the mangroves. Sci. Total Environ. 806, 150937. additive models in studies of species distributions: setting the scene. Ecol. Model.
Bhattacharyya, S., Pethick, J., Sarma, K.S., 2013. Managerial response to sea level rise in 157 (2–3), 89–100.
the tidal estuaries of the Indian Sundarbans: a geomorphological approach. Water Guisan, A., Lehmann, A., Ferrier, S., Austin, M., Overton, J.M.C., Aspinall, R., Hastie, T.,
Policy 15 (S1), 51–74. 2006. Making better biogeographical predictions of species’ distributions. J. Appl.
Breiman, L., 2001. Random forests. Mach. Learn. 45 (1), 5–32. Ecol. 43 (3), 386–392.
Brown, J.L., Anderson, B., 2014. SDMtoolbox: a python-based GIS toolkit for landscape Guisan, A., Thuiller, W., 2005. Predicting species distribution: offering more than
genetic, biogeographic and species distribution model analyses. Methods Ecol. Evol. simplehabitat models. Ecol. Lett. 8, 993–1009. https://doi.org/10.1111/j.1461-
5 (7), 694–700. 0248.2005.00792.x.
Buisson, L., Thuiller, W., Casajus, N., Lek, S., Grenouillet, G., 2010. Uncertainty in Hait, A.K., Behling, H., 2019. Responses of the mangrove ecosystem to Holocene
ensemble forecasting of species distribution. Glob. Chang. Biol. 16 (4), 1145–1157. environmental change in the Sundarban biosphere reserve. India. Acta Palaeobot. 59
Busby, J. R. (1991). BIOCLIM–A bioclimatic analysis and prediction system. In ‘Nature (2), 391–409.
Conservation: Cost Effective Biological Surveys and Data Analysis’.(Eds CR Margules Hama, A.A., Khwarahm, N.R., 2023. Predictive mapping of two endemic oak tree species
and MP Austin.) pp. 64–68. CSIRO, 64-68. under climate change scenarios in a semiarid region: Range overlap and implications
Cannon, C.H., Morley, R.J., Bush, A.B.G., 2009. The current refugial rainforests of for conservation. Ecol. Inform. 73, 101930.
sundaland are unrepresentative of their biogeographic past and highly vulnerable to Hastie, T., Tibshirani, R., 2004. Efficient quadratic regularization for expression arrays.
disturbance. Proc. Natl. Acad. Sci. 106 (27), 11188–11193. Biostatistics 5 (3), 329–340.
Cao, L., Fox, A.D., 2009. Birds and people both depend on China’s wetlands. Nature 460 Hastie, T., Tibshirani, R., Buja, A., 1994. Flexible discriminant analysis by optimal
(7252), 173. scoring. J. Am. Stat. Assoc. 89 (428), 1255–1270.
Coetzee, B.W., Robertson, M.P., Erasmus, B.F., Van Rensburg, B.J., Thuiller, W., 2009. Hijmans, R.J., Graham, C.H., 2006. The ability of climate envelope models to predict the
Ensemble models predict important bird areas in southern Africa will become less effect of climate change on species distributions. Glob. Chang. Biol. 12 (12),
effective for conserving endemic birds under climate change. Glob. Ecol. Biogeogr. 2272–2281.
18 (6), 701–710. Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., Jarvis, A., 2005. Very high
Dang, A.T.N., Kumar, L., Reid, M., Anh, L.N.T., 2021. Modelling the susceptibility of resolution interpolated climate surfaces for global land areas. International Journal
wetland plant species under climate change in the Mekong Delta. Vietnam. Ecol. of Climatology: A Journal of the R. Meteorol. Soc. 25 (15), 1965–1978.
Inform. 64, 101358. Hong, P., Wen, Y., Xiong, Y., Diao, L., Gu, X., Feng, H., Yang, C., Chen, L., 2021.
Das, S., Baumgartner, J.B., Esperon-Rodriguez, M., Wilson, P.D., Yap, J.-Y., Rossetto, M., Latitudinal gradients and climatic controls on reproduction and dispersal of the non-
Beaumont, L.J., 2019. Identifying climate refugia for 30 Australian rainforest plant native mangrove Sonneratia apetala in China. Estuar. Coast. Shelf Sci. 248, 106749.
species, from the last glacial maximum to 2070. Landsc. Ecol. 34 (12), 2883–2896. Hu, W., Wang, Y., Zhang, D., Yu, W., Chen, G., Xie, T., Liu, Z., Ma, Z., Du, J., Chao, B.,
de Sousa, L.M., Poggio, L., Batjes, N.H., Heuvelink, G.B., Kempen, B., Riberio, E., Lei, G., Chen, B., 2020. Mapping the potential of mangrove forest restoration based
Rossiter, D., 2020. SoilGrids 2.0: producing quality-assessed soil information for the on species distribution models: A case study in China. Sci. Total Environ. 748,
globe. Soil Discuss. 2020, 1–37. 142321.
De’Ath, G., 2007. Boosted trees for ecological modeling and prediction. Ecology 88 (1), Kaky, E., Nolan, V., Alatawi, A., Gilbert, F., 2020. A comparison between Ensemble and
243–251. MaxEnt species distribution modelling approaches for conservation: A case study
Dong, X., Ju, T., Grenouillet, G., Laffaille, P., Lek, S., Liu, J., 2020. Spatial pattern and with Egyptian medicinal plants. Ecol. Inform. 60, 101150.
determinants of global invasion risk of an invasive species, Kayranli, B., Scholz, M., Mustafa, A., Hedmark, Å., 2010. Carbon storage and fluxes
sharpbellyHemiculterleucisculus (Basilesky, 1855). Sci. Total Environ. 711, 134661. within freshwater wetlands: a critical review. Wetlands 30 (1), 111–124.
Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., Marquéz, J.R.G., Latimer, A.M., Wu, S., Gelfand, A.E., SilanderJr, J.A., 2006. Building statistical models to
Gruber, B., Lafourcade, B., Leitão, P.J., Münkemüller, T., McClean, C., Osborne, P.E., analyze species distributions. Ecol. Appl. 16 (1), 33–50.
Reineking, B., Schröder, B., Skidmore, A.K., Zurell, D., Lautenbach, S., 2013. Lawler, J.J., Shafer, S.L., White, D., Kareiva, P., Maurer, E.P., Blaustein, A.R., Bartlein, P.
Collinearity: a review of methods to deal with it and a simulation study evaluating J., 2009. Projected climate-induced faunal change in the Western Hemisphere.
their performance. Ecography 36 (1), 27–46. Ecology 90 (3), 588–597.
Dormann, C.F., Bobrowski, M., Dehling, D.M., Harris, D.J., Hartig, F., Lischke, H., Leong, R.C., Friess, D.A., Crase, B., Lee, W.K., Webb, E.L., 2018. High-resolution pattern
Moretti, M.D., Pagel, J., Pinkert, S., Schleuning, M., Schmidt, S.I., Sheppard, C.S., of mangrove species distribution is controlled by surface elevation. Estuar. Coast.
Steinbauer, M.J., Zeuss, D., Kraan, C., 2018. Biotic interactions in species Shelf Sci. 202, 185–192.

15
P. Samal et al. Ecological Indicators 154 (2023) 110631

Limaye, R.B., Kumaran, K.P.N., 2012. Mangrove vegetation responses to Holocene Reis, C.R.G., Nardoto, G.B., Oliveira, R.S., 2017. Global overview on nitrogen dynamics
climate change along Konkan coast of South-Western India. Quat. Inter. 263, in mangroves and consequences of increasing nitrogen availability for these systems.
114–128. Plant and Soil 410 (1-2), 1–19.
Limaye, R.B., Kumaran, K.P.N., Padmalal, D., 2014. Mangrove habitat dynamics in Robertson, A.I., Alongi, D.M., 1992. Tropical Mangrove Ecosystem. American geoscience
response to Holocene Sea level and climate changes along southwest coast of India. union, Washington.
Quat. Inter. 325, 116–125. Sahana, M., Areendran, G., Sajjad, H., 2022. Assessment of suitable habitat of mangrove
Lin, P., Shen, R.C., Shen, C.Y., 1994. The characteristics of chilling-resistance on six species for prioritizing restoration in coastal ecosystem of Sundarban Biosphere
mangrove plants. Journal of Xiamen University(Natural Science) 33, 249–252. Reserve. India. Sci. Rep. 12 (1), 20997.
McCullagh, P., Nelder, J.A. (Eds.), 1989. Generalized Linear Models. Springer US, Samal, P., Srivastava, J., Singarasubramanian, S.R., Saraf, P.N., Charles, B., 2022.
Boston, MA. Ensemble modeling approach to predict the past and future climate suitability for
McKerrow, A.J., Tarr, N.M., Rubino, M.J., Williams, S.G., Albright, T., 2018. Patterns of two mangrove species along the coastal wetlands of peninsular India. Ecol. Inform.
species richness hotspots and estimates of their protection are sensitive to spatial 72, 101819.
resolution. Divers. Distrib. 24 (10), 1464–1477. Senay, S.D., Worner, S.P., Ikeda, T., Davies, A., 2013. Novel three-step pseudo-absence
Middleton, B.A., (Ed.)., 2012. Global change and the function and distribution of selection technique for improved species distribution modelling. PLoS One 8 (8),
wetlands, Vol. 1. Springer Science and Business Media. e71218.
Nagelkerken, I., Dorenbosch, M., Verberk, W.C.E.P., De La Morinière, E.C., van Der Shabani, F., Kumar, L., Ahmadi, M., 2016. A comparison of absolute performance of
Velde, G., 2000. Importance of shallow-water biotopes of a Caribbean bay for different correlative and mechanistic species distribution models in an independent
juvenile coral reef fishes: patterns in biotope association, community structure and area. Ecol. Evol. 6 (16), 5973–5986.
spatial distribution. Mar. Ecol. Prog. Ser. 202, 175–192. Simard, M., Fatoyinbo, L., Smetanka, C., Rivera-Monroy, V.H., Castañeda-Moya, E.,
Nageswara Rao, K., Sadakata, N., 1993. Holocene evolution of deltas on the east coast of Thomas, N., Van der Stocken, T., 2019. Mangrove canopy height globally related to
India. In: In: Deltas of the World, pp. 1–15. precipitation, temperature and cyclone frequency. Nat. Geosci. 12 (1), 40–45.
Naimi, B., Araújo, M.B., 2016. sdm: a reproducible and extensible R platform for species Srivastava, J., Farooqui, A., 2017. Holocene climate and relative sea level changes in
distribution modelling. Ecography 39 (4), 368–375. Cauvery River delta, India based on pollen and sedimentary records. J. Palaeontol.
Naimi, B., Hamm, N.A., Groen, T.A., Skidmore, A.K., Toxopeus, A.G., 2014. Where is Soc. 62 (2), 193–204.
positional uncertainty a problem for species distribution modelling? Ecography 37 Stocker, T.F., Qin, D., Plattner, G.K., Alexander, L.V., Allen, S.K., Bindoff, N.L., Xie, S.P.,
(2), 191–203. 2013. Technical summary. In: ClImate Change 2013: the PhysIcal ScIence BasIs.
Norberg, A., Abrego, N., Blanchet, F.G., Adler, F.R., Anderson, B.J., Anttila, J., Araújo, M. ContrIbutIon of WorkIng Group I to the FIfth Assessment Report of the
B., Dallas, T., Dunson, D., Elith, J., Foster, S.D., Fox, R., Franklin, J., Godsoe, W., Intergovernmental Panel on ClImate Change. Cambridge University Press,
Guisan, A., O’Hara, B., Hill, N.A., Holt, R.D., Hui, F.K.C., Husby, M., Kålås, J.A., pp. 33–115.
Lehikoinen, A., Luoto, M., Mod, H.K., Newell, G., Renner, I., Roslin, T., Soininen, J., Svenning, J.C., Fløjgaard, C., Marske, K.A., Nógues-Bravo, D., Normand, S., 2011.
Thuiller, W., Vanhatalo, J., Warton, D., White, M., Zimmermann, N.E., Gravel, D., Applications of species distribution modeling to paleobiology. Quat. Sci. Rev. 30
Ovaskainen, O., 2019. A comprehensive evaluation of predictive performance of 33 (21–22), 2930–2947.
species distribution models at species and community levels. Ecol. Monogr. 89 (3), Svetnik, V., Liaw, A., Tong, C., Culberson, J.C., Sheridan, R.P., Feuston, B.P., 2003.
e01370. Random forest: a classification and regression tool for compound classification and
Ola, A., Schmidt, S., Lovelock, C.E., 2018. The effect of heterogeneous soil bulk density QSAR modeling. J. Chem. Inf. Comput. 43 (6), 1947–1958.
on root growth of field-grown mangrove species. Plant and Soil 432 (1-2), 91–105. Swets, J.A., 1988. Measuring the accuracy of diagnostic systems. Science 240 (4857),
Ola, A., Gauthier, A.R.G., Xiong, Y., Lovelock, C.E., 2019. The roots of blue carbon: 1285–1293.
responses of mangrove stilt roots to variation in soil bulk density. Biol. Lett. 15 (4), Therneau, T.M., Atkinson, E.J., 1997. An introduction to recursive partitioning using the
20180866. RPART routines (Vol. 61, p. 452). Mayo Foundation: Technical report.
Pearson, R.G., Dawson, T.P., 2003. Predicting the impacts of climate change on the Thuiller, W., 2003. BIOMOD–optimizing predictions of species distributions and
distribution of species: are bioclimate envelope models useful? Glob. Ecol. Biogeogr. projecting potential future shifts under global change. Glob. Chang. Biol. 9 (10),
12 (5), 361–371. 1353–1362.
Phillips, S.J., Dudík, M., 2008. Modeling of species distributions with Maxent: new Thuiller, W., Brotons, L., Araújo, M.B., Lavorel, S., 2004. Effects of restricting
extensions and a comprehensive evaluation. Ecography 31 (2), 161–175. environmental range of data to project current and future species distributions.
Phillips, S.J., Anderson, R.P., Schapire, R.E., 2006. Maximum entropy modeling of Ecography 27 (2), 165–172.
species geographic distributions. Ecol. Model. 190 (3–4), 231–259. Thuiller, W., Lafourcade, B., Engler, R., Araújo, M.B., 2009. BIOMOD–a platform for
Quisthoudt, K., Schmitz, N., Randin, C.F., Dahdouh-Guebas, F., Robert, E.M.R., ensemble forecasting of species distributions. Ecography 32 (3), 369–373.
Koedam, N., 2012. Temperature variation among mangrove latitudinal range limits Vanagas, G., 2004. Receiver operating characteristic curves and comparison of cardiac
worldwide. Trees 26, 1919–1931. https://doi.org/10.1007/s00468-012-0760-1. surgery risk stratification systems. Interact. Cardiovasc. Thorac. Surg. 3 (2),
Radha, K.O., Khwarahm, N.R., 2022. An Integrated Approach to Map the Impact of 319–322.
Climate Change on the Distributions of Crataegus azarolus and Crataegus monogyna Vapnik, V.N. (Ed.), 1995. The Nature of Statistical Learning Theory. Springer New York,
in Kurdistan Region. Iraq. Sustainability 14 (21), 14621. New York, NY.
Ravikumar, S., Nazar, S., Nuralshiefa, A., Abideen, S., 2005. Antibacterial activity of Verhoeven, J.T., Setter, T.L., 2010. Agricultural use of wetlands: opportunities and
traditional therapeutic coastal medicinal plants against some pathogens. J. Environ. limitations. Ann. Bot. 105 (1), 155–163.
Biol. 26 (2 Suppl), 383–386. Wu, Y., Ricklefs, R.E., Huang, Z., Zan, Q., Yu, S., 2018. Winter temperature structures
Ray, R., Majumder, N., Das, S., Chowdhury, C., Jana, T.K., 2014. Biogeochemical cycle of mangrove species distributions and assemblage composition in China. Glob. Ecol.
nitrogen in a tropical mangrove ecosystem, east coast of India. Mar. Chem. 167, Biogeogr. 27, 1492–1506. https://doi.org/10.1111/geb.12826.
33–43. Zhang, Y., Meng, X., Xia, P., Li, Z., 2021. Response of Mangrove Development to Air
Reef, R., Feller, I.C., Lovelock, C.E., 2010. Nutrition of mangroves. Tree Physiol. 30 (9), Temperature Variation Over the Past 3000 Years in Qinzhou Bay. Tropical China.
1148–1160. Front. Earth Sci. 9, 678189.
Zurada, J., 1992. Introduction to artificial neural systems. West Publishing Co.

16