You are on page 1of 17

Ecological Informatics 32 (2016) 107123

Contents lists available at ScienceDirect

Ecological Informatics
journal homepage: www.elsevier.com/locate/ecolinf

A review of Computational Intelligence techniques in coral


reef-related applications
S. Salcedo-Sanz a,, L. Cuadra a, M.J.A. Vermeij b,c
a
b
c

Department of Signal Processing and Communications, Universidad de Alcal, Alcal de Henares, Spain
CARMABI Research Station, Willemstad, Curaao
Department of Aquatic Microbiology, Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, Science Park 700, 1098 XH, Amsterdam, The Netherlands

a r t i c l e

i n f o

Article history:
Received 11 November 2015
Received in revised form 26 January 2016
Accepted 27 January 2016
Available online 6 February 2016
Keywords:
Computational Intelligence techniques
Coral reefs
Applications
Algorithms

a b s t r a c t
Studies on coral reefs increasingly combine aspects of science and technology to understand the complex dynamics and processes that shape these benthic ecosystems. Recently, the use of advanced computational algorithms
has entered coral reef science as new powerful tools that help solve complex coral reef related questions, which
were unsolvable just a decade earlier. Some of these advanced algorithms consist of Computational Intelligence
(CI) approaches, a branch of Articial Intelligence that uses intelligent systems to address complex real-world
problems yielding more robust, tractable and simpler solutions than those obtained by conventional mathematical techniques. This paper describes the most commonly used CI techniques related to coral reefs and the main
improvements obtained with these methods over classical algorithms in this eld. Some recommendations are
given for the application of CI techniques to complex coral reef related problems, and vice-versa, for the application of novel coral reef dynamics concepts to improve the Coral Reef Optimization (CRO) algorithm, an optimization method inspired by coral reef dynamics.
2016 Elsevier B.V. All rights reserved.

Contents
1.
2.

3.

4.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
A brief introduction to computational intelligence . . . . . . . . . . . . . . .
2.1.
Neural computation approaches . . . . . . . . . . . . . . . . . . .
2.1.1.
Multi-layer perceptron (MLP) . . . . . . . . . . . . . . . .
2.1.2.
Support vector machine . . . . . . . . . . . . . . . . . . .
2.2.
Evolutionary computation approaches . . . . . . . . . . . . . . . . .
2.2.1.
Genetic algorithms . . . . . . . . . . . . . . . . . . . . .
2.2.2.
Simulated annealing . . . . . . . . . . . . . . . . . . . . .
2.2.3.
Particle swarm optimization algorithm . . . . . . . . . . . .
2.2.4.
A coral reef-based algorithm for optimization problems . . . . .
2.3.
Fuzzy computation . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1.
Fuzzy logic . . . . . . . . . . . . . . . . . . . . . . . . .
Computational intelligence in coral reef applications . . . . . . . . . . . . . .
3.1.
Coral reef data. The need for CI techniques . . . . . . . . . . . . . . .
3.2.
Review based on the CI technique used . . . . . . . . . . . . . . . .
3.2.1.
Neural network applications in coral reef problems . . . . . . .
3.3.
SVM applications for coral reefs . . . . . . . . . . . . . . . . . . . .
3.4.
Evolutionary-type approaches in reef-related applications . . . . . . . .
3.5.
Fuzzy computation in coral reef applications . . . . . . . . . . . . . .
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.
The application of CI techniques in mapping-related problems is coral reefs
4.1.1.
Monitoring . . . . . . . . . . . . . . . . . . . . . . . . .
4.1.2.
Modeling . . . . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

108
109
110
110
110
110
110
111
111
112
113
113
113
113
114
114
115
115
116
116
116
117
117

Corresponding author at: Department of Signal Processing and Communications, Universidad de Alcal, 28871 Alcal de Henares, Madrid, Spain. Tel.: +34 91 885 6698; fax: +34 91
885 6699.
E-mail address: sancho.salcedo@uah.es (S. Salcedo-Sanz).

http://dx.doi.org/10.1016/j.ecoinf.2016.01.008
1574-9541/ 2016 Elsevier B.V. All rights reserved.

108

4.2.
Answering the research questions
5.
Conclusions . . . . . . . . . . . . .
Appendix A.
Supplementary data . . . . .
References . . . . . . . . . . . . . . . .

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

1. Introduction
Coral reefs are one of the ecologically, biologically and physically
most complex ecosystems on Earth (Ferrario et al., 2014; Kline et al.,
2015; Knowlton and Jackson, 2013; Lamy et al., 2015; Roberts, 2009).
Corals enlarge three dimensional habitat complexity (Schoening et al.,
2012) and play a key role in deep-sea epibenthic megafauna
(animals N 1 cm) (Dunlop et al., 2015), which capture carbon through
the redistribution of nutrients such as organic matter and oxygen
(Bett et al., 2001; Dunlop et al., 2015; Ruhl, 2007; Schoening et al.,
2012). Coral reefs are thus hot spots of marine biodiversity (Purser,
2015; Shantz et al., 2015; Tittensor et al., 2010), provide important
ecosystem services (Dulvy and Kindsvater, 2015; Purser et al., 2013a),
support the livelihood of millions of people (Ault et al., 2014; Burke
et al., 2011; Chen et al., 2015; Kittinger et al., 2015), and act as early
warning indicators of global climate change (Baker et al., 2008;
Freeman, 2015; Mooney et al., 2009; Woodroffe and Webster, 2014).
Climate change is probably the most dangerous threat to coral reefs
(Baker et al., 2004; Toth et al., 2015) because of the rise of thermal stress
events (which increase coral mortality (Eakin et al., 2010)) and the
increment of severe storms and ocean acidity (Anthony et al., 2008;
Hoegh-Guldberg et al., 2007) (which causes bleaching (Wooldridge
and Done, 2004) in tropical corals, damage reef structure and reduce
coral growth rates (Hoegh-Guldberg et al., 2007; Hooidonk et al.,
2014; Manzello et al., 2013)). In addition to climate change, anthropogenic local stressors such as overshing (which reduces and even
exhausts key species from the ecosystem (Januchowski-Hartley
et al., 2015; Mumby et al., 2006; Wilson et al., 2006)), mechanical
damage from shing (Clark and Rowden, 2009; Fossaa and Skjoldal,
2010; Orejas et al., 2009), offshore oil and gas industry (Allers et al.,
2013; Gates and Jones, 2012; Larsson and Purser, 2011; Larsson et al.,
2013; Pabortsava et al., 2011; Purser, 2015; Purser and Thomsen,
2012), sediments (Ban et al., 2014; Bartley et al., 2014; Kroon et al.,
2014; Larsson and Purser, 2011; Yamazaki et al., 2011), and marine
litter (Pham et al., 2014) are also degrading coral reef ecosystems.
These combined stressors, which have undermined the resilience
(Anthony et al., 2011; Hughes et al., 2003, 2007; McClanahan et al.,
2012; Putra et al., 2015) of coral reef-based ecosystems (Hughes et al.,
2010; Pratchett et al., 2014; Rowlands et al., 2015), operate at multiple
scales (Carilli et al., 2009; McClanahan et al., 2014), which range from
meters to thousands of kilometers (Hatcher, 1997). Monitoring the
evolution of coral reef ecosystems at a decadal scale is necessary to
understand and predict their dynamics and to design tools for coral
reef management (Scoplitis et al., 2009). Monitoring is carried out via
localized in situ observations, time-series of aerial photographs and
remotely sensed images.
Because of the combination of all these factors, the study of coral
reefs requires multi-disciplinary approaches that combine aspects of
eld observations (Andrew and Mapstone, 1987; Foster et al., 1991;
Leujak and Ormond, 2007; Pielou, 1974; Purser, 2015; Whorff and
Grifng, 1992), ecological theory (Fox and Bellwood, 2014), modeling
(Harborne et al., 2006; Webster et al., 2007) and simulation
(Langmead and Sheppard, 2004), and increasingly more often Computational Intelligence (CI) techniques (Bandyopadhyay et al., 2009; Beijbom
et al., 2012; Beneld et al., 2007; Chang et al., 2014; Collin and Planes,
2012; Elawady, 2015; Elith et al., 2006; Gao and Hailu, 2012, 2013;
Guinan et al., 2009; Halide and Ridd, 2002; Henriques et al., 2010;
Huang et al., 2011a; Johnson-Roberson et al., 2006, 2007; Juillet-Leclerc,
2006, 2007, 2009; Juillet-Leclerc and Thiria, 2007, 2008; Juillet-Leclerc

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

117
118
118
118

et al., 2006, 2007; Knudby et al., 2010, 2013; Leslie et al., 2003; Marcos
et al. 2005; Meesters et al., 1998; Mehta et al., 2007; Nagamani et al.,
2012; Naveau et al., 2004; OConnor, 2000; Padmavathi et al., 2010;
Pican et al., 1998; Pittman et al., 2009; Purser et al., 2008, 2009;
Ruitenbeek et al., 1999; Schoening et al., 2012; Shihavuddin et al., 2013;
Tong et al., 2013; Wahidin et al., 2015; Wang et al., 2004; Watts et al.,
2011; Wooldridge and Done, 2004; Yamamoto and Sugiura, 2002;
Zhang, 2015). Historically, data-driven approaches have been most
commonly used in coral reef research, but the rise of big data
approaches (Kemp and Sadler, 2014) has rendered some traditional
forms of data-analyses insufcient (Andrew and Mapstone, 1987;
Leujak and Ormond, 2007; Pielou, 1974) and novel computational
techniques are increasingly used to address this problem.
The current huge amount of coral reef data arises mainly from two
technologies whose capabilities are continuously being improved:
remote sensing and underwater vehicles. On the one, coral reef remote
sensing techniques (Goodman et al., 2013) both aerial imagery from
satellites or aircrafts (Shihavuddin et al., 2013; Xu and Zhao, 2014)
and bathymetric data via Airborne Light Detection and Ranging
(LiDAR) (Pittman et al., 2009), ship-based Multi-Beam Echo-Sounder
(MBES) and Sound Navigation and Ranging (SoNAR) (Costa et al.,
2009) are able to provide repeatable observations in large areas
(Lucas and Goodman, 2014; Maina et al., 2008; Rowlands, 2013;
Rowlands et al., 2012; Xu and Zhao, 2014) to quantify proxies of biodiversity (Shihavuddin et al., 2013; Turner et al., 2003), and has become a
relevant additional tool to in situ approaches (Mumby et al., 2004a;
Scoplitis et al., 2009, 2010; Shihavuddin et al., 2013; Wang et al.,
2007). On the other hand, underwater vehicles Remotely Operated
Vehicles (ROVs) (Neves et al., 2014; Purser, 2015), Deep Sea Crawler
(Purser et al., 2013b), autonomous underwater vehicles (AUVs), and
mobile robots (Thomsen et al., 2015) allow image acquisition at
depths and over spatial scales that are not possible with divers. Underwater vehicles provide large amounts of high spatial resolution imagery,
from which benthic organisms under study can be resolved
(Shihavuddin et al., 2013), (Turner et al., 2003). Both data acquisition technologies require increasingly the use of CI techniques, not
only because of the large amount of data and the difculty to classify
or detect biota from data (the classical manual analysis is both timeconsuming and labor-intensive) but also because of the need to
extract non-obvious relationships among the many factors that
interact with each other, and which arise from the inherent complexity of coral reefs mentioned in the rst paragraph.
CI is a branch of Articial Intelligence (AI), focused on the design of
robust and intelligent systems to tackle complex real-world problems
for which traditional approaches fail or are inefcient. For example, in
optimization problems, traditional techniques such as Newton-type
algorithms need continuous and derivable objective functions to be
applied, while CI techniques such as evolutionary computation (EC)
algorithms do not need this requisite. There are other problems such
as regression models (RMs) or classication approaches where CI techniques are known to obtain better results than traditional algorithms.
For example, in regression problems, traditional regression (LR) analysis
(Montgomery et al., 2012; Seber and Lee, 2012) or multi-linear regression
(MLR) models (Chatterjee and Hadi, 2015) are usually considered a baseline algorithm to be beaten by CI techniques. In classication problems,
the traditional nave k-nearest neighbors approach (Soriano et al., 2001)
is usually improved by CI techniques such as Support Vector Machines
(SVMs) or neural classiers. In this respect, note that CI is not a single
technique, but a set of techniques belonging to different subelds such

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

as neural computation (neural networks (Bishop, 1995, 2006; Haykin and


Neural Network, 2004), Extreme Learning Machines (Cambria et al.,
2013; Gao et al., 2015; Huang, 2014; Huang et al., 2006, 2010, 2011b,
2012, 2014), Support Vector Machines (Chang and Lin, 2011; Cristianini
and Shawe-Taylor, 2000; Hearst et al., 1998; Scholkopf and Smola, 1998,
2001, 2002; Smola and Schlkopf, 2004; Steinwart and Christmann,
2008), etc.), evolutionary computation (Eiben and Smith, 2003, 2008;
Yu and Gen, 2010) (genetic algorithms (Affenzeller et al., 2009; Man
et al., 2012; Mitchell, 1998; Randy, 2004; Simon, 2013), Genetic Programming (Banzhaf et al., 1998; Koza, 1992; Langdon, 2012; Poli and Koza,
2014; Poli et al., 2008), etc.) or Fuzzy Computation (Bezdek, 2013; Ross,
2009; Yager and Zadeh, 2012; Zadeh, 1965, 2005, 2008), to name a
few. These techniques have so far produced excellent results in dataanalytics applications (data-driven problems), including research questions addressing a variety of aspects of coral reefs, such as mapping reefs
(Knudby et al., 2010; Stockwell, 1999; Watts et al., 2011; Zhang, 2015),
coral detection (Dolan et al., 2008; Purser et al., 2009), coral reef classication (Bandyopadhyay et al., 2009; Henriques et al., 2010; Mehta et al.,
2007), modeling habitat suitability (Tong et al., 2013), model prediction
(Huang et al., 2011a), or design coral reef marine reserves (Leslie et al.,
2003), to name a few. In turn, the aforementioned extraordinary complexity of coral reefs has recently inspired a novel CI technique known
as Coral Reef Optimization (CRO) algorithm (Li et al., 2015; Medeiros
et al., 2015; Ortiz-Martin, 2014; Salcedo-Sanz et al., 2013, 2014a, 2014b,
2014c, 2014d, 2014e, 2014f, 2015a, 2015b; Yang et al., 2015), which has
been successfully applied to a number of optimization problems in energy
(Salcedo-Sanz et al., 2014b, 2014c, 2014d, 2015a) or telecommunications
(Salcedo-Sanz et al., 2014e, 2014f). The application of CI concepts to coral
reef related problems and, vice-versa, the use of coral reef concepts to
general-purpose optimization problems in other elds are on the rise.
This is shown in Fig. 1, which represents, as a function of the publication
year, the number of research works that use CI techniques in coral reef
problems (blue bars) and those that use coral reef concepts via CRO
algorithm (red bars) to solve general-purpose optimization problems.
Tables 1 and 2 list, respectively, the related publications and the acronyms
used in this paper.
Motivated by different research questions, as detailed in the following
paragraph, there are several compelling reasons why it could be useful to
apply CI techniques to coral reef-related problems:
(1) What are the fastest and most accurate methods to detect in images
different species hosted in coral reefs? Classical manual approaches
(Andrew and Mapstone, 1987; Pielou, 1974) have the advantage

109

of being very accurate due to the work of human experts, but


suffer from the disadvantage of being very time-consuming and
labor-intensive (Andrew and Mapstone, 1987; Foster et al.,
1991; Leujak and Ormond, 2007; Purser et al., 2009; Whorff
and Grifng, 1992). As will be shown throughout this survey,
CI methods can effectively solve these problems both in fully
automatic and semi-automated approaches. For instance, a
novel automated method based on CI techniques to estimate
coral coverage has been proposed in Purser et al. (2009), showing
its superior performance compared to manual image analysis.
(2) What are the most suitable methods for mapping and classifying coral
reef ecosystems? The estimation of live coral is difcult to map by
using remote sensing images because of the high spectral
similarity between corals and other biota such as algae and
seagrass (Hochberg and Atkinson, 2003; Xu and Zhao, 2014)
and due to the high dependence on factors such as the presence
of spectrally similar substrates, water depth, and turbidity, to
name a few (Hochberg et al., 2003a; Knudby et al., 2010;
Mumby et al., 2004b). Although some works (Isoun et al., 2003;
Joyce, 2005; Newman et al., 2007) have had partial success in
mapping live coral distribution by using classifying algorithms
based on spectral indices (Joyce, 2005) or on the percentage of
live coral cover (Isoun et al., 2003; Newman et al., 2007), nevertheless, CI techniques have been found to be useful to overcome
problems involving the detection of coral reefs from reectance
images (Bandyopadhyay et al., 2009; Xu and Zhao, 2014) or
from video images (Marcos et al., 2005).
(3) What are the most suitable methods for modeling (prediction, reconstruction, etc.) coral ecosystems? What are the most used methods to
model and design marine reserves? What are the best methods to
help decision making in this regard? As will be shown, CI methods,
in particular, fuzzy logic are very useful to model economic
policies (Ruitenbeek et al., 1999) and management options (Gao
and Hailu, 2012, 2013) in coral reefs, assisting in making more
informed policy decisions.
(4) What concepts of coral reefs can be applied to solve general-purpose
optimization problems?

Aiming at answering these research questions the purpose of this


paper is to provide an overview of those research studies that have
used CI techniques to study coral reefs and outline future research that
could further improve CI techniques for coral reef-related problems, and
vice versa, to improve the CRO algorithm by introducing more coral-reef
related concept. The structure of this paper is as follows: Section 2 starts
by overviewing some of the most important CI algorithms. Section 3
discusses the use of CI techniques on coral reef-related problems based
on the most recent literature. Finally, Section 5 completes the paper
with a summary of the main conclusions that can be drawn based
on this review.
2. A brief introduction to computational intelligence

Fig. 1. Time evolution of published works. Blue bars represent research contributions using CI
techniques to solve coral reef related problems. For the sake of clarity, references have been
listed in Table 1 along with the CI method used. Red bars represent the yearly number of
works that apply coral reef concepts via the CRO algorithm to general-purpose
optimization problems.

Computational Intelligence is a branch of Articial Intelligence (AI)


focused on the design of algorithms able to nd approximate solutions
for complex problems in science and engineering, which are unsolvable
by traditional techniques, or where traditional approaches produce low
quality solutions. Many of the CI algorithms are inspired by natural phenomena (bio-inspired) ranging from processes such as evolution
(leading to evolutionary computation (Simon, 2013)), social organisms
(inspiring for example Ant Colony Optimization Algorithms (Dorigo and
Birattari, 2010) or Swarm Intelligence (Yang et al., 2013)), the human
brain, resulting in the articial neural network (ANN) paradigm
(Bishop, 2006) or the way in which humans think (leading to Fuzzy
Computation (FC) (Ross, 2009)). For the sake of clarity Table 2 lists
the acronyms used in this survey. As will be shown in Section 3, CI

110

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

approaches have been successfully applied to an important number


of coral reef-related problems, in which techniques such as neural computation (Subsection 2.1), evolutionary computation (Subsection 2.2),
and Fuzzy Computation (Subsection 2.3) represent the main algorithms
that so far have been applied. The following subsections will introduce
these most commonly used algorithms and the underlying rationale.
For further mathematical details, the interested reader is referred to
the Supportive information Section 5.
2.1. Neural computation approaches
An articial neural network (ANN), or simply, a neural network
(NN) is a computation algorithm inspired by the way the human brain
works. The algorithm is able to generalize by learning from representative examples and subsequently solve a somewhat similar problem it
has never encountered before. In this respect, an ANN mimics human
brain connectivity as a parallel and distributed information processing
system based on single processing units that are called neurons for
analogy. Multi-Layer Perceptrons (MLPs), a particular implementation
of ANNs, are by far the most commonly used neural computation
techniques used for coral reef applications (e.g., (Bandyopadhyay
et al., 2009; Henriques et al., 2010)) together with kernel methods
(Elisseeff and Weston, 2001) such as Support Vector Machines (SVMs)
(Schoening et al., 2012).
2.1.1. Multi-layer perceptron (MLP)
An MLP consists of a number of computing single elements
(neurons) whose connections can be adjusted by a learning algorithm
taking into account pairs of input and output samples, in what is called
a learning process. This enables the algorithm to predict (or classify)
novel samples different from those used in the learning process. In its
design stage, an MLP is trained and validated by using, respectively, a
training set (for which the input and the corresponding outputs are
known), and a validation set (whose samples have not been used in
the training process). The nal performance is computed over a test
set that was not used during the design stage (Bishop, 1995; Haykin,
1998). The MLP is able to learn in the sense that, after the training
process, it is able to predict (or classify) samples different from those
used in the learning process, what makes MLPs universal approximators
of many functions. MLPs have been successfully applied to a great amount
of nonlinear prediction and classication problems (Bishop, 1995;
Haykin, 1998). For further details, the interested reader is referred
to the Supportive information Section 5.
The LevenbergMarquardt method is the most commonly used
algorithm to train the MLP (Hagan and Menhaj, 1994). A novel and
much faster learning method for a MLP structure is a so-called Extreme
Learning Machine (ELM) (Huang and Chen, 2007, 2008; Huang et al.,
2010, 2011b, 2012).
2.1.1.1. The ELM concept for neural network training. The key characteristic of ELM training is that it starts by randomly setting the network
weights and then computes the inverse of the hidden-layer output matrix (instead of following a procedure to tune the weights by minimizing
a given objective function). The main benets of this technique are its
simplicity, which makes the training algorithm extremely fast, and its
exceptional efciency, even superior, in some cases, to more conventional approaches such as classical MLPs or Support Vector Machines.
Additionally, the universal approximation capability of ELMs, along
with its classication ability, has been already proven (Cambria et al.,
2013; Huang et al., 2010, 2011b, 2012, 2014). See Section 5 for further
details.
2.1.2. Support vector machine
A Support Vector Machine (SVM) is a modern machine-learning
method that belongs to the general family of the kernel methods. These
techniques employ linear algorithms by constructing hyper-planes in

a high or even innite-dimensional space (Vapnik, 2013). When using


linear algorithms, well-established theory and efcient computational
solution methods are often available. Kernel methods exploit this fact
by embedding the initial data set S dened over an input or attribute
space X (SX ) into a higher (possibly innite) dimensional Hilbert
space H, or feature space. Then they build a linear algorithm therein
which results in an algorithm which is nonlinear relative to the input
data space. The mapping function is denoted as : X H. Linear algorithms benet from this mapping because of the higher dimensionality
of the feature space, but the computational loads dramatically increase
because sample coordinates should be computed in multidimensional
space. This computational load can be avoided through the use of the
so called kernel trick, by which, if an algorithm can be expressed with
dot products in the input space, its (nonlinear) kernel version only
needs the dot products among mapped samples. Kernel methods compute the similarity between training samples S = {xi}N
i=1 using pair-wise
inner products between mapped samples, and thus the so-called kernel
matrix Kij = K(xi, xj) = (xi), (xj) contains all the necessary information to perform many classical linear algorithms in the feature space. For
further details, the interested reader is referred to the Supportive
information that can be found in Section 5.
2.2. Evolutionary computation approaches
Evolutionary computation is inspired by the principles of Natural
Selection and the fact that only the ttest individuals survive. In coral
reef studies, the most commonly used approaches that belong to the
large family of evolutionary-type computation approaches are the
genetic algorithms (GAs) and simulated annealing (SA).
2.2.1. Genetic algorithms
A GA (Eiben and Smith, 2003; Golberg, 1989) is an optimization and
search technique that does not require derivative information. It is able
to deal with a large number of variables. It can provide a global solution
for multi-local extrema problems, optimize functions with continuous
or discrete variables and optimize variables with extremely complex
cost surfaces.
In essence, a GA is based on three main concepts (Eiben and Smith,
2003; Golberg, 1989):
(1) Encoding the candidate solutions (individuals).
(2) Generating an initial population of candidate solutions.
(3) Applying genetic operators (selection, recombination or crossover,
and mutation).

These concepts mimic the way organisms experience selective


forces in Nature. In Nature the creation of novel genetic coding (via
mutations) affects an organism's ability to survive. If mutations yield
better adapted organisms, their probability of survival increases as well
as the likelihood of passing their genetic codes on to a next generation.
In a GA, the selection operator evaluates the tness of individuals in a
population and selects for the one (the candidate solution to the problem)
that best solves the problem at hand (with the lowest error). Mutations
(random variations) also occur in the GA, leading to the appearance of
individuals with novel characteristics, different from those of their
parents. If the novel attribute(s) makes progeny better adapted to the
environment (i.e., it is more successful at solving the problem at hand),
the probabilities of survival and having descendants also increase. Part
of the offspring inherits the novel chromosomes, and thus the corresponding external characteristic codied by them. In this way, the initial
population of individuals evolves and, after a number of generations,
the aforementioned process results in the creation and multiplication of
individuals best adapted to the environment (better solutions to the
problem), whereas individuals representing bad solutions go extinct.

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

111

Table 1
List of works using CI in coral reef related problems. See Table 2 for acronyms.
Ref.

CI method

Goal

Input

Meesters et al. (1998)


Pican et al. (1998)
Ruitenbeek et al. (1999)
OConnor (2000)
Halide and Ridd (2002)
Yamamoto and Sugiura (2002)
Leslie et al. (2003)
Naveau et al. (2004)
Wooldridge and Done (2004)
Wang et al. (2004)
Marcos et al. (2005)
Juillet-Leclerc et al. (2006)
Juillet-Leclerc (2006)
Johnson-Roberson et al. (2006)
Mehta et al. (2007)
Juillet-Leclerc and Thiria (2007)
Juillet-Leclerc (2007)
Juillet-Leclerc et al. (2007)
Johnson-Roberson et al. (2007)
Beneld et al. (2007)
Juillet-Leclerc and Thiria (2008)
Purser et al. (2008)
Guinan et al. (2009)
Bandyopadhyay et al. (2009)
Purser et al. (2009)
Juillet-Leclerc (2009)
Pittman et al. (2009)
Henriques et al. (2010)
Knudby et al. (2010)
Padmavathi et al. (2010)
Huang et al. (2011a)
Watts et al. (2011)
Collin and Planes (2012)
Nagamani et al. (2012)
Gao and Hailu (2012)
Schoening et al. (2012)
Beijbom et al. (2012)
Shihavuddin et al. (2013)
Tong et al. (2013)
Gao and Hailu (2013)
Knudby et al. (2013)
Chang et al. (2014)
Zhang (2015)
Elawady (2015)
Wahidin et al. (2015)

FL
SOM
FL
FL
FL
NN
SA
NN
BBN
FL
NN
NN
NN
SVM
SVM
NN
NN
NN
SVM
FL
NN
SOM
GARP
NN
SOM
NN
BRT
SVM
SVM
PNN
SVM
NN
SVM
NN
AHP-fuzzy
SVM
SVM
NN
MaxEnt
FL
RF
GA
SVM
NN
SVM

Predict coral reef development


Texture analysis
Optimization of economic policies
Coral reef development
Modeling coral bleaching
Prediction of coral growth
Design of marine reserve networks
SST and SSS reconstructions
SST, SSS, and coral data
Biota classication
Classifying coral reef images
SST and SSS reconstructions
SST and SSS reconstructions
Coral segmentation and classication
Coral reef texture classication
SST and SSS reconstructions
SST and SSS reconstructions
SST and SSS reconstructions
Segmentation and classication
Mapping coral distribution
SST and SSS reconstructions
Detection of CWC habitats
CWC habitat
Coral cover classication
Detection of CWC habitats
Independent SST and SSS reconstructions
Predict diversity of sh and corals
Classication
Predictive mapping
Detection and classication
Predictive models
Mapping (locate reefs)
Coral health detection
Classication
Management strategies
Semi-automated image analysis
Mapping (coral coverage estimation)
Classication
Modeling habitat suitability
Modeling (management)
Mapping coral reef resilience
Mapping
Habitat mapping
Classication
Habitat mapping

Suspended particulate
ROV underwater imagery
Sedimentation, nutrient inux
Suspended particulate
Temperatures
Coral variables
26 habitat data
SST, SSS, and coral data
Input
Spectral information
Underwater images
SST, SSS, coral data
SST, SSS, and coral data
Images
Underwater imagery
SST, SSS, and coral data
SST, SSS, and coral data
SST, SSS, and coral data
Images
Images
Input
ROV images
Bathymetric data
Reectance patterns
Video transect
SST, SSS, and coral data
LiDAR bathymetry
Spectral data from images and bathymetric data
Satellite images
Underwater images
25 marine environmental variables
Bathymetry data
Multispectral image
Bathymetry
Benthos percentage
Underwater images
Underwater images
Underwater images
Terrain variables
Recreational shing data
High spatial resolution satellite data
Remote sensing reectance
Hyperspectral data and aerial photography
Underwater images
Underwater images

2.2.2. Simulated annealing


SA is also a meta-heuristic algorithm that has been widely applied
to solve combinatorial optimization problems (Kirkpatrick, 1984;
Kirkpatrick et al., 1983). It was inspired by the physical process of
heating a substance and then cooling it slowly till a strong crystalline
structure is obtained. In this sense, it is not truly a bio-inspired
approach, but a physics-based algorithm. This process is simulated by
lowering an initial temperature in slow stages until the system reaches
equilibrium and no more changes occur. SA yields a single solution
(conguration of the system), and each state of the algorithm consists
in changing the conguration several times, until a thermal equilibrium
is reached and a new stage starts, with a lower temperature. The
solution of the problem is the conguration obtained in the last stage.
In a standard SA, the changes in the conguration are performed as
follows:
A new conguration is built by random displacement of the current
one. If the new conguration is better, then it replaces the current one,
and if not, it can replace the current one probabilistically. This probability of accepting worse congurations is usually based on Boltzman-type
probabilistic functions. This probability of replacement is high in the
beginning of the algorithm and decreases in every stage, so at the last
stages of the algorithm it is quite unlikely to move towards worse

congurations (solutions to the problem). Note that this procedure


allows the system to move toward the best conguration, and the fact
that at a given point of the algorithm. Although SA is not guaranteed
to nd the global optima, it is still better than other algorithms in escaping
from local optima because it is able to accept worse congurations at
given points of the algorithm. The solution found by SA can be considered
a good enough solution, but it is not guaranteed to be the best, just like
what occurs in GA.
2.2.3. Particle swarm optimization algorithm
Particle swarm optimization (PSO) is a population-based stochastic
optimization technique developed by Kennedy and Eberhart (1995),
inspired by bird ocking and sh schooling. A PSO system is initialized
with a population of random solutions, and searches for the optimal
one by updating the population over several generations. PSO has
no evolution operators, such as crossover and mutation as genetic
algorithms do, but potential solutions instead, called particles, which
y through the problem search space to look for promising regions
according to its own experiences and experiences of the whole group.
Thus, social information is shared, and individuals prot from the
discoveries and previous experiences of other particles in the search.
The PSO is considered a global search algorithm.

112

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

Table 2
List of acronyms used in this paper.
Acronym

Meaning

AHP
AI
ANN
AUV
BBN
BRT
CI
CRO
CWC
DSS
EC
ELM
ENFA
ENSO
FC
FL
FS
GA
GAM
GARP
GM
H2SOM
k-NN
LiDAR
LR
MaxEnt
MBES
MLC
MLP
MPAs
MLR
NN
NTA
OBIA
PNN
POM
PSO
ROVs
RFs
RM
RSP
SA
SDM
SOM
SoNAR
SSS
SST
SVM
SVR

Analytical Hierarchy Process


Articial Intelligence
Articial neural network
Autonomous Underwater Vehicle
Bayesian Belief Network
Boosted regression tree
Computational Intelligence
Coral Reef Optimization algorithm
Cold-water coral
Decision Support Systems
Evolutionary computation
Extreme Learning Machine
Ecological Niche Factor Analysis
EI Nio Southern Oscillation
Fuzzy Computation
Fuzzy logic
Fuzzy set
Genetic algorithm
Generalized additive models
Genetic algorithm for Rule Set Production
Greedy Model
Hierarchically Organized Hyperbolic Self-Organizing Map
k-nearest neighbor
Light Detection and Ranging
Linear regression
Maximum Entropy
Multi-Beam Echo-Sounder
Maximum Likelihood Classication Algorithm
Multi-Layer Perceptron
Marine Protected Areas
Multi-linear regression
Neural network
Newton-type algorithms
Object-based image analysis
Probabilistic neural network
Particulate Organic Matter
Particle swarm optimization
Remotely Operated Vehicles
Random Forests
Regression model
Reserve Selection Problem
Simulated annealing
Species Distribution Model
Self-Organizing Map
Sound Navigation and Ranging
Sea surface salinity
Sea surface temperature
Support Vector Machine
Support Vector Machines for Regression

2.2.4. A coral reef-based algorithm for optimization problems


Coral reefs have recently inspired a new algorithm, the so called
Coral Reef Optimization (CRO) algorithm (Salcedo-Sanz et al., 2014a),
a kind of evolutionary-type approach with excellent convergence
properties, and potential to include many processes that occur in real
coral reefs to tackle different optimization processes. The CRO is an
optimization algorithm, based on the articial modeling of a coral reef
as a N  M square grid . Each of the squares (i, j) compounding is
able to allocate a coral (or colony of corals) i , j, which is encoded as a
string of numbers in a given alphabet I so as to represent a possible
solution to the optimization problem under consideration. This square
grid is rst initialized at random by establishing some squares in to
be occupied by corals (i.e. solutions to the problem) and some other
squares in the grid to be empty, namely, vacant holes in the reef
where new corals may settle and grow in the future. The rate between
free and occupied squares in at the beginning of the algorithm is an
important parameter of the CRO algorithm, which is denoted as
with 0 b b 1. Each coral is associated to a function f ij : IR that
models its health status and corresponds to the problem's objective

function. The CRO is based on the fact that reefs endure as long as
healthier corals better solutions to the problem at hand survive
even if less healthy corals perish.
Algorithm 1. Pseudo-code for the CRO algorithm

After the reef initialization described above, corals' reproduction


in the reef is articially simulated in the CRO algorithm by sequentially
applying different operators onto in an iterative basis until a given
stop criterion is met. These operators model the sexual reproduction
capability of corals (broadcast spawning and brooding), asexual reproduction (budding and fragmentation), as well as catastrophic events
in the reef such as polyp depredation. After the sexual reproduction
and asexual reproduction have taken place, the resulting set of larvae
(i.e. new solutions to the problem) try to allocate themselves a place
in the reef to settle and grow. This allocated space may be free or
busy, the latter by ghting against the coral actually located in that
place. If larvae do not succeed to be recruited on the coral after a given
number of attempts, they are assumed to be predated. Algorithm 1
illustrates the ow diagram of the CRO algorithm illustrating
the two CRO phases (reef initialization and reef formation), along
with all the operators described above and summarized next for
the sake of completion:
1. Broadcast spawning (external sexual reproduction), which consists of
the following steps:
1.a. At a given step k of the reef formation phase a fraction of the
existing corals is selected uniformly at random to be broadcast
spawners. This fraction will be denoted as Fb. Corals that are
not selected to be broadcast spawners (i.e. 1 -Fb) will reproduce
by brooding later within the algorithm execution.
1.b. Couples are selected from a pool of broadcast spawner corals at
step k. Each of such couples will breed a coral larva by sexual
crossover, which is then released out to the water. Once two
corals have been selected to procreate, they are not chosen anymore at step k (i.e. two corals are parents only once in a given
step). This couple selection can be done uniformly at random
or by resorting to any tness proportionate selection approach
(e.g. roulette wheel).
2. Brooding (internal sexual reproduction): as previously mentioned, at
each step k of the reef formation phase in the CRO algorithm, the
fraction of corals that will reproduce by brooding is 1-Fb. The modeling
consists of the formation of a coral larva by means of a random
mutation of the brooding-reproductive coral (self-fertilization considering hermaphrodite corals). The produced larva is then released out
to the water in a similar fashion than that of the larvae generated in
step 1.b.
3. Larvae setting: once all larvae are formed at step k either through
broadcast spawning (1.) or by brooding (2.), they will try to set and
grow in the reef. First, the health function of each coral larva is
computed. Second, each larva will randomly try to settle on a square
(i, j) of the reef. If the square is empty (free space in the reef), the
coral grows therein no matter the value of its health function. By

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

contrast, if a coral is already occupying the square at hand, the new


larva will set only if its health function is better than that of the
existing coral. We dene a number of attempts for a larva to set
in the reef: after unsuccessful tries, it will be preyed by other
animals in the reef.
4. Budding or fragmentation (asexual reproduction): in this process the
overall set of existing corals in the reef are sorted as a function of
their level of healthiness (given by f(ij)), from which a fraction Fa
duplicates itself and tries to settle in a different part of the reef by
following the setting process described in Step 3. Note that a maximum number of identical corals are allowed in the reef.
5. Depredation in polyp phase: corals may die during the reef formation
phase of the CRO algorithm. At the end of each reproduction step k, a
small number of corals in the reef can be preyed, thus liberating
space in the reef for next coral generation. The depredation operator
is applied with a very small probability Pd at each step k, and exclusively to a fraction Fd of the worse health corals in .
The CRO algorithm has been applied to a number of problems, in
different research areas in engineering, such as wind engineering
(Salcedo-Sanz et al., 2014c, 2015a), solar energy (Salcedo-Sanz et al.,
2014d), telecommunications (Salcedo-Sanz et al., 2014e, 2014f) or
education and computer games (Ortiz-Martin, 2014). In some of these
works, the CRO has been successfully hybridized with other CI
approaches, such as neural networks (Salcedo-Sanz et al., 2014b),
showing an excellent performance in the problem tackled. Variants of
the algorithm for multi-objective problems have also been proposed
(Salcedo-Sanz et al., 2013).
Regarding the next steps in CRO development, it is particularly interesting the possibility of including new processes that appear in real
reefs, such as multi-species or substrates to develop co-evolution
algorithms. The good convergence properties of the algorithm allow
its application in micro-evolution (algorithms with very low number
of generations), for problems with high requirements of computation
time.
2.3. Fuzzy computation
Fuzzy Computation (FC) is inspired by the fact that humans exhibit
the outstanding ability to reason in an environment of incomplete information, uncertainty, and partiality of class membership. Fuzzy logic (FL)
has already been used for coral reef related applications.
2.3.1. Fuzzy logic
Its original concept, proposed by Zadeh (1965), is based on the
concept of a fuzzy set (FS), which plays a key role in fuzzy logic. In
conventional set theory, an element either belongs to a set or it does
not. However, in FS theory, an element can belong to a set with a certain
degree (partial membership). The degree of membership is referred to
as the membership value, and is commonly represented by a real
value in [0, 1], where 0 and 1 correspond, respectively, to full nonmembership and membership. Predicates in FL can thus exhibit
partial degrees of truth, in the same way as elements can have partial
membership in a FS, the grade of truth being represented by a real
number between [0, 1]. These concepts assist us in introducing two
key ideas of FL: graduation and granulation (Zadeh, 2005, 2008).
In FL, everything is allowed to be graduated, i.e., be a matter of degree,
or granulated. To illustrate, the concept of size is granulated when
its values are described as small, medium and big. In this respect,
the key benets of FL are its inclusion of linguistic variables,
i.e., using words or qualitative descriptions instead of numbers)
(Zadeh, 2008), fuzzy if-then rules and the ability to compute with
information described in natural language. These are the fundamental
concepts of Fuzzy Inference Systems (FIS), whose details can be found
in a review (Zadeh, 2008).

113

At this point, we already have the knowledge needed to understand


better the next section, in which we review the use of CI algorithms in
coral reef-related problems.
3. Computational intelligence in coral reef applications
3.1. Coral reef data. The need for CI techniques
As mentioned in the Introduction, the study of coral reefs is a
challenging task not only because of their inherent complexity
(Ferrario et al., 2014; Kline et al., 2015; Knowlton and Jackson, 2013;
Lamy et al., 2015; Roberts, 2009; Schoening et al., 2012) but also
because of their crucial importance in global ecology (carbon capture
(Bett et al., 2001; Dunlop et al., 2015; Ruhl, 2007; Schoening et al.,
2012)) and as spots of marine biodiversity (Purser, 2015; Shantz et al.,
2015; Tittensor et al., 2010). Besides the undoubted scientic interest
of coral reefs and the need to preserve them, coral reef ecosystems provide also substantial economic value to human communities (Ban et al.,
2011; Brander et al., 2007; Cesar, 2002; De Groot et al., 2012; Hedley
et al., 2012; Moberg and Folke, 1999; Sarkis et al., 2013; Subade and
Francisco, 2014; TEEB, 2010; van Zanten et al., 2014). Studying,
protecting and managing coral reefs require information on their composition (spatial and temporal distribution of benthos and substrates)
(Hedley et al., 2012) and evolution. In particular, the adequate management of Marine Protected Areas (MPAs) having coral reefs demands not
only accurate and also relevant information to help scientists and managers predict how a coral reef will react to their strategies to conserve
them and/or to the exploitation of their resources (Lim et al., 2009)
(sh, tourism impact). Non-invasive monitoring of coral reefs and
other benthic communities can be carried out through remote sensing
and underwater images capture by divers and AUVs.
On the one hand, remote sensing is able to provide information
complementary to more traditional methods of data acquisition
(Klemas, 2001; Lim et al., 2009) and have several pros and cons why it
is convenient to use CI techniques: First of all, remote sensing techniques both aerial imagery from satellites or aircrafts (Shihavuddin
et al., 2013; Xu and Zhao, 2014) and bathymetric data via LiDAR
(Pittman et al., 2009) and ship-based multibeam SoNAR (Costa et al.,
2009) are able to provide non-invasive repeatable observations over
large areas (Lucas and Goodman, 2014; Maina et al., 2008; Rowlands,
2013; Rowlands et al., 2012; Xu and Zhao, 2014). Another important
point is the substantial potential of hyper-spectral airborne data
(Hedley et al., 2012) that has been demonstrated in several research
works that focused on discriminating live coral from dead coral
(Mumby et al., 2001; Mumby et al., 2004c), or on distinguishing coral
from macro-algae (Goodman and Ustin, 2007) using bathymetry and
on a radiative transfer model for optical characteristics of water column
(Hedley et al., 2009; Mobley et al., 2005). Moreover, occasionally, live
coral cover remains difcult to map using remote sensing, because of
sub-pixel heterogeneity and high spectral similarity between corals
and other substrate types such as algae and seagrass (Hochberg and
Atkinson, 2000, 2003; Hochberg et al., 2003a, 2004). This is because
the estimation of live coral cover is highly dependent on measurement
conditions such as water depth, turbidity, and the presence of spectrally
similar substrates (Knudby et al., 2010). Satellite imagery has limited
resolution and is less suitable for monitoring slight changes in both
color and shape of coral colonies (Mehta et al., 2007). At sub-meter
scale, images captured by current satellite sensors result in many
image pixels being made up of a blend of biota and substrate
(Leiper et al., 2012). This is one of the reasons why automatic or
semi-automatic approaches using CI are required. Furthermore, the
spatial complexity of the reef itself produces signicant challenges
for remote sensing techniques. Regarding this inherent ecological
complexity, some studies have compared the classication accuracy
of different coral reef maps generated from satellite and aerial
imagery under different measurement conditions or using distinct

114

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

sensor features (spectral or spatial resolution) (Andrfout et al.,


2003; Capolsini et al., 2003; Hochberg and Atkinson, 2003; Lim
et al., 2009; Mumby and Edwards, 2002; Mumby et al., 1997).
These research works have explored the different classication accuracies achieved from the use of hyper-spectral and multi-spectral
imagery (Lim et al., 2009) and between high resolution and lower
resolution images (Andrfout et al., 2003; Capolsini et al., 2003;
Hochberg and Atkinson, 2003; Mumby and Edwards, 2002; Mumby
et al., 1997), between uncorrected images and those that have
been pre-processed aiming at considering the attenuation of
the above water column (Mumby et al., 1997, 1998a, 1998b), and,
those works that have considered airwater interface effects
(Hedley et al., 2005; Hochberg and Atkinson, 2003; Hochberg et al.,
2003a, 2003b). However, in addition to these factors, the inherent
spatial complexity of the reefs themselves exhibits important
challenges for remote sensing. In this respect, (Lim et al., 2009) studies
how the accuracy of a maximum likelihood classier (MLC) is inuenced by the spatial complexity of Caribbean reefs appearing in high
spatial and spectral resolution imagery. However, the CI techniques
allow better solve these problems, as will be seen below.
On the other hand, images from underwater vehicles allow image acquisition at depths and over spatial scales that are not possible with divers (Shihavuddin et al., 2013). Underwater vehicles provide large
amounts of high spatial resolution imagery, from which actual individuals
being studied (benthic organisms) can be resolved (Shihavuddin et al.,
2013; Turner et al., 2003). These are key benets. The drawback is that
analysis and classication of coral reef underwater images are very challenging problems since coral reef images have complex patterns of colors,
textures and shapes (Mehta et al., 2007). A crucial problem in this regard
is that manual classication of species hosted in coral reefs based on these
images is very time-consuming and labor-intensive: The classical point
quadrat method (Pielou, 1974) is based on overlapping the image with
an array of points and quantifying the number of points intersecting the
different species (and/or substrates) within the image. The number of
points in the array is variable so that increasing the number of points improves the accuracy at the expense of boosting the processing time. Another manual approach consists in mapping all the species and
substrates of an image onto a digital overlay, and quantifying the overlap
percentage of each substrate or species (Andrew and Mapstone, 1987).
This second manual method is more accurate than that of Pielou
(1974)) but requires more processing time (Foster et al., 1991; Leujak
and Ormond, 2007; Purser et al., 2009; Whorff and Grifng, 1992).
Automatic methods for mapping and classing coral reef biota based
on underwater images are also a very challenging computer vision
problem: As will be shown throughout this survey, CI methods can
effectively solve this problem. For instance, a novel automated method
based on CI techniques to estimate coral coverage has been proposed
in Purser et al. (2009)), showing its superior performance compared
to manual image analysis. However, classifying coral reef images
taken in situ has inherent difculties because they represent twodimensional images from three-dimensional entities, whose information depends on perspective, and consists of a huge variety of shapes,
colors and textures. This complexity is sometimes a very challenge for
automated techniques, and is the reason why semi-automated image
analysis has been explored in Schoening et al. (2012) using a SVM
system and a small subset of images, in which taxa positions have
been previously labeled by experts.
Thus properly processing the aforementioned large amounts of complex data (photographs, spectra, etc.), extracting valuable information
from them (mapping, classifying organisms) or inferring non-obvious
relationships among the many factors arising from the inherent
complexity of coral reefs are some of the reasons why CI techniques
are being used increasingly. Regarding this, the following Subsection
3.2 reviews in detail the use of CI techniques in coral reef problems,
prior to their critical analysis, which we postpone to Section 4 for
reasons that will be better understood later on.

3.2. Review based on the CI technique used


3.2.1. Neural network applications in coral reef problems
A rst criterion to analyze the use of NN in coral problems is based
on the way these techniques are able to successfully work with data
exhibiting the difculties mentioned in Subsection 3.1. One of these
data-related problems is the similarity among reectance patterns of
coral and other biota and/or substrate from remote sensing satellite.
This can be tackled using NNs. This is the case of the work by
Bandyopadhyay et al. (2009), which used reectance patterns
generated from remote sensing satellite in the Gulf of Kachchh. The NN
was trained using a set of 300 samples per class (coral reefs, algae, mangroves, and seagrasses), these samples consisting of improved images
were correctly labeled by the authors. The NN was then trained using a
back propagation method, and tested by using a cross-validation method
with test images different from those used in the design stage. The probability of correct classication increased from PCC,kappa =0.84 (using conventional kappa coefcient statistics) to PCC , NN = 0.91, proving the
usefulness of NN in this application.
Another data-related problem stated in Subsection 3.1 is that the
classication of coral reef actual images taken in situ suffers form inherent difculties because they represent two-dimensional images from
three-dimensional entities, whose information depends on perspective,
and consists of a huge variety of shapes, colors and textures. This problem has been tackled using NN in Marcos et al. (2005)) to map the coral
reef (Great Barrier Reef) community composition into three benthic
classes: living coral, dead coral, and sand. The NN performance was
compared to that of a rule-based decision tree classier inspired by
the way in which marine scientists classify corals visually. The NN's
probability of classifying a habitat correctly was PCC,NN = 86.5%, which
was higher than that of the rule-based decision approach (PCC , rule =
79.7%). In a similar approach (in the sense that real images are used as
input data, although using a different unsupervised machine learning
approach), the research work by Purser et al. (2009)) has successfully
used a Self-Organizing Map (SOM) (Kohonen, 1990) to detect coldwater corals (CWCs) based on video transect data taken in deep-water
at Tisler Reef (Norway). The SOM in Purser et al. (2009)) learnt
to map similar textures to neighboring regions (forming clusters) on a
2-dimensional space (map), these being labeled by experts. The classication accuracy was PCC , NN = 91.4% , slightly better than that of the
Kappa coefcient (PCC, kappa = 90.0%). Purser has further deepened the
advantages of SOMs for detection and classication of CWCs based on
video transect data collected by a ROV from a number of Norwegian
CWC reefs (Purser et al., 2008). The novel contribution is that the
individual images were examined by CWC experts, who labeled them.
For each image a 30-dimensional texture features, being these feature
vectors those used to train the SOM.
One more data-related problem is that color measured from remote
sensing data is usually affected by the depth at which a community
is found as well as by the composition of that community. See
Subsection 3.1 for further details. To overcome this problem, the
framework adopted in Nagamani et al. (2012) used simulated
remote-sensing reectance data instead of color measures as inputs
to an MLP, which was used to classify four benthic bottom categories
(coral, sea grass, green algae and red algae), with a probability of
correctly classifying as high as 98%.
Alternatively, instead of using images (or computed data derived
from them), bathymetric data was used as some of the inputs to the
NN to locate reefs in Kangaroo Island (Australia)(Watts et al., 2011).
Specically, bathymetric measurements (the local slope and curvature
of the seabed) and data containing the location of known reefs (at a
bathymetric depth of less than 30 m) were combined to form the NN
inputs. The probability of accurately locating a reef was PCC, NN = 0.86,
which was much higher than the probability obtained using Cohen's
Kappa statistic (PCC , kappa = 0.63), which was used for comparative
purposes.

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

A second interesting criterion to analyze the use of CI methods to


coral reef problems is the application itself rather than the inuence of
the data processing. For instance, the works above aimed to map
habitats. However, there are many possible applications for which
NN are useful, such as mapping, as in the commented works
(Bandyopadhyay et al., 2009; Marcos et al., 2005; Watts et al.,
2011), coral detection in AUV (Elawady, 2015), or detection of
cold-water coral habitats (Purser et al., 2009). Also in prediction
problems: The prediction of coral growth as a function of underwater
solar radiation, substrate gradient, and current and previous coral
cover (8-year data) was tackled with an MLP (Yamamoto and
Sugiura, 2002), with a correlation coefcient of R = 0.83, better
than that of multiple regression predictor (R=0.80) used for comparative
purposes. The prediction of large-scale coral bleaching (Wooldridge and
Done, 2004) was explored using a Bayesian Belief Network (Siddique
and Adeli, 2013) based on sea surface temperature (SST), in situ data
(topographic and ecological attributes of reef sites), and expert opinion.
Reconstruction problems have also been tackled with NNs, for example
NNs have been used to reconstruct sea surface temperature (SST) and
sea surface salinity (SSS) using coral skeleton data collected in Yasawa
(Fiji Archipelago) and SST and SSS measured from 1960 and 1997
(Juillet-Leclerc, 2006, 2007, 2009; Juillet-Leclerc and Thiria, 2007,
2008; Juillet-Leclerc et al., 2006, 2007; Naveau et al., 2004). Successful
reconstructions were carried out from 1910 to 1960, and suggested
that only four strong EI Nio Southern Oscillation (ENSO) occurred
prior to 1960 (Juillet-Leclerc et al., 2006).
There are two main conclusions that can be inferred from these
works. The rst one is that NNs exhibit a higher accuracy than that of
alternative algorithms such as traditional Cohen's kappa statistics,
used for comparison purposes in different classication problems in
coral reef habitats. The second conclusion is related to the inuence of
input data types (reectance, video images, bathymetry data) on the
NN performance. The conclusion is that NNs exhibit a considerable
better accuracy than that of Cohen's kappa statistics when the inputs
are bathymetric data (PCC , NN = 0.86, much better than PCC , kappa =
0.63), or reectance patterns (PCC,NN = 0.91 vs. PCC,kappa = 0.84).
3.3. SVM applications for coral reefs
SVMs have been used for mapping (Beijbom et al., 2012; Wahidin
et al., 2015; Zhang, 2015), classication (Henriques et al., 2010;
Johnson-Roberson et al., 2006; Mehta et al., 2007; Padmavathi et al.,
2010; Shihavuddin et al., 2013; Zhang, 2015), segmentation and classication (Johnson-Roberson et al., 2007), detection (Collin and Planes,
2012) or semi-automated image analysis (Schoening et al., 2012). As
occurred in NNs, the problem of low quality underwater images in
coral reef classication can also been tackled using SVMs (Mehta et al.,
2007), aiming to classify low quality underwater images into 3 classes
of interest: corymbose Acropora, branching Acropora, and tabulate
Acropora. A key benet is that the classifying algorithm does not require
the estimation of any intermediary feature vector or histogram, as done
in In Kim et al. (2002) and Li et al. (2003). The SVM classier was trained
with 100 images (25 25 square pixels) for each coral type. The SVM
with radial basis function made use of raw pixel color data as input
vectors, and achieved a successful classication probability of 95%. The
difculty of extracting information from remote sense data is common
to (Henriques et al., 2010) and (Knudby et al., 2010). As in NNs,
the problem of distinguishing among classes with similar spectral
signatures have been tackled using the SVM approach. Specically
(Henriques et al., 2010) focused on producing a habitat map from
orbital images of the Coral Reef Environmental Protection Area at
Rio Grande do Norte (Brazil), and achieved an overall accuracy of
89.33%. Predictive maps of sh species richness, sh biomass, and
sh diversity in two different coral reefs in Zanzibar (Tanzania)
have been generated via SVMs using images provided by the
IKONOS satellite (Knudby et al., 2010). SVM performance was compared

115

to that of Bagging (Duda et al., 2012), Random Forests (Breiman, 2001),


Boosted regression trees (Duda et al., 2012), linear model (LM), and
generalized additive models (GAM) (Duda et al., 2012). Based on a 10fold cross-validation strategy repeated 100 times, the study revealed
that the SVM exhibited a very similar performance to that of a Random
Forest approach (in predicting sh diversity) and that of Bagging (in
sh biomass). For all the variables (sh species richness, sh biomass,
sh diversity), the Bagging model performed the best (lowest RMSE
values), and the LM, the worst. Finally, note that data fusion in combination with SVMs has reported very good results. In this respect, SVMs and
data fusion have been recently used by (Zhang, 2015) to map benthic
ecosystems in the Florida Keys. They used pre-processed hyperspectral
data and aerial photography to generate a rst set of fused data, S 1. Aerial
photography and bathymetry data were combined to generate statistics
describing water depth, and to form a second set of data, S 2 . The fused
dataset S F has been obtained based on the sets S 1 and S 2 , leading to
the nal dataset used as inputs for the classiers explored: SVMs,
Random Forest (RF), and k-nearest neighbor (k-NN) classiers. The
nal classication was carried out by an ensemble of the outputs from
SVM, RF, and k-NN classiers. The authors also explored different
strategies to combine the outputs from multiple classiers, such as the
majority vote, Bayesian average method, and fuzzy integral approach.
They showed that the ensemble of SVM, RF, and k-NN classiers fed
with the fused dataset S F (combination of hyperspectral imagery, aerial
photograph, and bathymetry data) obtained the best result, with an
overall accuracy of 89.6% (kappa value =0.84, for comparison), superior
to those achieved by the single classiers SVM (82.2%), RF (88.5%), and
k-NN (88.5%).
The application of SVMs leads, in general, to very good results in
coral reef texture and/or composition classication (95%), similar
to that of Random Forest approaches (in predicting sh diversity in
coral reef habitats), but with much better results that those derived
with kappa statistics alone (Henriques et al., 2010; Knudby et al.,
2010; Mehta et al., 2007; Zhang, 2015). The extent to which SVM
will be accurate depends not only on the problem itself but also on
the nature of SVM inputs (either row images or combinations of
bathymetric data and images). Specically, the combination of
bathymetric data and images result in excellent results when generating maps of coral reef environments. Support Vector Machines exhibit a number of benets when compared to classical supervised
classication methods such as maximum likelihood classication
(MLC) algorithm used in (Andrfout et al., 2003; Lim et al., 2009;
Warren et al., 2015). For instance, in texture classication, SVMs exhibit very good performance even when the number of training samples is limited (Mehta et al., 2007). Another benet is that SVMs
enable the use of raw image data as feature vectors, which is particularly transcendent for random natural textures where geometrical
features are difcult to get (In Kim et al., 2002; Li et al., 2003;
Mehta et al., 2007), as in reef images, which exhibit rich patterns of
textures, shapes and colors.
3.4. Evolutionary-type approaches in reef-related applications
Based on multi-scale terrain variables, a Genetic algorithm for Rule
Set Production (GARP) (Stockwell, 1999), along with Ecological Niche
Factor Analysis (ENFA) (Dolan et al., 2008; Hirzel et al., 2002), and
Maximum Entropy (MaxEnt) modeling (Elith et al., 2011; Phillips
et al., 2006) were used to predict the distribution of two gorgonian
species (Paragorgia arborea and Primnoa resedaeformis) across Rst
reef (Norwegian margin) (Tong et al., 2013). The main purpose of this
study was to elucidate whether or not these models were useful and
accurate enough to predict the habitats suitability for deep-water
gorgonians using terrain parameters derived from bathymetry data.
The rules were encoded as genes, which were mutated at random
to create other possible models (candidates), which described the
potential of a given location for each species to survive. The rules were

116

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

then mutated and evaluated by the GARP. Whether or not a single rule
should be included into the model population was decided by
comparing the predictive accuracy of each iteration to that of the next.
The main conclusions of this work were that the three methods were
able to predict the most adequate habitats for both gorgonian species,
and that these species colonize topographic highs.
The use of GARP has also been investigated, along with 14 other
predictive models (including MLP and SVM, regression models and
others (Huang et al., 2011a)), to predict the distribution of sponge
assemblies on the Australian continental shelf. The research made use
of 25 marine environmental variables (Huang et al., 2011a) and
concluded that MLP, SVM, and GARP algorithms successfully predict
the spatial distribution of sponge abundance, outperforming the results
obtained with regression models.
A novel approach that hybridizes a shallow water semi-analytical
(SSA) model and a GA to identify 5 bottom categories from subsurface
remote sensing reectance measurements was described in (Chang
et al., 2014). The algorithm is designed and tested based on a synthetic
database. The promising results (a probability of correct classication of
80% of the benthic community involved) suggest the viability of this
novel method, but the GA-SSA approach is yet to be tested for real
coral reefs.
Simulated annealing has also been successfully applied to a coral
reef-related problem aiming to design and implement marine reserves
to efciently conserve biodiversity (Leslie et al., 2003). Efciently
means here that the reserve design has to fulll a balance between
protecting a given percentage of habitat types and minimizing the
area and perimeter of the marine reserve to be designed. This is
a constrained optimization problem, the so-called Reserve Selection
Problem (RSP). In the RSP, the objective function is formed by a cost
term (related to the area of each site considered in the reserve) plus a
term related to boundary length. In turn, the problem's constraints are
related to the percentage of each habitat type that requires to be
protected. In particular, a RSP with 26 habitats was tackled using data
from the 9500 km 2 of the Florida Keys National Marine Sanctuary
(Leslie et al., 2003). A comparison between SA and a greedy approach
showed that the CI algorithm obtained excellent results to address this
problem. Specically, the SA approach was able to generate marine
reserves conserving the target habitats percentage while minimizing
the perimeter: 720 km (SA), much lower than that obtained by the
greedy approach (3376 km). Furthermore, from a practical viewpoint,
another key result obtained by (Leslie et al., 2003) was that the SA led
to reserve forms with less perimeter and larger overall areas, which is
supposed to be preferable for sociopolitical reasons.
The key ndings of this review of evolutionary-type approaches
in coral reef problems is that, in general, they work better than more
commonly used classical methods (greedy, regression approaches).
Evolutionary-type approaches have so far been useful to accurately
predict the distribution of species and to the identication of areas
that would be suitable to become designated as reserves.
3.5. Fuzzy computation in coral reef applications
Fuzzy Computation is profusely used to model different effect of
coral reefs. This ranges from the prediction of coral reef development
under nutrient and sediment stress (Meesters et al., 1998) and
modeling coral blenching (Halide and Ridd, 2002) to model economic
policies (Ruitenbeek et al., 1999) and management options (Gao and
Hailu, 2012, 2013) in coral reefs.
The degradation over time that coral reefs in Curaao (and their
corresponding diversity) could suffer, caused by eutrophication and
other anthropogenic stressors, has been also explored using a fuzzy
logic by (Meesters et al., 1998). The fuzzy logic approach made use of
seven input variables (dissolved inorganic nitrogen and phosphate,
suspended particulate matter, maximum colony size, substratum available for colonization, coral cover, and coral diversity) with possible

values (low, medium, and high) into three triangular fuzzy sets. The
boundaries between each fuzzy set along with the corresponding rules
were based on data and expert knowledge. The results showed that
fuzzy logic is useful to describe coral reef processes and the interactions
between variables and assist non experts in taking decisions for a better
management of coral reefs.
The prediction of occurrence of coral bleaching events on Magnetic Island (North Queensland, Australia) was also investigated
using a fuzzy logic framework (Halide and Ridd, 2002), based on
four input variables: 1) seawater temperatures, 2) temperatures
computed (for a specic year) subtracting temperatures of the former year from those of that year, 3) temperatures computed (for a
particular year) by subtracting the mean temperatures of the two
previous years, and 4) temperatures calculated by subtracting the
3 year average before.
Fuzzy logic models were used to evaluate the impact of various
stressors on coral reef health in Montego Bay (Jamaica), and to decide
the most adequate management interventions to be implemented
(Ruitenbeek et al., 1999). Specically, the goal was to identify and
decide among a variety of cost-efcient management approaches the
one that would be most likely to improve the health status of a coral
reef. The fuzzy model used comprised 14 ecological variables (effective
nutrient concentration, sediment deposition, % of algae on available
substrate, etc. (Ruitenbeek et al., 1999)), and classied reef quality
according to 5 possible fuzzy values (low, medium-low, medium,
medium high, and high). The study concluded that, when compared
to conventional economic approaches, the fuzzy logic approach leads
to more accurately models of the complex interactions between
natural and human-made factors, and assists in evaluating the effect
of different management scenarios on coral reef quality before they
had been implemented.
Although at a lesser extent, FL, within the context of Object-Based
Image Analysis (OBIA) (Leon and Woodroffe, 2009), has been used in
(Beneld et al., 2007; Wang et al., 2004) to map the distribution of
coral reefs.
The main conclusion, common to the application above, is that fuzzy
logic is useful, especially, to model different phenomena (development
under nutrient and sediment stress, coral blenching, etc.) and to help
make more informed policy decisions, and to design better management
strategies of coral reef ecosystems.
4. Discussion
As mentioned in Section 1, coral reefs are extremely complex
ecosystems (Bett et al., 2001; Dunlop et al., 2015; Ferrario et al.,
2014; Kline et al., 2015; Knowlton and Jackson, 2013; Lamy et al.,
2015; Purser, 2015; Roberts, 2009; Ruhl, 2007; Schoening et al.,
2012; Shantz et al., 2015; Tittensor et al., 2010) that play a key role
at both local and planetary scale (Baker et al., 2008; Freeman, 2015;
Mooney et al., 2009; Woodroffe and Webster, 2014). Study, recovery
(Gates and Jones, 2012; MacNeil et al., 2015; Morri et al., 2015) and
management (D'Angelo and Wiedenmann, 2014; Lim et al., 2009;
McClanahan et al., 2012; Weijerman et al., 2015) of coral reef ecosystems is a compulsory duty to both politicians and scientists. Aiming at
achieving these goals there are basically three types of generic groups
of tasks (Goodman et al., 2013; Phinn et al., 2006): mapping, monitoring
and modeling. In the following sub-sections we review the use of CI
techniques in these applications.
4.1. The application of CI techniques in mapping-related problems is coral
reefs
Mapping coral reef involves, in broad sense (Goodman et al., 2013),
many research activities such as mapping itself (Beijbom et al., 2012;
Wahidin et al., 2015; Zhang, 2015), habitat mapping (Andrfout
et al., 2003; Call et al., 2003; Capolsini et al., 2003; Maeder et al., 2002;

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

Mishra et al., 2006; Mumby and Edwards, 2002; Stumpf et al., 2003),
coral reef topography (Yamano, 2007), resource stocktaking
(Andrefouet et al., 2004), (Andrfout et al., 2009; Hochberg and
Atkinson, 2008), coral reef productivity (Andrfout and Payri, 2001;
Brock et al., 2006; Hochberg and Atkinson, 2008; Moses et al., 2009),
habitat diversity (Harborne et al., 2006; Mumby, 2001), biodiversity
(Dalleau et al., 2010; Mellin et al., 2009; Mumby et al., 2008), or MPA
planning and evaluation(Beger et al., 2006; Mora et al., 2006;
Rioja-Nieto and Sheppard, 2008). Many works focused on coral reef
mapping from remote sensing data have tackled the problem of coral
classication by using conventional algorithms, such as the k-means
or maximum likelihood classication (MLC) algorithms (Andrfout
et al., 2003; Warren et al., 2015), (Lim et al., 2009), as implemented in
software packages (Goodman et al., 2013) like ENVI (Exelis, 2012) or
ERDAS IMAGINE (Bertels et al., 2008; Capolsini et al., 2003; ERDAS.
ERDAS IMAGINE, 2015), (Harborne et al., 2006). However, as shown
in Table 1, CI techniques are now playing an increasingly important
role and, as will be shown, exhibit, in general terms, better performance.
The ranking is as follows: First, SVMs have been found to be the most CI
technique used for mapping (Beijbom et al., 2012; Wahidin et al., 2015;
Zhang, 2015), classication (Henriques et al., 2010; Johnson-Roberson
et al., 2006; Mehta et al., 2007; Padmavathi et al., 2010; Shihavuddin
et al., 2013; Zhang, 2015), segmentation and classication (JohnsonRoberson et al., 2007), detection (Collin and Planes, 2012), and semiautomated image analysis (Schoening et al., 2012). The second most
used CI techniques in mapping are NNs for mapping (Watts et al.,
2011), classication (Elawady, 2015; Marcos et al., 2005; Nagamani
et al., 2012; Padmavathi et al., 2010), classication and mapping
(Bandyopadhyay et al., 2009), and detection (Purser et al., 2009). FL,
in the context of Object-Based Image Analysis (OBIA) (Leon and
Woodroffe, 2009), has been used in (Beneld et al., 2007; Wang et al.,
2004) to map the distribution of coral reefs. Finally, RF (Knudby et al.,
2013) and GA (Chang et al., 2014) are the less widely used in mapping
and identication of bottom types.
4.1.1. Monitoring
Monitoring coral reefs include various activities such as exploring
habitat change (Dustan et al., 2001; LeDrew et al., 2004; Palandro
et al., 2003, 2008; Schuyler et al., 2006; Sharma et al., 2008), coral
decline (Abram et al., 2003; Hu et al., 2003; Shinn et al., 2000), or
coral reef productivity (Moses et al., 2008). The use of CI techniques is
lower in monitoring than in mapping and is limited to SVM (Zhang,
2015) and NN (Yamamoto and Sugiura, 2002) (coral growth).
4.1.2. Modeling
Modeling in coral reefs involves many different research tasks,
ranging from modeling population dynamics (Riegl and Purkis,
2005; Scoplitis et al., 2007) and sediment transport (Ouillon et al.,
2004; Yokoki et al., 2006), to prediction of coral reef development
under nutrient and sediment stress (Meesters et al., 1998), or even
optimization of cost-effectiveness models for coral reef management
(Gao and Hailu, 2012; Ruitenbeek et al., 1999). The most widely used
technique in coral reef modeling is the NN approach. NNs have been
used to model coral growth (Yamamoto and Sugiura, 2002), coral
skeleton growth (Juillet-Leclerc and Thiria, 2007, 2008), SST and
SSS reconstructions from coral skeleton (Juillet-Leclerc, 2009), and
cold-water coral habitats (Purser et al., 2008) (based on SOM), or
to predict large-scale coral bleaching (Wooldridge and Done, 2004)
(BBN). FL is the second most used technique in coral reef modeling.
This ranges from the prediction of coral reef development under
nutrient and sediment stress (Meesters et al., 1998) and modeling
coral blenching (Halide and Ridd, 2002) to model economic policies
(Ruitenbeek et al., 1999) and management options (Gao and Hailu,
2012, 2013) in coral reefs. Other CI techniques used to a lesser extent
are: GARP, which has been used to predict species' distributions
(Elith et al., 2006) or distribution of cold-water coral habitat

117

(Guinan et al., 2009); SVM, which has been used to model richness
and diversity of reef sh species (Knudby et al., 2010); SA, that has
been found useful to design marine reserve networks (Leslie et al.,
2003). Boosted regression trees, which were used to predict the
diversity and abundance of sh and corals (Pittman et al., 2009)
and nally MaxEnt, that has recently used for modeling the habitat
suitability for deep-water gorgonian corals (Tong et al., 2013).
4.2. Answering the research questions
(1) What is the fastest and most accurate method to detect corals from
images? The SVM method has been found to be the better method
to detect corals in images, specially, in underwater images. The
SVM approach in Mehta et al. (2007) succeeds in classifying low
quality underwater images. A key benet is that the classifying
algorithm does not require the estimation of any intermediary
feature vector or histogram: the SVM made use of raw image as
input vectors, and achieved a successful classication probability
of 95%. This is a very important benet of SVMs since they enable
the use of raw image data as feature vectors, which is particularly
transcendent for random natural textures where geometrical
features are difcult to get (Henriques et al., 2010; In Kim et al.,
2002 Li et al., 2003), as in reef images, which exhibit rich patterns
of textures, shapes and colors. The SOM approach (Purser et al.,
2009) have also successfully detected (PCC ,SOM = 91.4%) coldwater corals (CWCs) based on video transect data taken in deepwater at Tisler Reef (Norway).
(2) What are the most suitable method to map coral reef ecosystems?
Although mapping has been carried out in the literature by almost
all the CI techniques, SVMs seem to be the most suitable method to
map coral reef ecosystems. They have been used the most in
mapping (Beijbom et al., 2012; Wahidin et al., 2015; Zhang,
2015), classication (Henriques et al., 2010; Johnson-Roberson
et al., 2006; Mehta et al., 2007; Padmavathi et al., 2010;
Shihavuddin et al., 2013; Zhang, 2015), segmentation and classication (Johnson-Roberson et al., 2007), detection (Collin and
Planes, 2012), and semi-automated image analysis (Schoening
et al., 2012). NNs have also proven a good performance for
mapping in a good number of papers: mapping (Watts et al.,
2011), classication (Elawady, 2015; Marcos et al., 2005;
Nagamani et al., 2012; Padmavathi et al., 2010), classication and
mapping (Bandyopadhyay et al., 2009), detection (Purser et al.,
2009).
(3) What are the best methods for designing marine reserves? What
are the best methods to help decision making in this regard? FL
has been found to be the most feasible method to model economic policies (Ruitenbeek et al., 1999) and management options (Gao and Hailu, 2012, 2013) in coral reefs. The SA
approach (Leslie et al., 2003) was able to generate marine reserves conserving the target habitats percentage while minimizing the perimeter, and led to reserve forms with less
perimeter and larger overall areas, which is supposed to be
preferable for sociopolitical reasons.
(4) What concepts of coral reefs can be applied to solve generalpurpose optimization problems? The inclusion of real coral
reef processes into computation approaches thus provides
useful novel models, that can be combined with other CI approaches in the future to improve the solution of both science
and engineering problems. For example, it is possible to dene
new CRO models considering different coral species in the
reef. This is useful for dening new co-evolution approaches
in just one CRO population, to exploit competitive cooperation schemes. Another real characteristic of corals that can
be included in the CRO is the fact that corals grew at different
rates depending on the substrate they are xed to (Vermeij,

118

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

2005). This can be exploited for evolving several models of


problem encoding in a competitive way.
5. Conclusions
In this paper we have reviewed the use of Computational Intelligence
(CI) techniques (Neural, Fuzzy, and evolutionary computation) for coral
reef applications. We have summarized the main characteristics of some
of the most important CI algorithms, and described the application of
these approaches to ecological and management questions related to
coral reef ecosystems.
Neural networks (NNs), Support Vector Machines (SVMs), genetic
algorithms (GAs), and fuzzy logic (FL) are the most commonly used
CI algorithms in these problems. The review points out that all can be
useful, although NNs and SVMs are the CI approaches that generally
perform best compared to classical methods. In brief: NNs exhibit a
higher accuracy than that of other more used methods (the kappa
coefcient, in most of the cases) when applied to classication problems
in coral reef habitats. The extent to which the performance is superior
depends additionally on the NN input data nature (reectance, video
images, bathymetry data): NNs exhibit a probability of correct classication (PCC) that is considerable better than that of kappa coefcient when the inputs are bathymetric data (P CC , NN = 0.86, much
better than PCC , kappa = 0.63), or reectance patterns (PCC , NN = 0.91
vs. PCC , kappa = 0.84).
In general, the application of SVMs leads to: rst, very good results in
coral reef texture classication problems (95%); second, similar results
to that of other Machine Learning methods such as Random Forest (in
predicting sh diversity in coral reef habitats); and third, better results
that those only based on kappa statistics. As in NNs, the extent to
which the SVM approach has better or similar performance depends
not only on the problem in itself but also on the nature of the SVM
inputs (either row images or combinations of bathymetric data and
images). In this respect, the fusion of bathymetric data and images
make SVMs reach excellent results when generating maps of coral reef
environments.
Evolutionary-type approaches in coral reef problems work better
than more classical methods (greedy, regression approaches). In particular, the use of GA for Rule Set Production successfully predicts the
spatial distribution of sponge abundances on coral reefs, and outperforms classical regression approaches.
Finally, FL has been found to be useful to model the inuence of
anthropogenic stressors on coral reef ecosystems, and to predict the
interaction between these and possible human interventions to
improve their conservation. In this respect, FL can help politician make
more informed decisions, and design better management strategies
for coral reef ecosystems.
While some coral reef related problems are now addressed using CI
techniques, coral reef concepts can vice versa inspire novel CI methods,
such as the Coral Reef Optimization (CRO) algorithm (Salcedo-Sanz
et al., 2013, 2014a). This example of cross-fertilization between these
two research elds has been successfully applied to a variety of
problems in different engineering areas (Salcedo-Sanz et al., 2014b,
2014c, 2015a). The good convergence properties of the CRO algorithm
allow its application in micro-evolution (algorithms with very low
number of generations), which would be particularly useful in problems
with high requirements in terms of computational time, for instance, in
real time processes for mobile networks. The inclusion of real coral reef
processes into computation approaches thus provides useful novel
models, that can be combined with other CI approaches in the future
to improve the solution of both science and engineering problems. For
example, it is possible to dene new CRO models considering different
coral species in the reef. This is useful for dening new co-evolution
approaches in just one CRO population, to exploit competitive cooperation schemes. Another real characteristic of corals that can be included
in the CRO is the fact that corals grew at different rates depending on

the substrate they are xed to (Vermeij, 2005). This can be exploited
for evolving several models of problem encoding in a competitive way.

Appendix A. Supplementary data


Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.ecoinf.2016.01.008.

References
Abram, Nerilie J., Gagan, Michael K., McCulloch, Malcolm T., Chappell, John, Hantoro,
Wahyoe S., 2003. Coral reef death during the 1997 Indian Ocean dipole linked to
Indonesian wildres. Science 301 (5635), 952955.
Affenzeller, Michael, Wagner, Stefan, Winkler, Stephan, Beham, Andreas, 2009. Genetic
algorithms and genetic programming: modern concepts and practical applications.
Crc Press.
Allers, Elke, Abed, Raeid M.M., Wehrmann, Laura M., Wang, Tao, Larsson, Ann I., Purser,
Autun, de Beer, Dirk, 2013. Resistance of Lophelia pertusa to coverage by sediment
and petroleum drill cuttings. Mar. Pollut. Bull. 74 (1), 132140.
Andrfout, S., Payri, C., 2001. Scaling-up carbon and carbonate metabolism of coral reefs
using in-situ data and remote sensing. Coral Reefs 19 (3), 259269.
Andrfout, Serge, Kramer, Philip, Torres-Pulliza, Damaris, Joyce, Karen E., Hochberg, Eric J.,
Garza-Prez, Rodrigo, Mumby, Peter J., Riegl, Bernhard, Yamano, Hiroya, White, William
H., et al., 2003. Multi-site evaluation of IKONOS data for classication of tropical coral reef
environments. Remote Sens. Environ. 88 (1), 128143.
Andrefouet, Serge, Zubia, Mayalen, Payri, Claude, 2004. Mapping and biomass estimation
of the invasive brown algae Turbinaria ornata (Turner) J. Agardh and Sargassum
mangarevense (Grunow) Setchell on heterogeneous Tahitian coral reefs using 4-m
resolution IKONOS satellite data. Coral Reefs 23 (1), 2638.
Andrfout, Serge, Cabioch, Guy, Flamand, Benot, Pelletier, Bernard, 2009. A reappraisal of
the diversity of geomorphological and genetic processes of new caledonian coral reefs: a
synthesis from optical remote sensing, coring and acoustic multibeam observations.
Coral Reefs 28 (3), 691707.
Andrew, N.L., Mapstone, B.D., 1987. Sampling and the description of spatial pattern in
marine ecology. Oceanogr. Mar. Biol. 25, 3990.
Anthony, Kenneth R.N., Kline, David I., Diaz-Pulido, Guillermo, Dove, Sophie,
Hoegh-Guldberg, Ove, 2008. Ocean acidication causes bleaching and productivity
loss in coral reef builders. Proc. Natl. Acad. Sci. 105 (45), 1744217446.
Anthony, Kenneth, Maynard, Jeffrey A., Diaz-Pulido, Guillermo, Mumby, Peter J., Marshall,
Paul A., Cao, Long, Hoegh-Guldberg, Ove, 2011. Ocean acidication and warming will
lower coral reef resilience. Glob. Chang. Biol. 17 (5), 17981808.
Ault, J.S., Smith, S.G., Browder, J.A., Nuttle, W., Franklin, E.C., Luo, J., DiNardo, G.T.,
Bohnsack, J.A., 2014. Indicators for assessing the ecological dynamics and sustainability
of southern Florida's coral reef and coastal sheries. Ecol. Indic. 44, 164172.
Baker, Andrew C., Starger, Craig J., McClanahan, Tim R., Glynn, Peter W., 2004. Coral reefs:
corals' adaptive response to climate change. Nature 430 (7001), 741.
Baker, Andrew C., Glynn, Peter W., Riegl, Bernhard, 2008. Climate change and coral reef
bleaching: an ecological assessment of long-term impacts, recovery trends and future
outlook. Estuar. Coast. Shelf Sci. 80 (4), 435471.
Ban, Natalie C., Adams, Vanessa M., Almany, Glenn R., Ban, Stephen, Cinner, Josh E.,
McCook, Laurence J., Mills, Morena, Pressey, Robert L., White, Alan, 2011. Designing,
implementing and managing marine protected areas: emerging trends and opportunities for coral reef nations. J. Exp. Mar. Biol. Ecol. 408 (1), 2131.
Ban, Stephen S., Graham, Nicholas A.J., Connolly, Sean R., 2014. Evidence for multiple
stressor interactions and effects on coral reefs. Glob. Chang. Biol. 20 (3), 681697.
Bandyopadhyay, S., Sharma, S., Bahuguna, A., 2009. Articial neural network based coral
cover classiers using Indian Remote Sensing (IRS LISS-III) sensor data: a case
study in Gulf of Kachcch, India. Int. J. Geoinformatics 5 (3), 5562.
Banzhaf, Wolfgang, Nordin, Peter, Keller, Robert E., Francone, Frank D., 1998. Genetic
programming: an introduction vol. 1. Morgan Kaufmann, San Francisco.
Bartley, Rebecca, Bainbridge, Zoe T., Lewis, Stephen E., Kroon, Frederieke J., Wilkinson,
Scott N., Brodie, Jon E., Silburn, D. Mark, 2014. Relating sediment impacts on coral
reefs to watershed sources, processes and management: a review. Sci. Total Environ.
468, 11381153.
Beger, Maria, Jones, Geoffrey P., Possingham, Hugh P., 2006. A method of statistical
modelling of coral reef sh distribution: can it aid conservation planning in data poor
regions. Proceedings of 10th International Coral Reef Symposium, pp. 14451456.
Beijbom, Oscar, Edmunds, Peter J., Kline, David, Mitchell, B. Greg, Kriegman, David, et al.,
2012. Automated annotation of coral reef survey images. Computer Vision and
Pattern Recognition (CVPR), 2012 IEEE Conference on. IEEE, pp. 11701177.
Beneld, S.L., Guzman, H.M., Mair, J.M., Young, J.A.T., 2007. Mapping the distribution of
coral reefs and associated sublittoral habitats in Pacic Panama: a comparison of
optical satellite sensors and classication methodologies. Int. J. Remote Sens. 28
(22), 50475070.
Bertels, Luc, Vanderstraete, Tony, Van Coillie, Stijn, Knaeps, Els, Sterckx, Sindy, Goossens,
Rudi, Deronde, Bart, 2008. Mapping of coral reefs using hyperspectral casi data; a case
study: Fordata, Tanimbar, Indonesia. Int. J. Remote Sens. 29 (8), 23592391.
Bett, Brian J., Malzone, M. Gabriella, Narayanaswamy, Bhavani E., Wigham, Benjamin D.,
2001. Temporal variability in phytodetritus and megabenthic activity at the seabed
in the deep Northeast Atlantic. Prog. Oceanogr. 50 (1), 349368.
Bezdek, James C., 2013. Pattern Recognition with Fuzzy Objective Function Algorithms.
Springer Science & Business Media.

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123


Bishop, Christopher M., 1995. Neural Networks for Pattern Recognition. Oxford University
Press.
Bishop, Christopher M., 2006. Pattern Recognition and Machine Learning. Springer.
Brander, Luke M., Van Beukering, Pieter, Cesar, Herman S.J., 2007. The recreational value
of coral reefs: a meta-analysis. Ecol. Econ. 63 (1), 209218.
Breiman, Leo, 2001. Random forests. Mach. Learn. 45 (1), 532.
Brock, John, Yates, Kimberly, Halley, Robert, 2006. Integration of coral reef ecosystem
process studies and remote sensing. Remote Sensing of Aquatic Coastal Ecosystem
Processes. Springer, pp. 111131.
Burke, Lauretta Marie, Reytar, Kathleen, Spalding, Mark, Perry, Allison, et al., 2011. Reefs
at Risk Revisited. World Resources Institute, Washington, DC, p. 114.
Call, Katherine A., Hardy, John T., Wallin, David O., 2003. Coral reef habitat discrimination
using multivariate spectral analysis and satellite remote sensing. Int. J. Remote Sens.
24 (13), 26272639.
Cambria, Erik, Huang, Guang-Bin, Kasun, Liyanaarachchi Lekamalage Chamara, Zhou,
Hongming, Vong, Chi Man, Lin, Jiarun, Yin, Jianping, Cai, Zhiping, Liu, Qiang, Li,
Kuan, et al., 2013. Extreme learning machines [trends & controversies]. IEEE Intell.
Syst. 28 (6), 3059.
Capolsini, Patrick, Andrfout, Serge, Rion, Cdric, Payri, Claude, 2003. A comparison of
Landsat ETM+, SPOT HRV, Ikonos, ASTER, and airborne MASTER data for coral reef
habitat mapping in South Pacic islands. Can. J. Remote. Sens. 29 (2), 187200.
Carilli, Jessica E., Norris, Richard D., Black, Bryan A., Walsh, Sheila M., McField, Melanie,
2009. Local stressors reduce coral resilience to bleaching. PLoS One 4 (7), e6324.
Cesar, Herman S.J., 2002. Coral reefs: their functions, threats and economic value http://hdl.
handle.net/1834/557 International Oceanographic Data and Information Exchange
(IODE), UNESCO.
Chang, Chih-Chung, Lin, Chih-Jen, 2011. Libsvm: A library for support vector machines.
ACM Trans. Intell. Syst. Technol. 2 (3), 27.
Chang, Chih-Hua, Liu, Cheng-Chien, Chung, Hsiao-Wei, Lee, Long-Jeng, Yang, Wen-Chang,
2014. Development and evaluation of a genetic algorithm-based ocean color inversion model for simultaneously retrieving optical properties and bottom types in
coral reef regions. Appl. Opt. 53 (4), 605617.
Chatterjee, Samprit, Hadi, Ali S., 2015. Regression Analysis by Example. John Wiley & Sons.
Chen, Ping-Yu, Chen, Chi-Chung, Chu, LanFen, McCarl, Bruce, 2015. Evaluating the
economic damage of climate change on global coral reefs. Glob. Environ. Chang. 30,
1220.
Clark, Malcolm R., Rowden, Ashley A., 2009. Effect of deepwater trawling on the macroinvertebrate assemblages of seamounts on the Chatham Rise, New Zealand. DeepSea Res. I Oceanogr. Res. Pap. 56 (9), 15401554.
Collin, Antoine, Planes, Serge, 2012. Enhancing coral health detection using spectral
diversity indices from worldview-2 imagery and machine learners. Remote Sens. 4
(10), 32443264.
Costa, B.M., Battista, T.A., Pittman, S.J., 2009. Comparative evaluation of airborne LiDAR
and ship-based multibeam SoNAR bathymetry and intensity for mapping coral reef
ecosystems. Remote Sens. Environ. 113 (5), 10821100.
Cristianini, Nello, Shawe-Taylor, John, 2000. An Introduction to Support Vector Machines
and Other Kernel-based Learning Methods. Cambridge University Press.
Dalleau, Mayeul, Andrfout, Serge, Wabnitz, Colette C.C., Payri, Claude, Wantiez, Laurent,
Pichon, Michel, Friedman, K.I.M., Vigliola, Laurent, Benzoni, Francesca, 2010. Use of
habitats as surrogates of biodiversity for efcient coral reef conservation planning
in Pacic Ocean islands. Conserv. Biol. 24 (2), 541552.
D'Angelo, Cecilia, Wiedenmann, Joerg, 2014. Impacts of nutrient enrichment on coral
reefs: new perspectives and implications for coastal management and reef survival.
Curr. Opin. Environ. Sustain. 7, 8293.
De Groot, Rudolf, Brander, Luke, Van Der Ploeg, Sander, Costanza, Robert, Bernard,
Florence, Braat, Leon, Christie, Mike, Crossman, Neville, Ghermandi, Andrea, Hein,
Lars, et al., 2012. Global estimates of the value of ecosystems and their services in
monetary units. Ecosystem Serv. 1 (1), 5061.
Dolan, Margaret F.J., Grehan, Anthony J., Guinan, Janine C., Brown, Colin, 2008. Modelling
the local distribution of cold-water corals in relation to bathymetric variables: adding
spatial context to deep-sea video data. Deep-Sea Res. I Oceanogr. Res. Pap. 55 (11),
15641579.
Dorigo, Marco, Birattari, Mauro, 2010. Ant colony optimization. Encyclopedia of Machine
Learning. Springer, pp. 3639.
Duda, Richard O., Hart, Peter E., Stork, David G., 2012. Pattern Classication. John Wiley &
Sons.
Dulvy, Nicholas K., Kindsvater, Holly K., 2015. Ecology: recovering the potential of coral
reefs. Nature.
Dunlop, Katherine M., Kuhnz, Linda A., Ruhl, Henry A., Huffard, Christine L., Caress, David
W., Henthorn, Richard G., Hobson, Brett W., McGill, Paul, Smith, Kenneth L., 2015. An
evaluation of deep-sea benthic megafauna length measurements obtained with laser
and stereo camera methods. Deep-Sea Res. I Oceanogr. Res. Pap. 96, 3848.
Dustan, Phillip, Dobson, Eric, Nelson, George, 2001. Landsat thematic mapper: detection
of shifts in community composition of coral reefs. Conserv. Biol. 15 (4), 892902.
Eakin, C. Mark, Morgan, Jessica A., Heron, Scott F., Smith, Tyler B., Liu, Gang, Alvarez-Filip,
Lorenzo, Baca, Bart, Bartels, Erich, Bastidas, Carolina, Bouchon, Claude, et al., 2010.
Caribbean corals in crisis: record thermal stress, bleaching, and mortality in 2005.
PLoS One 5 (11), e13969.
Eiben, Agoston E., Smith, James E., 2003. Introduction to Evolutionary Computing. Springer
Science & Business Media.
Eiben, A.E., Smith, J.E., 2008. Introduction to Evolutionary Computing (Natural Computing
Series).
Mohamed Elawady. Sparse coral classication using deep convolutional neural networks.
arXiv preprint arXiv:1511.09067, 2015.
Elisseeff, Andr, Weston, Jason, 2001. A kernel method for multi-labelled classication.
Advances in Neural Information Processing Systems, pp. 681687.

119

Elith, Jane, Graham, Catherine H., Anderson, Robert P., Dudk, Miroslav, Ferrier, Simon,
Guisan, Antoine, Hijmans, Robert J., Huettmann, Falk, Leathwick, John R., Lehmann,
Anthony, et al., 2006. Novel methods improve prediction of species' distributions
from occurrence data. Ecography 29, 129151.
Elith, Jane, Phillips, Steven J., Hastie, Trevor, Dudk, Miroslav, Chee, Yung En, Yates, Colin J.,
2011. A statistical explanation of maxent for ecologists. Divers. Distrib. 17 (1), 4357.
ERDAS. ERDAS IMAGINE, 2015. ERDAS Imagine version 9.3. http://www.erdas.com.ar/
productos-imagine.htm.
Exelis, 2012. Exelis Visual Information Solutions. ENVIEnvironment for visualizing
images. http://www.exelisvis.com/Home.aspx, 2012. Software to process and analyze
all types of imagery and data such as multispectral, hyperspectral, LiDAR, and SAR.
Ferrario, Filippo, Beck, Michael W., Storlazzi, Curt D., Micheli, Fiorenza, Shepard, Christine C.,
Airoldi, Laura, 2014. The effectiveness of coral reefs for coastal hazard risk reduction and
adaptation. Nat. Commun. 5.
Fossaa, Jan Helge, Skjoldal, Hein Rune, 2010. Conservation of Cold-water Coral Reefs in
Norway. Oxford University Press, New York(USA).
Foster, M.S., Harrold, C., Hardin, D.D., 1991. Point vs. photo quadrat estimates of the cover
of sessile marine organisms. J. Exp. Mar. Biol. Ecol. 146 (2), 193203.
Fox, Rebecca J., Bellwood, David R., 2014. Herbivores in a small world: network theory
highlights vulnerability in the function of herbivory on coral reefs. Funct. Ecol. 28
(3), 642651.
Freeman, Lauren A., 2015. Robust performance of marginal pacic coral reef habitats in
future climate scenarios. PLoS One 10 (6), e0128875.
Gao, Lei, Hailu, Atakelty, 2012. Ranking management strategies with complex outcomes:
an AHP-fuzzy evaluation of recreational shing using an integrated agent-based
model of a coral reef ecosystem. Environ. Model. Softw. 31, 318.
Gao, Lei, Hailu, Atakelty, 2013. Identifying preferred management options: an integrated
agent-based recreational shing simulation model with an AHP-TOPSIS evaluation
method. Ecol. Model. 249, 7583.
Gao, Huang, Huang, Guang-Bin, Song, Shiji, You, Keyou, 2015. Trends in extreme learning
machines: a review. Neural Netw. 61, 3248.
Gates, Andrew R., Jones, Daniel O.B., 2012. Recovery of benthic megafauna from anthropogenic disturbance at a hydrocarbon drilling well (380 m Depth in the Norwegian Sea).
PLoS ONE 7 (10), e44114.
Golberg, David E., 1989. Genetic Algorithms in Search, Optimization, and Machine Learning. Addison-Wesley.
Goodman, James, Ustin, Susan L., 2007. Classication of benthic composition in a coral
reef environment using spectral unmixing. J. Appl. Remote. Sens. 1 (1), 011501.
Goodman, James A., Purkis, Samuel J., Phinn, Stuart R., 2013. Coral Reef Remote Sensing: a
Guide for Mapping, Monitoring and Management. Springer Science & Business Media.
Guinan, Janine, Brown, Colin, Dolan, Margaret F.J., Grehan, Anthony J., 2009. Ecological
niche modelling of the distribution of cold-water coral habitat using underwater
remote sensing data. Ecol. Inform. 4 (2), 8392.
Hagan, Martin T., Menhaj, Mohammad B., 1994. Training feedforward networks with the
Marquardt algorithm. IEEE Trans. Neural Netw. 5 (6), 989993.
Halide, Halmar, Ridd, Peter, 2002. Modelling coral bleaching events using a fuzzy logic
technique. Fuzzy Logic. Springer, pp. 352362.
Harborne, Alastair R., Mumby, Peter J., Zychaluk, Kamila, Hedley, John D., Blackwell, Paul
G., 2006. Modeling the beta diversity of coral reefs. Ecology 87 (11), 28712881.
Hatcher, B.G., 1997. Coral reef ecosystems: how much greater is the whole than the sum
of the parts? Coral Reefs 16 (1), S77S91.
Haykin, S., 1998. Neural Networks: a Comprehensive Foundation. Prentice Hall.
Haykin, Simon, Neural Network, 2004. A comprehensive foundation. Neural Netw. 2.
Hearst, Marti A., Dumais, Susan T., Osman, Edgar, Platt, John, Scholkopf, Bernhard, 1998.
Support vector machines. IEEE Intell. Syst. Appl. 13 (4), 1828.
Hedley, J.D., Harborne, A.R., Mumby, P.J., 2005. Simple and robust removal of sun glint for
mapping shallow-water benthos. Int. J. Remote Sens. 26 (10), 21072112.
Hedley, John, Roelfsema, Chris, Phinn, Stuart R., 2009. Efcient radiative transfer
model inversion for remote sensing applications. Remote Sens. Environ. 113
(11), 25272532.
Hedley, John, Roelfsema, Chris, Koetz, Benjamin, Phinn, Stuart, 2012. Capability of the sentinel 2 mission for tropical coral reef mapping and coral bleaching detection. Remote
Sens. Environ. 120, 145155.
Henriques, Antnio P.M., Neto, Adrio D. Dria, Amaral, Ricardo F., 2010. Classication of
multispectral images in coral environments using a hybrid of classier ensembles.
Neurocomputing 73 (7), 12561264.
Hirzel, Alexandre H., Hausser, J., Chessel, D., Perrin, N., 2002. Ecological-niche factor
analysis: how to compute habitat-suitability maps without absence data? Ecology
83 (7), 20272036.
Hochberg, Eric J., Atkinson, M.J., 2000. Spectral discrimination of coral reef benthic
communities. Coral Reefs 19 (2), 164171.
Hochberg, Eric J., Atkinson, Marlin J., 2003. Capabilities of remote sensors to classify coral,
algae, and sand as pure and mixed spectra. Remote Sens. Environ. 85 (2), 174189.
Hochberg, E.J., Atkinson, M.J., 2008. Coral reef benthic productivity based on optical
absorptance and light-use efciency. Coral Reefs 27 (1), 4959.
Hochberg, Eric J., Atkinson, Marlin J., Andrfout, Serge, 2003a. Spectral reectance of
coral reef bottom-types worldwide and implications for coral reef remote sensing.
Remote Sens. Environ. 85 (2), 159173.
Hochberg, Eric J., Andrefouet, Serge, Tyler, Misty R., 2003b. Sea surface correction of
high spatial resolution IKONOS images to improve bottom mapping in near-shore
environments. IEEE Trans. Geosci. Remote Sens. 41 (7), 17241729.
Hochberg, Eric J., Atkinson, Marlin J., Apprill, Amy, Andrefouet, Serge, 2004. Spectral
reectance of coral. Coral Reefs 23 (1), 8495.
Hoegh-Guldberg, Ove, Mumby, P.J., Hooten, A.J., Steneck, R.S., Greeneld, P., Gomez, E.,
Harvell, C.D., Sale, P.F., Edwards, A.J., Caldeira, K., et al., 2007. Coral reefs under
rapid climate change and ocean acidication. Science 318 (5857), 17371742.

120

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

Hooidonk, Ruben, Maynard, Jeffrey Allen, Manzello, Derek, Planes, Serge, 2014. Opposite
latitudinal gradients in projected ocean acidication and bleaching impacts on coral
reefs. Glob. Chang. Biol. 20 (1), 103112.
Hu, Chuanmin, Hackett, Keith E., Callahan, Michael K., Andrfout, Serge, Wheaton,
Jennifer L., Porter, James W., Muller-Karger, Frank E., 2003. The 2002 ocean color
anomaly in the Florida Bight: a cause of local coral reef decline? Geophys. Res. Lett.
30 (3).
Huang, Guang-Bin, 2014. An insight into extreme learning machines: random neurons,
random features and kernels. Cogn. Comput. 6 (3), 376390.
Huang, Guang-Bin, Chen, Lei, 2007. Convex incremental extreme learning machine.
Neurocomputing 70 (16), 30563062.
Huang, Guang-Bin, Chen, Lei, 2008. Enhanced random search based incremental extreme
learning machine. Neurocomputing 71 (16), 34603468.
Huang, Guang-Bin, Zhu, Qin-Yu, Siew, Chee-Kheong, 2006. Extreme learning machine:
theory and applications. Neurocomputing 70 (1), 489501.
Huang, Guang-Bin, Ding, Xiaojian, Zhou, Hongming, 2010. Optimization method based
extreme learning machine for classication. Neurocomputing 74 (1), 155163.
Huang, Zhi, Brooke, Brendan, Li, Jin, 2011a. Performance of predictive models in marine
benthic environments based on predictions of sponge distribution on the australian
continental shelf. Ecol. Inform. 6 (3), 205216.
Huang, Guang-Bin, Wang, Dian Hui, Lan, Yuan, 2011b. Extreme learning machines: a
survey. Int. J. Mach. Learn. Cybern. 2 (2), 107122.
Huang, Guang-Bin, Zhou, Hongming, Ding, Xiaojian, Zhang, Rui, 2012. Extreme learning
machine for regression and multiclass classication. IEEE Trans. Syst. Man Cybern. B
Cybern. 42 (2), 513529.
Huang, Gao, Song, Shiji, Gupta, Jatinder N.D., Wu, Cheng, 2014. Semi-supervised and
unsupervised extreme learning machines. IEEE Trans. Cybern. 44 (12), 24052417.
Hughes, Terry P., Baird, Andrew H., Bellwood, David R., Card, Margaret, Connolly, Sean R.,
Folke, Carl, Grosberg, Richard, Hoegh-Guldberg, Ove, Jackson, J.B.C., Kleypas, Janice, et
al., 2003. Climate change, human impacts, and the resilience of coral reefs. Science
301 (5635), 929933.
Hughes, Terence P., Rodrigues, Maria J., Bellwood, David R., Ceccarelli, Daniela, HoeghGuldberg, Ove, McCook, Laurence, Moltschaniwskyj, Natalie, Pratchett, Morgan S.,
Steneck, Robert S., Willis, Bette, 2007. Phase shifts, herbivory, and the resilience of
coral reefs to climate change. Curr. Biol. 17 (4), 360365.
Hughes, Terry P., Graham, Nicholas A.J., Jackson, Jeremy B.C., Mumby, Peter J., Steneck,
Robert S., 2010. Rising to the challenge of sustaining coral reef resilience. Trends
Ecol. Evol. 25 (11), 633642.
In Kim, Kwang, Jung, Keechul, Park, Se Hyun, Kim, Hang Joon, 2002. Support vector
machines for texture classication. IEEE Trans. Pattern Anal. Mach. Intell. 24 (11),
15421550.
Isoun, E., Fletcher, C., Frazer, N., Gradie, J., 2003. Multi-spectral mapping of reef bathymetry
and coral cover; Kailua Bay, Hawaii. Coral Reefs 22 (1), 6882.
Januchowski-Hartley, Fraser A., Graham, Nicholas A.J., Cinner, Joshua E., Russ, Garry R.,
2015. Local shing inuences coral reef sh behavior inside protected areas of the
Indo-Pacic. Biol. Conserv. 182, 812.
Johnson-Roberson, Matthew, Kumar, Suresh, Pizarro, Oscar, Willams, Stefan, 2006. Stereoscopic imaging for coral segmentation and classication. OCEANS 2006. IEEE, pp. 16.
Johnson-Roberson, Matthew, Kumar, Suresh, Willams, Stefan, 2007. Segmentation and
classication of coral for oceanographic surveys: a semi-supervised machine learning
approach. OCEANS 2006-Asia Pacic. IEEE, pp. 16.
Joyce, Karen E., 2005. A ethod for mapping live coral cover using remote sensing PhD
thesis School of Geography, Planning and Architecture.
Juillet-Leclerc, A., 2006. A new generation of climatic proxies required to complement
instrumental records. Geophysical Research Abstracts vol. 8, p. 06308.
Juillet-Leclerc, A., 2007. The conversion into environmental parameters of multi-proxies
derived from coral skeleton. Geophysical Research Abstracts vol. 9, p. 03306.
Juillet-Leclerc, A., 2009. Independent sst and sss reconstructions from coral skeleton, by
using neural network. AGU Fall Meeting Abstracts vol. 1, p. 1433.
Juillet-Leclerc, A., Thiria, S., 2007. Neural networks applied on multi-proxies from coral
skeleton. Geophysical Research Abstracts vol. 9, p. 03332.
Juillet-Leclerc, A., Thiria, S., 2008. Neural network: a statistical treatment well suited to
coral skeleton proxies. AGU Fall Meeting Abstracts vol. 1, p. 1514.
Juillet-Leclerc, A., Thiria, S., Naveau, P., Delcroix, T., Le Bec, N., Blamart, D., Correge, T.,
2006. Spcz migration and enso events during the 20th century as revealed by climate
proxies from a Fiji coral. Geophys. Res. Lett. 33 (17).
Juillet-Leclerc, A., Thiria, S., Peron, C., 2007. Independent temperature and salinity
reconstruction from coral skeleton. AGU Fall Meeting Abstracts vol. 1, p. 05.
Kemp, David B., Sadler, Peter M., 2014. Climatic and eustatic signals in a global compilation of shallow marine carbonate accumulation rates. Sedimentology 61 (5),
12861297.
Kennedy, J., Eberhart, R., 1995. Particle swarm optimization. Proceedings of IEEE International Conference of Neural Networks. IEEE, pp. 19421948.
Kirkpatrick, Scott, 1984. Optimization by simulated annealing: Quantitative studies.
J. Stat. Phys. 34 (56), 975986.
Kirkpatrick, Scott, Gelatt, C. Daniel, Vecchi, Mario P., et al., 1983. Optimization by simulated
annealing. Science 220 (4598), 671680.
Kittinger, John N., Teneva, Lida T., Koike, Haruko, Stamoulis, Kostantinos A., Kittinger,
Daniela S., Oleson, Kirsten L.L., Conklin, Eric, Gomes, Mahana, Wilcox, Bart, Friedlander,
Alan M., 2015. From reef to table: Social and ecological factors affecting coral reef
sheries, artisanal seafood supply chains, and seafood security. PLoS One 10 (8),
e0123856.
Klemas, Victor V., 2001. Remote sensing of landscape-level coastal environmental indicators.
Environ. Manag. 27 (1), 4757.
Kline, David I., Teneva, Lida, Hauri, Claudine, Schneider, Kenneth, Miard, Thomas, Chai,
Aaron, Marker, Malcolm, Dunbar, Rob, Caldeira, Ken, Lazar, Boaz, et al., 2015. Six

month in situ high-resolution carbonate chemistry and temperature study on a


coral reef at reveals asynchronous PH and temperature anomalies. PLoS One 10 (6).
Knowlton, N., Jackson, J., 2013. Corals and coral reefs. Encyclopedia of Biodiversity, second
ed., pp. 330346.
Knudby, Anders, LeDrew, Ellsworth, Brenning, Alexander, 2010. Predictive mapping
of reef sh species richness, diversity and biomass in Zanzibar using IKONOS imagery and machine-learning techniques. Remote Sens. Environ. 114 (6),
12301241.
Knudby, Anders, Jupiter, Stacy, Roelfsema, Chris, Lyons, Mitchell, Phinn, Stuart, 2013.
Mapping coral reef resilience indicators using eld and remotely sensed data. Remote
Sens. 5 (3), 13111334.
Kohonen, Teuvo, 1990. The self-organizing map. Proc. IEEE 78 (9), 14641480.
Koza, John R., 1992. Genetic programming: on the programming of computers by means
of natural selection vol. 1. MIT Press.
Kroon, Frederieke J., Schaffelke, Britta, Bartley, Rebecca, 2014. Informing policy to protect
coastal coral reefs: Insight from a global review of reducing agricultural pollution to
coastal ecosystems. Mar. Pollut. Bull. 85 (1), 3341.
Lamy, Thomas, Legendre, Pierre, Chancerelle, Yannick, Siu, Gilles, Claudet, Joachim, 2015.
Understanding the spatio-temporal response of coral reef sh communities to
natural disturbances: insights from beta-diversity decomposition. PLoS One 10
(9), e0138696.
Langdon, William B., 2012. Genetic programming and data structures: genetic
programming + data structures = automatic programming! vol. 1. Springer Science &
Business Media.
Langmead, O., Sheppard, C., 2004. Coral reef community dynamics and disturbance: a
simulation model. Ecol. Model. 175 (3), 271290.
Larsson, Ann I., Purser, Autun, 2011. Sedimentation on the cold-water coral lophelia
pertusa: cleaning efciency from natural sediments and drill cuttings. Mar. Pollut.
Bull. 62 (6), 11591168.
Larsson, Ann I., van Oevelen, Dick, Purser, Autun, Thomsen, Laurenz, 2013. Tolerance to
long-term exposure of suspended benthic sediments and drill cuttings in the coldwater coral Lophelia pertusa. Mar. Pollut. Bull. 70 (1), 176188.
LeDrew, Ellsworth F., Holden, Heather, Wulder, Michael A., Derksen, Chris, Newman,
Candace, 2004. A spatial statistical operator applied to multidate satellite imagery
for identication of coral reef stress. Remote Sens. Environ. 91 (3), 271279.
Leiper, Ian, Phinn, Stuart, Dekker, Arnold G., 2012. Spectral reectance of coral reef benthos
and substrate assemblages on Heron Reef, Australia. Int. J. Remote Sens. 33 (12),
39463965.
Leon, J., Woodroffe, C.D., 2009. Geomorphological classication of coral reef in torres strait
using obia. 7th International Conference on Geomorphology (ANZIAG).
Leslie, Heather, Ruckelshaus, Mary, Ball, Ian R., Andelman, Sandy, Possingham, Hugh P.,
2003. Using siting algorithms in the design of marine reserve networks. Ecol. Appl.
13 (sp1), 185198.
Leujak, W., Ormond, R.F.G., 2007. Comparative accuracy and efciency of six coral
community survey methods. J. Exp. Mar. Biol. Ecol. 351 (1), 168187.
Li, Shutao, Kwok, James T., Zhu, Hailong, Wang, Yaonan, 2003. Texture classication using
the support vector machines. Pattern Recogn. 36 (12), 28832893.
Li, Ming, Miao, Chunyan, Leung, Cyril, 2015. A coral reef algorithm based on learning
automata for the coverage control problem of heterogeneous directional sensor
networks. Sensors 15 (12), 3061730635.
Lim, Alan, Hedley, John D., LeDrew, Ellsworth, Mumby, Peter J., Roelfsema, Chris, 2009.
The effects of ecologically determined spatial complexity on the classication accuracy
of simulated coral reef images. Remote Sens. Environ. 113 (5), 965978.
Lucas, Matthew Q., Goodman, James, 2014. Linking coral reef remote sensing and eld
ecology: it's a matter of scale. J. Mar. Sci. Eng. 3 (1), 120.
MacNeil, M. Aaron, Graham, Nicholas A.J., Cinner, Joshua E., Wilson, Shaun K., Williams,
Ivor D., Maina, Joseph, Newman, Steven, Friedlander, Alan M., Jupiter, Stacy,
Polunin, Nicholas V.C., et al., 2015. Recovery potential of the world's coral reef shes.
Nature 520 (7547), 341344.
Maeder, Jill, Narumalani, Sunil, Rundquist, Donald C., Perk, Rlchard L., Schalles, John,
Hutchins, Kevin, Keck, Jennifer, 2002. Classifying and mapping general coral-reef
structure using ikonos data. Photogramm. Eng. Remote. Sens. 68 (12), 12971306.
Maina, Joseph, Venus, Valentijn, McClanahan, Timothy R., Ateweberhan, Mebrahtu, 2008.
Modelling susceptibility of coral reefs to environmental stress using remote sensing
data and gis models. Ecol. Model. 212 (3), 180199.
Man, Kim-Fung, Tang, Kit Sang, Kwong, Sam, 2012. Genetic Algorithms: Concepts and
Designs. Springer Science & Business Media.
Manzello, Derek, Enochs, Ian, Musielewicz, Sylvia, Carlton, Rene, Gledhill, Dwight, 2013.
Tropical cyclones cause CACO3 undersaturation of coral reef seawater in a high-CO2
world. J. Geophys. Res. Oceans 118 (10), 53125321.
Marcos, Ma Shiela Angeli, Soriano, Maricor, Saloma, Caesar, 2005. Classication of coral
reef images from underwater video using neural networks. Opt. Express 13 (22),
87668771.
McClanahan, Tim R., Donner, Simon D., Maynard, Jeffrey A., MacNeil, M. Aaron, Graham,
Nicholas A.J., Maina, Joseph, Baker, Andrew C., Alemu, Jahson B.I., Beger, Maria,
Campbell, Stuart J., et al., 2012. Prioritizing key resilience indicators to support coral
reef management in a changing climate. PLoS One 7 (8), e42884.
McClanahan, Timothy R., Graham, Nicholas A.J., Darling, Emily S., 2014. Coral reefs in a
crystal ball: predicting the future from the vulnerability of corals and reef shes to
multiple stressors. Curr. Opin. Environ. Sustain. 7, 5964.
Medeiros, Inacio G., Xavier-Junior, Joao C., Canuto, Anne M.P., 2015. Applying the coral
reefs optimization algorithm to clustering problems. Neural Networks (IJCNN),
2015 International Joint Conference on. IEEE, pp. 18.
Meesters, Erik H., Bak, Rolf P.M., Westmacott, Susie, Ridgley, Mark, Dollar, Steve, 1998. A
fuzzy logic model to predict coral reef development under nutrient and sediment
stress. Conserv. Biol. 12 (5), 957965.

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123


Mehta, Anand, Ribeiro, Eraldo, Gilner, Jessica, van Woesik, Robert, 2007. Coral reef texture
classication using support vector machines. Proceedings of International Conference
on Computer Vision Theory and Applications, Barcelona, Spain, pp. 302310.
Mellin, Camille, Andrfout, Serge, Kulbicki, Michel, Dalleau, Mayeul, Vigliola, Laurent,
2009. Remote sensing and shhabitat relationships in coral reef ecosystems: review
and pathways for multi-scale hierarchical research. Mar. Pollut. Bull. 58 (1), 1119.
Mishra, Deepak, Narumalani, Sunil, Rundquist, Donald, Lawson, Merlin, 2006. Benthic
habitat mapping in tropical marine environments using quickbird multispectral
data. Photogramm. Eng. Remote. Sens. 72 (9), 10371048.
Mitchell, Melanie, 1998. An introduction to genetic algorithms. MIT Press.
Moberg, Fredrik, Folke, Carl, 1999. Ecological goods and services of coral reef ecosystems.
Ecol. Econ. 29 (2), 215233.
Mobley, Curtis D., Sundman, Lydia K., Davis, Curtiss O., Bowles, Jeffrey H., Downes,
Trijntje Valerie, Leathers, Robert A., Montes, Marcos J., Bissett, William Paul,
Kohler, David D.R., Reid, Ruth Pamela, et al., 2005. Interpretation of
hyperspectral remote-sensing imagery by spectrum matching and look-up tables. Appl. Opt. 44 (17), 35763592.
Montgomery, Douglas C., Peck, Elizabeth A., Vining, G. Geoffrey, 2012. Introduction to
Linear Regression Analysis. vol. 821. John Wiley & Sons.
Mooney, Harold, Larigauderie, Anne, Cesario, Manuel, Elmquist, Thomas, HoeghGuldberg, Ove, Lavorel, Sandra, Mace, Georgina M., Palmer, Margaret, Scholes,
Robert, Yahara, Tetsukazu, 2009. Biodiversity, climate change, and ecosystem
services. Curr. Opin. Environ. Sustain. 1 (1), 4654.
Mora, Camilo, Andrfout, Serge, Costello, Mark J., Kranenburg, Christine, Rollo, Audrey,
Veron, John, Gaston, Kevin J., Myers, Ransom A., 2006. Coral reefs and the global
network of marine Protected Areas. Science 312 (5781), 17501751.
Morri, Carla, Montefalcone, Monica, Lasagna, Roberta, Gatti, Giulia, Rovere, Alessio,
Parravicini, Valeriano, Baldelli, Giuseppe, Colantoni, Paolo, Bianchi, Carlo Nike, 2015.
Through bleaching and tsunami: coral reef recovery in the maldives. Mar. Pollut.
Bull. 98 (1), 188200.
Moses, C.S., Palandro, D.A., Andrfout, S., Muller-Karger, F., 2008. Remote sensing of
changes in carbonate production on coral reefs: the Florida Keys. Proceedings of
11th international coral reef symposium, pp. 6266.
Moses, Christopher S., Andrefouet, Serge, Kranenburg, Christine J., Muller-Karger, Frank E.,
2009. Regional estimates of reef carbonate dynamics and productivity using Landsat
7 ETM+, and potential impacts from ocean acidication. Mar. Ecol. Prog. Ser. 380,
103.
Mumby, Peter J., 2001. Beta and habitat diversity in marine systems: a new approach to
measurement, scaling and interpretation. Oecologia 128 (2), 274280.
Mumby, Peter J., Edwards, Alasdair J., 2002. Mapping marine environments with ikonos
imagery: enhanced spatial resolution can deliver greater thematic accuracy. Remote
Sens. Environ. 82 (2), 248257.
Mumby, P.J., Green, E.P., Edwards, A.J., Clark, C.D., 1997. Coral reef habitat mapping: how
much detail can remote sensing provide? Mar. Biol. 130 (2), 193202.
Mumby, P.J., Green, E.P., Clark, C.D., Edwards, A.J., 1998a. Digital analysis of multispectral
airborne imagery of coral reefs. Coral Reefs 17 (1), 5969.
Mumby, P.J., Clark, C.D., Green, E.P., Edwards, A.J., 1998b. Benets of water column
correction and contextual editing for mapping coral reefs. Int. J. Remote Sens. 19
(1), 203210.
Mumby, Peter J., Chisholm, John R.M., Clark, Chris D., Hedley, John D., Jaubert, Jean, 2001.
Spectrographic imaging: a bird's-eye view of the health of coral reefs. Nature 413
(6851), 36.
Mumby, Peter J., Skirving, William, Strong, Alan E., Hardy, John T., LeDrew, Ellsworth F.,
Hochberg, Eric J., Stumpf, Rick P., David, Laura T., 2004a. Remote sensing of coral
reefs and their physical environment. Mar. Pollut. Bull. 48 (3), 219228.
Mumby, Peter J., Edwards, Alasdair J., Arias-Gonzlez, J. Ernesto, Lindeman, Kenyon C.,
Blackwell, Paul G., Gall, Angela, Gorczynska, Malgosia I., Harborne, Alastair R.,
Pescod, Claire L., Renken, Henk, et al., 2004b. Mangroves enhance the biomass of
coral reef sh communities in the Caribbean. Nature 427 (6974), 533536.
Mumby, Peter J., Hedley, J.D., Chisholm, J.R.M., Clark, C.D., Ripley, H., Jaubert, J., 2004c. The
cover of living and dead corals from airborne remote sensing. Coral Reefs 23 (2),
171183.
Mumby, Peter J., Dahlgren, Craig P., Harborne, Alastair R., Kappel, Carrie V., Micheli,
Fiorenza, Brumbaugh, Daniel R., Holmes, Katherine E., Mendes, Judith M., Broad,
Kenneth, Sanchirico, James N., et al., 2006. Fishing, trophic cascades, and the process
of grazing on coral reefs. Science 311 (5757), 98101.
Mumby, Peter J., Broad, Kenneth, Brumbaugh, Daniel R., Dahlgren, Craig, Harborne,
Alastair R., Hastings, Alan, Holmes, Katherine E., Kappel, Carrie V., Micheli, Fiorenza,
Sanchirico, James N., et al., 2008. Coral reef habitats as surrogates of species,
ecological functions, and ecosystem services. Conserv. Biol. 22 (4), 941951.
Nagamani, P.V., Chauhan, Prakash, Sanwlani, Nivedita, Ali, M.M., 2012. Articial Neural
Network (ANN) based inversion of benthic substrate bottom type and bathymetry
in optically shallow watersinitial model results. J. Indian Soc. Remote Sens. 40 (1),
137143.
Naveau, P., Juillet-Leclerc, A., Blamart, D., Correge, T., 2004. Sea surface salinity reconstruction in Fiji during the last century from multi-proxies of coral skeleton using neural
network: preliminary results. AGU Fall Meeting Abstracts vol. 1, p. 1345.
Neves, Brbara M., Du Preez, Cherisse, Edinger, Evan, 2014. Mapping coral and sponge
habitats on a shelf-depth environment using multibeam sonar and ROV video
observations: Learmonth Bank, northern British Columbia, Canada. Deep-Sea Res. II
Top. Stud. Oceanogr. 99, 169183.
Newman, Candace M., Knudby, Anders J., LeDrew, Ellsworth F., 2007. Assessing the effect
of management zonation on live coral cover using multi-date ikonos satellite imagery.
J. Appl. Remote. Sens. 1 (1), 011504.
OConnor, Raymond J., 2000. Expert systems, fuzzy logic, and coral reef development
under environmental stress. Conserv. Biol. 14 (3), 904906.

121

Orejas, Covadonga, Gori, Andrea, Iacono, Claudio Lo, Puig, Pere, Gili, Josep Maria, Dale,
Mark R.T., et al., 2009. Cold-water corals in the Cap de Creus canyon, northwestern
Mediterranean: spatial distribution, density and anthropogenic impact. Mar. Ecol.
Prog. Ser. 397, 3751.
Ortiz-Martin, A., 2014. A Coral Reefs Optimization Algorithm for Lemmings Game Resolution
Levels Master's thesis Universidad Autnoma de Madrid (Supervisor: Dr. David
Camacho).
Ouillon, Sylvain, Douillet, Pierre, Andrfout, Serge, 2004. Coupling satellite data with in
situ measurements and numerical modeling to study ne suspended-sediment
transport: a study for the lagoon of New Caledonia. Coral Reefs 23 (1), 109122.
Pabortsava, Katsiaryna, Purser, Autun, Wagner, Hannes, Thomsen, Laurenz, 2011. The
inuence of drill cuttings on physical characteristics of phytodetritus. Mar. Pollut.
Bull. 62 (10), 21702180.
Padmavathi, G., Muthukumar, M., Thakur, Suresh Kumar, 2010. Kernel principal
component analysis feature detection and classication for underwater images.
Image and Signal Processing (CISP), 2010 3rd International Congress on vol. 2. IEEE,
pp. 983988.
Palandro, David, Andrfout, Serge, Muller-Karger, Frank E., Dustan, Phillip, Hu,
Chuanmin, Hallock, Pamela, 2003. Detection of changes in coral reef communities
using Landsat-5 TM and Landsat-7 ETM+ data. Can. J. Remote. Sens. 29 (2), 201209.
Palandro, David A., Andrfout, Serge, Hu, Chuanmin, Hallock, Pamela, Mller-Karger,
Frank E., Dustan, Phillip, Callahan, Michael K., Kranenburg, Christine, Beaver, Carl R.,
2008. Quantication of two decades of shallow-water coral reef habitat decline in
the Florida Keys National Marine Sanctuary using Landsat data (19842002). Remote
Sens. Environ. 112 (8), 33883399.
Pham, Christopher K., Ramirez-Llodra, Eva, Alt, Claudia H.S., Amaro, Teresa, Bergmann,
Melanie, Canals, Miquel, Davies, Jaime, Duineveld, Gerard, Galgani, Franois,
Howell, Kerry L., et al., 2014. Marine litter distribution and density in european
seas, from the shelves to deep basins.
Phillips, Steven J., Anderson, Robert P., Schapire, Robert E., 2006. Maximum entropy
modeling of species geographic distributions. Ecol. Model. 190 (3), 231259.
Phinn, Stuart, Joyce, Karen, Scarth, Peter, Roelfsema, Chris, 2006. The role of integrated
information acquisition and management in the analysis of coastal ecosystem c.
Remote Sensing of Aquatic Coastal Ecosystem Processes. Springer, pp. 217249.
Pican, N., Trucco, E., Ross, M., Lane, D.M., Petillot, Y., et al., 1998. Texture analysis for
seabed classication: co-occurrence matrices vs. self-organizing maps. OCEANS98
Conference Proceedings vol. 1. IEEE, pp. 424428.
Pielou, Evelyn C., 1974. Population and Community Ecology: Principles and Methods. CRC
Press.
Pittman, Simon J., Costa, Bryan M., Battista, Tim A., 2009. Using LiDAR bathymetry and
boosted regression trees to predict the diversity and abundance of sh and corals.
J. Coast. Res. 2738.
Poli, Riccardo, Koza, John, 2014. Genetic Programming. Springer.
Poli, Riccardo, Langdon, William B., McPhee, Nicholas F., Koza, John R., 2008. A Field Guide
to Genetic Programming. Lulu.com.
Pratchett, Morgan S., Hoey, Andrew S., Wilson, Shaun K., 2014. Reef degradation and the
loss of critical ecosystem goods and services provided by coral reef shes. Curr. Opin.
Environ. Sustain. 7, 3743.
Purser, Autun, 2015. A time series study of Lophelia pertusa and reef megafauna responses
to drill cuttings exposure on the Norwegian Margin. PLoS One 10 (7), e0134076.
Purser, Autun, Thomsen, Laurenz, 2012. Monitoring strategies for drill cutting discharge
in the vicinity of cold-water coral ecosystems. Mar. Pollut. Bull. 64 (11), 23092316.
Purser, Autun, Bergmann, Melanie, Lundlv, Tomas, Nattkemper, Tim, Ontrup, Jrg, 2008.
The applicability of machine-learning algorithms for automated detection and
classication of cold-water coral habitats from video transect data. 4th International
Symposium on Deep-Sea Corals, Wellington, New Zealand, December 2008.
Purser, Autun, Bergmann, Melanie, Lundlv, Tomas, Ontrup, Jrg, Nattkemper, Tim W.,
2009. Use of machine-learning algorithms for the automated detection of
cold-water coral habitats: a pilot study. Mar. Ecol. Prog. Ser. 397, 241251.
Purser, Autun, Orejas, Covadonga, Gori, Andrea, Tong, Ruiju, Unnithan, Vikram, Thomsen,
Laurenz, 2013a. Local variation in the distribution of benthic megafauna species associated with cold-water coral reefs on the Norwegian Margin. Cont. Shelf Res. 54,
3751.
Purser, Autun, Thomsen, Laurenz, Barnes, Chris, Best, Mairi, Chapman, Ross, Hofbauer,
Michael, Menzel, Maik, Wagner, Hannes, 2013b. Temporal and spatial benthic data
collection via an internet operated deep sea crawler. Methods Oceanogr. 5, 118.
Putra, Mochamad Iqbal Herwata, Afatta, Siham, Wilson, Joanne, Muljadi, Andreas,
Yusidarta, Isai, 2015. Coral reef resilience in 17 islands marine recreation park,
riungan assessment of functional groups of herbivorous sh and benthic substrate.
Prog. Environ. Sci. 23, 230239.
Randy, L., 2004. Haupt and Sue Ellen Haupt. Practical Genetic Algorithms. John Wiley &
Sons.
Riegl, Bernhard M., Purkis, Samuel J., 2005. Detection of shallow subtidal corals from
IKONOS satellite and QTC View (50, 200 kHz) single-beam sonar data (Arabian
Gulf; Dubai, UAE). Remote Sens. Environ. 95 (1), 96114.
Rioja-Nieto, Rodolfo, Sheppard, Charles, 2008. Effects of management strategies on
the landscape ecology of a marine protected area. Ocean Coast. Manag. 51 (5),
397404.
Roberts, J. Murray, 2009. Cold-Water Corals: the Biology and Geology of Deep-sea Coral
Habitats. Cambridge University Press.
Ross, Timothy J., 2009. Fuzzy Logic with Engineering Applications. John Wiley & Sons.
Rowlands, Gwilym, 2013. Remote Sensing the Diversity, Distribution and Resilience of
Coral Reef Environments PhD thesis Nova Southeastern University.
Rowlands, Gwilym, Purkis, Sam, Riegl, Bernhard, Metsamaa, Liisa, Bruckner, Andrew,
Renaud, Philip, 2012. Satellite imaging coral reef resilience at regional scale. A
case-study from Saudi Arabia. Mar. Pollut. Bull. 64 (6), 12221237.

122

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123

Rowlands, Gwilym, Purkis, Sam, Bruckner, Andrew, 2015. Tight coupling between coral
reef morphology and mapped resilience in the Red Sea. Mar. Pollut. Bull.
Ruhl, Henry A., 2007. Abundance and size distribution dynamics of abyssal epibenthic
megafauna in the northeast Pacic. Ecology 88 (5), 12501262.
Ruitenbeek, Jack, Ridgley, Mark, Dollar, Steve, Huber, Richard, 1999. Optimization of
economic policies and investment projects using a fuzzy logic based cost-effectiveness
model of coral reef quality: empirical results for Montego Bay, Jamaica. Coral Reefs 18
(4), 381392.
Salcedo-Sanz, Sancho, Pastor-Snchez, A., Gallo-Marazuela, D., Portilla-Figueras, Antonio,
2013. A novel Coral Reefs Optimization algorithm for multi-objective problems. Intelligent Data Engineering and Automated LearningIDEAL 2013. Springer, pp. 326333.
Salcedo-Sanz, S., Del Ser, J., Landa-Torres, I., Gil-Lpez, S., Portilla-Figueras, J.A., 2014a. The
Coral Reefs Optimization algorithm: a novel metaheuristic for efciently solving
optimization problems. Sci. World J. 2014.
Salcedo-Sanz, S., Pastor-Snchez, A., Prieto, L., Blanco-Aguilera, A., Garca-Herrera, R.,
2014b. Feature selection in wind speed prediction systems based on a hybrid coral
reefs optimization extreme learning machine approach. Energy Convers. Manag.
87, 1018.
Salcedo-Sanz, S., Gallo-Marazuela, D., Pastor-Snchez, A., Carro-Calvo, L., Portilla-Figueras,
A., Prieto, L., 2014c. Offshore wind farm design with the Coral Reefs Optimization
algorithm. Renew. Energy 63, 109115.
Salcedo-Sanz, S., Casanova-Mateo, C., Pastor-Snchez, A., Snchez-Girn, M., 2014d. Daily
global solar radiation prediction based on a hybrid Coral Reefs OptimizationExtreme
Learning Machine approach. Sol. Energy 105, 9198.
Salcedo-Sanz, S., Sanchez-Garcia, J.E., Portilla-Figueras, J.A., Jimenez-Fernandez, S.,
Ahmadzadeh, A.M., 2014e. A coral-reef optimization algorithm for the optimal
service distribution problem in mobile radio access networks. Trans. Emerg. Telecomm.
Technol. 25 (11), 10571069.
Salcedo-Sanz, S., Garca-Daz, P., Portilla-Figueras, J.A., Del Ser, J., Gil-Lpez, S., 2014f. A
Coral Reefs Optimization algorithm for optimal mobile network deployment with
electromagnetic pollution control criterion. Appl. Soft Comput. 24, 239248.
Salcedo-Sanz, S., Pastor-Snchez, A., Del Ser, J., Prieto, L., Geem, Z.W., 2015a. A Coral Reefs
Optimization algorithm with Harmony Search operators for accurate wind speed
prediction. Renew. Energy 75, 93101.
Salcedo-Sanz, S., Pastor-Snchez, A., Portilla-Figueras, J.A., Prieto, L., 2015b. Effective
multi-objective optimization with the coral reefs optimization algorithm. Eng.
Optim. 119.
Sarkis, Samia, van Beukering, Pieter J.H., McKenzie, Emily, Brander, Luke, Hess, Sebastiaan,
Bervoets, Tadzio, der Putten, Lois Looijenstijn-van, Roelfsema, Mark, 2013. Total economic value of Bermuda's coral reefs: A summary. Coral Reefs of the United Kingdom
Overseas Territories. Springer, pp. 201211.
Schoening, T., Bergmann, M., Ontrup, J., Taylor, J., Dannheim, J., et al., 2012. Semiautomated image analysis for the assessment of megafaunal densities at the Arctic
deep-sea observatory Hausgarten. PLoS One 7 (6), e38179.
Schlkopf, Bernhard, Smola, Alex, 1998. Support vector machines. Encyclopedia of
Biostatistics.
Schlkopf, Bernhard, Smola, Alexander J., 2001. Learning with Kernels: Support Vector
Machines, Regularization, Optimization, and Beyond. MIT Press.
Schlkopf, Bernhard, Smola, Alexander J., 2002. Learning with Kernels: Support Vector
Machines, Regularization, Optimization, and Beyond. MIT Press.
Schuyler, Qamar, Dustan, Phillip, Dobson, Eric, 2006. Remote sensing of coral reef
community change on a remote coral atoll: Karang Kapota, Indonesia. Proceedings
of 10th International Coral Reef Symposium vol. 1770.
Scoplitis, Julie, Andrfout, Serge, Largout, Christine, 2007. Modelling coral reef habitat
trajectories: evaluation of an integrated timed automata and remote sensing
approach. Ecol. Model. 205 (1), 5980.
Scoplitis, J., Andrfout, Serge, Phinn, S., Chabanet, Pascale, Naim, O., Tourrand,
Catherine, Done, T., 2009. Changes of coral communities over 35 years: integrating
in situ and remote-sensing data on Saint-Leu Reef (La Runion, Indian Ocean). Estuar.
Coast. Shelf Sci. 84 (3), 342352.
Scoplitis, Julie, Andrfout, Serge, Phinn, Stuart, Arroyo, Lara, Dalleau, Mayeul, Cros,
Annick, Chabanet, Pascale, 2010. The next step in shallow coral reef monitoring: combining remote sensing and in situ approaches. Mar. Pollut. Bull. 60 (11), 19561968.
Seber, George A.F., Lee, Alan J., 2012. Linear Regression Analysis. vol. 936. John Wiley & Sons.
Shantz, Andrew A., Ladd, Mark C., Shrack, Elizabeth, Burkepile, Deron E., 2015. Fishderived nutrient hotspots shape coral reef benthic communities. Ecol. Appl.
Sharma, S., Bahuguna, A., Chaudhary, N.R., Nayak, S., Chavan, S., Pandey, C.N., 2008. Status
and monitoring the health of coral reef using multi-temporal remote sensing a case
study of Pirotan Coral Reef Island, Marine National Park, Gulf of Kachchh, Gujarat,
India. Proceedings of 11th International Coral Reef Symposium, pp. 647651.
Shihavuddin, A.S.M., Gracias, Nuno, Garcia, Rafael, Gleason, Arthur C.R., Gintert, Brooke,
2013. Image-based coral reef classication and thematic mapping. Remote Sens. 5
(4), 18091841.
Shinn, Eugene A., Smith, Garriet W., Prospero, Joseph M., Betzer, Peter, Hayes, Marshall L.,
Garrison, Virginia, Barber, Richard T., 2000. African dust and the demise of Caribbean
coral reefs. Geophys. Res. Lett. 27 (19), 30293032.
Siddique, Nazmul, Adeli, Hojjat, 2013. Computational Intelligence: Synergies of Fuzzy
Logic, Neural Networks and Evolutionary Computing. John Wiley & Sons.
Simon, Dan, 2013. Evolutionary Optimization Algorithms. John Wiley & Sons.
Smola, Alex J., Schlkopf, Bernhard, 2004. A tutorial on support vector regression. Stat.
Comput. 14 (3), 199222.
Soriano, Maricor, Marcos, Sheila, Saloma, Caesar, Quibilan, Miledel, Alino, Porrio, 2001.
Image classication of coral reef components from underwater color video. Oceans,
2001. MTS/IEEE Conference and Exhibition vol. 2. IEEE, pp. 10081013.
Steinwart, Ingo, Christmann, Andreas, 2008. Support vector machines. Springer Science &
Business Media.

Stockwell, David, 1999. The GARP modelling system: problems and solutions to automated
spatial prediction. Int. J. Geogr. Inf. Sci. 13 (2), 143158.
Stumpf, Richard P., Holderied, Kristine, Sinclair, Mark, 2003. Determination of water
depth with high-resolution satellite imagery over variable bottom types. Limnol.
Oceanogr. 48 (1part2), 547556.
Subade, Rodelio F., Francisco, Herminia A., 2014. Do non-users value coral reefs?:
economic valuation of conserving Tubbataha Reefs, Philippines. Ecol. Econ. 102,
2432.
TEEB, 2010. The economics of ecosystems and biodiversity: mainstreaming the economics of
nature: a synthesis of the approach, conclusions and recommendations of TEEB. The
Economics of Ecosystems and Biodiversity (TEEB) global initiative (http://www.
teebweb.org).
Thomsen, Laurenz, Purser, Autun, Schwendner, Jakob, Duda, Alexander, Flogen, Sascha,
Kwasnitschka, Tom, Pfannkuche, Olaf, Wilde, Detlef, Rosta, Roland, 2015. Temporal
and spatial benthic data collection via mobile robots: present and future applications.
OCEANS 2015-Genova. IEEE, pp. 15.
Tittensor, Derek P., Mora, Camilo, Jetz, Walter, Lotze, Heike K., Ricard, Daniel, Vanden
Berghe, Edward, Worm, Boris, 2010. Global patterns and predictors of marine biodiversity across taxa. Nature 466 (7310), 10981101.
Tong, Ruiju, Purser, Autun, Guinan, Janine, Unnithan, Vikram, 2013. Modeling the habitat
suitability for deep-water gorgonian corals based on terrain variables. Ecol. Inform.
13, 123132.
Toth, Lauren T., Aronson, Richard B., Cobb, Kim M., Cheng, Hai, Edwards, R. Lawrence,
Grothe, Pamela R., Sayani, Hussein R., 2015. Climatic and biotic thresholds of coralreef shutdown. Nat. Clim. 5 (4), 369374.
Turner, Woody, Spector, Sacha, Gardiner, Ned, Fladeland, Matthew, Sterling, Eleanor,
Steininger, Marc, 2003. Remote sensing for biodiversity science and conservation.
Trends Ecol. Evol. 18 (6), 306314.
van Zanten, Boris T., van Beukering, Pieter J.H., Wagtendonk, Alfred J., 2014. Coastal protection by coral reefs: a framework for spatial assessment and economic valuation.
Ocean Coast. Manag. 96, 94103.
Vapnik, Vladimir, 2013. The Nature of Statistical Learning Theory. Springer Science &
Business Media.
Vermeij, M.J.A., 2005. Substrate composition and adult distribution determine recruitment
patterns in a caribbean brooding coral. Mar. Ecol. Prog. Ser. 295, 123133.
Wahidin, Nurhalis, Siregar, Vincentius P., Nababan, Bisman, Jaya, Indra, Wouthuyzen,
Sam, 2015. Object-based image analysis for coral reef benthic habitat mapping with
several classication algorithms. Prog. Environ. Sci. 24, 222227.
Wang, L., Sousa, W.P., Gong, P., 2004. Integration of object-based and pixel-based
classication for mapping mangroves with ikonos imagery. Int. J. Remote Sens.
25 (24), 56555668.
Wang, Y.Y., Liu, Z.G., Li, Jing, Chen, Y.H., 2007. A review of coral reef remote sensing. Adv.
Earth Sci. 12 (4), 396402.
Warren, Christopher, Dupont, Jennifer, Abdel-Moati, Mohamed, Hobeichi, Sanaa,
Palandro, David, Purkis, Sam, 2015. Remote sensing of qatar nearshore habitats
with perspectives for coastal management. Mar. Pollut. Bull.
Watts, Michael J., Li, Yuxiao, Russell, Bayden D., Mellin, Camille, Connell, Sean D.,
Fordham, Damien A., 2011. A novel method for mapping reefs and subtidal rocky
habitats using articial neural networks. Ecol. Model. 222 (15), 26062614.
Webster, Jody M., Wallace, Laura M., Clague, David A., Braga, Juan Carlos, 2007. Numerical
modeling of the growth and drowning of hawaiian coral reefs during the last two
glacial cycles (0250 kyr). Geochem. Geophys. Geosyst. 8 (3).
Weijerman, Mariska, Fulton, Elizabeth A., Kaplan, Isaac C., Gorton, Rebecca, Leemans,
Rik, Mooij, Wolf M., Brainard, Russell E., 2015. An integrated coral reef ecosystem
model to support resource management under a changing climate. PLoS One 10
(12).
Whorff, J.S., Grifng, L., 1992. A video recording and analysis system used to sample
intertidal communities. J. Exp. Mar. Biol. Ecol. 160 (1), 112.
Wilson, Shaun K., Graham, Nicholas A.J., Pratchett, Morgan S., Jones, Geoffrey P., Polunin,
Nicholas V.C., 2006. Multiple disturbances and the global degradation of coral reefs:
are reef shes at risk or resilient? Glob. Chang. Biol. 12 (11), 22202234.
Woodroffe, Colin D., Webster, Jody M., 2014. Coral reefs and sea-level change. Mar. Geol.
352, 248267.
Wooldridge, Scott, Done, Terry, 2004. Learning to predict large-scale coral bleaching from
past events: a Bayesian approach using remotely sensed data, in-situ data, and
environmental proxies. Coral Reefs 23 (1), 96108.
Xu, Jingping, Zhao, Dongzhi, 2014. Review of coral reef ecosystem remote sensing. Acta
Ecol. Sin. 34 (1), 1925.
Yager, Ronald R., Zadeh, Lot A., 2012. An Introduction to Fuzzy Logic Applications in
Intelligent Systems vol. 165. Springer Science & Business Media.
Yamamoto, Takaaki Maekawa Hidekazu, Sugiura, Norio, 2002. Coral growth processes using multiple regression analysis and neural network model. Ecol. Eng. 14
(3), 311.
Yamano, H., 2007. The use of multi-temporal satellite images to estimate intertidal reefat topography. J. Spat. Sci. 52 (1), 7379.
Yamazaki, Atsuko, Watanabe, Tsuyoshi, Tsunogai, Urumu, 2011. Nitrogen isotopes of organic nitrogen in reef coral skeletons as a proxy of tropical nutrient dynamics.
Geophys. Res. Lett. 38 (19).
Yang, Xin-She, Cui, Zhihua, Xiao, Renbin, Gandomi, Amir Hossein, Karamanoglu, Mehmet,
2013. Swarm Intelligence and Bio-Inspired Computation: Theory and Applications.
Newnes.
Yang, Zhiyong, Zhang, Taohong, Zhang, Dezheng, 2015. A novel algorithm with differential
evolution and coral reef optimization for extreme learning machine training. Cogn.
Neurodyn. 111.
Yokoki, H., Yamano, H., Kayanne, H., Sato, D., Shimazaki, H., Yamaguchi, T., Chikamori, M.,
Ishoda, A., Takagi, H., 2006. Numerical calculations of longshore sediment transport

S. Salcedo-Sanz et al. / Ecological Informatics 32 (2016) 107123


due to wave transformation in the lagoon of Majuro Atoll, Marshall Islands. Proceedings
of the 10th International Coral Reef Symposium vol. 1570, p. 1576.
Yu, Xinjie, Gen, Mitsuo, 2010. Introduction to Evolutionary Algorithms. Springer Science &
Business Media.
Zadeh, Lot A., 1965. Fuzzy sets. Inf. Control. 8 (3), 338353.
Zadeh, Lot A., 2005. From imprecise to granular probabilities. Fuzzy Sets Syst. 154 (3),
370374.

123

Zadeh, Lot A., 2008. Is there a need for fuzzy logic? Inf. Sci. 178 (13), 27512779.
Zhang, Caiyun, 2015. Applying data fusion techniques for benthic habitat mapping and
monitoring in a coral reef ecosystem. ISPRS J. Photogramm. Remote Sens. 104,
213223.