Assessing the role of climate induced sea-level changes on the evolution of body size in insular mammals in the Mediterranean Sea since the late Pliocene

(1)

level changes on the evolution of body size

in insular mammals in the Mediterranean

Sea since the late Pliocene

__________________________________________________

Keywords: biogeography, species distribution modelling, ArcGIS, MaxEnt, island rule, evolution, isolation, body size, sea-level, last glacial maximum, last interglacial period, Pliocene warm period, mid-Holocene

Jasper Steenvoorden

UvA ID: 10552510

Supervisor: Kenneth Rijsdijk

Co-supervisor: John de Vos

(2)

J. Steenvoorden | Bachelor Thesis Future Planet Studies (Earth Sciences) | 4 July 2016

I. Abstract

Earlier research has shown that islands in the Mediterranean Sea contained an extremely unbalanced insular mammal fauna, consisting mainly of species of deer, elephants and hippo’s, while lacking carnivores. Furthermore, a striking characteristic of these insular mammals was their deviation in size from their mainland ancestor, in which large mammals became significantly smaller while small rodents evolved to be larger. Many hypotheses have been brought forward to explain this trend, including the importance of parameters such as isolation, ecological interactions, predation and also climate. However, there is still controversy on which of these factors are influential and which are not. This research assessed the influence of climate induced sea-level changes on the evolution of body size in insular mammals in the Mediterranean Sea by reconstructing the physical characteristics, geographic configuration and species distribution of the islands of Crete, the Cyclades, Cyprus, Lesbos and Malta during the Pliocene Warm Period, Last Interglacial Period, Last Glacial Maximum and Mid-Holocene. It was found out that if climate change in the form of sea-level has a direct influence on the evolution of body size of insular mammals on these islands, short-term interglacial high sea-level periods and their influence on island physical characteristics and resources are important in determining body size. Furthermore, computational research with the help of ArcGIS and MaxEnt has yielded some interesting and promising results for use in future research on the influence of climate on island environments and their biota. Despite, improvements have to be made in acquisition of more and higher resolution data.

(3)

II. III. Introduction

The Aegean archipelago and various other islands in the Mediterranean Sea share an interesting and complex paleogeographic history. As a result, their biogeography has been of interest to many researchers in past decades (Sfenthourakis, 1996). Especially since the release of Wilson & MacArthur’s publication on the equilibrium theory of island biogeography in 1967, the study of island flora and fauna has increased drastically (De Vos et al., 2007).

The paleo- and biogeographic complexity of islands in the Mediterranean Sea originate especially from drastic changes in tectonic movement and eustatic sea level change during the late Pleisto- and Holocene(Sfenthourakis, 1996; Perissoratis & Conispoliatis, 2003). For instance, based on global sea level curves and local geological and geoarchaeological data, it was found out that many islands varied significantly in size and were connected to the mainland and one another during various periods ranging from 21500BP-8000BP (Perissoratis & Conispoliatis, 2003). Consequently, these alternating periods of emergence and submergence of Mediterranean islands have had a notable impact on extinctions and the insular evolution of mammals in the region (De Vos et al. 2007).

Even in the early 20th century, paleontologists such as Dorothea Bate found out that there is a similar presence of specific mammal fossils on the islands in the Mediterranean Sea, such as elephants, deer, hippo’s and rats (De Vos et al., 2007). Sondaar (1986) discussed the causes of past mammal distribution on islands in the Mediterranean Sea and the unbalanced nature of fauna on these islands in general. It was found out that these presently extinct mammals, which were considered good swimmers, arrived on these islands swimming by means of a concept called sweepstakes dispersal (Simpson, 1940). Sweepstakes dispersal means mammals arrived on these relatively distant islands on rare occasions, as the geographical route that had to be taken was impossible for many other species and was most often a one-way affair (Sondaar, 1986). As a result, the large bodies of water between the mainland and the islands functioned as a filter for only the strongest of swimmers, leading to a strong selective event that may explain the unbalanced fauna on these islands (Simpson, 1940). Furthermore, this suggested the fauna on the island remained nearly isolated (Sondaar, 1986). This insight contrasted the theory of land-bridges as a means of dispersal to these islands.

Furthermore, a striking feature that was observed from research on many of these islands is that mammals existed in an extremely wide variety of sizes and also differed significantly in size from their mainland ancestor (De Vos et al., 2007). These processes of substantial body mass change are called dwarfism and gigantism (Van der Geer et al., 2010). For example, the elephant species found on Naxos (Paleoloxodon Antiquus), one of the Aegean Islands, had a body size of only 10% as compared to its mainland ancestor (Van Der Geer et al., 2014). Furthermore, De Vos (1979) distinguished six size groups of cervids (Candiacervus sp.) on Crete that differed in withers size ranging from 40 centimeters to 165 centimeters, and the elephant found on Sicily (Paleooloxodon

(5)

Falconeri) is the most extreme event of dwarfism, shrinking to just 2% of the

size of its mainland ancestor (Van Der Geer et al., 2016). However, there are also many cases of gigantism. Even though these changes are not as spectacular in absolute size, they may well be large in relative terms (Van Der Geer et al., 2010). For instance, the moon-rat (Deinogalerix Koenigswaldi) of Gargano increased in size by approximately 207 times (Lomolino et al., 2013). Figure 1 shows some of the remarkable cases of gigantism and dwarfism in mammals in the Mediterranean compared to their mainland ancestor.

Lastly, De Vos (2006) and Sondaar (1986) argue that this unbalanced mammal fauna and wide variety in size does not only occur in the Mediterranean Sea, but rather that there are many parallels in the evolution of cervids and other mammals on islands around the world, such as on the Japanese Islands (the Ryukyu-islands), the Philippines (Masbate), Indonesia (the Lesser Sunda Islands) and the Channel Islands off California.

Figure 3.1: a variety of cases of body size evolution in insular mammals as compared to their mainland ancestor (from Lomolino et al., 2013)

As a result of these findings, many hypotheses have been brought forward to try and explain the trend of change in body size typically observed in these mammals (elephants, deer and hippos). This trend was coined ‘the Island Rule’ by Van Valen (1973), and has been used as the general term for this process by subsequent authors (Van Der Geer et al., 2016). According to Lomolino (2005), body size variation is one of the most fundamental responses to island environments, as it influences and represents a wide variety of significant physiological and ecological characteristics of mammals such as their immigration potential, ecological interactions and resource requirements. Consequently, body size variation in insular mammals may be a variable that can relate to the geographic characteristics and dynamics of the islands these species

(6)

evolved on. However, body size is not a perfect, nor the sole consequence of immigration potential, ecological interactions and resource requirements. Of equal importance are an island's climate, geographical complexity such as ruggedness of terrain, resource availability and isolation (Lomolino, 2005). Most likely, the situation that influences this evolution differs from island to island, and from taxon to taxon (Van Der Geer et al., 2010). Furthermore, the abovementioned factors are all closely correlated and affected by each other. Therefore, discriminating one of these factors as the primary factor for body size variation may be difficult (Heaney, 1978). On the other hand, Lomolino (2005) does state that studies that focus on within-archipelago patterns in body size variation of a particular species may provide especially important clues to the factors influencing evolution and assembly of biotic communities.

Accordingly, this research looked into the role that past climatic conditions in the form sea level change during glacial-interglacial cycles may have had on physical island characteristics (land area, isolation and connectivity) and the subsequent distribution and evolution of insular mammals on specific islands in the Mediterranean and Aegean Sea (Crete, Cyprus, Malta, Lesbos and the Cyclades). This was done by choosing five scenarios of relevant climatic conditions in past time (Present, Last Glacial Maximum, Last Interglacial Period, Pliocene Warm Period and the Mid-Holocene), which were then thoroughly researched on their (climatic) parameters such as temperature, precipitation but predominantly sea-level to allow for reconstruction. Thereafter, these reconstructed scenarios were used to assess and explain the effects of climatic parameters on the potential distribution of mammal species and suitable habitat on the selected islands. In particular it was focused on whether under certain former climatic regimes and their sea-level stand, populations could become separated from each other or if island characteristics would significantly influence factors such as island area and thus resource availability, resulting in different body sizes.

This research was done by using Geographical Information Systems (ArcGIS) in which relevant maps were created that could be analysed themselves and be used for the modelling of past species distribution in the computer model MaxEnt, based on literature findings of past biota in this region. These results were then used to more quantitatively assess the influence of climate and climate induced sea-level changes on body size evolution of insular mammals on these islands. The main research question for this thesis is therefore:

“How can ArcGIS and species distribution modelling of deer, elephants and hippo’s in MaxEnt during five past climate scenarios be used to assess the role of climate induced changes in physical island characteristics and geographic configuration on the evolution of body size typically observed in insular mammals in the Mediterranean Sea?”

Doing this is important, as ascertaining the effect of past climatic conditions on the evolution and distribution of past mammal populations may provide much

(7)

purposeful data for the assessment of current species responses to climate change. These records obtained from paleo ecological research may provide a great resource for conservation strategies and the future management of ecological and evolutionary processes on the planet (Willis et al., 2010). Furthermore, the use ArcGIS and the modelling of past extinct species distribution may provide novel insights in reconstructing past climates and the effects on the biota of these islands.

IV. Study Area

The study area of this research are the islands of Crete, Cyprus, Malta, Lesbos and the Cyclades, located in the Mediterranean and Aegean Sea. Figure 2 shows an overview of the researched islands and their locations.

Figure 4.1: Study area showing Europe on the left image and Crete (1), Cyclades (2), Cyprus (3), Malta (4) and Lesbos (5) on the right image

The numerous outcropping islands in the Aegean Sea, of which Crete, Lesbos and the Cyclades are included in this research, result from the complex geomorphology and thus paleogeography in this region (Kapsimalis et al., 2009 & Lykousis, 2009). As a consequence of the shallow depth (<250 meters) of this so called Cycladic Plateau, that divides the northern Aegean Sea from the southern Aegean Sea, it has been subject to several stages of emergence and submergence during the biggest part of the Pleistocene. During periods of low sea level in the Pleistocene, presently submerged parts of the Cycladic Plateau would emerge and form connected islands, or in periods of extensive sea level drop would form one big island. Furthermore, during some parts of the Pleistocene, it was connected to both the Greek and Eurasian mainland (Kapsimalis et al., 2009 & Lykousis, 2009). According to Kapsimalis et al. (2009), breakup of this mega island started around the start of the Holocene (circa 12.000BP).

It is unfortunate, however, that most of the islands in the Aegean Sea consist of mainly metamorphic and igneous rocks (Hejl et al., 2002). As a result of the lack of sedimentary deposits, this region only contains a few sporadic findings of endemic mammal fossils as compared to the stunning amount of fossils known to be found on other islands in the Mediterranean Sea (Van Der Geer et al., 2014). On the other hand, as according to Papazachos (1990), the

(8)

Cycladic Plateau was essentially aseismic during the Late Pleistocene and still is now, sea level changes may attribute as a sole factor to assessing its paleogeography up to the Late Pleistocene, which may make reconstructing past island configuration more easy. As a result, underwater surveys may also reveal interesting information regarding the biogeography of the Cycladic Plateau during this period in time (Kapsimalis et al., 2009).

Malta is the smallest of the investigated islands, compromising a land area of only 316 km2 _{(Vogiatzakis et al., 2008). However, despite its small size and}

relatively low relief and amount of habitats, Vogiatzakis et al. (2008) state Malta is an interesting biogeographic area due to its history of land bridges with Sicily. Lastly, Cyprus is the largest of the researched islands with a land area of 9251 km2_{. Consequently, it varies sufficiently in relief, is relatively isolated to support}

high habitat and species diversity, but still close enough to the surrounding continent to be influenced by continental biotas (Vogiatzakis et al., 2008).

V. Methodology

At the center of this research was an analysis of island physical characteristics (land area, isolation and connectivity), geographic configuration, climate conditions and the species distribution of insular mammals in the Mediterranean Sea during four past scenarios and a full glacial cycle in order to explain, with the help of literature, the role of climate driven sea-level variability on the dispersal and evolution of body size in these mammals. To do this analysis, the computer programmes ArcGIS and MaxEnt were used.

However, before this research could be done, understanding sufficiently how these programmes work and how climate evolved in the Mediterranean Sea since the start of the Pliocene (3.1 million years ago) was of utmost importance. As a result, this research consisted of two parts. Firstly, a literature review was conducted to build a background which helped in understanding and comparing the results of the analysis in ArcGIS and MaxEnt, which was the second part of the research.

The first paragraph of this chapter will go into more detail about the five scenarios used for reconstruction and the functionalities of ArcGIS and MaxEnt. Thereafter, the second paragraph will represent the utilized data and explain how this data was processed to be able to be used in the ArcGIS and species distribution analysis.

a. Literature review

In order to understand climate variability and its influence on physical island configuration and biota in the Mediterranean Sea since the Pliocene, literature was investigated to help select five relevant reconstructions in paleoclimatic and physical paleogeographic history. These scenarios were then assessed individually with GIS on specifically the mentioned islands in chapter 4 to understand precisely how these islands have reacted to climatic changes in the

(9)

past. The five chosen scenarios are described chronologically in more detail below.

Scenarios

Present

The first scenario that has been determined is the present. The present is an important time scenario to consider in researching the biogeography of this region for a number of reasons. Firstly, present environmental and geographical data of this region is available in large amounts, and as a result, the current configuration of these islands can be represented easily and serve as a baseline for the scenarios of the past. Secondly, researching the past is by no means useful if data on the present configuration of these islands is not known, because the significance of change in paleo-climatological variables such as precipitation during a past scenario is not known if the present precipitation of that area is not available. Lastly, it is important to understand the present configuration of the study area to know with what kind of an environment is dealt with during the research. Concluding, the present was used as a baseline that can serve to understand the significance of changes in configuration of these islands during the past.

Mid-Holocene (MH) – circa 6000 years BP

The second scenario that is reconstructed is the MH, which differs from the last three scenarios considering it plays in more recent times. The period around the MH is known as a period of profound changes in climate (Steig, 1999). During this short interval, Steig (1999) stated that land air temperatures appeared to have declined across much of the globe. Furthermore, some areas experienced a dry period followed by increasingly wet and cool conditions. Lastly, this period in time also involved significant changes in atmospheric and oceanic circulation (Steig, 1999). On the other hand, sea-level changes due to ice melting from the major ice sheets during this period in time was suggested to be stabilized, with research from Siddall et al. (2003) stating that sea-level was approximately 6 to 8 meters lower than today.

Last Glacial Maximum (LGM) - circa 20.000 years BP

The third scenario that was reconstructed in paleoclimatic and paleogeographic history is the LGM. During this period in time, sea level had reached its lowest point, after which it started rising up until the present (Kapsimalis et al., 2009). According to a model developed by Lambeck & Purcell (2005), sea level was occurring at depths around 140 meters lower than present. Furthermore, global annual surface temperature was estimated to be 4.5 degrees Celsius colder and precipitation was estimated to be 18% lower as compared to pre-industrial conditions (Otto-Bliesner et al., 2006). The main characteristic of the Aegean Sea

(10)

during this time is that the Cycladic Plateau formed a single large island, containing relatively flat plains with an average slope of only 1.58° (Kapsimalis et al., 2009). According to Van Der Geer et al. (2014), this may have formed suitable habitats for elephants and other mammals and their subsequent evolution. Just as with the Cycladic Plateau, the flat plains in between Malta and Sicily emerged, resulting in a land bridge for species to disperse (Vogiatzakis et al., 2008).

Last Interglacial Period (LIP)- circa 125.000 years BP

The LIP or the Eemian was the last interglacial period, in which sea level was indicated to peak 5.5 to 9 meters above present sea level. Furthermore, there was a substantial variability for peak sea level at a variety of different geographical sites (Dutton & Lambeck, 2012). According to the climate record of ice cores taken from a Greenland ice sheet during research from Andersen et al. (2004), temperature around the globe was approximately 5 degrees Celsius warmer than today. It is suggested that this configuration and especially the significant change of sea level and thus physical characteristics of islands and the mainland between the Last Interglacial Period and LGM has had an important influence on the dispersal of mammals on islands in the Mediterranean Sea.

Pliocene Warm Period (PWP) - circa 3.100.000 years BP

According to Salzmann et al. (2011), the Pliocene epoch, which spanned from approximately 5.33-2.58 million years ago, was a generally warmer and wetter period with slightly higher atmospheric CO2 concentrations than the modern climate. Temperature was estimated to be approximately 2 to 3 degrees Celsius higher than present, predominantly during the late Pliocene that spanned from 3.6-2.58 million years BP (Salzmann et al., 2011). Furthermore, Pliocene sea levels are estimated at a value of approximately 25 meters above present (Raymo et al., 2011). The main reason why studying the PWP as a past scenario is especially interesting is due to the fact that it had the highest sea-level until the present and it shares many similarities with the current climate. As a result, it may be used as a sea-level baseline and an analogue to gain a better understanding of future climate warming and thus geographic island configuration (De Boer et al., 2015).

ArcGIS and MaxEnt

ArcGIS is a computer programme that can create, display and investigate maps which can be used to answer geographical questions (Ormsby, 2004). With geographic information such as the digital elevation model (DEM) and paleogeographic and paleoclimatic information (i.e sea level, temperature and precipitation) of a specific island and during a specific scenario, ArcGIS may be used to produce several maps that will then be implemented in the species distribution model (SDM) MaxEnt.

(11)

A species distribution model is a model that can predict species distribution based on the environmental conditions of sites of known occurrences of a specific species. MaxEnt, which comes from maximum entropy method, uses this specific way of modelling species distribution based on presence-only data (Phillips et al., 2005). Here, fossil findings were used rather than occurrence data. The SDM that results from the implementation of maps from ArcGIS yields a new map that predicts the environmental suitability for the specific researched species as a function of the given environmental variables and the locations of fossil findings. As a result, examples of relevant maps that had to be created by ArcGIS and implemented in MaxEnt to provide accurate results are climatic, geomorphological, hydrological and vegetation maps such as precipitation, temperature, climate anomalies, slope, aspect and vegetation classes maps (Phillips et al., 2005).

MaxEnt was chosen and implemented as the SDM programme for this research as according to Merow et al. (2013), it is one of the most popular tools for species distribution modelling due to the fact that it outperforms many other methods on its predictive accuracy and is also very easy to use. It was especially interesting in this research as it can produce a SDM map by taking a set of environmental maps with predictors such as precipitation and temperature across a user defined landscape that is divided into grids, which is exactly the format that was created by the computer programme ArcGIS (Merow et al., 2013).

b. GIS and species distribution analysis

As the quantitative research in the form of a GIS and MaxEnt species distribution analysis was the core of this project, large datasets of environmental variables had to be gathered and processed in ArcGIS and MaxEnt to be able to reconstruct past climates and subsequently assess the influence of climate on the dispersal and evolution of mammals in the Mediterranean Sea. Below, the used programmes and data are shown, together with their sources and a short explanation of the content of the used datasets.

Programmes and datasets

Programmes

- ArcMap 10.1 and Arcmap 10.4 (obtained from

http://desktop.arcgis.com/en/arcmap/ with a university license) - MaxEnt version 3.3.3k (obtained from

https://www.cs.princeton.edu/~schapire/maxent/ and free for download)

Environmental Variables

- Bioclimatic variables 1950-2000 (obtained from www.worldclim.org)

(12)

J. Steenvoorden | Bachelor Thesis Future Planet Studies (Earth Sciences) | 4 July 2016 - Bioclimatic variables LGM (obtained from www.worldclim.org)

- Bioclimatic variables Last Interglacial Period (obtained from www.worldclim.org)

Bioclimatic variables are climatic parameters that are derived from the monthly temperature and rainfall values in order to generate more biologically meaningful variables such as seasonality and extreme or limiting environmental factors (e.g temperature of the coldest month). The bioclimatic variables from 1950-2000 that have been used for this research were obtained from ‘WorldClim – Global Climate Data’ (free for download) and created from a research by Hijmans et al. (2005). In this research, high resolution climate surfaces for global land areas were developed by interpolating annual data from many weather stations around the globe. For stations with records of multiple years, the average for the years 1960-1990 have been calculated, and in some cases, data was available for 1950-2000. The spatial resolution for both these datasets is 1 km2_.

For the development of the Mid-Holocene, LGM and Last Interglacial Period bioclimatic variables datasets, the present data from Hijmans et al. (2005) was downscaled and calibrated by using the Community Climate System Model 4 (CCSM4). In the case of the Last Interglacial Period, this was done in a research by Otto-Bliesner et al. (2008). The spatial resolution for the LGM data is 2.5 arc-minutes and the spatial resolution for the Mid-Holocene and Last Interglacial Period is 30 arc-seconds.

Global Relief Model (DEM and bathymetric data)

- The Global Multi-Resolution Topography Synthesis (GMRT) (obtained from

http://www.marine-geo.org/index.php)

The GMRT was used to obtain the DEM and bathymetric data for this research. This dataset was explicitly chosen as it was the highest resolution of combined DEM and bathymetric that was able to be found. The GMRT is maintained as a multi-resolution gridded global DEM that includes cleaned processed ship-based multi beam sonar data at their full spatial resolution. The GMRT was produced from the synthesis of a variety of elevation sources by Ryan et al. (2009). The spatial resolution of the DEM spans from 10m in portions of the USA to 5 km resolution in the Antarctic. For the research area, spatial resolution is between 100-200 meters.

Species Localities

(13)

J. Steenvoorden | Bachelor Thesis Future Planet Studies (Earth Sciences) | 4 July 2016 - Mammuthus Creticus localities map (obtained from Poulakakis et al.,

2002)

- Hippopotamus Creutzburgi localities map (obtained from Van Huut, 2007) - Hippopotamus Minor localities map (obtained from Van Huut, 2007)

- Elephas Mnaidriensis localities map (obtained from Van Huut, 2007)

- Hippopotamus Pentlandi localities map (obtained from Van Der Geer et al.,

2010)

The species localities maps were obtained from various researches during the 20th_{and 21}st_{century, in which fossil mammal findings were collected, identified}

and then represented in a map. Data processing

Before the obtained datasets could be used for ArcGIS analysis and species distribution modelling in MaxEnt, these first had to be processed in ArcGIS to reconstruct the past scenarios and to create the format that is required for use in MaxEnt. In order to do so, a variety of steps had to be undertaken. This paragraph will deal with the undertaken steps in short. However, appendix 1 contains a detailed manual of how the steps taken and how the data is processed in ArcGIS and MaxEnt for this research. It is suggested that appendix 1 will be looked at in more detail, as this may greatly help in understanding and visualizing the used data.

Reconstructing past climate scenarios

For the reconstruction of each past climate scenario, several maps were created that represent the physical characteristics and climate and habitat conditions of every researched island during that specific scenario. In order to do so, the first step was to use DEM from GMRT in combination with literature to reconstruct the physical characteristics and geographic configuration of these islands. This can be done when sea-level at a particular scenario is known. For example, it was stated in research by Lambeck & Purcell (2005) that sea-level during the LGM was occurring approximately 140 meters below present. In ArcGIS, it could then be calculated which grid cells occur above -140 meters and thus form the land area of each island during the LGM. Thereafter, the environmental variables datasets that were obtained from ‘WorldClim – Global Climate Data’ were interpolated to the extent of every researched island during that specific scenario with the help of kriging (based on spatial autocorrelation) to represent the specific environmental variables. The environmental variables that were used can be related to three principal traits: temperature, precipitation and topography. The total list of used environmental variables for the ArcGIS analysis and MaxEnt species distribution modelling is given in table 5.1 below.

(14)

Nr. Layer Name Type of data Source 1 Annual Precipitation Continuous WorldClim 2 Precipitation of Driest Quarter Continuous WorldClim 3 Precipitation of Wettest Quarter Continuous WorldClim 4 Annual Mean Temperature Continuous WorldClim 5 Temperature Annual Range Continuous WorldClim 6 Digital Elevation Model Continuous GMRT

7 Aspect Map Continuous GMRT

8 Slope Map Continuous GMRT

Table 5.1: table showing total list of used environmental variables

Georeferencing of species localities

Before the species localities maps could be used to accurately represent the locations of the fossil findings, these first had to be georeferenced. This is important, as the maps from the research these were obtained from do not contain spatial referencing. With georeferencing, these maps were processed so that they precisely represent the present DEM of the island and thus could be utilized to add XY coordinates to the fossil findings.

Editing formats for MaxEnt

MaxEnt is extremely strict regarding its compatible formats and this required the environmental variables and species localities to be edited to the format that is compatible for MaxEnt. For example, all datasets need the same geographic bounds, cell size and projection system. Fortunately, ArcGIS could be used to convert the created environmental variables and species localities maps to this format. Species localities had to be converted to a .csv file and comma delimited. The environmental variable maps had to be converted a .asc file. Furthermore, the spatial reference for all environmental layers was set to be WGS_1984_UTM_Zone_35N.

Deliverables

Besides the availability of useful maps in ArcGIS, much more deliverables became available after the data was implemented in MaxEnt. As was mentioned before, MaxEnt for instance yields a map that predicts the environmental suitability for the specific researched species as a function of the given environmental variables and the location of fossil findings. A cumulative output was used, and can be interpreted as the sum of the probabilities of all grid cells with no higher probability than the grid cell times 100. For example, the grid cell that is predicted as having the best conditions for the species, according to the model, will have cumulative value 100, while cumulative values close to 0 indicate predictions of unsuitable conditions (Phillips et al., 2005). As a result, habitat was concluded to be suitable with a cumulative value over 1.6. MaxEnt offers significantly more information that may be used to analyse results. For

(15)

example, it offers an analysis of variable contributions and jackknife plots, which were analysed in the results. A variable contribution table show the importance of a specific environmental variables in determining suitable habitat conditions, and a jackknife plot results from a jackknife resampling method, a statistical method in which observations are left out systematically to estimate the importance of a specific observation. If the regularized training gain is low without the variable or high as the only variable, it means it is of high importance to determining suitable habitat conditions for that specific species (Efron & Stein, 1981).

Despite, there were no sufficient species localities data for Lesbos and the Cyclades. As a result, no species distribution analysis was conducted for these islands. In addition to that, the Pliocene Warm Period and LGM were first chosen as the two scenarios for species distribution modelling as they are the two most extreme cases of sea-level changes. However, during the Pliocene Warm Period, some of the species localities of Candiacervus and Hippopotamus Minor were situated in an area that was flooded during that scenario, leading to an error. As a result, the Last Interglacial Period has been chosen instead of the Pliocene Warm Period. Lastly, a species distribution analysis has not been conducted for the species Elephas Mnaidriensis, as there were only two fossil findings for these species, which is too few to yield reliable results.

VI. Anticipated results and hypotheses

Literature suggests that climate induced sea-level changes do have an influence on the evolution of body size variation of insular mammals. This is due to the fact that sea-level changes in general have a significant effect on a variety of factors that may influence body size of insular mammals (Lomolino, 2005). For example, Meiri et al. (2005) state that island area, which is influenced by climate induced sea-level changes, is theoretically the most important factor in determining the evolution of body size, as it notably influences resource availability and ecological interactions on an island. Moreover, isolation can also influence the size evolution of insular mammals, as larger individuals are more likely survive immigration attempts to islands (Meiri et al., 2005). Lastly, connection with the mainland may inhibit body size evolution through predation, an increase in resources and gene exchange with larger mainland species (Van Der Geer et al., 2010). Table 6.1 schematically shows the four types of changes in island physical characteristics and geographic configuration that are typically observed as a result of sea-level decrease during glacials and sea-level increase during interglacials, together with their potential associated biogeographical consequences.

Evolution of body size on insular mammals in the Mediterranean Sea most likely also differs per island and per taxon (Van Der Geer et al., 2010). The difference per island may be caused by for instance factors such as competition and predation pressures, which are largely a result of island size and thus may differ per island (Melton, 1982). Furthermore, due to differences in bathymetry,

(16)

a 2 meter sea level rise may for example create geographic barriers on one island while on another it will not, an example that may also cause evolution to differ per island. According to Heaney (1978), size evolution may also differ per taxon, because it is large animals that are most affected by factors such as resource limitation, while small mammals are most affected by reduced levels of predation and competition. Therefore, the hypotheses for this research were:

1. Climate induced changes in sea-level directly influence body size evolution of insular mammals in the Mediterranean Sea through its influence on land area and isolation and thus resource availability and ecological interactions.

2. Evolution of body size in insular mammals in the Mediterranean Sea differs per island due the heterogeneous influence of sea-level changes on an islands physical characteristics and geographic configuration and due to differences in an islands general physical characteristics and geographic configuration.

Table 6.1: four types of changes in physical island characteristics as a result of sea-level changes and their expected influence on body size evolution (from Hammoud, 2016) edited for this research

(17)

VII. Results

In this chapter, the results of the ArcGIS analysis and species distribution analysis will be dealt with. This will be done in such a way that insight is gained in the dynamics of the researched islands and their specific insular mammals since the Pliocene Warm Period, so that this information can be elaborated on in the discussion in the next chapter. The results consist of figures that represent the DEM or physical island characteristics and species distribution during each scenario, together with a detailed explanation. The results from the other environmental variables will not be displayed and elaborated on in this section, but rather will be dealt with in combination with the literature in the discussion in the next chapter.

a. Crete

ArcGIS analysis

On Crete, changes in island size as a result of sea-level changes seem relatively insignificant at first glance (figure 7.1). However, the land area difference between the highest sea-level during the Pliocene Warm Period and the lowest sea-level during the LGM still yields approximately a 41.3% increase in size from 8780 km2_{during the Pliocene Warm Period to 12405 km}2 _{during the LGM. The}

increase in area occurs mostly on the northern part of the island, where a shallow plateau is situated. After the LGM, the land area of Crete gradually decreased to 9420 km2_{during the present.}

Furthermore, Crete remains rather unchanged regarding geographic isolation as a result of sea-level change, a parameter that is of significant importance for species to evolve. Apart from an increase in size of smaller islands surrounding Crete and the formation of a peninsula in the north-western part of Crete since the PWP, the island does not appear to create important geographically isolated regions. It’s distance to the mainland also does not decrease notably. However, the islands of Kythira to the north-west and Karpathos to the north-east increase in size significantly, which may have allowed for easier use as a stepping stone for insular mammals. Despite, these characteristics are in sharp contrast with the islands of the Cyclades, Lesbos and Malta, which will be dealt with in the next paragraph.

(18)

Figure 7.1: five maps showing the Digital Elevation Model of Crete during (A) the Pliocene Warm Period, (B) the Last Interglacial Period, (C) the LGM, (D) the Mid-Holocene and (E) the Present Species distribution analysis

Candiacervus sp.

As a consequence of the fact that for the Candiacervus sp., or Cretan deer, fossil occurrence data is predominantly near the coast, predicted suitable habitat conditions in both scenarios are near the coast as well (figure 7.2).

As a result of this, and from a comparison between the variable importance tables 7.1 and 7.2 and the jackknifes in figure 7.3 and 7.4 it can be concluded that during both scenarios, the DEM is the highest contributing variable to determine the suitable habitat conditions. However, during the Last Interglacial Period, the second and third highest contributing variables are the temperature annual range and annual precipitation, while during the LGM just the annual precipitation is of importance.

Moreover, the amount of suitable habitat conditions for Candiacervus sp. increase significantly during the LGM as compared to the Last interglacial Period. The suitable habitat conditions were approximately 1737 km2_{during the Last}

Interglacial Period, which increased to about 3494 km2_{during the LGM – an}

increase of 101.15%. This may be partly because of the different climatic conditions, but also greatly due to a general increase in land area.

(19)

Figure 7.2: predicted suitable habitat conditions for Candiacervus sp. during (A) the Last Interglacial Period and (B) the LGM

(20)

Variable Percent contribution Permutation importance

DEM 60.7 39.3

Temperature Annual Range 12.7 3.3

Annual Precipitation 11.5 12.5

Annual Mean Temperature 8.3 24.3

Aspect 5.7 0.2

Precipitation Wettest

Quarter 0.5 0

Precipitation Driest Quarter 0.5 20.4

Slope 0.1 0

Table 7.1: table showing variable contributions for Candiacervus sp. during the Last Interglacial Period

Figure 7.4: graph of the jackknife for Candiacervus sp. during the LGM Variable Percent contribution Permutation

importance DEM 69.1 66.3 Annual Precipitation 14 20.9 Aspect 5.2 0.7 Precipitation Wettest Quarter 4.5 4.6 Slope 4.2 3.4

Annual Mean Temperature 1.4 4.1 Temperature Annual Range 1.3 0 Precipitation Driest Quarter 0.2 0

(21)

J. Steenvoorden | Bachelor Thesis Future Planet Studies (Earth Sciences) | 4 July 2016 Hippopotamus Creutzburgi

As opposed to the Cretan deer, the Hippopotamus Creutzburgi could potentially be found in a more widespread amount of habitats, including more and higher elevated areas than the Cretan deer, which may partly be caused by the two findings in the eastern part of Crete that were situated in a more elevated part of the island (figure 7.5).

However, if the jackknifes and variable importance tables in figure 7.6 and 7.7 and table 7.3 and table 7.4 are analysed and compared, it can be seen that instead of the DEM it are precipitation and precipitation anomalies that are of biggest significance towards determining suitable habitat conditions for the

Hippopotamus Creutzburgi.

Furthermore, just as with the Cretan deer, suitable habitat conditions increase from approximately 5673 km2_{during the Last Interglacial Period to}

6959 km2_{during the LGM – an increase of about 22.7%. This is significantly}

lower than the increase in suitable habitat conditions for the Cretan deer, which may be a consequence of drier conditions during the LGM.

Figure 7.5: predicted suitable habitat conditions for Hippopotamus Creutzburgi during (A) the Last Interglacial Period and (B) the LGM

(22)

Figure 7.6: graph of the jackknife for Hippopotamus Creutzburgi during the Last Interglacial Period

Variable Percent

contribution Permutation importance

Precipitation Driest Quarter 10.1 54.9

Slope 5.3 0

Aspect 0.5 3

Temperature Annual Range 0 0

Precipitation Wettest Quarter 0 0

DEM 0 0

Annual Mean Temperature 0 0

Table 7.3: table showing variable contributions for Hippopotamus Creutzburgi during the Last Interglacial Period

(23)

Variable Percent contribution Permutation importance

Precipitation Driest Quarter 16.4 18.6 Precipitation Wettest

Quarter 6.4 18

Aspect 6.1 3

Slope 0.4 2.6

DEM 0 1.1

Table 7.4: table showing variable contributions for Hippopotamus Creutzburgi during the LGM

Mammuthus Creticus

Just as with the Candiacervus sp. and Hippopotamus Creutzburgi, many of the known fossil occurrences of the Mammuthus Creticus occur on the northern and eastern part of the island (figure 7.8). As a result, much of the suitable habitat is situated close to the coast. However, due to a higher elevation finding in the middle-eastern part of the island, higher elevated areas are also represented as having suitable habitat conditions.

If the jackknife graphs and variable importance tables of figure 7.9 and 7.10 and table 7.5 and 7.6 are analysed and compared, it can be seen that just as with the Cretan deer, the dominant contributing environmental variable is the DEM, followed by the annual precipitation during the Last Interglacial Period and the annual precipitation and temperature annual range during the LGM.

In addition to that, the total area of suitable habitat conditions hardly increases from the Last Interglacial Period to the LGM with an increase from 5924 km2_{to 5960 km}2_{– an increase of not even 1%.}

(24)

Figure 7.8: predicted suitable habitat conditions for Mammuthus Creticus during (A) the Last Interglacial Period and (B) the LGM

(25)

DEM 72.7 70.5

Aspect 5.5 1.9

Slope 2.7 2.5

Temperature Annual Range 1.5 0 Precipitation Wettest Quarter 0.6 1.4 Precipitation Driest Quarter 0 0

Table 7.5: table showing variable contributions for Mammuthus Creticus during the Last Interglacial Period

Figure 7.10: graph of the jackknife for Mammuthus Creticus during the LGM Variable Percent contribution Permutation

importance

DEM 53.7 67.1

Annual Precipitation 12.7 8

Aspect 6.1 2.2

Slope 5.9 9.8

Precipitation Wettest

Quarter 5.3 9.3

Annual Mean Temperature 0.5 0.9 Precipitation Driest Quarter 0.5 0.3

(26)

b. Cyclades ArcGIS analysis

The Cyclades are an extremely interesting island group regarding their physical island characteristics and isolation and their subsequent influence on species dispersal. During glacial periods when sea-level dropped, the Cyclades formed one single island as the island group is surrounded by a flat plateau with an average slope of only 1.58° (Kapsimalis et al., 2009).

Interestingly, the ArcGIS analysis can confirm this. Especially during a glacial maximum, the Cyclades experience significant changes in their physical characteristics and geographic configuration (figure 7.11). The once isolated islands start to form a mega island when sea-level is approximately 95 meters lower than present sea-level. As sea-level was about -140 meters during the LGM, the extent of the so called Cycladic Plateau was extremely broad. Its size increased from 3806 km2_{during the Pliocene Warm Period to 14754 km}2_during

the LGM, which is an increase in size by approximately 3.9 times. Even so, the distance to the mainland reduced from approximately 12 to 4 kilometres on the north-western part of the Cyclades during the LGM, while on the eastern part of the Cyclades during the LGM, the mainland was only 20 kilometres away due to a land bridge formation between the island of Icaria and present day Turkey, possibly providing options for mainland mammals to cross over to the island.

Furthermore, it is interesting to note that in the case of the Cyclades, but also the other researched islands, an increase in sea-level does not influence island area and its physical characteristics as significantly as a decrease in sea-level. For example, the present land area of the Cyclades compromises 4611 km2, which means that a 25 meter increase in sea-level only decreases land area by 17.5%. In addition to that, despite the lowering of total land-area, a comparison between the Pliocene Warm Period and other scenarios results in the understanding that a 25 sea-level increase does not create or break-up geographic barriers.

(27)

Figure 7.11: five maps showing the Digital Elevation Model of the Cyclades during (A) the Pliocene Warm Period, (B) the Last Interglacial Period, (C) the LGM, (D) the Mid-Holocene and (E) the

(28)

c. Cyprus ArcGIS analysis

The situation of Cyprus regarding physical island characteristics such as island size is quite similar to Crete (figure 7.12). This is because just as Crete, Cyprus also does not completely lay on a flat plateau like the Cyclades. Consequently, it can be seen that from the Pliocene Warm Period to the LGM with a sea-level change of approximately 165 meters, physical island characteristics do not change extremely, with the retreating coastline being the main feature. Island size increases from 12931,770464 km2_{in the Pliocene Warm Period to}

18485,607938 km2_{during the LGM, which is an increase of approximately 43%.}

However, the most important physical characteristic for Cyprus might be its distance from the mainland. As compared to the other researched islands such as for example the Cyclades, which has a distance of 4.5 kilometres from the mainland during the LGM, Cyprus is situated 60 kilometres from the mainland during the LGM and 71.5 kilometres during a high sea-level period like the Pliocene Warm Period.

(29)

Figure 7.12: five maps showing the Digital Elevation Model of Cyprus during (A) the Pliocene Warm Period, (B) the Last Interglacial Period, (C) the LGM, (D) the Mid-Holocene and (E) the Present

(30)

Species distribution analysis

Hippopotamus Minor

For the Hippopotamus Minor of Cyprus, many parts of the island have suitable predicted habitat conditions, despite the very highly elevated areas in the center of the island (figure 7.13). This may be caused by the size of Cyprus and an ample distributed amount of species occurrences.

Furthermore, if the jackknifes and variable contribution tables in figure 7.14 and 7.15 and table 7.7 and 7.8 are analysed and compared, it can be seen that in contrast with Crete, Cyprus has a bigger variety of highly contributing environmental variables. During the Last Interglacial Period, the main contributing variables are annual precipitation, precipitation during the driest quarter and temperature. However, during the LGM, the precipitation during the wettest quarter, the aspect and the DEM also become of importance. The reason the DEM and aspect might become of importance is due to the increases in land area with relatively unsuitable habitat conditions on the coastal areas.

Besides, the area of suitable habitat conditions (with a predicted suitable conditions of above 1,6) increases from about 9566 km2_{during the Last}

Interglacial Period to 11155 km2_{during the LGM – an increase of approximately}

(31)

Figure 7.13: predicted suitable habitat conditions for Hippopotamus Minor during (A) the Last Interglacial Period and (B) the LGM

(32)

Precipitation Driest Quarter 34.9 0 Annual Mean Temperature 10.4 69.7

Aspect 8.9 9.2

Precipitation Wettest Quarter 4.3 7.2

DEM 2.9 0

Slope 2.5 0.8

Table 7.7: table showing variable importance for Hippopotamus Minor during the Last Interglacial Period

Figure 7.15: graph of the jackknife for Hippopotamus Minor during the LGM

contribution Permutation importance Precipitation Driest Quarter 29.3 37.3

DEM 21.9 19.5

Precipitation Wettest Quarter 13.4 5.8

Aspect 13.2 11

Slope 4 1.6

(33)

d. Lesbos ArcGIS analysis

Lesbos is a small island that presently has a land area of approximately 1738 km2_{. As a result, it is generally highly influenced by sea-level changes in relative}

terms. Furthermore, it is relatively close to the mainland with a distance of about 12-14 kilometres at present and the seabed in between the island and the mainland is comparable to the Cyclades as it is flat and shallow.

It can be seen that the shallow seabed in between the Lesbos and the mainland has significant implications for the islands’ connectivity (figure 7.16). It forms a land bridge during glacial periods with low sea-level which results in Lesbos being a peninsula rather than an island. This land bridge formation starts when sea-level is approximately 40 meters below present. Furthermore, at some periods in time such as the Mid-Holocene and the period in between the LGM and Mid-Holocene, big lakes might have developed in parts of the island.

In addition to that, it can yet again be seen that a sea-level rise as compared to a sea-level decline has less implications on an islands’ physical characteristics, which in the case of the researched islands has to do with the presence of shallow sea-beds that are found nearly everywhere in the Aegean Sea.

(34)

Figure 7.16: five maps showing the Digital Elevation Model of Lesbos during (A) the Pliocene Warm Period, (B) the Last Interglacial Period, (C) the LGM, (D) the Mid-Holocene and (E) the Present

(35)

e. Malta ArcGIS analysis

Lastly, Malta is interesting as the presence of a shallow flat between Sicily and Malta that forms land bridge during times of low sea-level which was already mentioned by Vogiatzakis et al. (2008). The ArcGIS analysis can confirm this. A lowering of sea-level allows for the flat in between Malta and Sicily to emerge which results in Malta becoming a peninsula instead of an island just as with Lesbos (figure 7.17). This land bridge starts forming at a sea-level of approximately -100 meters and exists for approximately 15.000 years if Lambeck & Chappell (2001) is consulted. Furthermore, it can be seen that yet again, there is no significant formation or deformation of geographic barriers as a result of sea-level. Moreover, distance to the mainland remains rather constant over time with a distance of approximately 85 kilometres. However, the land bridge that forms between Malta and Sicily during the LGM and potentially other glacial cycles has an area of approximately 6667 km2_{, which is extremely large}

(36)

Figure 7.17: five maps showing the Digital Elevation Model of Malta during (A) the Pliocene Warm Period, (B) the Last Interglacial Period, (C) the LGM, (D) the Mid-Holocene and (E) the Present Species distribution analysis

Hippopotamus Pentlandi

From the species distribution analysis for the Hippopotamus Pentlandi, it can concluded that a big area of land on Malta and Sicily formed suitable habitat conditions for Hippopotamus Pentlandi. Furthermore, it can be concluded that the whole of Malta formed suitable habitat conditions for Hippopotamus Pentlandi. In addition to that, it can be seen that the low lying flats that form due to low

(37)

sea-J. Steenvoorden | Bachelor Thesis Future Planet Studies (Earth Sciences) | 4 July 2016

level in the LGM form relatively suitable habitat for Hippopotamus Pentlandi, but also big parts of unsuitable habitat.

If the jackknifes and variable importance tables of figure 7.19 and 7.20 and table 7.9 and 7.10 are compared and analysed, it can be seen that during both scenarios, different variables are important. During the Last Interglacial Period, it is mainly temperature, aspect and precipitation of the driest quarter that determine suitable habitat. However, during the LGM, it is predominantly the DEM and aspect that decide which habitat is suitable. This may be cause by the formation of a land bridge with low lying flats and the extension of the coastal areas around Sicily.

Lastly, the islands’ suitable habitat conditions (with a predicted suitable condition of above 1.6) increase from approximately 17632 km2_{during the Last}

Interglacial Period to 21348 km2_{during the LGM – an increase of about 22%.}

Figure 7.18: predicted suitable habitat conditions for Hippopotamus Pentlandi during (A) the Last Interglacial Period and (B) the LGM

(38)

contribution Permutation importance Annual Mean Temperature 45.6 30.1

Aspect 29.8 34.5

Precipitation Driest Quarter 21.3 35.4 Temperature Annual Range 3.3 0

Slope 0 0

Precipitation Wettest Quarter 0 0

DEM 0 0

Annual Precipitation 0 0

Table 7.9: table showing variable importance for Hippopotamus Pentlandi during the Last Interglacial Period

Figure 7.20: graph of the jackknife for Hippopotamus Pentlandi during the LGM

DEM 50.8 73.2

Aspect 19.2 12

Precipitation Driest Quarter 12.6 2.3 Precipitation Wettest Quarter 7.5 5

Slope 6.9 5.7

Temperature Annual Range 2.8 1

(39)

VIII. Discussion

The discussion will have largely the same structure as the results chapter, in which each island is first dealt with separately. This part is then summarized with table 8.1, which summarizes the influence climate induced sea-level decline on the physical island characteristics of the investigated islands. Thereafter, a general discussion is held to discuss the methodology, deal with general discussion points and to integrate the separate island discussion.

Chapter six and table 6.1 explained that the physical characteristics and geographic configuration of an island were considered to be important in the evolution of body size of insular mammals, as these notably influences resource availability and ecological interactions through for example isolation and connectivity (Meiri et al., 2005). However, it can be found that in all researched islands, many of the physical characteristics that were hypothesized to be influenced by climate induced sea-level changes are not or that this differs significantly per island.

a. Crete

For Crete, sea-level increase and decrease only influence land area, but geographic barriers and isolated areas are not formed. Furthermore, distance to the mainland remains rather large and a land bridge is not formed during any stage since the Pliocene Warm Period. This is in sharp contrast with Poulakakis et al. (2002) and Van Der Geer et al. (2006), who suggested that high sea-level stands in interglacial periods up to the early Pliocene submerged and subdivided the island into smaller separate islands. Cornell et al. (2016) further tackle these hypotheses, as they state that sea-level since the Pliocene has hardly been above 25 meters and thus Crete would never have been able to be submerged or subdivided.

However, this reduction in land area and thus reduction in resource availability during glacial-interglacial cycles may have been an important factor of influence for the evolution of body size. An interesting discovery made by Sondaar (1977) may support this as he found that many of the deer on Crete exhibited osteoporosis (a condition in which bones become fragile and sometimes deformed as a result of the loss of mineral matter), a phenomena that is linked to malnutrition (Dermitzakis et al., 2006).

Furthermore, it appears that Crete contains a significantly larger amount of fossil findings than on the other researched islands. In the cases of the Cyclades and Lesbos, it was even restricted to a few sporadic findings that were not even accurately assigned to a specific location, preventing this research from conducting a species distribution analysis for these islands. Van Der Geer et al. (2014) mentioned that the absence of fossil findings may be caused by the lack of sedimentary deposits in this area which may inhibit preservability. However, this is not necessarily true, as the lack of fossil findings may be caused by the lack of or unsuccessful paleontological surveys, and thus, many fossils may potentially still be underground or on the seabed. Kapsimalis et al. (2009)

(40)

already suggested the interesting results that underwater surveys of all researched islands might provide due to the shallow flat seabed’s that were land during various glacial periods.

Moreover, the species localities of all researched species on Crete were generally clustered towards caves in the northern and eastern part of the island, as can also be seen from the species distribution maps in the results chapter. Furthermore, the fact that fossils were found at a specific location does not necessarily imply it was the habitat for the specific species. Consequently, this is something that may have significantly biased the results from the species distribution analysis, as suitable habitat conditions will trend towards the environmental conditions of these highly clustered findings alone.

Lastly, an interesting note can be made on the body structure and osteology (study of bones) of the Cretan deer and its implications on the reliability of the species distribution analysis. It was found in the species distribution analysis of the Candiacervus sp. that the coastal regions were the most suitable of habitat, based on the fossil findings. However, Van Der Geer et al. (2006) stated that the Cretan deer shared a body structure and ecological niche similar to the wild goat of Crete (Capra Aegagrus) today: barren rocks and thorny bushes. Consequently, it is likely that the Cretan deer would occupy the higher elevated and less densely forested areas as well, something that does clearly does not result from the species distribution analysis.

b. Cyclades

Despite a lack of fossil findings, the physical characteristics of the Cyclades may remain the most excellent example of paleo-island formation. The ArcGIS analysis resulted in the understanding that during glacial periods and their subsequent low sea-levels, the shallow flat seabed in between the islands in the Aegean Sea forms an island often called the paleo-Cyclades that according to Van Der Geer et al. (2014) contains flat plains that may have formed suitable habitats for elephants and other mammals. This relatively flat mega island is also located close to the mainland with a distance of only approximately 4 kilometers. There is information about dwarf elephants and elephants on islands in the Aegean Sea in general. For example, in Poulakakis et al. (2002), many papers are brought forward that present (dwarf) elephant findings on the Cyclades and other Aegean islands like Tilos, Imvros and Rhodos. Furthermore, a paper by Van Der Geer et al. (2014) described the finding of a pygmy elephant on the island of Naxos with only 10% the size of its mainland ancestor.

On the other hand, it remains unfortunate that the findings in the Aegean Sea are rather sporadic and not phylogenetically accommodated (Van Der Geer et al., 2014), as the availability of fossil findings in this biogeographically interesting region may have provided significant evidence for the possible influence of physical island characteristics on the evolution of insular mammals in this region. Despite, the presence of pygmy elephant fossils on some of the islands in the Cyclades suggest at least some evidence of the influence of insularity on body size evolution.