Linköping University | Department of Physics, Chemistry and Biology Master thesis, 60 hp | Educational Program: Ecology and the Environment Spring 2016 | ISRN: LITH-IFM-A-EX--16/3218--SE
Dispersive trait expression of
Asellus aquaticus from a rare
cave habitat
Martin Brengdahl
Examinator, Bo Ebenman
Datum
Date
2016-05-27
Avdelning, institution Division, Department
Department of Physics, Chemistry and Biology Linköping University
URL för elektronisk version
ISBN
ISRN: LITH-IFM-A-EX--16/3218--SE
_________________________________________________________________ Serietitel och serienummer ISSN
Title of series, numbering ______________________________ Språk Language Svenska/Swedish Engelska/English ________________ Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport _____________ Titel Title
Dispersive trait expression of Asellus aquaticus from a rare cave habitat
Författare Author Martin Brengdahl Nyckelord Sammanfattning Abstract
Dispersal influences several ecological and evolutionary processes, such as intraspecific competition, genetic drift and inbreeding. It can lead to phenotypic mismatch with the habitat when a locally adapted individual winds up in an environment with a divergent selection regime compared to the source habitat. The aim of this project was to compare dispersive traits in the freshwater isopod Asellus aquaticus from a cave habitat, with surface dwelling isopods collected upstream and downstream from the cave system. The subterranean stream (cave) represents a rare, geographically limited habitat which has a divergent selective pressure compared to the surrounding habitats. Experiments on dispersal were performed in the laboratory, in darkness with IR-equipment for visualization. Displacement was measured using one-dimensional test arenas. Compared to the surface phenotype, the cave phenotype was expected to have reduced fitness outside of the cave and unlikely to successfully disperse to new areas of similar suitable conditions. The results did not follow my main hypothesis that isopods from the cave would be less dispersive than individuals from the surface. The inconclusive results might derive from large variation in the data and divergent adaptations which yield similar expression of dispersal.
Content
1 Abstract ... 2
2 Introduction ... 2
3 Material & methods ... 5
3.1 Study organism ... 5
3.2 Field collection ... 5
3.3 Displacement experiment ... 8
3.4 Morphological measurement ... 9
3.5 Data analysis & statistical model ... 10
4 Results ... 14
4.1 Common garden isopods ... 15
4.2 Wild caught isopods ... 17
4.2.1 Summer & Autumn batch ... 17
4.2.2 Autumn batch... 19 4.2.3 Summer batch ... 20 5 Discussion ... 21 5.1 Conclusions ... 25 6 Acknowledgement ... 25 7 References ... 26
1 Abstract
Dispersal influences several ecological and evolutionary processes, such as intraspecific competition, genetic drift and inbreeding. It can lead to phenotypic mismatch with the habitat when a locally adapted individual winds up in an environment with a divergent selection regime compared to the source habitat. The aim of this project was to compare dispersive traits in the freshwater isopod Asellus aquaticus from a cave habitat, with surface dwelling isopods collected upstream and downstream from the cave system. The subterranean stream (cave) represents a rare,
geographically limited habitat which has a divergent selective pressure compared to the surrounding habitats. Experiments on dispersal were performed in the laboratory, in darkness with IR-equipment for
visualization. Displacement was measured using one-dimensional test arenas. Compared to the surface phenotype, the cave phenotype was expected to have reduced fitness outside of the cave and unlikely to successfully disperse to new areas of similar suitable conditions. The results did not follow my main hypothesis that isopods from the cave would be less dispersive than individuals from the surface. The inconclusive results might derive from large variation in the data and divergent adaptations which yield similar expression of dispersal.
2 Introduction
Dispersive traits are crucial for the ecology and evolution of most
species, affecting survival and fecundity, as well as gene flow across the landscape. Dispersion affects a wide array of scientific topics, such as population dynamics, spatial distribution of species’, responses to change in the environment, morphology and life history (Bowler & Benton 2005; Travis et al 2012). Selection of dispersive traits occur in response to landscape characteristics, including habitat availability, habitat
persistence and landscape structure (Travis & Dytham 1999; Bonte et al 2010). The landscape structure may therefore have a pronounced impact on the evolution of dispersal, as short and long distance dispersal
strategies are affected by separate properties of the landscape (Bonte et al 2010).
There might be several evolutionary stable attractors for dispersal in a fitness landscape, leading to dispersal dimorphism (Kisdi 2002) or even polymorphism. However, during certain circumstances, such as when migration rates are high and have a higher impact on a population’s gene
pool than selection pressure, local adaptations will not appear and polymorphism will be lost. The allele with the highest average
reproductive success over the entire metapopulation will thus become fixed over time (Lenormand 2002). However, Kirkpatrick and Barton (2006) highlights that evolution can be strongly driven by local
adaptation mechanisms even when migration occurs between populations.
High dispersal rates counter kin competition and inbreeding (Gandon 1999). It also counters extinction risks due to environmental fluctuation, genetic drift and negative mutations in small populations as well as escaping competition from conspecifics (Ronce 2007). There are several costs for dispersal which create trade-offs between dispersal capability and other traits (Bonte et al 2012). This explains why high dispersal rates are not a universal solution.
In some cases, as at an expanding range, high dispersal is favoured even though it increases mortality and reduce lifetime reproductive success in a short time perspective (Barton et al 2012). It might take several
generations before the high dispersal strategy has repaid its costs through a higher number of descendants. This illustrates the importance of not measuring fitness solely by lifetime reproductive success (Travis et al 2009), or draw preconceived conclusions about selection on dispersal. Benard & McCauley (2008) discusses how natal habitat can affect the probability of dispersal through phenotypic plasticity, which can conceal genotypic adaptation (Baquedano et al 2008). Experiments with a
common garden approach might counter such plasticity by introducing a common life history between phenotypes.
Divergent selection can create reproductive isolation due to both
decreased fitness in migrants themselves but also in hybrid offspring (Via et al 2000; Hendry 2004; Nosil et al 2005). Divergent selection even creates heterogeneous divergence on the genome itself, creating variation in the properties and occurrences of differentiated genomic regions (Nosil et al 2009). Because plasticity is costly (DeWitt et al 1998, but see Auld et al 2009) it is not likely that plasticity would fully counter the
disadvantage of migration to a contrasting habitat (Marshall et al 2010) nor be favourable in a stable environment. However, plasticity can act as a counter to directional maladaptive gene flow in marginal habitats (Chevin & Lande 2011).
It is fair to state that caves are reasonable rare and unique environments and as Culver & Pipan (2015) concludes, the evolution of cave
occur in cave animals (Protas et al 2011; Pipan & Culver 2012) which puts cave animals at quite a disadvantage in surface environments. There are examples of both obligate (Culver et al 2000; Voituron et al 2011) and facultative species’ (Protas & Jeffery 2012; Manenti & Ficetola 2013) which are used to study cave adaptation. A few which have had recent attention is the olm (Roteus anguinus) (Mail & Bulog 2015), the Mexican tetra (Astyanax mexicanus) (Gross et al 2015) and the water louse (Asellus aquaticus) (Konec et al 2016). A. aquaticus is widespread in lakes, streams and ponds throughout Europe (see Sworobowicz et al 2015)
In this study I compare dispersive traits in two phenotypes of A.
aquaticus, cave and surface dwelling. There are many examples of
regressive and constructive phenotypic changes which occur in cave phenotypes, compared to the surface morph (Gross 2012). For instance, elongation of pereiopod (leg) length have been found in the cave
phenotype of Asellus (Turk et al 1996) and relative length of legs is coupled to movement speed in both ants and toads (Phillips et al 2006; Pearce-Duvet et al 2011). Because dispersal rate is connected to habitat availability (Travis & Dytham 1999) it is reasonable that the differences between phenotypes affect dispersive trait expression.
However, there is little information on dispersive traits in A. aquaticus and knowledge about variation in morphological traits such as body size and length of legs influence on movement have not been well studied. Eroukhmanoff & Svenssson (2009) show that escape speed is correlated to body size.
There have been some studies on different aspects of movement in A.
aquaticus (Englund & Hambäck 2004; Van den Brink et al 2007). Often
they have another main focus than dispersion alone such as recovery after pollution events (Galic et al 2012) and colonization after glaciation
periods (Verovnik et al 2005). Even though published knowledge about the subterranean Asellus population in the Swedish region is scarce, the subject have had lot of attention in southern part of Europe, mainly the Balkans and southeastern part of central Europe (Turk et al 1996; Verovnik et al 2004, 2005; Prevorčnik et al 2009; Protas et al 2011; Konec et al 2015).
The main hypothesis of the present study was that animals in the cave would be less dispersive than individuals collected outside of the cave. Compared to the outside phenotype, the cave phenotype was expected to have reduced fitness outside of the cave and unlikely to successfully disperse to new areas of similar suitable conditions.
3 Material & methods 3.1 Study organism
Asellus aquaticus is a common freshwater isopod. It feeds primarily on
bacteria and fungi associated with detritus, as well as periphyton ( Marcus et al 1978; Arakelova 2001) but it can also feed on both
macrophytes themselves as well as dead conspecifics (Marcus et al 1978). In the study region of southern Scandinavia A. aquaticus reproduce only once and have one or two generations per year (Økland 1978).
There are studies of A. aquaticus where recently divergent selection pressures have created new ecotypes (Hargeby et al 2004; Hargeby et al 2005; Eroukhmanoff et al 2009; Eroukhmanoff et al 2011; Harris et al 2011; Karlsson Green et al 2016) but also studies which are looking at cave and surface habitats where the selection pressure have led to phenotypes with such differences that they sometimes are discussed as subspecies’ (Turk et al 1996; Verovnik et al 2004; Prevorčnik et al 2009; Konec et al 2015).
3.2 Field collection
Collection of test animals was conducted in the Lummelunda stream running through the karst system in Cave Lummelunda (Gotland) at several sites upstream, inside and downstream of the cave system during the 16th through 18th of June 2015. Nine sampling locations were used and were distributed with three upstream of the cave, three inside and three downstream of the cave system (Figure 1 & 2), for coordinates for the locations see Table 1. Henceforth the locations will be referred to as belonging to either the upstream (US), cave (CA) or downstream (DS) area if not referred to directly.
Figure 1. The sampling locations Up- (US) and Downstream (DS) of Cave Lummelunda, with the number indexing ascending with the water current (GOOGLE EARTH 2016).
Figure 2. The sampling locations in Cave Lummelunda (CA), with the number indexing ascending with the water current. Map from SSF (Swedish
Table 1. The sample locations in downstream order, with the coordinates in SWEREF99 TM (no available coordinates for the cave locations, see Figure 2). Sediment description is very broad and should only be viewed as a rough description.
Stream
Order Site name Coordinates Sediment
1 Upstream 1 US1 6404890.03 N 702639.5 E gravel 2 Upstream 2 US2 6405014.39 N, 702560.2 E gravel 3 Upstream 3 US3 6404997.57 N, 702572.9 E gravel 4 Cave 1 CA1 Sandbanken gravel 5 Cave 2 CA2 Upstream Lerhamnen rocks 6 Cave 3 CA3 Turistgrottan, Sal 6 gravel 7 Downstream 1 DS1 6404628.10 N, 704171.5 E fresh detritus
8 Downstream 2 DS2 6404636.03 N, 704020.7 E knee-waist high detritus 9 Downstream 3 DS3 6404696.88 N, 703835.5 E 2-3 dm detritus
Isopods were collected with hand nets using the standardized kick sampling method (mesh size 0.5 mm, SS-EN 27 828) and traps which were emptied after 24 hours. The traps, designed as in Pehrsson (1984), were created out of 200ml polypropylene boxes with the bottom
exchanged with a conically shaped net (mesh size 1.1 mm, PVC reinforced with fibre glass), faced inwards with a 1 cm opening in the centre. The lids central part was also exchanged with a net of the same type, thus enabling water flow through the trap. Because collection using different methods is likely to result in different bias (Biro & Dingemanse 2009) isopods collected with traps were treated separately from the ones collected by kick sampling.
A weekbefore the experiments started isopods were moved to an
experimental constant room with air temperature of approximately 10 °C. The isopods were stored in polypropylene cages, which partly consisted of net walls, which in turn was kept within larger containers, sharing water between four cages. The cages were arranged so that at least one location (US1, US2, US3, CA1…) from each area (US, CA, DS) was represented in each container. Oxygenation was ensured using air pumps connected to diffusers and the water, which was collected downstream of Lake Rosenkällasjön (6472399.06 N, 533831.44 E SWEREF99 TM) was changed weekly. The isopods were fed with decaying alder (Alnus
glutinosa) leaves. There was no light, except for a few minutes per day
Due to high mortality rates in the initial batch (above 60 %), probably caused by a Saprolegnia spp outbreak (personal observation),
complementary sampling using only the kick sampling method was conducted during the 3rd and 4th of October 2015. During this occasion, a
high number of isopods had to be released, especially in the downstream area, due to their small size making video recording of them futile. A lab reared, common garden, first generation of both the cave and surface phenotype (parental generation collected with kick sampling at site CA 3 and between sites US 2 & 3 respectively) was also used in the experiment. These animals were reared in open containers with identical handling.
3.3 Displacement experiment
Displacement experiments were conducted in above mentioned experimental constant room in grey polypropylene pipes (7.5 cm in diameter, 2 m in length), which were cleaved lengthwise. The ends were covered with transparent polypropylene which was sealed tight with aquarium silicone glue. To stabilize the half-pipes and protect the plastic ends from mechanical stress two plastic strips was attached underneath each pipe, in longitudinal direction. The pipes were filled with boiled sand (1-2 mm Haley-Shining black 136.0169), covering the bottom with a 0.5-1 cm layer. Water was subsequently added to a total depth of 2.5 cm (sand included). For experimental trials, the water used was collected downstream of Lake Rosenkällasjön. All water used, both during experimental trials and in storage containers was collected at least 96 hours before usage in order to prevent interaction from predatory cues, which can have a large impact on A. aquaticus behaviour (Harris et al 2013).
The experimental trials was conducted in darkness, using IR illuminators with a wavelength of 850 nm (ECOLINE TV6700, Security Center ABUS Group) and a surveillance camera with IR capabilities (DCS-2330L, D-Link) for recording (Figure 3). Video analysis was done subsequently and not in realtime.
Figure 3. Picture of the experimental set up, taken as a freeze frame function with the surveillance camera. Four additional IR lamps which themselves are not visible are in use, two above each short end. The bright lines,
perpendicular to the pipes, represent the cut off for the trials, with a length of 140 cm in between. Note that the full length of the pipes could not be utilized due to technical limitations.
Trials were done for five individuals simultaneously, with one individual in each pipe. Each individual was placed in the middle of a pipe,
thereafter allowed to explore the pipe for one minute before extracting a picture and the current position counted as the starting point in the trial. Thereafter either the time to passing the cut off line (70 cm from the centre of the pipes, see Figure 3) was recorded or, if not passing the cut off 15 minutes after the start, the current position recorded. After
maximal test time (16 minutes) immediate reruns where conducted three times making each individual run in the same pipe four times.
Trials were semi-randomized, so that individuals were picked
systematically from locations, but randomly allocated to pipe. When occurring, kick sampling and trap catches were treated as separate
locations. After one individual from all locations had completed the trials the next individual from the first location was picked and so on.
3.4 Morphological measurement
After the displacement experiment each individual was photographed alive in a water filled petri dish, placed upon a graph paper. All photos were taken with a Nikon Coolpix 4500, which was placed on a tripod
above the petri dish. Thereafter the photographed individual was placed in carbonated water for euthanization. Two legs, one from the third and one from the seventh pair, were removed from each individual and photographed. The isopods were subsequently stored in -20° C, which allowed for sex determination later on. This was done using the hallmarks on the first pair of legs and pleopods (Enckell 1980).
Body length was measured from the pictures of living individuals as the distance between the tip of the head and the tip of the pleotelson
(anterior). The angle between the legs of the first segment and the tip of the head was measured as an approximation of hydrodynamic
characteristics (Eroukhmanoff & Svenssson 2009). Measurement error on head morphology can be severe, where within individual variance can be higher than between individuals (Bertin et al 2002). A subset consisting of 20 randomly chosen individuals from the summer batch was
investigated for measurement errors, but both the within and between picture variance for individuals were lower than variance between individuals. The length of ischium, merus and carpus were used as a composite distance in representation of leg length. Measurement of all morphological traits was done in ImageJ (Rasband 2013).
3.5 Data analysis & statistical model
Bayesian analysis are used in a wide range of applications and is a field constantly undergoing methodological advancements. Most applications are based on Bayes’ Theorem, which include three components. These are: i) the current knowledge or beliefs, incorporated as the prior ii) the observed data, incorporated via the likelihood and iii) the updated belief based on the previous two, i.e. the posterior. There are several detailed introductions to Bayesian inference. See e.g. Van Oijen et al (2005) for a good description in an ecological context or Bolker et al (2009) for a comparison with other statistical methods on a practical level.
In this study both population and individual parameters were estimated using a Bayesian hierarchical model approach, where the test animal’s dispersal was viewed as a one dimensional diffusion process. The
population effect (β0) from the resulting models was the main interest as
credible differences in β0 could indicate differences in expression of
dispersal between phenotypes. Effects from individual traits (β1) would
indicate effects from corresponding covariate (morphological traits, sex or catchment period).
In the hierarchical model used in this study the squared distance from starting to end point, net square displacement (NSD or Yij2), and the time
for said displacement (tij) to occur were used in all analysed models. Sex
and morphological measurements were added as covariates and analysed individually or all together, other combinations of covariates were
omitted in order to reduce run time for the model. The covariates used were relative leg length (to body length) of the seventh and third pair, head angle, body length and sex. In the analysis of the wild caught animals the catchment period (June or October) was also used as a covariate. For the wild caught animals analysis were done both joint and separate in regard of catchment period. Analysis were done separate for common garden animals. Isopods caught with traps were omitted in the Bayesian analysis due too few caught individuals.
All possible groupings of common garden animals were tested (i.e. cave & surface are alike and cave & surface are different) for model selection. For the wild caught animals the groupings which were tested where
constrained to the ecologically most reasonable outcomes in order to limit run time for the model analysis. These were: i) all nine sampling
locations are alike, ii) all locations are unique, iii) all surface locations are alike, but different from the cave locations, which themselves are alike, iv) all upstream, cave and downstream locations are alike within, but different between areas.
Following the assumptions that the movement can be approximated with a diffusion process a probability density function derived for one
dimensional Brownian particles was used on the individual level. The distance measured in an individual’s trial (𝑌𝑖𝑗), the observed data, was distributed with a normal distribution based on the starting point and the individual’s variance (Vij).
𝑌𝑖𝑗~𝑁𝑜𝑟𝑚𝑎𝑙(0, 𝑉𝑖𝑗) Eq. 1
The latter was calculated using the function g(Di,tij), which was based on
the diffusion constant for the individual (𝐷𝑖) and the time since start of the trial for the individual (𝑡𝑖𝑗).
𝑉𝑖𝑗 = 𝑔(𝐷𝑖, 𝑡𝑖𝑗) Eq. 2
The variance (Vij) for the normal distribution of Yij was estimated as two
𝑉𝑖𝑗 = 2𝐷𝑖𝑡𝑖𝑗 Eq. 3
At the population level, the model assume that population shape (α) was distributed as a gamma distribution with hyperpriors 𝐴𝛼 and 𝛽𝛼 (Eq. 4). The hyperprior shape for population shape (𝐴𝛼) was set to 1, thus making it exponential, and the hyperprior rate for population shape (𝛽𝛼) was set to 0.0295, making the mean of the resulting gamma distribution for population shape 33.898 and 95 % of the population shape values (α) to be below 100.
𝛼~𝐺𝑎𝑚𝑚𝑎(𝐴𝛼, 𝛽𝛼) Eq. 4
The diffusion constant (Di) for individual i was assumed to be distributed
as an inverse gamma distribution based on the population shape (α) and the population scale βi.
𝐷𝑖~𝐼𝑛𝑣𝐺𝑎𝑚𝑚𝑎(𝛼, 𝛽𝑖) Eq. 5 The latter was calculated with the function f(β0, β1, Xi), based on the
parameters 𝜷𝒐 (group intercept values, “population effect”), 𝜷𝟏 (covariate
effects, “individual effect”) and 𝑿𝒊 (covariate values for the individual) These parameters were sometimes vectors, depending on number of covariates and groupings in the proposed model.
𝛽𝑖 = 𝑓(𝜷𝒐, 𝜷𝟏, 𝑿𝒊) Eq. 6
The priors for 𝛽0 and 𝛽1 were set to be normally distributed with mean 0 and standard deviation of 100, chosen to be uninformative and therefore likely flat over their respective posterior distributions, limiting the impact of the respective prior.
The function f(β0, β1, Xi) was built in two components (Eq. 7 and 8). The mean (m), which was dependent on the parameters β0, β1, Xi and a
transformation of m into the scale parameter for the inverse gamma distribution. The calculation of m (Eq. 7) consists of an exponential function of both β0 and the product of β1 and Xi.
𝑚 = 𝑒(𝜷𝟎+𝑿𝒊𝜷𝟏) Eq. 7
The diffusion constant (Di) had a conditional distribution which was
known and therefore sampled with Gibbs sampling in the Markov-chain Monte Carlo (MCMC) algorithm. The conditional distribution of
population parameter shape (α) was unknown and was therefore sampled with a Metropolis-Hastings algorithm within the MCMC. The MCMC-chain in the study were in all analysis run over 2 million iterations, with an adaptive proposal algorithm, which optimizes the acceptance rate towards 0.234.
To analyse results, deviance information criterion (DIC) (Spiegelhalter et al 2002) was used for model selection. The model with lowest DIC value indicates the best parsimony of the model parameters’ likelihood of producing the observed data and relatively low complexity of the model. However, as McCarthy & Masters (2005) explains, differences of less than two indicates indistinguishable models and only when the difference is greater than ten can the poorer model (with higher DIC value) be seen as having virtually no support. This is done one the same basis as for AIC, as described by Burnham & Anderson (2002).
Mean square displacement per time unit was used to illustrate an overview of the observed data (Eq. 9),
𝑀𝑆𝐷 =∑ 𝑌𝑖𝑗2
𝑡𝑖𝑗
𝑗𝑚𝑎𝑥 Eq. 9
where Yij is the observed dispersion of an individual’s trial, tij is the time
for Yij and jmax is the number of trials for individual i.
Computation of the model analysis was done in R (R Core Team 2014) with the additional packages; circular (Agostinelli & Lund 2013), MCMCpack (Martin et al 2011), partitions (Hankin 2006), matrixStats (Bengtsson 2015), Hmisc (Harrell & Dupont 2016) and undocumented functions to a large extent written by Tom Lindström, but modified by the author.
4 Results
Sex determination revealed a bias of males over females in the wild
caught batches, but a higher occurrence of females in the common garden animals (Table 2.) The high mortality in the summer batch and release of small individuals in the autumn batch may have distorted the true
sampling proportion. When comparing catch per unit effort (CPUE) from net sampling in the summer batch, based on original catch data, the
surface sample locations had a higher CPUE compared to the cave
locations. Trap sampling showed the opposite, but often included possible predators in the surface locations, which may have reduced the efficiency of these traps.
Table 2. Distribution of isopods in regard to sampling site, sex, sampling period and method. Only individuals which were used in the displacement experiment are listed. However, individuals caught in traps were omitted in the Bayesian analysis due to potential sampling bias between methods.
Summer Autumn Common
Net Trap Net garden
Site Female Male Female Male Female Male Site Female Male
US1 1 11 3 1 15 15 US 22 7 US2 1 12 0 0 20 10 CA 17 12 US3 3 7 1 2 15 14 CA1 4 5 0 2 14 16 CA2 3 11 0 9 11 19 CA3 6 15 2 5 11 19 DS1 8 5 0 0 7 23 DS2 5 9 0 0 9 20 DS3 1 5 0 1 4 26 Sum 32 80 6 20 106 162 39 19
As a representation and summary of the observed data Yij, MSD per time
unit (Figure 4) revealed large variation within sample location and big overlap between sample locations.
Figure 4. Overview of dispersion as MSD per time unit (cm2s-1). Boxplots with
the median as thick horizontal line, 25th and 75th percentile as lower and upper
box limits, whiskers as either the extreme ends of the data or, when outlying data points are present, 1.5 times the inter quartile range (the distance
between the 25th and 75th percentile). For abbreviations of locations see Table
1.
4.1 Common garden isopods
The model selection for common garden individuals did not favour a single model over another as there was less in DIC-value than two
between the lowest scoring model and the second (Table 3). As these two models were contradictory, there was no evident result regarding
differences in dispersal between the cave and surface phenotype.
However, when analysing model 2 (M = 2 in Table 3), the probability of
β0Cave (the effect on dispersal from belonging to the cave phenotype)
being higher than β0Surface was 0.85. As model 1 (M = 1 in Table 3) group
all individuals as the same group, there can be no likelihood comparison of β0 for that model. Although there was a trend of β0Cave being larger than
β0Surface, the credibility intervals largely overlap (Figure 5), suggesting
autocorrelation with α. In both top candidate models (M = 1 and 2) body length was used as a covariate and it had positive effect on individual dispersion (through Di)in both models, with 0.95 credibility intervals not
Table 3. Model selection with DIC for common garden isopods. Grouping indicates if the model regarded the cave phenotype and surface phenotype as one homogenous group (1, 1) or separate (1, 2). The covariates length of seventh and third leg respectively, head angle, body length and sex are either active (1) or inactive (0). M is just an index number for the models.
Figure 5. Intercept for the population effect with β0Cave, β0Surface from model 2
and β0 from model 1 in the common garden analysis. Points are the median
and error bars show the 0.95 credibility interval for the posterior distribution.
M Grouping Leg 7 Leg 3 Angle Body L Sex DIC
1 1 1 0 0 0 1 0 114.64 2 1 2 0 0 0 1 0 115.46 3 1 2 0 0 1 0 0 117.23 4 1 1 0 0 0 0 0 117.66 5 1 2 0 0 0 0 0 117.71 6 1 1 1 0 0 0 0 117.91 7 1 2 1 0 0 0 0 118.29 8 1 1 0 0 0 0 1 118.35 9 1 1 0 1 0 0 0 118.39 10 1 2 0 0 0 0 1 118.58 11 1 1 0 0 1 0 0 118.67 12 1 2 0 1 0 0 0 119.07 13 1 1 1 1 1 1 1 119.66 14 1 2 1 1 1 1 1 120.17
Figure 6. Effects of body length as a covariate, β1, for the common garden analysis. Credibility intervals did not overlap zero in neither of the models, which suggests that body length always had a positive effect on dispersion. Points are the median and error bars show the 0.95 credibility interval for the posterior distribution.
4.2 Wild caught isopods
4.2.1 Summer & Autumn batch
Model selection for wild caught isopods with both catchment batches in the same analysis did not favour a single model as there were other
models within two DIC scores of the one with the lowest value (Table 4). As with the analysis of the common garden animals, the two top
candidate models were contradictory and therefore no evident grouping of locations could be done. When analysing model 2 (M = 2 in Table 4), the probability of β0Surface (the effect on dispersal from belonging to the
surface phenotype)being higher than β0Cave was 0.82, but the credibility
intervals largely overlap (Figure 7). In the top candidate models the catchment period was the single active covariate and it had a positive effect on Di in both models, indicating that individuals from the autumn
batch disperse faster than individuals in the summer batch (because catchment was analysed as summer = 0 and autumn = 1) (Figure 8, see also Figure 4).
Table 4. Model selection with DIC for wild caught animals. Numbers for respective sample location (see Table 1 for abbreviations of locations) corresponds to the grouping in the model, thus the locations with equal
number were grouped together within the same model. The covariates; length of seventh and third leg respectively, catchment period, body length and sex are either active (1) or inactive (0). M is just an index number for the models. Models within ten in DIC distance to M = 1 are shown. See section 8.1 (Appendix 1) for extended table.
M US1 US2 US3 CA1 CA2 CA3 DS1 DS2 DS3 7leg 3leg Catch BodyL Sex DIC
1 1 1 1 1 1 1 1 1 1 0 0 1 0 0 682.07 2 1 1 1 2 2 2 1 1 1 0 0 1 0 0 683.22 3 1 1 1 2 2 2 3 3 3 0 0 1 0 0 684.85 4 1 1 1 1 1 1 1 1 1 1 1 1 1 1 686.03 5 1 1 1 2 2 2 1 1 1 1 1 1 1 1 686.92 6 1 1 1 2 2 2 3 3 3 1 1 1 1 1 688.55
Figure 7. Intercept for the population effect with β0Cave, β0Surface from model 2
and β0 from model 1 in the analysis of all wild caught animals. Points are the
median and error bars show the 0.95 credibility interval for the posterior
Figure 8. Effects of catchment period as a covariate, β1, for the analysis of wild caught animals in model 1 and 2. Credibility intervals did not overlap zero in neither of the models, which suggests that individuals in the autumn batch disperse faster than individuals from the summer batch. Points are the median and error bars show the 0.95 credibility interval for the posterior distribution.
4.2.2 Autumn batch
Model selection in analysis of only the autumn batch follows the pattern from Table 3 and 4, but in here as many as eight models were ranked as being inseparable (Table 5, Appendix 1). Most of the top candidate models group all sample locations together, but two models group cave and surface locations separately. No covariate of particular impact was found, as model selection did not seem to favour any particular covariate and all credibility intervals for covariate effects overlapped zero
(credibility intervals not presented). When analysing model 6 and 7 (the top candidate models with more than one grouping, M = 6 and 7 in Table 5), the probability of β0Surface (the effect on dispersal from belonging to the
surface phenotype)being higher than β0Cave was 0.72 and 0.76
Figure 9. Intercept for the population effect with β0Surface and β0Cave from model
6 and 7. Also β0 for models 1 through 5 and 8, which group all locations
together. Analysis done on wild caught animals from the autumn batch. Points are the median and error bars show the 0.95 credibility interval for the
posterior distribution.
4.2.3 Summer batch
Model selection in analysis of only the summer batch follows the same pattern as shown previously for the autumn batch, with the top eight candidate models being inseparable (Table 6, Appendix 1). However, the exact same ranking does not appear. Most of the top candidate
alternatives group all sample locations together, but model 4 and 8 group cave and surface locations separately. No covariate of particular impact was found, as model selection did not seem to favour any particular
covariate and all credibility intervals for covariate effects overlapped zero (credibility intervals not presented). When analysing model 4 and 8 (the top candidate models with more than one grouping, M = 4 and 8 in Table 6), the probability of β0Surface (the effect on dispersal from belonging to the
surface phenotype) being higher than β0Cave was 0.79 and 0.81
Figure 10. Intercept for the population effect with β0Surface and β0Cave from
model 4 and 8. Also β0 for models 1 through 3 and 5 through 7, which group
all locations together. Analysis done on wild caught animals from the summer batch. Points are the median and error bars show the 0.95 credibility interval for the posterior distribution.
5 Discussion
There is an urgent need to understand dispersal in order to predict
responses to changes in environment as well as fragmentation (Bowler & Benton 2005), where local adaptations can influence the dynamics of range shifting (Atkins & Travis 2010). Dispersal dimorphism often evolve (Kisdi 2002) and a wide array of traits are connected to high or low expression of dispersal (Hudina et al 2014). It is therefore important to get a better understanding of the differentiation of dispersal traits in order to connect these traits to other ecological mechanisms. As more suitable habitat select for increased dispersal distance (smaller risk of migrating to an unfavourable habitat) (Bonte et al 2010), it is reasonable to assume a significant divergence in dispersive traits between cave dwelling and surface populations of A. aquaticus. Habitat clumping generates high dispersal success even among weak dispersers, especially for individuals constrained to one habitat type (King & With 2002). The cave habitat is aggregated and it is fair to assume that the cave phenotype is constrained to the cave. Because of the ambiguous results, there were however no results from my study to support this statement. The dispersal trait expression could be alike, but it might just as well differ between phenotypes, as the differences in DIC-value between models was too small.
However, there is a big difference in predator abundance between the habitats. Both fish and predatory macroinvertebrates are frequently found in the surface habitat and scarcely in the cave habitat (personal
observation; Hargeby, personal communication) and predation from birds does not exist in the cave. My results might to some degree derive from divergent selection on foraging rate, as predatory evasion and resource competition likely have different impact in surface and cave habitats. Foraging rate is likely to respond to shifts between minimizing of
predation risk and intraspecific competition, depending on their relative importance (Urban & Richardson 2015). The combined effects of
selection from habitat availability and optimization of foraging rate might be a cause for similar dispersal trait expression between the phenotypes. Under the assumption that the selected models which group cave and surface animals separately is correct, there were opposing trends between the analysis of common garden and wild caught animals. In common garden, the cave animals were more likely to disperse longer, while the opposite tendencies were seen in the wild caught animals. This might simply be due to stochasticity, but might also be explained by the risk allocation hypothesis (Lima & Bednekoff 1999). According to this hypothesis animals which are exposed to short periods of high risk can afford to forage at a moderate pace while unthreatened and show strong vigilance (antipredator response) while exposed to the threat. Individuals which are exposed to high risk over an extended period of time must instead maximize foraging rate while the threat is low. If foraging under those short periods does not meet energy demands, individuals
experiencing high predation threat should forage at a moderate pace even when exposed to high risk (Ferrari et al 2009) and under environmental variability (Higginson et al 2012). However, the energetic state of the prey greatly affects how strong risk allocation is expressed (Matassa & Trussell 2014). This could explain why the surface phenotype showed a tendency for higher dispersal in the wild caught batch but not in the common garden, even when taking into account that the cave phenotype optimize for foraging rate. Because the common garden isopods lacked exposure to predators, the surface phenotype would not show the same response to utilize the period of low risk. For the wild caught isopods of surface phenotype, lab conditions embodies low predation risk. The
above reasoning assumes both divergent selection and adaptive behaviour in the individual to current conditions. It is also possible that such an adaptive response affect the surface phenotype so that the initial response differs from the cave phenotype, but with time changes to similar levels (McNamara et al 2013).
My comparison of the phenotypes might increase the insight into differentiation of dispersal traits due to local adaptations, but further studies are needed. Beyond the obvious fact that more data might be beneficial, such studies might benefit from using experimental setups which enables analysis as a random walk (e.g. Kareiva & Shigesada 1983; Westerberg et al 2008) rather than diffusion. Codling et al (2008) discusses several models for movement which might be more beneficial to the study of dispersal in relation to local adaptation, but likely require usage of two dimensional test arenas and auto tracking software to work properly on A. aquaticus. Approaches with informed movement might also see success (e.g. Fronhofer et al 2013), but are probable harder to test empirically on Asellus.
Body size seems to have some impact on dispersal in A. aquaticus, as both top candidate models in the common garden analysis show positive effects from body length on dispersion. Interestingly, sex did not, even though males are known to be more active than females (Huang & Sih 1990). Sexual selection favour large males (Bertin et al 2002), and males are generally larger than females (eg Harris et al 2011).However, for the isopods used in the common garden experiment size difference was not evident (mean body length and SD: females 7.2 ± 0.67 mm, males 7.9 ± 1.04 mm).
There are several possible explanations for dispersive difference between the autumn and summer batch. Different stages in life is a likely
explanation, as age might influence dispersal (Altwegg et al 2000). It is also reasonable that the Saprolegnia outbreak in the summer batch might have weakened the surviving individuals. Another factor which might contribute to the difference is slightly longer storage of individuals from the summer batch.
Studies have proposed that darkness is the major selective force
responsible for subterranean adaptation (Culver et al 2010), but Culver & Pipan (2012, 2015) highlights the need to look at cave habitats as a
continuum of selective forces with both convergent and divergent
selection. It is reasonable to assume that the combination of these forces have an effect on the dispersive traits in subterranean Asellus populations. According to long term time series of phosphorous and nitrogen,
measured just after the stream leaves the cave system, the nutrient levels are very high (Abrahamnsson 2010). This suggests that cave adaptations in Cave Lummelunda does not derive from contrasting selection due to low nutrient levels. In a recent study Konec et al (2015) compare
concordant and contrasting evolution. However, there are many other aspects which differ between those populations, such as different ancestry and sulphidic water, making general assumptions about selective forces due to nutrient levels on Asellus in caves problematic. The concordant evolution found in the above mentioned study is, according to Konec et al, probably due to darkness as a selective force.
At least some of the cave and surface phenotypes of A. aquaticus is inter-fertile (Protas et al 2011) and although hybridization between the specific phenotypes have been successful in vitro (Hargeby, personal
communication) it is yet unclear if it occurs in vivo. Assortative mating with regard to phenotype occur in recently diverged phenotypes of the species (Hargeby & Erlandsson 2006; Eroukhmanoff et al 2011), but limited gene flow between very diversified populations have been found (Verovnik et al 2003). However, the latter study also points towards existence of barriers against gene flow between populations. It is also possible that asymmetric isolation barriers occur, limiting gene flow to one direction, which in some cases can be caused by such small
differences in habitat as microclimatic conditions (Gosden et al 2015). However, genetic analysis can increase the insight into the particular gene flow in vivo, as proved in other geographic areas by for instance
Verovnik et al (2003, 2004, 2005), Sworobowicz et al (2015) or Konec et al (2016). With such knowledge it would be possible to investigate gene flow as supplement to studies on phenotypic level.
It might be hard to generalize findings about dispersal traits in regard to cave habitats as the apparent mixture of divergent and convergent selective forces in caves likely acts on dispersal traits in interaction. Quantification of the selective intensity of these forces is therefore key to predict dispersal trait expression. It is likely that the population effect (β0)
on dispersal in the different habitats in this study have been influenced more intensely by different selective forces. It is possible that these forces result in a contrasting adaptation, which yet resemble similar effect on dispersal, which in such a case would lead to difficulties when comparing the phenotypes directly.
5.1 Conclusions
The main hypothesis that isopods from the cave phenotype would be less dispersive than isopods from the surface phenotype could not be
supported by the results. Neither could support for similar dispersal be obtained or higher dispersal for the cave phenotype. However, it is reasonable to assume either similarity or difference between the phenotypes occur. The inconclusive results might derive from large variation in dispersal data on both individual and population level, but also on divergent adaptations which yield similar expression of dispersal. Further study with experimental design which is more adapted to the discontinuous and variable movement of the isopods may bring clarity to the question. That body size had positive effects on dispersal in the
common garden analysis highlights the potential of the concept behind this study.
6 Acknowledgement
I would like to thank my supervisors Anders Hargeby and Tom
Lindström for all their help throughout the project. Anders for assistance with field collection, rearing of common garden animals and crucial discussions about A. aquaticus. Tom for introducing and guiding me through the Bayesian world. I also thank Björn Eriksson for assisting with the second field collection and Hanne Ödin for helping out with practicalities at the collection site.
7 References
Abrahamnsson I (2010) Vattenkemiska tillstånd och trender i vattendrag
på Gotland - En utvärdering av resultat från den samordnade recipientkontrollen: Rapporter om natur och miljö nr 2010:16
Agostinelli C, Lund U (2013) R package “circular”: Circular Statistics. Retrieved from https://r-forge.r-project.org/projects/circular/
Altwegg R, Ringsby T H, Sather B E (2000) Phenotypic correlates and
consequences of dispersal in a metapopulation of house sparrows Passer domesticus. Journal of Animal Ecology 69(5), 762–70
http://doi.org/10.1046/j.1365-2656.2000.00431.x
Arakelova K S (2001) The evaluation of individual production and scope
for growth in aquatic sow bugs (Asellus aquaticus). Aquatic Ecology 35(1), 31–42 http://doi.org/10.1023/A:1011446224456
Atkins K E, Travis J M J (2010) Local adaptation and the evolution of
species’ ranges under climate change. Journal of Theoretical Biology 266(3), 449–57 http://doi.org/10.1016/j.jtbi.2010.07.014
Auld J R, Agrawal A A, Relyea R (2009) Re-evaluating the costs and
limits of adaptive phenotypic plasticity. Proceedings Biological
Sciences / The Royal Society 277, 503–11 http://doi.org/10.1098/rspb.2009.1355
Baquedano F J, Valladares F, Castillo F J (2008) Phenotypic plasticity
blurs ecotypic divergence in the response of Quercus coccifera and Pinus halepensis to water stress. European Journal of Forest
Research 127(6), 495–506 http://doi.org/10.1007/s10342-008-0232-8 Barton K a, Hovestadt T, Phillips B L, Travis J M J (2012) Risky
movement increases the rate of range expansion. Proceedings of the
Royal Society B: Biological Sciences 279(1731), 1194–202 http://doi.org/10.1098/rspb.2011.1254
Benard M F, McCauley S J (2008) Integrating across life-history stages:
consequences of natal habitat effects on dispersal. The American
Naturalist 171(5), 553–67 http://doi.org/10.1086/587072
Bengtsson H (2015) matrixStats: Functions that Apply to Rows and
Columns of Matrices (and to Vectors). Retrieved from
http://cran.r-project.org/package=matrixStats
Bertin A, David B, Cézilly F, Alibert P (2002) Quantification of sexual
approaches. Biological Journal of the Linnean Society 77(4), 523–33
http://doi.org/10.1046/j.1095-8312.2002.00125.x
Biro P, Dingemanse N (2009) Sampling bias resulting from animal
personality. Trends in Ecology and Evolution 24(2), 67
http://doi.org/10.1016/j.tree.2008.10.004
Bolker B M, Brooks M E, Clark C J, Geange S W, Poulsen J R, Stevens M H H, White J S S (2009) Generalized linear mixed models: a
practical guide for ecology and evolution. Trends in Ecology and
Evolution 24(3), 127–35 http://doi.org/10.1016/j.tree.2008.10.008 Bonte D, Hovestadt T, Poethke H J (2010) Evolution of dispersal
polymorphism and local adaptation of dispersal distance in spatially structured landscapes. Oikos 119(3), 560–6
http://doi.org/10.1111/j.1600-0706.2009.17943.x
Bonte D, Van Dyck H, Bullock J M, Coulon A, Delgado M, Gibbs M, … Travis J M J (2012) Costs of dispersal. Biological Reviews 87(2), 290–312 http://doi.org/10.1111/j.1469-185X.2011.00201.x
Bowler D E, Benton T G (2005) Causes and consequences of animal
dispersal strategies: relating individual behaviour to spatial dynamics. Biological Reviews of the Cambridge Philosophical
Society 80(2), 205–25 http://doi.org/10.1017/s1464793104006645 Burnham K, Anderson D (2002) Model selection and multimodel
inference: a practical information-theoretic approach. Springer
Science & Business Media.
Chevin L M, Lande R (2011) Adaptation to marginal habitats by
evolution of increased phenotypic plasticity. Journal of Evolutionary
Biology 24(7), 1462–76 http://doi.org/10.1111/j.1420-9101.2011.02279.x
Codling E A, Plank M J, Benhamou S, Interface J R S (2008) Random
walk models in biology. Journal of the Royal Society, Interface / the
Royal Society 5(25), 813–34 http://doi.org/10.1098/rsif.2008.0014 Culver D C, Holsinger J R, Christman M C, Pipan T (2010)
Morphological Differences Among Eyeless Amphipods in the Genus Stygobromus Dwelling in Different Subterranean Habitats. Journal
of Crustacean Biology 30(1), 68–74 http://doi.org/10.1651/09-3156.1 Culver D C, Master L L, Christman M C, Hobbs H H (2000) Obligate
cave fauna of the 48 contiguous United States. Conservation Biology 14(2), 386–401 http://doi.org/10.1046/j.1523-1739.2000.99026.x
Life. Acta Carsologica 44(3), 415–25
DeWitt T J, Sih A, Wilson D S (1998) Cost and limits of phenotypic
plasticity. Trends in Ecology & Evolution 13(97), 77–81
http://doi.org/10.1111/j.1558-5646.2009.00647.x
Enckell P (1980) Kräftdjur. Odense: Bokförlaget Signum i Lund AB Englund G, Hambäck P a (2004) Scale dependence of emigration rates.
Ecology 85(2), 320–7 http://doi.org/10.1890/03-3039
Eroukhmanoff F, Hargeby A, Arnberg N N, Hellgren O, Bensch S, Svensson E I (2009) Parallelism and historical contingency during
rapid ecotype divergence in an isopod. Journal of Evolutionary
Biology 22(5), 1098–110 http://doi.org/10.1111/j.1420-9101.2009.01723.x
Eroukhmanoff F, Hargeby A, Svensson E I (2011) The role of different
reproductive barriers during phenotypic divergence of isopod ecotypes. Evolution 65(9), 2631–40
http://doi.org/10.1111/j.1558-5646.2011.01327.x
Eroukhmanoff F, Svenssson E I (2009) Contemporary parallel
diversification, antipredator adaptations and phenotypic integration in an aquatic isopod. PLoS ONE 4(7), 1–11
http://doi.org/10.1371/journal.pone.0006173
Ferrari M C O, Sih A, Chivers D P (2009) The paradox of risk allocation:
a review and prospectus. Animal Behaviour 78(3), 579–85
http://doi.org/10.1016/j.anbehav.2009.05.034
Fronhofer E A, Hovestadt T, Poethke H J (2013) From random walks to
informed movement. Oikos 122(6), 857–66
http://doi.org/10.1111/j.1600-0706.2012.21021.x
Galic N, Baveco H, Hengeveld G M, Thorbek P, Bruns E, Van Den Brink P J (2012) Simulating population recovery of an aquatic isopod:
Effects of timing of stress and landscape structure. Environmental
Pollution 163, 91–9 http://doi.org/10.1016/j.envpol.2011.12.024 Gandon S (1999) Kin Competition , the Cost of Inbreeding and the
Evolution of Dispersal. Journal of Theoretical Biology 200, 345–64
GOOGLE EARTH (2016) “Lummelunda” 57°44’1915“N,
18°24’5792”E 2016 Accessed 3 March 2016.
Gosden T P, Waller J T, Svensson E I (2015) Asymmetric isolating
barriers between different microclimatic environments caused by low immigrant survival. Proceedings Biological Sciences / The Royal
Gross J B (2012) Cave Evolution Joshua. eLS
http://doi.org/10.1002/9780470015902.a0023628
Gross J B, Stahl B a, Powers A K, Carlson B M (2015) Natural bone
fragmentation in the blind cave-dwelling fish, Astyanax mexicanus : candidate gene identification through integrative comparative
genomics. Evolution & Development 18, 7 – 18
http://doi.org/10.1111/ede.12131
Hankin R K S (2006) Additive integer partitions in R. Journal of Statistical Software, Code Snippets 16(1)
Hargeby A, Erlandsson J (2006) Is size-assortative mating important for
rapid pigment differentiation in a freshwater isopod?. Journal of
Evolutionary Biology 19(6), 1911–9 http://doi.org/10.1111/j.1420-9101.2006.01170.x
Hargeby A, Johansson J, Ahnesjö J (2004) Habitat-specific pigmentation
in a freshwater isopod: adaptive evolution over a small
spatiotemporal scale. Evolution; International Journal of Organic
Evolution 58(1), 81–94 http://doi.org/10.1554/02-698
Hargeby A, Stoltz J, Johansson J (2005) Locally differentiated cryptic
pigmentation in the freshwater isopod Asellus aquaticus. Journal of
Evolutionary Biology 18(3), 713–21 http://doi.org/10.1111/j.1420-9101.2004.00837.x
Harrell F E, Dupont C (2016) Hmisc: Harrell Miscellaneous. Retrieved from http://cran.r-project.org/package=Hmisc
Harris S, Eroukhmanoff F, Green K K, Svensson E I, Pettersson L B (2011) Changes in behavioural trait integration following rapid
ecotype divergence in an aquatic isopod. Journal of Evolutionary
Biology 24(9), 1887–96 http://doi.org/10.1111/j.1420-9101.2011.02322.x
Harris S, Karlsson Green K, Pettersson L B (2013) Predator faunas past
and present: Quantifying the influence of waterborne cues in divergent ecotypes of the isopod Asellus aquaticus. Oecologia 173(3), 791–9 http://doi.org/10.1007/s00442-013-2667-y
Hendry A P (2004) Selection Against Migrants Contributes to the Rapid
Evolution of Reproductive Isolation. Evolutionary Ecology Research 6, 1219–36
Higginson A D, Fawcett T W, Trimmer P C, McNamara J M, Houston A I (2012) Generalized Optimal Risk Allocation: Foraging and
Naturalist 180(5), 589–603 http://doi.org/10.1086/667885
Huang C, Sih A (1990) Experimental Studies on Behaviorally Mediated ,
Indirect Interactions through a Shared Predator. 71(4), 1515–22
Hudina S, Hock K, Žganec K (2014) The role of aggression in range
expansion and biological invasions. Current Zoology 60(3), 401–9
Kareiva P M, Shigesada N (1983) Analysing insect movement as a
correlated random walk. Oecologia 56(2), 234–8
Karlsson Green K, Eroukhmanoff F, Harris S, Pettersson L B, Svensson E I (2016) Rapid changes in genetic architecture of behavioural
syndromes following colonization of a novel environment. Journal of
Evolutionary Biology 29(1), 144–52 http://doi.org/10.1111/jeb.12769 King A W, With K A (2002) Dispersal success on spatially structured
landscapes: When do spatial pattern and dispersal behavior really matter?. Ecological Modelling 147(1), 23–39
http://doi.org/10.1016/S0304-3800(01)00400-8
Kirkpatrick M, Barton N (2006) Chromosome Inversions, Local
Adaptation and Speciation. Genetics 173, 419–34 http://doi.org/DOI:
10.1534/genetics.105.047985
Kisdi E (2002) Dispersal: risk spreading versus local adaptation. The American Naturalist 159(6), 579–96 http://doi.org/10.1086/339989 Konec M, Delić T, Trontelj P (2016) DNA barcoding sheds light on
hidden subterranean boundary between Adriatic and Danubian drainage basins. Ecohydrology (Accepted)
http://doi.org/10.1002/eco.1727
Konec M, Prevorčnik S, Sarbu S M, Verovnik R, Trontelj P (2015)
Parallels between two geographically and ecologically disparate cave invasions by the same species, Asellus aquaticus
(Isopoda, Crustacea). Journal of Evolutionary Biology (28), 864 – 75
http://doi.org/10.1111/jeb.12610
Lenormand T (2002) Gene flow and the limits to natural selection. Trends in Ecology and Evolution 17(4), 183–9
http://doi.org/10.1016/S0169-5347(02)02497-7
Lima S L, Bednekoff P (1999) Temporal Variation in Danger Drives
Antipredator Behavior: The Predation Risk Allocation Hypothesis.
The American Naturalist 153(6), 649–59 http://doi.org/10.1086/303202
Mail L, Bulog B (2015) The Morphology of Male Gonads of the Neotenic
Manenti R, Ficetola G F (2013) Salamanders breeding in subterranean
habitats: Local adaptations or behavioural plasticity?. Journal of
Zoology 289(3), 182–8 http://doi.org/10.1111/j.1469-7998.2012.00976.x
Marcus J, Sutcliffe D, Willougbby L (1978) Feeding and growth of
Asellus aquaticus (Isopoda) on food items from the littoral of Windermere, including green leaves of Elodea canadensis.
Freshwater Biology 8, 505–19
http://doi.org/10.1017/CBO9781107415324.004
Marshall D J, Monro K, Bode M, Keough M J, Swearer S (2010)
Phenotype-environment mismatches reduce connectivity in the sea.
Ecology Letters 13(1), 128–40 http://doi.org/10.1111/j.1461-0248.2009.01408.x
Martin A, Quinn K, Park J (2011) Markov Chain Monte Carlo (MCMC)
Package. Journal of Statistical Software 42(9), 1–21 Retrieved from
http://www.jstatsoft.org/v42/i09/
Matassa C M, Trussell G C (2014) Prey state shapes the effects of
temporal variation in predation risk. Proceedings of the Royal
Society: Biological Sciences 281(1796), 20141952 http://doi.org/10.1098/rspb.2014.1952
McCarthy M A, Masters P (2005) Profiting from prior information in
Bayesian analyses of ecological data. Journal of Applied Ecology 42(6), 1012–9 http://doi.org/10.1111/j.1365-2664.2005.01101.x
McNamara J, Fawcett T, Houston A (2013) An Adaptive Response to
Uncertainty Generates Positive and Negative Contrast Effects. 340(May), 1084–6 http://doi.org/10.1126/science.1230599
Nosil P, Funk D J, Ortiz-Barrientos D (2009) Divergent selection and
heterogeneous genomic divergence. Molecular Ecology 18(3), 375–
402 http://doi.org/10.1111/j.1365-294X.2008.03946.x
Nosil P, Vines T, Funk D (2005) PERSPECTIVE: REPRODUCTIVE
ISOLATION CAUSED BY NATURAL SELECTION AGAINST
IMMIGRANTS FROM DIVERGENT HABITATS. Evolution 59, 705–
19
Økland K (1978) Life history and growth of Asellus aquaticus in relation
to environment in a eutrophic lake in Norway. Hydrobiologia 59,
243–60
Pearce-Duvet J M C, Elemans C P H, Feener D H (2011) Walking the
Behavioral Ecology 22(3), 501–9 http://doi.org/10.1093/beheco/arr001
Pehrsson O (1984) Relationships of Food to Spatial and Temporal
Breeding Strategies of Mallards in Sweden. The Journal of Wildlife
Management 48(2), 322–39
Phillips B L, Brown G P, Webb J K, Shine R (2006) Invasion and the
evolution of speed in toads. Nature 439(7078), 803
http://doi.org/10.1038/439803a
Pipan T, Culver D C (2012) Convergence and divergence in the
subterranean realm: A reassessment. Biological Journal of the
Linnean Society 107(1), 1–14 http://doi.org/10.1111/j.1095-8312.2012.01964.x
Prevorčnik S, Jugovic J, Sket B (2009) Geography of morphological
differentiation in asellus aquaticus (crustacea: Isopoda: Asellidae).
Journal of Zoological Systematics and Evolutionary Research 47(2), 124–31 http://doi.org/10.1111/j.1439-0469.2008.00493.x
Protas M E, Trontelj P, Patel N H (2011) Genetic basis of eye and
pigment loss in the cave crustacean, Asellus aquaticus. Proceedings
of the National Academy of Sciences of the United States of America
108(14), 5702–7 http://doi.org/10.1073/pnas.1013850108
Protas M, Jeffery W R (2012) Evolution and development in cave
animals: From fish to crustaceans. Wiley Interdisciplinary Reviews:
Developmental Biology 1(6), 823–45 http://doi.org/10.1002/wdev.61 R Core Team (2014) R: A language and environment for statistical
computing. Vienna, Austria: R Foundation for Statistical Computing
Retrieved from http://www.r-project.org/.
Ronce O (2007) How Does It Feel to Be Like a Rolling Stone? Ten
Questions About Dispersal Evolution. Annual Review of Ecology,
Evolution, and Systematics 38(1), 231–53
http://doi.org/10.1146/annurev.ecolsys.38.091206.095611 Spiegelhalter D J, Best N G, Carlin B P, van der Linde A (2002)
Bayesian Measures of Model Complexity anf Fit. Journal of the
Royal Statistical Society Series B (Statistical Methodology) 64(4), 583–639 http://doi.org/10.1111/1467-9868.00353
Sworobowicz L, Grabowski M, Mamos T, Burzyński A, Kilikowska A, Sell J, Wysocka A (2015) Revisiting the phylogeography of Asellus aquaticus in Europe: insights into cryptic diversity and
http://doi.org/10.1111/fwb.12613
Travis J M J, Dytham C (1999) Habitat persistence, habitat availability
and the evolution of dispersal. Proceedings of the Royal Society B:
Biological Sciences 266(1420), 723 http://doi.org/10.1098/rspb.1999.0696
Travis J M J, Mustin K, Bartoń K a, Benton T G, Clobert J, Delgado M M, … Bonte D (2012) Modelling dispersal: An eco-evolutionary
framework incorporating emigration, movement, settlement
behaviour and the multiple costs involved. Methods in Ecology and
Evolution 3(4), 628–41 http://doi.org/10.1111/j.2041-210X.2012.00193.x
Travis J M J, Mustin K, Benton T G, Dytham C (2009) Accelerating
invasion rates result from the evolution of density-dependent dispersal. Journal of Theoretical Biology 259(1), 151–8
http://doi.org/10.1016/j.jtbi.2009.03.008
Turk S, Sket B, Sarbu ?Erban (1996) Comparison between some epigean
and hypogean populations of Asellus aquaticus (Crustacea: Isopoda: Asellidae). Hydrobiologia 337(1-3), 161–70
http://doi.org/10.1007/BF00028517
Urban M C, Richardson J L (2015) The Evolution of Foraging Rate
across Local and Geographic Gradients in Predation Risk and Competition. The American Naturalist 186(1), E16–E32
http://doi.org/10.1086/681716
Van den Brink P J, Baveco J M H, Verboom J, Heimbach F (2007) An
individual-based approach to model spatial population dynamics of invertebrates in aquatic ecosystems after pesticide contamination.
Environmental Toxicology and Chemistry / SETAC 26(10), 2226–36 http://doi.org/10.1897/07-022R.1
Van Oijen M, Rougier J, Smith R (2005) Bayesian calibration of
process-based forest models: bridging the gap between models and data. Tree
Physiology 25(7), 915–27 http://doi.org/10.1093/treephys/25.7.915 Verovnik R, Sket B, Prevorčnik S, Trontelj P (2003) Random amplified
polymorphic DNA diversity among surface and subterranean populations of Asellus aquaticus (Crustacea: Isopoda). Genetica 119(2), 155–65 http://doi.org/10.1023/A:1026085716777
Verovnik R, Sket B, Trontelj P (2004) Phylogeography of subterranean
and surface populations of water lice Asellus aquaticus (Crustacea: Isopoda). Molecular Ecology 13(6), 1519–32
http://doi.org/10.1111/j.1365-294X.2004.02171.x
Verovnik R, Sket B, Trontelj P (2005) The colonization of Europe by the
freshwater crustacean Asellus aquaticus (Crustacea: Isopoda) proceeded from ancient refugia and was directed by habitat connectivity. Molecular Ecology 14(14), 4355–69
http://doi.org/10.1111/j.1365-294X.2005.02745.x
Via S, Bouck A C, Skillman S (2000) Reproductive Isolation between
Divergent Races of Pea Aphids on Two Hosts II Selection against Migrants and Hybrids in the Parental Environments Published by : Society for the Study of Evolution Stable URL :
http://www.jstor.org/stable/2640661 Accesse. 54(5), 1626–37
Voituron Y, de Fraipont M, Issartel J, Guillaume O, Clobert J (2011)
Extreme lifespan of the human fish (Proteus anguinus): a challenge for ageing mechanisms. Biology Letters 7(1), 105–7
http://doi.org/10.1098/rsbl.2010.0539
Westerberg L, Lindström T, Nilsson E, Wennergren U (2008) The effect
on dispersal from complex correlations in small-scale movement.
Ecological Modelling 213(2), 263–72
8 Appendix 1 - DIC Tables
Table 4 extended. Model selection with DIC for wild caught animals. Numbers for respective sample location (see Table 1 for abbreviations of locations) corresponds to the grouping in the model, thus the locations with equal
number were grouped together within the same model. The covariates; length of seventh and third leg respectively, catchment period, body length and sex are either active (1) or inactive (0). M is just an index number for the models.
M US1 US2 US3 CA1 CA2 CA3 DS1 DS2 DS3 7leg 3leg Catch BodyL Sex DIC
1 1 1 1 1 1 1 1 1 1 0 0 1 0 0 682.07 2 1 1 1 2 2 2 1 1 1 0 0 1 0 0 683.22 3 1 1 1 2 2 2 3 3 3 0 0 1 0 0 684.85 4 1 1 1 1 1 1 1 1 1 1 1 1 1 1 686.03 5 1 1 1 2 2 2 1 1 1 1 1 1 1 1 686.92 6 1 1 1 2 2 2 3 3 3 1 1 1 1 1 688.55 7 1 1 1 1 1 1 1 1 1 1 0 0 0 0 692.08 8 1 1 1 2 2 2 1 1 1 1 0 0 0 0 692.44 9 1 1 1 1 1 1 1 1 1 0 0 0 0 0 692.51 10 1 1 1 1 1 1 1 1 1 0 1 0 0 0 692.79 11 1 1 1 2 2 2 1 1 1 0 0 0 0 0 693.39 12 1 1 1 2 2 2 1 1 1 0 1 0 0 0 693.50 13 1 1 1 1 1 1 1 1 1 0 0 0 1 0 693.95 14 1 1 1 2 2 2 3 3 3 1 0 0 0 0 694.34 15 1 1 1 1 1 1 1 1 1 0 0 0 0 1 694.35 16 1 1 1 2 2 2 1 1 1 0 0 0 0 1 694.95 17 1 2 3 4 5 6 7 8 9 0 0 1 0 0 695.04 18 1 1 1 2 2 2 1 1 1 0 0 0 1 0 695.14 19 1 1 1 2 2 2 3 3 3 0 0 0 0 0 695.17 20 1 1 1 2 2 2 3 3 3 0 1 0 0 0 695.19 21 1 1 1 2 2 2 3 3 3 0 0 0 0 1 696.81 22 1 1 1 2 2 2 3 3 3 0 0 0 1 0 696.85 23 1 2 3 4 5 6 7 8 9 1 1 1 1 1 698.77 24 1 2 3 4 5 6 7 8 9 1 0 0 0 0 703.31 25 1 2 3 4 5 6 7 8 9 0 0 0 0 0 704.25 26 1 2 3 4 5 6 7 8 9 0 1 0 0 0 704.60 27 1 2 3 4 5 6 7 8 9 0 0 0 0 1 705.57 28 1 2 3 4 5 6 7 8 9 0 0 0 1 0 705.65