ORIGINAL PAPER

Modelling the risk of invasion by the red-swamp crayfish(Procambarus clarkii): incorporating local variablesto better inform management decisions

Francisco D. Moreira • Fernando Ascensao •

Cesar Capinha • Diana Rodrigues • Pedro Segurado •

Margarida Santos-Reis • Rui Rebelo

Received: 14 September 2013 / Accepted: 20 May 2014

� Springer International Publishing Switzerland 2014

Abstract The correct modelling of the potential

distribution of an invasive species is crucial to define

effective management and monitoring strategies. Here

we compared the results of models built at different

spatial scales to identify the areas at risk of invasion by

the red swamp crayfish (Procambarus clarkii) in the

northwest of Portugal. Firstly, we surveyed crayfish at

97 locations. Secondly, we used presence–absence

data and local variables to model its current distribu-

tion (local variables model) and identified slope and

river width as the best explanatory factors. Thirdly, we

integrated these two local variables into a former

model built for the Iberian Peninsula (regional model)

increasing considerably its predictive power. Finally,

we compared both models focusing on the area

predicted to be invaded. The local model showed a

considerably narrower extent of suitable areas for

crayfish in the study area than the regional model.

These results show that the refinement of regional

scale predictions through the incorporation of species-

environment relationships at local scales may be

important for supporting management decisions. By

not integrating the effects of local variability, regional

bioclimatic models may overlook the potential distri-

bution of this invader at manageable extents. Results

further suggest that a wide range of native ecosystems

of conservation value are probably unsuitable for this

invasive species.

Keywords Biological invasions � Red swamp

crayfish � Risk assessment � Spatial Scale �Species distribution models

Introduction

Invasive species are a major threat to biodiversity

worldwide (Sala et al. 2000; Vie et al. 2009) and an

important part of current research in invasion ecology

aims at predicting the areas at risk of invasion.

Available predictions have been largely based on

correlative approaches that compare environmental

conditions in which species are observed in their

native or introduced ranges and those found in the new

areas (Guisan and Thuiller 2005), usually on a regional

or continental scale (e.g. Ficetola et al. 2007; Capinha

and Anastacio 2011). However, these large-scale

predictions may be insufficient to forecast local-scale

F. D. Moreira (&) � F. Ascensao � C. Capinha �D. Rodrigues

Centro de Biologia Ambiental, Faculdade de Ciencias,

Universidade de Lisboa, 1749-016 Lisbon, Portugal

e-mail: franciscodmoreira@gmail.com

P. Segurado

Centro de Estudos Florestais, Universidade Tecnica de

Lisboa, Tapada da Ajuda, 1349-017 Lisbon, Portugal

M. Santos-Reis � R. Rebelo

Centro de Biologia Ambiental, Departamento de Biologia

Animal, Faculdade de Ciencias, Universidade de Lisboa,

Ed. C2, Campo Grande, 1749-016 Lisbon, Portugal

123

Biol Invasions

DOI 10.1007/s10530-014-0725-y

distribution (Guisan and Thuiller 2005; Loo et al.

2009; Reino et al. 2013), which is often the scale at

which management options are feasible. On the other

hand, using local variables alone to predict the

invasibility can lead to erroneous assumptions of

suitability because factors acting at coarser scales (e.g.

climate) may hamper the species’ establishment. This

duality emphasizes the importance of using analyses at

both local and regional/continental scales when esti-

mating the invasibility of an area. Nevertheless, local

analyses have been frequently ignored despite their

potential to refine information about invasibility, an

essential step for defining eradication/contention

strategies of invasive species or for mitigating their

effects in the native ecosystems (Loo et al. 2009).

The red swamp crayfish (Procambarus clarkii,

Girard 1852, hereafter ‘crayfish’) is native to south-

eastern North America and is currently an invasive

species in five continents (Gherardi 2006). In the

Iberian Peninsula the species was introduced in 1973

in southwestern Spain and was first recorded in SE

Portugal in 1979 (Ramos and Pereira 1981). Since

then, and as a result of natural dispersion and

anthropogenic transport, its distribution is expanding

to the north. In fact, humans have been playing a key

role in this invasion, transporting crayfish to new sites

and/or altering the landscape in a way that facilitates

its spread and establishment (Siesa et al. 2011;

Capinha et al. 2013).

The negative impacts of this invader are largely

known, affecting many taxa from aquatic plants

(Correia 2003; Geiger et al. 2005; Carreira et al.

2013) to invertebrates (e.g. Correia 2003; Geiger et al.

2005; Correia and Anastacio 2008) and amphibians

(e.g. Cruz et al. 2006, 2008; Nunes et al. 2010; Ficetola

et al. 2011). Conversely, there may be positive impacts

on higher trophic levels as a result of the role of this

crayfish as an abundant and easy prey for many

mammalian and avian predators (Correia 2001; Geiger

et al. 2005). Either way, its establishment disrupts the

structure and function of native ecosystems by

changing the relations between species; as well as

the flow of matter and energy (Geiger et al. 2005;

Ficetola et al. 2012).

During the last 40 years P. clarkii established itself

in all of Portugal’s main river basins. Capinha and

Anastacio (2011), using climatic and physiographic

predictors, suggested that one of the distribution limits

for this species will be the Northwest of the Iberian

Peninsula; however these authors also suggested that

most of the Northwest is suitable for this species.

Moreira (2011) conducted a bibliographical analysis

of more than 2,500 articles (scientific and grey

literature including news in local journals) and found

that: (1) the first records of the species in Northwest

Portugal date back to the year 2000; (2) four years later

the species was already present in all of the main river

basins of this region (Fig. 1). However, i n spite of

being present for at least a decade in this area this

crayfish was apparently not able to colonize all the

available freshwater habitats. This absence may be due

to local limiting factors, such as steep slope and high

water velocity, factors that have already been identi-

fied by other authors as limitative of P. clarkii

expansion in other areas of the Iberian Peninsula

(Cruz and Rebelo 2007; Banha and Anastacio 2011;

Gil-Sanchez and Alba-Tercedor 2002; Maceda-Veiga

et al. 2013).

In this study we aimed to identify the most

influential factors that are conditioning P. clarkii

expansion in the fringe of its distribution. We first

assessed the species presence and modelled its current

distribution. We then evaluated how the inclusion of

local variables improved Capinha and Anastacio

(2011) model. With this information we expected to

improve the identification of the areas at most risk of

being invaded.

Methods

Study area

The study was conducted in the northwest of Portugal,

a mountainous area (maximum altitude of 1.507 m)

within the Atlantic climate region, with an annual

mean temperature ranging between 7.8 and 15.6 �C

and an annual precipitation of 981 mm to 1,705 mm.

Along a SW–NE axis this area shows a gradient of

increasing elevation, as well as a change from silt to

rocky soils. There are just a few large lentic water

bodies; the landscape is dominated by large extensions

of different sized rivers (Fig. 1).

The region is one of the most biologically diverse in

Portugal, with approximately 30 % of its area under

some type of legal protection (Fig. 1), which may

constitute an opportunity to implement management

strategies concerning the invasion by P. clarkii.

F. D. Moreira et al.

123

Field survey

To assess crayfish presence we sampled 97 river

stretches spread throughout the study area, covering

all of the main river basins (Fig. 1). Crayfish distri-

bution data was collected from July to October 2010,

matching the time of year when the species is more

active (Gherardi et al. 2000). Sampling sites corre-

spond to different sized rivers. Within this paper,

‘‘small rivers’’ refers to rivers with strahler levels of

1–2 (that is, with either one or no tributaries of their

own) and which are \10 m wide; ‘‘medium sized

rivers’’ are those with strahler 2–5, and which usually

are between 10 and 30 m wide; and ‘‘main rivers’’ are

those that flow directly to the sea and are more than

30 m wide. At each sampling site we performed a

100 m transect along the river bank and searched for

signs of crayfish presence (live specimens or pieces of

recently predated individuals). When no signs of

crayfish were found, we installed five funnel traps (5 L

containers with a 43 mm diameter opening) baited

with fish-tasted cat food. Traps were set until the

presence of crayfish was confirmed or for a maximum

of three consecutive nights. For each basin, the

sampling was carried upstream, starting from the

vicinity of the river mouth.

Environmental variables

To characterize each 100 m river stretch we used ten

local environmental variables (for the complete list

and methodological details see Table 1) plus the

occurrence probability of P. clarkii according to the

Capinha and Anastacio (2011) model (hereafter

referred to as ‘‘regional model’’) which included the

following variables compiled at 1 km resolution:

annual mean temperature; mean temperature of

warmest quarter; mean temperature of coldest quarter;

annual precipitation; precipitation of wettest quarter

and precipitation of driest quarter; altitude; slope;

distance to ocean and a compound topographical

index.

Fig. 1 Study area within

the Iberian Peninsula; Grey

areas are under

environmental protection—

Natura 2000. Black dots

represent dates when

crayfish was first recorded in

each basin before our

sampling effort in 2010.

Black open square indicates

the area where the sampling

efficiency was tested.

Adapted from Moreira

(2011)

The red-swamp crayfish (Procambarus clarkii)

123

Slope was extracted for each river segment from the

Catchment Characterisation and Modelling database

(Vogt et al. 2007) as a surrogate of water current

velocity which has been pointed by several authors as

having a negative impact on this species (Gil-Sanchez

and Alba-Tercedor 2002; Cruz and Rebelo 2007). This

slope is actually quite different from the one used by

Capinha and Anastacio (2011) as it refers to the slope

of the riverbed and not to the average slope of a one

square kilometre cell. Additionally, we measured

width and depth of the river on site since they can

also influence current velocity.

The variable Human impacts assessed the level of

anthropological disturbance as it has been shown to

positively influence the distribution of crayfish (Siesa

et al. 2011; Capinha et al. 2013). The dominant habitat

surrounding each site was characterized within a 10 m

buffer along the river margin, since the surrounding

habitat can influence river characteristics (Allan

2004).

Soil and vegetation of the river margins and bed

were characterized by identifying soil and vegetation

types in ten points spaced out by 10 cm in three 1-m

transects distributed randomly along each sampling

point; counts of each type of soil and vegetation were

then converted into percentages. Soil was character-

ized according to its granulometry as the crayfish has

been referred to prefer silt surfaces because of its

burrowing behaviour (Correia and Ferreira 1995).

Margin vegetation was also characterized on the basis

of its cover using an increasing magnitude of human

intervention: from natural riparian forest to clear river

Table 1 Explanatory variables for modelling

Variable Type Unit Description Total range and

range of values

used for modelling

Regional model Continuous – Probability according to Capinha and Anastacio (2011) 0–0.82

0.56–0.81

Altitude Continuous m GPS 4–369

4–250

Slope Continuous % Steepness of the river segment 0–26.9

0–7.7

Width Continuous m Distance between the two margins (wet region) 2–315

3–315

Depth Continuous cm Water depth at each sampling site 20–130

30–120

Human impact Ordinal – Anthropogenic disturbance

1) no disturbance; 1–3

2) small impacts (such as for leisure activities); 1–3

3) high human impact and river bed modification

(e.g. channelling)

Surrounding area Ordinal – Land cover in the first 10 m of margin

1) natural vegetation; 1–3

2) afforestation or traditional small-scale agriculture; 1–3

3) human structures (houses, paved roads, etc.).

Margin soil Categorical – Soil type at the margin according to granulometry, from

1 (softest) to 4 (hardest)

1–4

1–3

River bed soil Categorical – Soil type at the river bed according to granulometry,

from 1 (softest) to 4 (hardest)

1–4

1–3

Margin vegetation Categorical – Vegetation type at the margin classified in a landscape

with increasingly more human impact

1–4

1–4

River bed vegetation Categorical – Vegetation type at the river bed classified by the

percentage of cover

1–4

1–4

F. D. Moreira et al.

123

banks. High cover of margin vegetation can influence

microclimate and deprive riverbed vegetation of light

(Allan 2004). The same procedure was applied to river

bed vegetation, one of the most important components

of crayfish’s diet (Gutierrez-Yurrita et al. 1998).

Although water flow has been found to influence

crayfish distribution (Gil-Sanchez and Alba-Tercedor

2002; Cruz and Rebelo 2007; Banha and Anastacio

2011), it was not considered in our models because

frequent and unpredictable discharges from dams

located upstream in all the main rivers of NW Portugal

cause a high temporal variability.

Data analysis

Sampling efficiency

Before further analysis, we assessed the accuracy of

our method in detecting the real absence of crayfish in

sampled sites, by calculating the probability of not

finding crayfish where the species is known to be

present (Type II error). This test was performed in a

small protected area (‘‘Paisagem Protegida da Lagoa

de Bertiandos e S. Pedro d’Arcos’’) and nearby rivers

(Fig. 1), in five different sites where crayfish is

present, but with contrasting abundances—two sites

with abundant populations (89 % occupied traps; 2.46

individuals caught per unit effort—CPUE) and three

sites where crayfish is captured only occasionally

(16 % occupied traps; 0.17 CPUE) (Rodrigues,

unpublished data). We used 15 traps per site that were

set for five consecutive nights and checked daily for

crayfish presence.

We built occupancy models (MacKenzie et al.

2002) separately for each type of site, using PRE-

SENCE software (Hines 2006). We randomly selected

five traps (out of the 15 that were used each night) to

simulate one sampling unit, and results were then

transformed to presence/absence. This random proce-

dure of choosing five out of the 15 traps was repeated

1,000 times for each of the 5 days of sampling. We

then estimated the detectability of the species using

one to five sampling days, in both low and high

abundance sites.

Modelling procedures

To investigate the relative influence of environmental

variables in the crayfish distribution and predict the

potential area of expansion we fitted logistic regres-

sion models to our data. A logistic regression model

was used to evaluate relationships between presence/

absence in our sampling sites and the set of environ-

mental predictors depicted in Table 1. Presence/

absence data were defined as follows: for each main

river basin, the invaded area was considered to

encompass the entire basin starting from the most

upstream site with confirmed presence. From this area,

we selected all presences (n = 46) for model building.

Absences were also chosen within the invaded area,

based on whether or not the site was located nearby a

site known to have crayfish; only those sites where

crayfish was not recorded and were distanced\10 km

from a known ‘presence’ were considered ‘absences’

(n = 16) (see Fig. 3). This is an important step to

insure that the equilibrium assumption is met (Guisan

and Zimmermann 2000).

Before model building we evaluated the collinear-

ity between the variables, using the Variation Inflation

Factor, which shows how much the regression coef-

ficient is increased due to collinearity. Variables with a

Variation Inflation factor above five were removed

from the subsequent analysis (Craney and Surles

2002). This procedure removed Altitude and Depth

from further analysis.

We constrained our models to include the variable

representing the probabilities of species occurrence

projected by the regional model for each sampling site,

while the other remaining eight variables were com-

bined in all possibilities in a total of 255 models

(hereafter referred to as ‘combined models’). Model

selection procedure was based on the Akaike Infor-

mation Criteria adjusted for small sample sizes (AICc,

Burnham and Anderson 2004). In order to identify

those models with larger support, Akaike weights (wi)

were computed for each candidate model (dAICc \ 4,

Richards 2005).

A Moran’s I test was performed to check for spatial

autocorrelation in the residuals of the best fitted

model. For that purpose a distance matrix among all

sites was computed based on distances along the river

network. Dispersion between river basins was consid-

ered to be negligible.

We assessed the classification accuracy of the

different models using a leave-one-out cross-valida-

tion to compute two commonly used measures

(Allouche et al. 2006): area under the ROC (AUC)

and true skill statistics (TSS). For TSS computation

The red-swamp crayfish (Procambarus clarkii)

123

and presence/absence predictions, we used species

prevalence as the cut-off probability.

To map the probability of colonization of the river

stretches that were not directly sampled, we first

divided the best combined model into two simpler

models: (1) a ‘Regional model’ represents the proba-

bility of distribution according to Capinha and Anasta-

cio (2011), as explained previously; and (2), a ‘local

model’ that was built with the two local variables

included in the combined model. This local model does

not represent the result of the modelling selection

procedure. Using the partial responses of river width,

we could determine, to any given slope, the critical

value for the model to change its prediction from

absence to presence, based on the cut-off probability.

River width was calculated with the aerial images from

the software Google Earth (version 7.1.2.2041) to any

given site. We used this function to classify all the river

stretches of the study area as susceptible of being

invaded, or protected from invasion. This separation

between presence and absence sites was clear for most

stream sections, either with steep slopes (where the

species would require wide rivers to be present) or

moderate slopes (where the species would still be able

to establish in narrow rivers).We then overlapped the

map resulting from this approach with the regional

model in order to trace the species potential distribu-

tion boundaries in each basin.

All statistical analyses were performed in R soft-

ware version 2.12 (R Development Core Team 2012).

Fig. 2 Species distribution

model by Capinha and

Anastacio (2011) and

current distribution of

crayfish. Black to white

gradient represents a

decrease in suitability of the

area. Black and white dots

represent crayfish presence

and absence sites,

respectively. Dashed line

represents the separation of

the invaded (left-side) and

non-invaded area (right-

side) (see ‘‘Methods’’ for

details)

F. D. Moreira et al.

123

AICc and wi were computed using the R package

MuMIn (Barton 2009). Moran’ I values and tests were

computed with the R package spdep (Bivand 2007).

Results

Sampling efficiency

Sampling efficiency was very high in sites with high

crayfish abundance, with 100 % probability of detec-

tion within the first day with any given set of five traps.

At low abundance sites the probability of detecting

crayfish within 1 day of sampling was also high

(78.6 %) and after 3 days of sampling it increased to

97.4 %. On the basis of these results, we set 3 days as

the minimum number of days necessary to confirm

crayfish presence at any given site.

Field survey

We found crayfish in 46 out of 97 sampling sites. The

species is mostly present at lower altitudes and is

apparently more widespread in the southern part of the

study area, where it has supposedly arrived earlier

(Fig. 2).

Combined suitability models

The overlap of the current distribution of crayfish with

the regional model highlighted many river segments

located in the non-invaded area with high environmental

suitability for the species, and conversely absences in

river segments showing high suitability index values in

the invaded area (Fig. 2).

When considering only the combined models, eight

models best fitted our data (Table 2). All of these

represent a considerable increase in predictive power

when compared with the regional model. Furthermore,

all the models include a common set of three variables

(regional, width and slope) and therefore we consid-

ered the simpler nested model as the final model

(Richards et al. 2011).

Slope and river width were the most important

variables being included in all of the best combined

models. We used these variables to represent a

probability of colonization of a given site. Plotting

the results of width versus slope, two types of sites

emerged: wide rivers in flatlands (crayfish presence)

and narrow rivers and streams in moderate to steep

slopes (crayfish absence). The results from our model

were then mapped for the whole study area, indicating

that crayfish will likely invade most of the extent of the

main rivers, as well as some of the medium sized

rivers. Interestingly, most of the small rivers seem to

have a lower risk of invasion (Fig. 3).

Both the local model and the regional model agreed

in 61.9 % of the river segments, with 28.2 % agree-

ment in predicted presences and 33.7 % in predicted

absences. The remaining corresponded almost exclu-

sively (37.4 %) to segments with a high suitability

index according to the regional model and a low

suitability index according to the local model, with

\1 % for the opposite condition (Fig. 4).

Table 2 Candidate combined models and the best possible model with local variables

AICc dAICc Akaike weight TSS AUC

Combined models

Regional (?); Slope (-); Width (?) 32.68 0 0.16 0.64 0.94

Regional (?); Slope (-); Width (?); Riverbed veg. (-) 34.61 1.93 0.06 0.64 0.94

Regional (?); Slope (-); Width (?); Human imp (?) 34.70 2.02 0.06 0.64 0.93

Regional (?); Slope (-); Width (?); Margin veg. (-) 34.78 2.10 0.06 0.64 0.94

Regional (?); Slope (-); Width (?); Margin soil (-) 34.85 2.17 0.05 0.64 0.94

Regional (?); Slope (-); Width (?); Surround. area (?) 34.98 2.30 0.05 0.64 0.94

Regional (?); Slope (-); Width (?); Riverbed soil (-) 34.99 2.31 0.05 0.64 0.94

Regional (?); Slope (-); Width (?); Human imp (?); Margin veg. (-) 36.45 3.77 0.02 0.66 0.93

Regional (?) 62.44 29.76 0.00 0.21 0.72

The red-swamp crayfish (Procambarus clarkii)

123

Discussion

The current distribution of P. clarkii in the NW of

Portugal is clearly the result of the interplay between

regional and local variables. In fact, there was a

significant improvement in predictive power by com-

bining the local variables with the regional model. The

combined model suggests that some invaded rivers

apparently maintain large stretches that are not likely

to be invaded. The same applies to the non-invaded

area where it seems that the risk of invasion is lower

than predicted when considering only the regional

scale. Most agreements depicted in Fig. 3 were

restricted to the primary rivers (in predicting presence)

and to small rivers at high altitudes (in predicting

absence). On the other hand, disagreement between

models was mostly associated with medium sized

rivers at intermediate to low altitudes. Bioclimatic

models usually predict the potential distribution of a

species at coarse to medium scales but cannot

incorporate local variability that plays a major role

in the distribution of a species in a given area (Loo

et al. 2009; Reino et al. 2013; Wearne et al. 2013).

Thus, the inclusion of local variables is expected to

refine such predictions at finer scales, a critical step

when assessing future routes of invasion, identifying

areas where to concentrate conservation efforts and

support management decisions (Fig. 4).

The local variables that help predict the distribution

of crayfish in NW Portugal were generally similar with

Fig. 3 Species distribution

model in the study area.

Grey areas under

environmental protection—

Natura 2000. Dashed line

represents the edge of

invasion (see ‘‘Methods’’ for

details). Black lines are

segments of river predicted

as presence and the grey

ones predicted as absences.

Black and white dots

indicate presence and

absence sites, respectively,

while the bigger white dots

indicate the absences that

were used for modelling

F. D. Moreira et al.

123

variables included in other published models (Gil-

Sanchez and Alba-Tercedor 2002; Cruz and Rebelo

2007; Banha and Anastacio 2011). River width and

slope were chosen for their positive and negative

relationship, respectively, on the potential distribution

of the crayfish. This result was expected since crayfish

preference for wide and slow streams is well docu-

mented (e.g. Gil-Sanchez and Alba-Tercedor 2002;

Cruz and Rebelo 2007). It should be noted that not all

the variability range of these and other variables was

used for calibrating models, because some the sites

that were not exposed to the crayfish invasion could

not be considered as true absences. This may have

cause some problems when extrapolating the model

for the whole study area. However, we think the range

of conditions that was omitted in model calibration is

the most unsuitable for the species (e.g. very steep

sites) and therefore will have little implications on

model transferability.

In contrast with the results of Siesa et al. (2011), we

found no significant association between crayfish

presence and human dominated areas. This may result

from a lower degree of human impact in NW Portugal

than in Lombardy in Italy, as well as from the type of

aquatic habitat considered which is mainly lentic in

Italy and entirely lotic in this study.

Mapping the predictions of two models, one based

on width and slope (local variables) and another based

on macroecological factors (regional model) revealed

a similar prediction for 62 % of the river stretches

Fig. 4 Comparison

between the species

distribution models of

Capinha and Anastacio

(2011) and of the present

study. Black both the

regional and local models

show high suitability index;

Green both the regional and

local models show low

suitability; Blue the regional

model indicates a high

suitability and the local

model a low suitability; Red

the regional model indicates

a high suitability and the

local model a low

suitability. Grey areas are

under environmental

protection—Natura 2000.

Dashed line represents the

edge of invasion

The red-swamp crayfish (Procambarus clarkii)

123

considered, while the remaining ones seemed to

almost exclusively show a high suitability according

to the regional model and a low suitability according

to the local model. This is the result of a weak

prediction ability of the regional model for medium

size tributaries. Since temperature was the most

important variable for the regional model (Capinha

and Anastacio 2011), we assume that these habitats

may have an adequate temperature for crayfish, but

nevertheless are steep (and water flow will probably be

strong), which makes them unsuitable for this species.

On the contrary, stretches for which the regional

model indicated a low suitability and the local model a

high suitability were rare as high altitude plateaus with

large and less steep rivers stretches are very uncom-

mon. This too might be explained by the fact that low

temperature habitats will mostly be located at high

altitudes, where rivers will be both narrower and

steeper.

Since the variables chosen in both models are

related to altitude, one could expect crayfish to be

present at low altitudes and slowly disappear as

altitude increases. However, this may not always be

the case as the current edges of crayfish distribution in

the most northern basins of the study area are located

at an altitude of approximately 25 m a.s.l. albeit this

species is known to occur up to varying altitudes such

as 190 m a.s.l in the SW Portugal (Cruz and Rebelo

2007), 572 m a.s.l. in the NE Portugal (Bernardo et al.

2011) and 750 m a.s.l. in southern Spain (Maceda-

Veiga et al. 2013). The restriction of crayfishes to low

altitude sites in the northern edge of the study area may

be due to insufficient propagule pressure due to

limitations to natural dispersal. However, the river

located just north of the area where we tested sampling

efficiency (river Estoraos) had a nearby (3.3 km) high

density site for more than 10 years, and yet the

crayfish altitudinal limit in this river is still very low

(51 m a.s.l.). Individuals of this species are known to

disperse upstream up to 1.74 km/year in other edge

areas (Bernardo et al. 2011); the fact that they were not

able to disperse over 3.3 km during a 10 years period

suggests that low altitude regions of NW Portugal are

less suitable to this species. This may be due to a

stronger stream flow, lower overall temperature or

even to other unassessed variables such as the stream’s

chemical properties (e.g. calcium concentration) at

each site (Cairns and Yan 2009; Favaro et al. 2010).

One of the assumptions for building a presence/

absence model is that the species is at equilibrium and

violating this assumption may have severe conse-

quences on the results (Guisan and Zimmermann

2000; Gallien et al. 2012; Vaclavik and Meentemeyer

2012). This is especially important for invasive

species during the invasion process since these are

the ones that violate this assumption more severely

while being also the ones for which predictions of

potential distributions are most desirable. An impor-

tant assumption of modelling approaches based on

presence/absence data is that locations where the

species is absent are truly unsuitable. To ascertain this,

all the sites in the three northernmost rivers that had no

crayfish in the 2010 survey were sampled again in

2012 and were found to be still devoid of crayfish.

However, overlaying the current distribution of cray-

fish with the results of any of the models, it is clear that

there should be suitable habitat for the species in the

non-invaded area and lack of crayfish in some places

may be just the result of a slow expansion. Another

hypothesis is that environmental variables other than

the ones accounted for in these models are preventing

the species from expanding its distribution. A possible

effect of physical barriers is not plausible, considering

that no physical barriers were found at the current edge

of invasion in any of the rivers. We therefore assume

that the sites we considered as true absences still are or

have been unsuitable for crayfish in the last

5–10 years. However, climate change may consider-

ably change this picture, particularly if precipitation/

river flow are affected (Capinha et al. 2013).

Conservation implications

Approximately 30 % of NW Portugal is under some

sort of environmental protection. Therefore progres-

sion of crayfish in this region may increase the impacts

on several endemic species and communities. Most

feared are those on native amphibians (e.g. Cruz and

Rebelo 2005) and fishes (e.g. Correia 2003) but also

important are those on macroinvertebrates (e.g. Geiger

et al. 2005; Correia and Anastacio 2008) and,

indirectly, on macroinvertebrate-feeding species like

the threatened Pyrenean desman (Galemys pyrenai-

cus). Crayfish may also have a direct influence on

riverbed vegetation by damaging soil structure and

increasing water turbidity through its burrowing

F. D. Moreira et al.

123

behavior (Correia and Ferreira 1995; Angeler et al.

2001). Last but not least, its recognised impacts on the

higher trophic levels that will influence the whole

ecosystem. Particularly important is the role it may

play in supporting populations of native species (such

as the European otter Lutra lutra); it may also be a

facilitator of the expansion of another new invader in

NW Portugal, the American mink (Neovison vison)

(Rodrigues et al. 2014), that in turn exerts predation

over the local biodiversity, strengthening the negative

impacts of the crayfish.

Due to crayfish biological and ecological traits and

continuing human mediated dispersion, most of the

Iberian Peninsula is seen as a lost cause in terms of

preventing its expansion. However, this work suggests

that crayfish might not colonize as many habitats as

one would expect.

The current abundance levels and the extent to

which the species has spread indicate that an eradica-

tion program is no longer feasible. Instead, a manage-

ment program, coupled with regular monitoring

efforts, is urgent and critical to understand and prevent

its further progression. New insights on the needs of

this species at a local scale are essential as these are

easier to act on through management programs. The

use of physical barriers on rivers (Kerby et al. 2005;

Rosewarne et al. 2013) to control the numbers of

crayfish that progress upstream is a management

option that has already shown promising results in

another region of the Iberian Peninsula (Dana et al.

2011). These barriers would function as a high slopes

that would increase water flow. However, as a barrier

will impact multiple other species, its use might be

controversial and will impose mitigation measures,

meaning that it cannot be applied at a widespread scale

and needs a detailed and scientifically sound planning.

Today P. clarkii is the most widely introduced

crayfish in the world. While it is an abundant and

widespread species in southwestern Europe, it is

currently spreading in central and even northern

Europe, as well as in other hilly or mountainous

temperate areas of the world (Gherardi 2006). In these

regions the physiographic and climatic conditions

encountered by this invasive are relatively similar to

those of NW Portugal and the constraints on its

dispersal identified in this work may be used to help to

contain its invasion.

Also important is the recognition that the high

invasion rate of the crayfish is a human-mediated

process, already referred by Capinha et al. (2013) at

the European level, but particularly evident in our

study area (similar years of arrival in distant basins).

This evidence reinforces the urgent need of informing

local stakeholders of the consequences of the species

expansion to new areas, developing local solutions and

promoting social awareness of this environmental

problem at the national level.

Acknowledgments This study was financed by FCT—

Fundacao para a Ciencia e Tecnologia (Portugal) through

project DILEMA—Alien species and conservation dilemmas:

the effects of native competitors and alien prey species on the

spread of American mink in Portugal (PTDC/BIA-BECD/

102433/2008). The authors would like to thank the staff at

‘‘Paisagem Protegida da Lagoa de Bertiandos e S. Pedro

d’Arcos’’ for logistic support and everyone involved in the

fieldwork. The authors would also like to thank G.F. Ficetola

and an anonymous reviewer for their overall contribution and

consequent improvement of the manuscript.

References

Allan JD (2004) Landscapes and riverscapes: the influence of

land use on stream ecosystems. Annu Rev Ecol Evol Syst

35:257–284

Allouche O, Tsoar A, Kadmon R (2006) Assessing the accuracy

of species distribution models: prevalence, kappa and the

true skill statistic (TSS). J Appl Ecol 43:1223–1232

Angeler DG, Sanchez-Carillo S, Garca G, Alvarez-Cobelas M

(2001) The influence of Procambarus clarkii (Decapoda:

Cambaridae) on water quality and sediment characteristics

in Spanish floodplain wetland. Hydrobiologia 464:89–98

Banha F, Anastacio PM (2011) Interactions between invasive

crayfish and native river shrimp. Knowl Manag Aquat

Ecosyst 401(17):1–12

Barton K (2009) Bivand: multi-model inference. R package,

version 1.7.2. http://r-forge.r-project.org/projects/mumin/

Bernardo JM, Costa AM, Bruxelas S, Teixeira A (2011) Dis-

persal and coexistence of two non-native crayfish species

(Pacifastacus leniusculus and Procambaru clarkii) in NE

Portugal over a 10-year period. Knowl Manag Aquat

Ecosyst 401:28

Bivand R (2007) spdep: Spatial dependence: weighting

schemes, statistics and models. R package version 0.5-43.

http://r-forge.r-project.org/projects/spdep/

Burnham KP, Anderson DR (2004) Multimodel inference:

understanding AIC and BIC in model selection. Sociol

Methods Res 33(2):261–304

Cairns A, Yan N (2009) A review of the influence of low

ambient calcium concentrations on freshwater daphniids,

gammarids, and crayfish. Environ Rev 17:67–79

Capinha C, Anastacio P (2011) Assessing the environmental

requirements of invaders using ensembles of distribution

models. Divers Distrib 17:13–24

Capinha C, Brotons L, Anastacio P (2013) Geographical vari-

ability in propagule pressure and climatic suitability

The red-swamp crayfish (Procambarus clarkii)

123

explain the European distribution of two highly invasive

crayfish. J Biogeogr 40:548–558

Carreira BM, Dias MP, Rebelo R (2013) How consumption and

fragmentation of macrophytes by the invasive crayfish

Procambarus clarkii shape the macrophyte communities of

temporary ponds. Hydrobiologia 721:89–98

Correia AM (2001) Seasonal and interspecific evaluation of

predation by mammals and birds on the introduced red

swamp crayfish Procambarus clarkii (Crustacea, Cam-

baridae) in a freshwater marsh (Portugal). J Zool

255(4):533–541

Correia AM (2003) Food choice by the introduced crayfish

Procambarus clarkii. Ann Zool Fenn 40:517–528

Correia AM, Anastacio PM (2008) Shifts in aquatic macroin-

vertebrate biodiversity associated with the presence and

size of an alien crayfish. Ecol Res 23:729–734

Correia AM, Ferreira O (1995) Burrowing behavior of the

introduced red swamp crayfish Procambarus clarkii

(Decapoda: Cambaridae) in Portugal. J Crustacean Biol

15:248–257

Craney TA, Surles JG (2002) Model-dependent variance infla-

tion factor cutoff values. Qual Eng 14(3):391–403

Cruz MJ, Rebelo R (2007) Colonization of freshwater habitats

by an introduced crayfish, Procambarus clarkii, in South-

west Iberian Peninsula. Hydrobiologia 575:191–201

Cruz MJ, Rebelo R, Crespo EG (2006) Effects of an introduced

crayfish, Procambarus clarkii, on the distribution of south-

western Iberian amphibians in their breeding habitats.

Ecography 29:329–338

Cruz MJ, Segurado P, Sousa M, Rebelo R (2008) Collapse of the

amphibian community of the Paul do Boquilobo Natural

Reserve (central Portugal) after the arrival of the exotic

American crayfish Procambarus clarkii. Herpetol J

18:197–204

Dana ED, Garcia-de-Lomas J, Gonzalez R, Ortega F (2011)

Effectiveness of dam construction to contain the invasive

crayfish Procambarus clarkii in a Mediterranean mountain

stream. Ecol Eng 37:1607–1613

Favaro L, Tirelli T, Pessani D (2010) The role of water chem-

istry in the distribution of Austropotamobius pallipes

(Crustacea Decapoda Astacidae) in Piedmont (Italy). C R

Biol 333:68–75

Ficetola GF, Thuiller W, Miaud C (2007) Prediction and vali-

dation of the potential global distribution of a problematic

alien invasive species—the American bullfrog. Divers

Distrib 13:476–485

Ficetola GF, Siesa ME, Manenti R, Bottoni L, De Bernardi F,

Padoa-Schioppa E (2011) Early assessment of the impact

of alien species: differential consequences of an invasive

crayfish on adult and larval amphibians. Divers Distrib

17:1141–1151

Ficetola GF, Siesa ME, De Bernardi F, Padoa-Schioppa E

(2012) Complex impact of an invasive crayfish on fresh-

water food webs. Biodivers Conserv 21:2641–2651

Gallien L, Douzet R, Pratte S, Zimmermann NE, Thuiller W

(2012) Invasive species distribution models—how violat-

ing the equilibrium assumption can create new insights.

Glob Ecol Biogeogr 21:1126–1136

Geiger W, Alcorlo P, Baltanas A, Montes C (2005) Impact of an

introduced crustacean on the trophic webs of Mediterra-

nean wetlands. Biol Invasions 7:49–73

Gherardi F (2006) Crayfish invading Europe: the case study of

Procambarus clarkii. Mar Freshw Behav Physiol

39:175–191

Gherardi F, Barbaresi S, Salvi G (2000) Spatial and temporal

patterns in the movement of Procambarus clarkii, an

invasive crayfish. Aquat Sci 62:179–193

Gil-Sanchez JM, Alba-Tercedor J (2002) Ecology of the native

and introduced crayfishes Austropotamobius pallipes and

Procambarus clarkii in southern Spain and implications for

conservation of the native species. Biol Conserv 105:75–80

Guisan A, Thuiller W (2005) Predicting species distribution:

offering more than simple habitat models. Ecol Lett

8:993–1009

Guisan A, Zimmermann NE (2000) Predictive habitat distri-

bution models in ecology. Ecol Model 135:147–186

Gutierrez-Yurrita PJ, Sancho G, Bravo MA, Baltanas A, Montes

C (1998) Diet of the red swamp crayfish Procambarus

clarkii in natural ecosystems of the Donana National Park

temporary freshwater marsh (Spain). J Crustacean Biol

18(1):120–127

Hines JE (2006) PRESENCE—Software to estimate patch

occupancy and related parameters. USGS-PWRC. http://

www.mbr-pwrc.usgs.gov/software/presence.html

Kerby JL, Riley SPD, Kats LB, Wilson P (2005) Barriers and

flow as limiting factors in the spread of an invasive crayfish

(Procambarus clarkii) in southern California streams. Biol

Conserv 126:402–409

Loo SE, Nally RM, O’Dowd DJ, Thomson JR, Lake PS (2009)

Multiple scale analysis of factors influencing the distribu-

tion of an invasive aquatic grass. Biol Invasions

11:1903–1912

Maceda-Veiga A, De Sostoa A, Sanchez-Espada S (2013)

Factors affecting the establishment of the invasive crayfish

Procambarus clarkii (Crustacea, Decapoda) in the Medi-

terranean rivers of the northeastern Iberian Peninsula.

Hydrobiologia 703:33–45

MacKenzie DI, Nichols JD, Lachman GB, Droege S, Royle JA,

Langtimm A (2002) Estimating site occupancy rates when

detection probabilities are less than one. Ecology

83(8):2248–2255

Moreira F (2011) Propagacao de um invasor: passado, presente e

futuro de P. clarkii no noroeste de Portugal. MSc Disser-

tation. University of Lisbon

Nunes AL, Cruz MJ, Tejedo M, Laurila A, Rebelo R (2010)

Nonlethal injury caused by an invasive alien predator and

its consequences for an anuran tadpole. Basic Appl Ecol

11(7):645–654

R Development Core Team (2012) R: A language and envi-

ronment for statistical computing. R Foundation for Sta-

tistical Computing, Vienna, Austria. ISBN 3-900051-07-0.

http://www.R-project.org/

Ramos MA, Pereira TM (1981) Um novo Astacidae para a fauna

portuguesa: Procambarus clarkii (Girard, 1852). Bol Inst

Nac Invest Pescas 6:37–47

Reino L, Beja P, Araujo MB, Dray S, Segurado P (2013) Does

local habitat fragmentation affect large-scale distributions?

The case of a specialist grassland bird. Divers Distrib

19(4):423–432

Richards SA (2005) Testing ecological theory using the Infor-

mation-Theoretic approach: examples and cautionary

results. Ecology 86(10):2805–2814

F. D. Moreira et al.

123

Richards SA, Whittingham MJ, Stephens PA (2011) Model

selection and model averaging in behavioural ecology: the

utility of the IT-AIC framework. Behav Ecol Sociobiol

65:77–89

Rodrigues D, Simoes L, Mullins J, Lampa S, Mendes RC,

Fernandes C, Rebelo R, Santos-Reis M (2014) Tracking the

expansion of the American mink (Neovison vison) range in

NW Portugal. Biol Invasions. doi:10.1007/s10530-014-

0706-1

Rosewarne P, Wright RM, Dunn AM (2013) Do low-head riv-

erine structures hinder the spread of invasive crayfish?

Case study of signal crayfish (Pacifastacus leniusculus)

movements at a flow gauging weir. Manag Biol Inv

4(4):273–282

Sala OE, Chapin FS III, Armesto JJ, Berlow E, Bloomfield J,

Dirzo R, Huber-Sanwald E, Huenneke LF, Jackson RB,

Kinzig A, Leemans R, Lodge DM, Mooney HA, Oester-

held M, Poff NLR, Sykes MT, Walker BH, Walker M, Wall

DH (2000) Global biodiversity scenarios for the year 2100.

Science 287:1770–1774

Siesa ME, Manenti R, Padoa-Schioppa E, De Bernardi F,

Ficetola GF (2011) Spatial autocorrelation and the analysis

of invasion processes from distribution data: a study with

the crayfish Procambarus clarkii. Biol Invasions

13:2147–2160

Vaclavik R, Meentemeyer RK (2012) Equilibrium or not?

Modelling potential distribution of invasive species in

different stages of invasion. Divers Distrib 18(1):73–83

Vie J, Hilton-Taylor C, Stuart SN (2009) Wildlife in a changing

world: An analysis of the 2008 IUCN red list of threatened

species 1-184. IUCN

Vogt J, Soille P, De Jager A, Rimaviciute E, Mehl W, Foisneau

S, Bodis K, Dusart J, Paracchini ML, Haastrup P (2007) A

pan-European River and Catchment Database. JRC Ref-

erence Reports

Wearne LJ, Ko D, Hannan-Jones M, Calvert M (2013) Potential

distribution and risk assessment of an invasive plant spe-

cies: a case study of Hymenachne amplexicaulis in Aus-

tralia. Hum Ecol Risk Assess Int J 19:53–79

The red-swamp crayfish (Procambarus clarkii)

123