Modelling the rise of invasive lionfish in the Mediterranean

doi:10.21203/rs.3.rs-4835003/v1

Download PDF

Research Article

Modelling the rise of invasive lionfish in the Mediterranean

https://doi.org/10.21203/rs.3.rs-4835003/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

The spread of the Indo-Pacific lionfish Pterois miles into the Atlantic Ocean represents a well-known example of a successful invasion. Lionfish have successfully invaded the Atlantic, aided by biological traits such as high thermal tolerance, adaptability to various salinities, high fecundity, conspicuous defenses, and generalist feeding habits. The Mediterranean Sea is now experiencing an early-stage lionfish invasion, spreading westward from the eastern Mediterranean, likely through the Suez Canal. Using ecological niche modelling under various climate scenarios, we predict the potential spread of lionfish, identifying regions from low to high habitat suitability. Predictions indicate significant expansion, especially in the eastern and central Mediterranean, under greater warming scenarios. This expansion could lead to substantial declines in native species, decimation of commercially important fish stocks, and trophic cascades, severely impacting local economies and marine biodiversity. Urgent understanding and management of lionfish impacts in the Mediterranean are essential, given their established presence in the western Atlantic and documented ecological consequences.

Biological invasion

Mediterranean Sea

Climate change

Species distribution model

Pterois miles

Invasive species pose a major threat to global biodiversity, with invasions rising at an unprecedented rate (Mainka and Howard, 2010; Simberloff et al. 2013; Bariche et al. 2017; Diagne et al., 2021; IPBES, 2023). A changing and variable climate combined with human induced factors, such as increasing global marine trade, fishing, tourism, and coastal development, facilitate the accelerated spread of invasive species (Rahel and Olden, 2008; Phillips and Kotrschal, 2021). These species often exhibit high dispersal tendencies, short generation times, generalist niche tolerance, and diverse diet preferences, enhancing their survival and colonization of new habitats (Catford et al., 2009; Poland et al. 2021). Species invasions represent a mounting threat to flora and fauna, causing a host of problems for native biodiversity across all trophic levels (Simberloff et al. 2013; Gámez-Virués et al. 2015; Savva et al. 2020).

Marine invasions of vertebrates from a high trophic level are rare, most aquatic invasive species occupy low trophic levels such as algae and sea grass, often distinctly influenced by competitive interactions (Côté et al. 2013). The competitive nature of invasive species imposes disruption of ecosystem structures, alteration of habitats and species interactions through the reduction or homogenization of biodiversity, and often plays a part in native species extinction (Hitt et al., 2003; Matsuzaki et al., 2009; Frazer et al. 2012; Crystal-Ornelas et al. 2020; Leroy et al. 2023). In addition to ecological costs, invasive species also generate large economic expenses, estimated to be in the region of hundreds of billions of dollars globally, and include effects such as reduction in crop yields, damage to infrastructure, and loss of ecosystem services, as well as the added economic cost of their mitigation and management (Cuthbert et al. 2021; Diagne et al. 2021; Haubrock et al. 2021; Fantle-Lepczyk et al. 2022).

In the Mediterranean, warming sea surface temperatures combined with an increase in shipping trade are believed to be some of the main causes of the arrival and spread of aquatic alien species (i.e. non-native with potential to become invasive; Soto et al. 2024) (Marras et al. 2015). The main route of origin for these alien species in the Mediterranean is the trade route via the Suez Canal from the Red Sea into the Mediterranean. The movement of species between these seas in the Red Sea to Mediterranean direction is known as Lessepsian migration, with the most recent and particularly problematic of these migrant species being the lionfish (Por, 2012; Dimitriou et al. 2019). The lionfish, particularly the Pterois volitans and Pterois miles subspecies are some of the world’s most successful marine invaders, demonstrating superior behavioural and life history traits that enable rapid proliferation and substantial population establishment (Côté, 2013; Côté and Smith, 2018). The species is native to the Indo-Pacific Ocean but has already staged one of the most impressive and well documented aquatic invasions on record, colonizing a huge swath of the western Atlantic coastline, including the Caribbean (Côté and Smith, 2018; Dimitriou et al. 2019). There is now increasing cause for concern based on the impacts of the Atlantic invasion over the newer invasion within the Mediterranean, where lionfish appear to be within the early stages of population establishment and settlement (Savva et al. 2020), observed for the first time here in 2012 (Poursanidis, 2020). The literature indicates a rapid westward geographic expansion in the decade since 2012, sparking worries about the ecological impacts the species could have on the Mediterranean given the well-studied harmful effects caused by lionfish invading the Atlantic (Morris, 2009té, 2013; Anton et al. 2014; Phillips and Kotrschal, 2021).

Recently, there has been a growing number of studies forecasting the risk of invasive fish as a result of a substantial increase in suitable habitats in the Mediterranean for warm water adapted fish shown through the modelling of current and future conditions (Marras et al. 2015; Coro et al. 2018). Species distribution models (SDMs) are becoming increasingly popular as an influential tool in estimating the actual or current spatial distribution of suitable habitats for marine invasive species (Robinson et al. 2017). On the other hand, ecological niche models (ENMs) characterize a species’ fundamental niche, a set of environmental variables that create positive population increase, and are able to predict invasion risk (Peterson, 2003; Anderson et al. 2011; Jiménez-Valverde et al. 2011). ENMs are applied when the aim is to estimate potential distribution, rather than the actual distribution like SDMs, but the terms SDMs and ENMs are often used interchangeably within the literature (Ahmad et al. 2020). ENMs are particularly useful for studying invasive potential of alien species and their interaction with future climate warming scenarios (Melo-Merino, 2020).

The aim of this study is to assess the current spatial distribution of the lionfish P. Miles invasion into the Mediterranean using SDMs and its projected future invasion potential using ENMs. The future predictions depend on two climate warming scenarios: relative concentration path (RCP) 4.5, modelling a moderate scenario of greenhouse gas emissions, and RCP 8.5, using the worst-case scenario of greenhouse gas emissions (Romero-Alvarez et al. 2017; Costa et al. 2023). The estimates using both RCP are used to predict occurrence for a mid-term (2040–2050) and a long-term (2090–2100) time scale.

2.1. Study area

The Mediterranean is an intercontinental sea almost entirely enclosed by land, found at the crossroads of three continents: Europe, Africa, and Asia. The Mediterranean Sea covers approximately 2.5 million km², with an average depth of 1500 m and a deepest recorded point of 5267 m in the Ionian Sea (Danovaro et al. 2010). The Atlantic Ocean is connected to the Mediterranean in the very far west by the strait of Gibraltar, and the Suez Canal in the southeast connects the Mediterranean to the red sea and Indian ocean (Coll et al. 2010). As an incredibly diverse marine ecosystem, estimated to host between 4 and 18% of global marine biodiversity, the Mediterranean is one of the 25 hotspots for biodiversity in the world despite its small dimensions representing less than 0.8% of the world’s ocean area (Bianchi and Morri, 2000; Cuttelod et al. 2009; Coll et al. 2010). Additionally, this sea hosts a uniquely high percentage of endemic species, thought to be almost 30% (Myers et al. 2000).

2.2. Species occurrence and environmental data

Occurrence records for lionfish were derived from the Global Biodiversity Information Facility (GBIF; http://www.gbif.org). The presence data contained a total of 1,964 results for all locations including native and non-native ranges. Records without location data (latitudes and longitudes), coordinate uncertainty over 10,000m, and other dubious records such as land records, duplicates, and outliers were removed. An uncertainty threshold of 10,000m was chosen since this is above the resolution of the environmental variable layers that were chosen for the model (Loya-Cancino et al. 2023). The presences were clipped according to the extent of the Mediterranean Sea using ArcGIS Pro, with a total of 162 records included within the models.

Environmental marine layers that might explain lionfish distribution were obtained from BioOracle (www.bio-oracle.org; Tyberghein et al. 2012; Assis et al. 2017), providing a range of climatic, biotic, and abiotic data. Each marine environmental predictor was downloaded in ASCII format at a resolution of 5 arcmin, which works out as approximately 9.2 km at the equator (Turan, 2020). Each environmental variable was selected for its relevance to lionfish distribution, resulting in 28 candidate variables (Table S1). Variables with high collinearity (correlation coefficients > 0.7) were excluded to avoid redundancy and improve model accuracy (Melo-Merino et al. 2020; Raghavan et al. 2019).

2.3. Modelling development

In a review of ENMs and SDMs in marine environments by Melo-Merino (2020), machine learning is the most popular method of modelling species niches or distributions. The Maximum Entropy (MaxEnt) method has robust accuracy even with low sample sizes and is designed to work with presence only data (Phillips et al. 2006). Statistical methods were second against machine learning, including Generalized Linear models (GLMs) and Generalized Additive models (GAMs) (Melo-Merino, 2020). Using more than one modelling approach is thought to improve predictive performance (Marmion et al. 2009); however, Zhu and Peterson (2017) found that ensemble modelling provides no better predictive performance than individual models, and predictive performance can often be situation dependent (Hao et al. 2020).

Here, three modelling methods were used to predict current lionfish distribution in the Mediterranean, each being compared for performance. The methods were Maximum Entropy (MaxEnt) developed with the MaxEnt software; generalised linear model (GLM) and Random Forest (RF), both built within R version 4.3.0 (R Core team, 2023). Each method used the same occurrence data and environmental raster layers when looking at the same predicted scenario. All were tested for correlation and clipped to the extent of a Mediterranean Sea within ArcGIS Pro. Five models per modelling technique for a total of 15 were developed, to predict current suitable habitat (3 models) and suitable habitat for 2040–2050 relative concentration pathways (RCP; See section 2.3.1) 4.5 (3 models), 2090–2100 RCP 4.5 (3 models), 2040–2050 RCP 8.5 (3 models), and 2090–2100 RCP 8.5 (3 models).

As MaxEnt is a presence-only species distribution modelling technique, only the presence data downloaded from GBIF was needed to produce a model run. Often, marine species absence data is sparsely available and usually both economically and logistically costly to obtain, so the use of a presence-only method is becoming increasingly popular to aid decision making in conservation planning (Bombosch et al. 2014; Gomes et al. 2018; Smith et al. 2021). However, as MaxEnt provides estimates of relative suitability and not absolute occupancy probabilities, robust presence-absence model techniques are needed to validate the model outputs (Guillera-Arroita et al. 2014). Efforts need to be taken to consider sampling biases related to sampling bias and a higher probability of commission errors leading to false positive results (Bird et al. 2014; Guillera-Arroita et al. 2015). Here, a GLM and a RF algorithm were applied to provide cross validation for the results from the MaxEnt models.

2.3.1. Maximum Entropy (MaxEnt)

Five variables (surface temperature °C, surface salinity range PSS, benthic salinity range PSS, surface productivity mean mg/m³ and surface chlorophyll mean mg/m³) alongside the lionfish presence records were used to predict current spatial distribution using MaxEnt. A randomly selected 25% of the species occurrences were used to test the model and the remaining 75% to train the model (Loya-Cancino et al. 2023).

RCPs specify a set of four climate scenarios useful for modelling future climate evolution and projecting influences on future economic, environmental, and social factors (Van Vuuren et al. 2011). These scenarios include changes in greenhouse gas emissions, radiative forcings, air pollution estimates, and land use changes. The four scenarios range from RCP 2.6 (least warming) to RCP 8.5 (most warming). In this study, we used RCP 4.5 (1.8°C warming; D’Amen and Azzuro 2020) and RCP 8.5 (3.7°C warming; Jubb et al. 2013) to predict the future potential suitability for lionfish populations for (i) 2040 to 2050 and (ii) 2090 to 2100.

A total of 8 environmental variables, consisting of temperature and salinity measurements for future expansion of invasive lionfish, were downloaded from BioOracle (Table S2). The available data only included surface level predictions of temperature and salinity, so there are no variables of benthic level data or measures of nutrient and productivity to include in the future projections. However, temperature and salinity are often linked to being the most crucial to marine species in alteration of physiological processes, and are essential to building ENMs (D’Amen and Azzuro, 2020). After running a correlation analysis, two variables for each of the four future scenarios were uncorrelated (Table 1).

Table 1

Variables included in future predictions according to RCP and associated time period
RCP	Time period	Uncorrelated variables
4.5	2040–2050	Surface minimum salinity
4.5	2040–2050	Surface mean temperature
4.5	2090–2100	Surface range salinity
4.5	2090–2100	Surface range temperature
8.5	2040–2050	Surface range salinity
8.5	2040–2050	Surface range temperature
8.5	2090–2100	Surface range salinity
8.5	2090–2100	Surface range temperature

2.3.2. Generalised linear models (GLMs)

GLMs are a commonly used and regression-based method that use maximum likelihood to estimate species distributions (Evangelista et al. 2016). The GLM method was selected to fit a model that involves both presence and absence data to provide validation to the MaxEnt method. Since the data collected from GBIF includes exclusively presences, pseudo absence data in the form of generated background points were incorporated using R version 4.3.0 (R Core team, 2023) and the package ‘dismo’ (Barbet-Massin, Albert, and Thuiller, 2012, Hijmans and Elith, 2017). The environmental predictor variables, presence data, and study site were kept the same across the two time-periods.

2.3.3. Random Forest (RF)

RF is a machine learning technique that predicts distributions with minimal tuning, small sample sizes, and has been increasingly used to model marine or aquatic ecosystems (Luan et al. 2020); however, there are instances where distributions were fitted poorly (Valavi et al. 2021). Poor predictive performances in RF are thought to be linked with the use of presence only data and a large number of background samples, but this is not a general problem as other methods including MaxEnt perform well using these data (Elith et al. 2006). RF provides a second validation to MaxEnt and GLMs, and also required background of pseudo absence data. The environmental predictor variables, presence data, and study site were kept the same.

2.4. Assessing predictions

The performance of each model, MaxEnt, GLM, and RF, was measured in terms of the area under the receiver operating characteristic curve (AUC) which was used to assess the quality of the model prediction (ranging from 0 to 1), where the higher the AUC the better the model. Models with an AUC of greater than 0.9 are highly accurate, models with AUC of 0.7–0.9 moderately accurate, and those with less than 0.7 are considered to have a poor accuracy (Bučas et al. 2013). AUC is a very common method to assess model performance and despite being criticised when used with background points, its reliability is robust when comparing models of the same species, same environmental variables, and same study area (Foucade et al. 2014). Additionally, response curves generated by the MaxEnt software are used to show the changing probability of presence depending on each of the environmental variables chosen.

3.1. Predicted suitability under current climatic conditions

The models provide spatially specific evaluations of the likelihood of lionfish to invade particular areas across the Mediterranean, quantified in metres and a gradient of invasion likelihood. As a result of a reasonably large physiological tolerance, the predicted suitability under current climate conditions of the lionfish modelled showed potential for wide geological coverage of the Mediterranean Sea. Although the locations and sizes of the likelihood ranges vary from model to model, aspects of each model tend to agree with each other.

MaxEnt and RF models show similar patterns, whereas the GLM model predicts broader high-risk zones. The Aegean Sea shows large areas of high likelihood of lionfish in all three models, with MaxEnt and RF being more agreeable with each other than the GLM model. Similarly, areas off the coast of southern and eastern Spain and southern France indicate suitable habitat for lionfish under current climate scenarios. Where the Mediterranean narrows between the southernmost part of Italy and northern part of Tunisia, the models indicate high likelihood of distribution, potentially due to warmer and shallower conditions (Fig. 2). The spatial predictions from MaxEnt and RF have generally comparable results, whereas the GLM for the current scenario has a much more pessimistic prediction. The GLM indicates large and connected areas along the south of the Mediterranean, with a larger range of high likelihood (i.e. range between 0.8 to 1), with 5,737 m suitability compared to only 760 m for MaxEnt and 3,406 m for RF (Table 2).

Unsuitable locations for the lionfish species are indicated by all three models most often in the middle sections of the Mediterranean, suggesting that depth along with other predictors associated with open water create unsuitable niches for this species in these areas (Fig. 2). MaxEnt and RF show higher estimations of 0–0.2 likelihood, with 21,424 m and 16,522 m respectively. Whereas GLM only predicts 9,969 m within this lowest likelihood, significantly less than the other two (Table 2). The models of current climate conditions show a more detailed prediction output than the future climate scenario models, potentially lending from the higher number of environmental variables or predictors used in these models.

Table 2

Likelihood range (m) for the current projected distribution across the three models.
Likelihood range	Current (m)
	Max Ent	GLM	RF
0–0.2	21424	9969	16522
0.2–0.4	2530	4578	3202
0.4–0.6	1661	2667	2543
0.6–0.8	1404	4827	2106
0.8–1	760	5737	3406

3.2. Predicted suitability under future climate scenarios

Future projections of lionfish distribution show increases in suitable areas in both time-periods (i.e. 2040 to 2050 and 2090 to 2100). Overall, a general increase in suitable conditions spreads outwards from mainland Greece across both RCP scenarios (i.e. RCP 4.5 and RCP 8.5). According to the models, the lionfish is expected to extend across the entire Aegean and connect across open waters to increasingly suitable waters along the coast of Tunisia and Algeria. The northernmost sections of the Adriatic Sea show low to no potential to be occupied (Fig. 3, 4).

For 2040 to 2050, RF shows a more detailed prediction of various likelihoods across both RCP scenarios, particularly when compared to the outputs from the GLM. The GLM’s outputs are more homogenous, showing large and connected areas of each suitability level without much variation in likelihood across huge areas of the Mediterranean (Fig. 3). Similarly, the GLM’s range estimates indicate low areas of 0–0.2 likelihood, showing 110 m to 548 m, which is often significantly lower than the ranges estimated by MaxEnt and RF (Table 3).

For 2090 to 2100 RCP 8.5, there are large changes to suitability for lionfish, with lower likelihoods in the range of 0–0.2 sparse, and an increase in likelihoods of 0.4 and above. RF has the lowest area of 0–0.2 likelihood of all models across the two RCP climate scenarios, with only 7 m range. Furthermore, RF is also having the lowest 0.8–1 range, with only 7 m range (Fig. 4; Table 3). Since the warming effects become intensified under RCP 8.5, increases in 0.8–1 range is expected as the sea becomes warmer and potentially more suitable for the lionfish, which exist within a wide range of surface temperatures.

Table 3

Likelihood range (m) for the future climatic projection per RCP scenario (4.5 and 8.5) and time-period (2040 to 2050 and 2090 to 2100) across the three models
Likelihood range	2040 to 2050 (m)						2090 to 2100 (m)
Likelihood range	RCP 4.5			RPC 8.5			RCP 4.5			RCP 8.5
	Max Ent	GLM	RF	Max Ent	GLM	RF	Max Ent	GLM	RF	Max Ent	GLM	RF
0–0.2	11504	548	16328	3333	274	13744	4040	110	11470	2604	284	7
0.2–0.4	7545	2624	3189	11165	970	1871	14524	457	4131	5083	408	5418
0.4–0.6	4709	8824	2242	7915	12585	1826	5043	8786	4231	5717	9044	21625
0.6–0.8	2207	11440	2039	3404	10784	2659	1888	16713	3588	14482	15633	829
0.8–1	1921	4450	4088	2069	3273	7786	2391	1820	4466	N/A	2517	7

3.3. Assessing predictions

Each of the single species distribution models were cross validated using evaluation scores of AUC (Table 4). The predictive performance of the models is generally good, with all the models showing an AUC of at least 0.6. RF performs best in every climate scenario, with an AUC from 0.82–0.98. This is followed by MaxEnt, with AUC ranging from 0.66 to 0.94. Lastly, GLM shows the lowest AUC in every scenario, ranging from 0.63 to 0.88 (Table 4).

Under the current climate conditions, the models show moderately accurate predictions (AUC score of at least 0.8 for all three modelling methods). RF and MaxEnt provides excellent predictive performance, with an AUC greater than 0.9. GLM shows 0.88, with this being the only projection where the GLM’s AUC exceeded 0.7 (threshold for a well performing model), suggesting poor predictive performance overall for the GLMs here (Table 4). The results for the 2040 to 2050 scenarios of RCP 4.5 and 8.5 show excellent predictive performance when using RF, moderate performance for MaxEnt, and poor performance when using GLM (Table 4). However, the predictive performance of the models shows sequentially declining detail as predictions increase in time-period. For the climate scenario 2090 to 2100 RCP 8.5, both MaxEnt and RF show the lowest AUC values of the study, with the GLM very close to these low values (Table 4).

Table 4

The area under the curve (AUC) evaluation metric per RCP scenario (4.5 and 8.5) and time-period (2040 to 2050 and 2090 to 2100) across the three models
	Current			2040 to 2050						2090 to 2100
	Current			RCP 4.5			RPC 8.5			RCP 4.5			RCP 8.5
	Max Ent	GLM	RF	Max Ent	GLM	RF	Max Ent	GLM	RF	Max Ent	GLM	RF	Max Ent	GLM	RF
AUC	0.94	0.88	0.98	0.88	0.63	0.93	0.81	0.68	0.91	0.82	0.64	0.90	0.66	0.64	0.82

Rising ocean temperatures as a result of climate change are expected to enable the range expansion of non-native species, with the potential of having knock on effects on fragile ecosystems and becoming nuisance invaders (Poloczanska et al. 2016). Our predicted current spatial distribution of the lionfish P. miles raised the key question on why we have not started monitoring closely the coasts around the Mediterranean Sea. There are high probabilities (> 0.8 likelihood) of having this invasive species already present across all coasts, except for Libya and north coast of Egypt (based on Random Forest with highest AUC). This should pitch the urgency of dealing with a potential problematic based on the negative effects reported in the Atlantic (Côté and Smith, 2018; Dimitriou et al. 2019). Our future predicted scenarios indicate that the lionfish is expected to spread by the end of the century from the eastern Mediterranean towards the western coastlines and central areas of the sea. A worrying RCP 8.5 scenario with 3.7°C warming shows nearly the whole of the Mediterranean Sea as suitable for population establishment (predominantly likelihood of 0.4 to 0.6 based on Random Forest with highest AUC). The predictions call for a prompt management in the Mediterranean Sea, especially as we may still be in the mid-stage of the invasion (i.e. eradication unlikely, but control feasible; Ahmed et al. 2022) and could potentially still be able to prevent any further range expansion.

4.1. Prediction of current spatial distribution

The predictions indicate potential suitability for lionfish particularly in the southern Aegean Sea, off the eastern coast of Greece in the Ionian Sea, and in the stretch of ocean between southern Sicily and the northern coast of Tunisia. These estimates are consistent with georeferenced occurrences published by Dimitriadis et al (2020), recording well-established populations in the southern and central Aegean, the Greek Ionian Sea, and sightings between Sicily and Tunisia. Ulman et al (2020) report that lionfish sightings in the Turkish Aegean have been on the rise, with areas near to Datça reporting a first sighting in 2016 and subsequent group sightings of up to 50 lionfish by summer 2019.

Here, all three models agree that the strait of Sicily between Sicily and Tunisia currently has a high likelihood of lionfish invading, with Azzurro et al (2017) reporting a rising number of sightings of lionfish off the coast of Sicily and supporting the model predictions. Similarly, Turan (2020) indicates the most suitable areas for lionfish are currently found on the coast of northeast African countries and in central Mediterranean countries of Greece, Italy, and Malta, which agree with our results. This is further supported by lionfish caught by fishermen that have been reported in Tunisian waters in 2015 and in 2020 (Amor et al. 2022). However, unlike Turan et al. (2020), the coastlines of Libya and Egypt in our results appear to have low likelihood. This is consistent with Poursanidis et al. (2020), where even if temperature and salinity are optimal, the lack of a least cost path of dispersal by ocean currents to these coastlines might be a contributing factor to this. Contrary to the low likelihood predictions however, two sightings of lionfish have been reported by recreational fishermen in December 2018 in the coastal waters of Libya (Al Mabruk and Rizgalla, 2019) and another two off the Mediterranean coast of Egypt in August 2018 (Al Mabruk et al. 2020). These sightings imply that the model may have underfit the niche predictions of the species.

The south coast of France is likely to host lionfish currently, despite water temperatures thought to be outside of the thermal tolerance of lionfish of below 10°C (Kimball et al. 2004). As of yet, lionfish have not been reported in French waters, but modelling by Loya-Cancino et al. (2023) agrees with our findings. D’Amen and Azzurro (2020) predicted under current climatic conditions that the western Mediterranean, the northern Adriatic, and the northern Aegean are likely to demonstrate low amounts of risk for lionfish invasion as a result of unfavourable climates, but forecasting these invasive fish might gain new suitable areas by 2050. This assessment agrees with our results, where each of the areas of predicted absence mentioned by D’Amen and Azzurro (2020) align.

In contrast to the predictions found here, Johnston and Purkis (2014) produced estimations consistent with an unlikeliness of a complete invasion in the Mediterranean, forecasting low connectivity of larval dispersal when compared to the Atlantic lionfish invasion. These findings are contradictory to more recent studies, including Huseyinoglu et al (2021) which mentions that the increased frequency of which lionfish have been reportedly sighted signifies a strong geographical pattern of expansion concerning the western Mediterranean. However, when Poursanidis et al (2020) modelled the probability of lionfish occurrence, the results showed the Adriatic and Ionian Seas to be unfavourable, as well as most of the western part of the Mediterranean such as the southern French coast. These results not only disagree with our study but are also not well matched to those studies mentioned earlier, with the exception of D’Amen and Azzurro (2020). This difference in results might have stemmed from the use of a different set of environmental variables in their analysis, such as Poursanidis et al (2020) being derived from MARSPEC rather than BioOracle, potentially causing a different and more conservative modelled scenario.

Poursanidis et al (2020) suggest that local adaptation and niche unfilling should be considered as the reason for an expansion towards the west of the Mediterranean, with further studies exploring niche dynamics of lionfish expansion (D’Amen and Azzurro, 2020). Phillips and Kotrschal (2021) suggested that lionfish could have evolved mutations that have given rise to physiologically better cold resistance within invasive populations, allowing individuals to reside closer to their supposed temperature barrier. Similar examples were studied in other species, for example, marine sticklebacks Gasterosteus aculeatus can evolve cold tolerance under climate pressures (Barret et al. 2011) and the Chinese minnow Rhynchocypris oxycephalus shows different levels of heat sensitivity depending on the geographical location of the population (Yu et al. 2018). Changes in physiology could be difficult for models to account for, leaving potential areas of suitability out of the risk map (Parravicini et al. 2015). However, as climate conditions continue to warm, lionfish range expansion will be anyway possible into areas previously outside of thermal tolerance without accounting for improved physiology.

4.2. Looking into the potential future distribution

The predictions for the 2040–2050 period indicate a more extensive spread of lionfish into large ranges of the south-eastern parts of the Mediterranean and the beginnings of spread in the western extents. Under the milder climate change scenario of RCP 4.5, the southern coasts of France and Spain appear to be at either very low or low risk of lionfish invasion. However, the predictions for the warmer RCP 8.5 indicate spots in these regions where lionfish are highly likely to become able to persist. This is further supported by other studies, which considered coasts of Gibraltar, Spain, Portugal and France suitable habitats for lionfish (Poursanidis et al. 2020; Loya-Cancino et al. 2023). Other areas of high likelihood are around the southern coast of Greece and the south coast of Turkey, in some cases showing large areas of connectivity between areas of very high likelihood.

Our findings for 2040–2050 also show low distribution patterns of likelihood around the coastal areas of northern Italy, especially in the most northern parts of the Adriatic Sea, which could be explained by a thermal tolerance limit (Johnston and Purkis, 2014). In scenarios of a warming sea surface temperature, especially RCP 8.5 it is likely that even tropical species could become closer to their upper physiological temperature limits, potentially forcing species to more northern, temperate latitudes (e.g. Buisson et al. 2008; Helaouet et al. 2013; Ruiz-Navarro et al. 2016). The scenarios projected for the end of the century (2090–2100), highlight strong north and east shifts in moderate to high likelihood ranges, particularly under the RCP 8.5 scenario, where the MaxEnt and GLM show agreement in large ranges of 0.6–0.8 likelihood. There is a high potential suitability for lionfish around Greece, Turkey, and the bottleneck at the strait of Sicily. Our findings also show most regions around France and Spain predicted to have moderate to high suitability for the invasive species, which goes in line with some studies (e.g. Schickele et al. 2021) but not so well with others (e.g. Dimitriadis et al. 2020). Similar happens with disagreements about the suitability of the east coast of Italy (Schickele et al. 2021). The use of a single (or different) environmental parameter, modelling approaches and data collection sources could explain these differences.

4.3. To model or not to model?

SDMs and ENMs can often fail to predict the expansion of a species range. Invasive species can violate the assumptions where the species is at range (environmental equilibrium) and occupy all regions suitable for presence, sometimes leading to current distribution not being reliably predicted (Briscoe Runquist et al. 2021). Niche shifts are common, inaccurate climatic tolerances or unanticipated evolutionary fluctuations in populations could lead to the future of underlying environmental relationships being different depending on the region of the species’ global range (Araujo and Rahbek, 2006). Furthermore, uneven species sampling can cause bias and is frequently known to hinder modelling accuracy (e.g. Bystriakova et al. 2012; Fourcade et al. 2014; Lodge et al. 2016).

The different modelling techniques used here returned clearly distinct geographic predictions, when using the same data, with the potential to result in different conservation strategies depending on which is chosen for decision making. Ensemble modelling, a method of averaging the prediction outputs of several different methods, could be used to reduce the uncertainty produced from the use of a single model (Araújo and New, 2007). Additionally, our study used only Mediterranean occurrence data (i.e. only data within the invasive range) to avoid poor fit from the use of native occurrences to predict distribution in non-native locations. This can happen due to a tendency of species to alter their climatic niche when arriving into new non-previously habited areas (Phillips et al. 2009).

The ecology of a species can affect the model itself (Hernandez et al. 2006; Syphard and Franklin, 2010). Species with a larger body size and/or a more striking body colour are likely to be easier to record, thus increasing potential data availability (McPherson and Jetz, 2007). Range size can also influence the availability of data, for example, if a species has a large range and exhibits local adaptation, spatial variability in the relationships between occurrence and environmental variables can cause an overemphasis of a species’ ecological niche (Stockwell and Peterson, 2002). Furthermore, potential invasive species might experience limited duration of residence and constraints on their dispersal (Hui and Richardson, 2017), with pockets of habitats (accessible and suitable) not occupied due to the short period of time the species has been present. Hui (2023) suggests species distribution models could be reconceptualised to become more dynamic and inclusive of the detailed ecological and biological processes specific to the invasive species in question. Where possible, these biological considerations should be accounted for in future models, especially given the potential for the projection range to be broader than the sampled space sometimes stemming from niche shifts under the influences of climate change (Mainali et al. 2015; Dominguez Almela et al. 2020; Botella et al. 2022).

Overall, modelling species distribution is a difficult task, as models are often an oversimplification of the problem in hands relying on assumptions that may not fully capture the dynamics of species-environment interactions (Dominguez Almela et al. 2022). For example, Dimitriadis et al (2020) considered temperature the only limiting factor for the expansion of lionfish, predicting relatively minimal areas that were unsuitable for lionfish. In contrast, Evangalista et al (2020) states that lionfish are unlikely to find suitable habitats outside of the eastern sector of the Mediterranean by 2050, but discusses that this might not be entirely correct without considering the niche dynamics of the species (Azzurro et al. 2017; Parravicini et al. 2015). With that being said, lionfish have been observed outside of these future suitability projections already, with records currently spreading westwards to the central Mediterranean (Azzurro et al. 2017).

In our future scenario models, just temperature and salinity were used as the environmental variables to assess the fundamental Grinnellian niche (Melo-Merino et al. 2020), not including key elements of nutrient concentrations, habitat types, and life history tolerances (Poloczanska et al. 2016). By simply using temperature and salinity data, the realism of the model findings may have been reduced since the effects of climate change on the interactions occurring between a species and its niche depend on food web interactions, phenology, competition for food and habitat, all of which are difficult to quantify but are important to consider in the study of a species’ distribution (Loya-Cancino et al. 2020). Without considering these ecological variables, it may not be possible to fully infer the effect that each of the current and future climate scenarios have or could have on the lionfish species, so careful interpretation of the results might be necessary. Where the habitat preferences of a species are difficult to summarise within predictor variables, models may misjudge the relationship between possible occurrence and the environment (McPherson and Jetz, 2007; Dominguez t al. 2022). Furthermore, invasive species are dispersed through water, so representing water movement (i.e. currents, flow) and potential physical barriers (e.g. locks, dams) to dispersal in the models would produce more realistic and reliable predictions (Melo-Marino et al. 2020; Dominguez Almela et al. 2020).

4.4. Implications for lionfish eradication, emergent uses and management

In regions such as the Mediterranean, where invasive species are considered one of the main threats to biodiversity, preventative measures guided by accurate spatial and temporal species distribution models can provide significant value for managing biological invasions (Srivastava et al. 2019). The foremost example of the impacts that invasive lionfish can cause have been seen in the western Atlantic and have been extensively modelled to reveal the highly destructive traits of these conspicuous fish and the resultant extensive damage to native marine communities and ecosystems (Schiekele et al. 2021). Although the current threat of invasive lionfish in the Mediterranean appears to be limited to the south-eastern areas, our model projections highlight a strong likelihood of this species spreading ubiquitously throughout the Mediterranean Sea by the end of the century. Steps towards eradication, emergent uses and management are vital to avoid similar destruction to that seen in ecosystems of the western Atlantic (Dominguez Almela et al. 2023) and achieve improved and fast adaptation to the risk of lionfish invasion in the Mediterranean Sea.

Raising awareness and executing monitoring efforts in the Mediterranean will be tremendously important during these early stages of colonisation (Morris and Whitfield, 2009; Gülenç, 2019). If lionfish are allowed to further establish populations, control can quickly become unfeasible (Bariche et al. 2013). To date, Cyprus appears to be the only country in the Mediterranean initiating policy change in favour of substantial lionfish removal efforts (Kleitou et al. 2019; Candelmo et al. 2022). Spearfishing permits have been granted on the northern coasts alongside the organisation of lionfish derbies meaning in the period of 2018–2021 over 35,000 lionfish were removed due to the program (Ulman et al. 2022). The capture and hunting of lionfish alone is unlikely to control invasive populations as these fish can endure hundreds of metres of depth, but combining capture with methods of predation, competition, and natural pathogens could create regional control (Albins and Hixon, 2008).

Imposing restrictions on the fishing of species capable of controlling lionfish, such as the Mediterranean dusky group, could be put in place to further help control the spread of lionfish, particularly in areas forecast in the present and the future to be suitable habitats for the invasive lionfish (Mumby et al. 2011). New techniques for the removal and fishery of lionfish are being explored, but some requires a high investment cost (e.g. Sutherland et al. 2017; Kleitou et al. 2022). However, these costs may be leverage by establishing a commercial market amongst the seafood and jewellery industries for the consumption of lionfish. This is a current management strategy in the Western Atlantic that can be promoted in the Mediterranean by advertising the edibility and taste of lionfish (Bilecenoğlu, 2018). By promoting the capture of this species for market consumption, targeting and removal actions will continue to be sustained, creating long term management of the species that is beneficial for the environment but also socioeconomic trade (Kleitou et al. 2022).

The spread of lionfish within the Mediterranean basin, although in early stage of an invasion, could be very difficult to realistically be prevented/stopped (Ahmed et al. 2022). However, some preventative measures could still limit that spread of the populations in the Mediterranean (Bilecenoğlu, 2018). One example would be installing a high salinity area, reinstating a salinity barrier, or adding locks in the Suez Canal areas. These could decrease the movement and larval dispersal of Lessepsian migrant species but may come at a high cost unless combined with other construction initiatives (Galil et al. 2017). Management efforts facilitating systematic observation at sentinel locations, such as ports where species are likely to be ejected from ballast water, are crucial to stay up to date with the status, impacts, and interactions of invading lionfish, Cyprus is an appropriate choice as it is close to the Suez Canal and is a hotspot for the settlement of invasive Lessepsian species before the invasion moves any further west (Kleitou et al. 2019). Given the obstacles associated with preventing and eradicating already established invasive species, flexible management efforts assisted by mapping of current species distributions and modelling both present and future scenarios for informed decisions (Dominguez Almela et al. 2021) are fundamental in addressing the threat of invasive lionfish, through appropriate public awareness schemes, informed targeted removal, and continuous monitoring (Giakoumi et al. 2019).

Growing global environmental change and increasing global trade heighten the speed and intensity in which marine biological invasions occur, accentuating the need to manage and monitor such species movements. Species distribution modelling is being more broadly applied across the sphere of biogeography, taking into account range dynamics, global changes, and various complex habitat types. These models are continuing to leverage renewing spatial data to provide up to date insights, especially through the combination of process models and parallel lines of evidence (Franklin, 2023). Progress where methodological limitations are concerned are continuously being addressed, such as an improved understanding and inclusion of niche theory in nonequilibrium species interactions, enhanced model selection and evaluation, and tackling of sampling bias and issues (Melo-Merino et al. 2020). New methods and ways to approach species modelling can be applied such as genetically informed species distribution modelling and joint distribution modelling where community dynamics are considered. Improvements to species modelling tools will benefit future work in considering marine invasive species and several other applications, through the methods for prevention, eradication, and management of invasive species mentioned above, whist hopefully facilitating steps towards securing the future of biodiversity around the world.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Competing Interests

The authors have no relevant financial or non-financial interests to disclose.

Author Contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Mitchell. The first draft of the manuscript was written by Mitchell and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript

Data availability

All data used are available in the main text.

Ahmad S, Yang L, Khan TU, Wanghe K, Li M, Luan X (2020) Using an ensemble modelling approach to predict the potential distribution of Himalayan gray goral (Naemorhedus goral bedfordi) in Pakistan. Glob Ecol Conserv 21:pe00845. https://doi.org/10.1016/j.gecco.2019.e00845
Ahmed DA, Hudgins EJ, Cuthbert RN et al (2022) Managing biological invasions: the cost of inaction. Biol Invasions 24:1927–1946. https://doi.org/10.1007/s10530-022-02755-0
Al Mabruk SA, Rizgalla J (2019) First record of lionfish (Scorpaenidae: Pterois) from Libyan waters. J Black Sea/Mediterranean Environ 25(1):108–114
Al Mabruk SA, Rizgalla J, Giovos I, Bariche M (2020) Social media reveals the first records of the invasive lionfish P miles (Bennett, 1828) and parrotfish Scarus ghobban Forsskål, 1775 from Egypt (Mediterranean Sea). BioInvasions Rec 9(3):574–579. https://doi.org/10.3391/bir.2020.9.3.13
Albins MA, Hixon MA (2008) Invasive Indo-Pacific lionfish Pterois volitans reduce recruitment of Atlantic coral-reef fishes. Mar Ecol Prog Ser 367:233–238. https://doi.org/10.3354/meps07620
Amor MMB, Ounifi BAK, Bdioui M, Hasni E (2022) Additional occurrence of common lionfish, Pterois miles in the Northern Tunisian Coast. In 2nd Mediterranean Symposium on the Non-Indigenous Species (p 89)
Anderson RP, Martínez-Meyer E, Nakamura M, Araújo MB, Peterson AT, Soberón J, Pearson RG (2011) Ecological niches and geographic distributions (MPB-49). Princeton University Press. https://doi.org/10.1515/9781400840670
Anton A, Simpson MS, Vu I (2014) Environmental and biotic correlates to lionfish invasion success in Bahamian coral reefs. PLoS ONE 9(9):pe106229. https://doi.org/10.1371/journal.pone.0106229
Araújo MB, New M (2007) Ensemble forecasting of species distributions. Trends Ecol Evol 22(1):42–47. https://doi.org/10.1016/j.tree.2006.09.010
Assis J, Tyberghein L, Bosch S, Verbruggen H, Serrão EA, De Clerck O (2017) Bio-ORACLE v2 0: Extending marine data layers for bioclimatic modelling. Glob Ecol Biogeogr 27(3):277–284. https://doi.org/10.1111/geb.12693
Azzurro E, Stancanelli B, Di Martino V, Bariche M (2017) Range expansion of the common lionfish Pterois miles (Bennett, 1828) in the Mediterranean Sea: an unwanted new guest for Italian waters. BioInvasions Rec 6(2):95–98. https://doi.org/10.3391/bir.2017.6.2.01
Bariche M, Kleitou P, Kalogirou S, Bernardi G (2017) Genetics reveal the identity and origin of the lionfish invasion in the Mediterranean Sea. Sci Rep 7(1):1–6. https://doi.org/10.1038/s41598-017-07326-1
Bianchi CN, Morri C (2000) Marine biodiversity of the Mediterranean Sea: situation, problems and prospects for future research. Mar Pollut Bull 40(5):367–376. https://doi.org/10.1016/S0025-326X(00)00027-8
Bilecenoğlu M (2018) Controlling the lionfish invasion in the eastern Mediterranean Sea. In Hüseyinoğlu MF, Öztürk B (Eds.) 2018. Lionfish Invasion and Its Management in the Mediterranean Sea. Turkish Marine Research Foundation (TUDAV) Publication no: 49, Istanbul, Turkey
Bird TJ, Bates AE, Lefcheck JS, Hill NA, Thomson RJ, Edgar GJ, Stuart-Smith RD, Wotherspoon S, Krkosek M, Stuart-Smith JF, Pecl GT (2014) Statistical solutions for error and bias in global citizen science datasets. Biol Conserv 173:144–154. https://doi.org/10.1016/j.biocon.2013.07.037
Bombosch A, Zitterbart DP, Van Opzeeland I, Frickenhaus S, Burkhardt E, Wisz MS, Boebel O (2014) Predictive habitat modelling of humpback (Megaptera novaeangliae) and Antarctic minke (Balaenoptera bonaerensis) whales in the Southern Ocean as a planning tool for seismic surveys. Deep Sea Res Part I: Oceanogr Res Pap 91:101–114. https://doi.org/10.1016/j.dsr.2014.05.017
Botella C, Bonnet P, Hui C, Joly A, Richardson DM (2022) Dynamic species distribution modeling reveals the pivotal role of human-mediated long-distance dispersal in plant invasion. Biol 11(9):p1293. https://doi.org/10.3390/biology11091293
Briscoe Runquist RD, Lake TA, Moeller DA (2021) Improving predictions of range expansion for invasive species using joint species distribution models and surrogate co-occurring species. J Biogeogr 48(7):1693–1705. https://doi.org/10.1111/jbi.14105
Bučas M, Bergström U, Downie AL, Sundblad G, Gullström M, Von Numers M, Šiaulys A, Lindegarth M (2013) Empirical modelling of benthic species distribution, abundance, and diversity in the Baltic Sea: evaluating the scope for predictive mapping using different modelling approaches. ICES J Mar Sci 70(6):1233–1243. https://doi.org/10.1093/icesjms/fst036
Buisson L, Thuiller W, Lek S, Lim PUY, Grenouillet G (2008) Climate change hastens the turnover of stream fish assemblages. Glob Change Biol 14(10):2232–2248. https://doi.org/10.1111/j.1365-2486.2008.01657.x
Bystriakova N, Peregrym M, Erkens RH, Bezsmertna O, Schneider H (2012) Sampling bias in geographic and environmental space and its effect on the predictive power of species distribution models. Syst Biodivers 10(3):305–315. https://doi.org/10.1080/14772000.2012.705357
Çinar ME, Arianoutsou M, Zenetos A, Golani D (2014) Impacts of invasive alien marine species on ecosystem services and biodiversity: a pan-European review. Aquat Invasions 9(4):391–423. http://dx.doi.org/10.3391/ai.2014.9.4.01
Coll M, Piroddi C, Steenbeek J, Kaschner K, Ben Rais Lasram F, Aguzzi J, Ballesteros E, Bianchi CN, Corbera J, Dailianis T, Danovaro R (2010) The biodiversity of the Mediterranean Sea: estimates, patterns, and threats. PLoS ONE 5(8):pe11842. https://doi.org/10.1371/journal.pone.0011842
Candelmo AC, Ulman A, Ali FA, Gittings SR, Huber SR, Magno LE (2022) Information on Marine Management Policies for Invasive Lionfish in the Western Atlantic Ocean and Mediterranean Sea. Dryad Data Repository
Catford JA, Jansson R, Nilsson C (2009) Reducing redundancy in invasion ecology by integrating hypotheses into a single theoretical framework. Divers Distrib 15:22–40. https://doi.org/10.1111/j.1472-4642.2008.00521.x
Coro G, Vilas LG, Magliozzi C, Ellenbroek A, Scarponi P, Pagano P (2018) Forecasting the ongoing invasion of Lagocephalus sceleratus in the Mediterranean Sea. Ecol Modell 371:37–49. https://doi.org/10.1016/j.ecolmodel.2018.01.007
Costa EF, Encarnação J, Teodósio MA, Morais P (2023) Aquatic species shows asymmetric distribution range shifts in native and non-native areas. Front Mar Sci 10:p1158206. https://doi.org/10.3389/fmars.2023.1158206
Côté IM, Green SJ, Hixon MA (2013) Predatory fish invaders: insights from Indo-Pacific lionfish in the western Atlantic and Caribbean. Biol Conserv 164:50–61. https://doi.org/10.1016/j.biocon.2013.04.014
Côté IM, Smith NS (2018) The lionfish Pterois sp invasion: Has the worst-case scenario come to pass? J Fish Biol 92(3):660–689. https://doi.org/10.1111/jfb.13544
Crystal-Ornelas R, Lockwood JL (2020) Cumulative meta‐analysis identifies declining but negative impacts of invasive species on richness after 20 year. Ecology 101(8):pe03082. https://doi.org/10.1002/ecy.3082
Cuthbert RN, Pattison Z, Taylor NG, Verbrugge L, Diagne C, Ahmed DA, Leroy B, Angulo E, Briski E, Capinha C, Catford JA (2021) Global economic costs of aquatic invasive alien species. Sci Total Environ 775:p145238. https://doi.org/10.1016/j.scitotenv.2021.145238
Danovaro R, Company JB, Corinaldesi C, D'Onghia G, Galil B, Gambi C, Gooday AJ, Lampadariou N, Luna GM, Morigi C, Olu K (2010) Deep-sea biodiversity in the Mediterranean Sea: the known, the unknown, and the unknowable. PLoS ONE 5(8):pe11832. https://doi.org/10.1371/journal.pone.0011832
Diagne C, Leroy B, Vaissière AC, Gozlan RE, Roiz D, Jarić I, Salles JM, Bradshaw CJ, Courchamp F (2021) High and rising economic costs of biological invasions worldwide. Nature 592(7855):571–576. https://doi.org/10.1038/s41586-021-03405-6
Dimitriadis C, Galanidi M, Zenetos A, Corsini-Foka M, Giovos I, Karachle PK, Fournari–konstantinidoy I, Kytinou E, Issaris Y, Azzurro E, Castriota L (2020) Updating the occurrences of Pterois miles in the Mediterranean Sea, with considerations on thermal boundaries and future range expansion. Mediterr Mar Sci 21(1):62–69. https://doi.org/10.12681/mms.21845
Dimitriou AC, Chartosia N, Hall-Spencer JM, Kleitou P, Jimenez C, Antoniou C, Hadjioannou L, Kletou D, Sfenthourakis S (2019) Genetic data suggest multiple introductions of the lionfish (Pterois miles) into the Mediterranean Sea. Divers 11(9):p149. https://doi.org/10.3390/d11090149
Dominguez Almela V, Palmer SC, Gillingham PK, Travis JM, Britton JR (2020) Integrating an individual-based model with approximate Bayesian computation to predict the invasion of a freshwater fish provides insights into dispersal and range expansion dynamics. Biol Invasions 22:1461–1480. https://doi.org/10.1007/s10530-020-02197-6
Dominguez Almela V, Palmer SC, Andreou D, Gillingham PK, Travis JM, Britton JR (2021) Predicting the outcomes of management strategies for controlling invasive river fishes using individual-based models. J Appl Ecol 58(11):2427–2440. https://doi.org/10.1111/1365-2664.13981
Dominguez Almela V, Palmer SCF, Andreou D, Gillingham PK, Travis JMJ, Britton JR (2022) Predicting the influence of river network configuration, biological traits and habitat quality interactions on riverine fish invasions. Divers Distrib 28:257–270. https://doi.org/10.1111/ddi.13459
Dominguez Almela V, Addo KA, Corbett J, Cumberbatch J, Dash J, Marsh R, Oxenford H, Tonon T, Van Der Plank S, Webber M, Tompkins EL (2023) Science and policy lessons learned from a decade of adaptation to the emergent risk of sargassum proliferation across the tropical Atlantic. Env Res Com 5(6):061002. https://iopscience.iop.org/article/10.1088/2515-7620/acd493/meta
Elith J, Graham H, Anderson CP, Dudík R, Ferrier M, Guisan S, Hijmans AJ, Huettmann R, Leathwick FR, Lehmann J, A and, Li J (2006) Novel methods improve prediction of species’ distributions from occurrence data. Ecography 29(2):129–151. https://doi.org/10.1111/j.2006.0906-7590.04596.x
Fourcade Y, Engler JO, Rödder D, Secondi J (2014) Mapping species distributions with MAXENT using a geographically biased sample of presence data: a performance assessment of methods for correcting sampling bias. PLoS ONE 9(5):pe97122. https://doi.org/10.1371/journal.pone.0097122
Franklin J (2023) Species distribution modelling supports the study of past, present and future biogeographies. J Biogeogr. https://doi.org/10.1111/jbi.14617
Frazer TK, Jacoby CA, Edwards MA, Barry SC, Manfrino CM (2012) Coping with the lionfish invasion: can targeted removals yield beneficial effects? Rev Fish Sci 20(4):185–191. https://doi.org/10.1080/10641262.2012.700655
Galil B, Marchini A, Occhipinti-Ambrogi A, Ojaveer H (2017) The enlargement of the Suez Canal—Erythraean introductions and management challenges. Manag Biol Invasions 8(2):141–152
Gámez-Virués S, Perović DJ, Gossner MM, Börschig C, Blüthgen N, De Jong H, Simons NK, Klein AM, Krauss J, Maier G, Scherber C, Steckel J, Rothenwohrer C, Steffan-Dewenter I, Weiner CN, Weisser W, Werner M, Tscharntke T, Westphal C (2015) Landscape simplification filters species traits and drives biotic homogenization. Nat Commun 6:8568. https://doi.org/10.1038/ncomms9568
Giakoumi S, Katsanevakis S, Albano PG, Azzurro E, Cardoso AC, Cebrian E, Deidun A, Edelist D, Francour P, Jimenez C, Mačić V (2019) Management priorities for marine invasive species. Sci Total Environ 688:976–982. https://doi.org/10.1016/j.scitotenv.2019.06.282
Gomes VH, IJff SD, Raes N, Amaral IL, Salomão RP, de Souza Coelho L, de Almeida Matos FD, Castilho CV, de Lima Filho A, López D, DC and, Guevara JE (2018) Species Distribution Modelling: Contrasting presence-only models with plot abundance data. Sci Rep 8(1):p1003. https://doi.org/10.1038/s41598-017-18927-1
Hao T, Elith J, Lahoz-Monfort JJ, Guillera‐Arroita G (2020) Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models. Ecography 43(4):549–558. https://doi.org/10.1111/ecog.04890
Haubrock PJ, Turbelin AJ, Cuthbert RN, Novoa A, Taylor NG, Angulo E, Ballesteros-Mejia L, Bodey TW, Capinha C, Diagne C, Essl F (2021) Economic costs of invasive alien species across Europe, vol 67. NeoBiota, pp 153–190. https://doi.org/10.3897/neobiota.67.58196
Helaouet P, Beaugrand G, Edwards M (2013) Understanding long-term changes in species abundance using a niche-based approach. PLoS ONE 8(11):pe79186. https://doi.org/10.1371/journal.pone.0079186
Hernandez PA, Graham CH, Master LL, Albert DL (2006) The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography 29(5):773–785
Hitt NP, Frissell CA, Muhlfeld CC, Allendorf FW (2003) Spread of hybridization between native westslope cutthroat trout, Oncorhynchus clarki lewisi, and nonnative rainbow trout, Oncorhynchus mykiss. Can J Fish Aquat Sci 60:1440–1451. https://doi.org/10.1139/f03-125
Hui C (2023) The dos and don’ts for predicting invasion dynamics with species distribution models. Biol Invasions 25(4):947–953. https://doi.org/10.1007/s10530-022-02976-3
Hui C, Richardson DM 2017 Invasion dynamics. Oxford University Press
IPBES, Summary for Policymakers of the Thematic Assessment Report on Invasive Alien Species and their Control of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services, Roy HE, Pauchard A, Stoett P, Renard Truong T, Bacher S, Galil BS, Hulme PE, Ikeda T, Sankaran KV (2023) McGeoch, MA, Meyerson, LA, Nuñez, MA, Ordonez, A, Rahlao, SJ, Schwindt, E, Seebens, H, Sheppard, AW, & Vandvik, V (eds) IPBES secretariat, Bonn, Germany. https://doi.org/10.5281/zenodo.7430692
Jiménez-Valverde A, Peterson AT, Soberón J et al (2011) Use of niche models in invasive species risk assessments. Biol Invasions 13:2785–2797. https://doi.org/10.1007/s10530-011-9963-4
Johnston MW, Purkis SJ (2014) Are lionfish set for a Mediterranean invasion? Modelling explains why this is unlikely to occur. Mar Pollut Bull 88(1–2):138–147. https://doi.org/10.1016/j.marpolbul.2014.09.013
Kimball ME, Miller JM, Whitfield PE, Hare JA (2004) Thermal tolerance and potential distribution of invasive lionfish (Pterois volitans/miles complex) on the east coast of the United States. Mar Ecol Prog Ser 283:269–278
Kleitou P, Savva I, Kletou D, Hall-Spencer JM, Antoniou C, Christodoulides Y, Chartosia N, Hadjioannou L, Dimitriou AC, Jimenez C, Petrou A (2019) Invasive lionfish in the Mediterranean: Low public awareness yet high stakeholder concerns. Mar Policy 104:66–74. https://doi.org/10.1016/j.marpol.2019.02.052
Kleitou P, Hall-Spencer J, Rees S, Kletou D (2022) Guide to Lionfish Management in the Mediterranean. Univ Plymouth. https://doi.org/10.24382/KYCK-J558
Leroy B, Bellard C, Dias MS, Hugueny B, Jézéquel C, Leprieur F, Tedesco PA (2023) Major shifts in biogeographic regions of freshwater fishes as evidence of the Anthropocene epoch. Sci Adv 9(46):eadi5502. https://doi.org/10.1126/sciadv.adi5502
Lodge DM, Simonin PW, Burgiel SW, Keller RP, Bossenbroek JM, Jerde CL, Kramer AM, Rutherford ES, Barnes MA, Wittmann ME, Chadderton WL (2016) Risk analysis and bioeconomics of invasive species to inform policy and management. Annu Rev Environ Resour 41:453–488. https://doi.org/10.1146/annurev-environ-110615-085532
Loya-Cancino KF, Ángeles-González LE, Yañez-Arenas C, Ibarra-Cerdeña CN, Velázquez-Abunader I, Aguilar-Perera A, Vidal-Martínez VM (2023) Predictions of current and potential global invasion risk in populations of lionfish (Pterois volitans and Pterois miles) under climate change scenarios. Mar Biol 170(3):p27. https://doi.org/10.1007/s00227-023-04174-8
Luan J, Zhang C, Xu B, Xue Y, Ren Y (2020) The predictive performances of random forest models with limited sample size and different species traits. Fish Res 227:p105534. https://doi.org/10.1016/j.fishres.2020.105534
Mainali KP, Warren DL, Dhileepan K, McConnachie A, Strathie L, Hassan G, Karki D, Shrestha BB, Parmesan C (2015) Projecting future expansion of invasive species: comparing and improving methodologies for species distribution modeling. Glob Change Biol 21(12):4464–4480. https://doi.org/10.1111/gcb.13038
Mainka SA, Howard GW (2010) Climate change and invasive species: double jeopardy. Integ Zool 5(2):102–111. https://doi.org/10.1111/j.1749-4877.2010.00193.x
Marmion M, Parviainen M, Luoto M, Heikkinen RK, Thuiller W (2009) Evaluation of consensus methods in predictive species distribution modelling. Divers Distrib 15(1):59–69. https://doi.org/10.1111/j.1472-4642.2008.00491.x
Marras S, Cucco A, Antognarelli F, Azzurro E, Milazzo M, Bariche M, Butenschön M, Kay S, Di Bitetto M, Quattrocchi G, Sinerchia M (2015) Predicting future thermal habitat suitability of competing native and invasive fish species: from metabolic scope to oceanographic modelling. Conserv Physiol 3(1):p059. https://doi.org/10.1093/conphys/cou059
Matsuzaki S-I, Usio N, Takamura N, Washitani I (2009) Contrasting impacts of invasive engineers on freshwater ecosystems: An experiment and meta-analysis. Oecologia 158:673–686. https://doi.org/10.1007/s00442-008-1180-1
McPherson JM, Jetz W (2007) Effects of species’ ecology on the accuracy of distribution models. Ecography 30(1):135–151. https://doi.org/10.1111/j.0906-7590.2007.04823.x
Morris JA Jr (2009) The biology and ecology of the invasive Indo-Pacific lionfish. North Carolina State University
Morris JA Jr, Whitfield PE (2009) Biology, ecology, control and management of the invasive Indo-Pacific lionfish: an updated integrated assessment
Mumby PJ, Harborne AR, Brumbaugh DR (2011) Grouper as a natural biocontrol of invasive lionfish. PLoS ONE 6(6):pe21510. https://doi.org/10.1371/journal.pone.0021510
Myers N, Mittermeier RA, Mittermeier CG, Da Fonseca GA, Kent J (2000) Biodiversity hotspots for conservation priorities. Nature 403(6772):853–858. https://doi.org/10.1038/35002501
Parravicini V, Azzurro E, Kulbicki M, Belmaker J (2015) Niche shift can impair the ability to predict invasion risk in the marine realm: an illustration using Mediterranean fish invaders. Ecol Lett 18(3):246–253. https://doi.org/10.1111/ele.12401
Peterson AT (2003) Predicting the geography of species’ invasions via ecological niche modeling. Q Rev Biol 78:419–433
Phillips EW, Kotrschal A (2021) Where are they now? Tracking the Mediterranean lionfish invasion via local dive centers. J Environ Manage 298:p113354. https://doi.org/10.1016/j.jenvman.2021.113354
Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecol Modell 190(3–4):231–259. https://doi.org/10.1016/j.ecolmodel.2005.03.026
Phillips SJ, Dudík M, Elith J, Graham CH, Lehmann A, Leathwick J, Ferrier S (2009) Sample selection bias and presence-only distribution models: implications for background and pseudo‐absence data. Ecol Appl 19(1):181–197. https://doi.org/10.1890/07-2153.1
Poland TM, Patel-Weynand T, Finch DM, Miniat CF, Hayes DC, Lopez VM (2021) Invasive species in forests and rangelands of the United States: A comprehensive science synthesis for the United States forest sector (p 455). Springer Nat. https://doi.org/10.1007/978-3-030-45367-1
Poloczanska ES, Burrows MT, Brown CJ, García Molinos J, Halpern BS, Hoegh-Guldberg O, Kappel CV, Moore PJ, Richardson AJ, Schoeman DS, Sydeman WJ (2016) Responses of marine organisms to climate change across oceans. Front Mar Sci 62. https://doi.org/10.3389/fmars.2016.00062
Por FD (2012) Lessepsian migration: the influx of Red Sea biota into the Mediterranean by way of the Suez Canal, vol 23. Springer Science & Business Media
Poursanidis D, Kalogirou S, Azzurro E, Parravicini V, Bariche M, Zu Dohna H (2020) Habitat suitability, niche unfilling and the potential spread of Pterois miles in the Mediterranean Sea. Mar Pollut Bull 154:p111054. https://doi.org/10.1016/j.marpolbul.2020.111054
R Core Team (2023) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/
Raghavan RK, Barker SC, Cobos ME, Barker D, Teo EJM, Foley DH, Nakao R, Lawrence K, Heath ACG, Peterson AT (2019) Potential spatial distribution of the newly introduced long-horned tick, Haemaphysalis longicornis in North America. Sci Rep 9(1):p498. https://doi.org/10.1038/s41598-018-37205-2
Rahel FJ, Olden JD (2008) Assessing the effects of climate change on aquatic invasive species. Conserv Biol 22(3):521–533. https://doi.org/10.1111/j.1523-1739.2008.00950.x
Robinson NM, Nelson WA, Costello MJ, Sutherland JE, Lundquist CJ (2017) A systematic review of marine-based species distribution models (SDMs) with recommendations for best practice. Front Mar Sci 4:p421. https://doi.org/10.3389/fmars.2017.00421
Romero-Alvarez D, Escobar LE, Varela S, Larkin DJ, Phelps NB (2017) Forecasting distributions of an aquatic invasive species (Nitellopsis obtusa) under future climate scenarios. PLoS ONE 12(7):pe0180930. https://doi.org/10.1371/journal.pone.0180930
Ruiz-Navarro A, Gillingham PK, Britton JR (2016) Predicting shifts in the climate space of freshwater fishes in Great Britain due to climate change. Biol Conserv 203:33–42. https://doi.org/10.1016/j.biocon.2016.08.021
Savva I, Chartosia N, Antoniou C, Kleitou P, Georgiou A, Stern N, Hadjioannou L, Jimenez C, Andreou V, Hall-Spencer JM, Kletou D (2020) They are here to stay: the biology and ecology of lionfish (Pterois miles) in the Mediterranean Sea. J Fish Biol 97(1):148–162. https://doi.org/10.1111/jfb.14340
Simberloff D, Martin JL, Genovesi P, Maris V, Wardle DA, Aronson J, Courchamp F, Galil B, García-Berthou E, Pascal M, Pyšek P, Sousa R, Tabacchi E, Vila M (2013) Impacts of biological invasions: what’s what and the way forward. Trends Ecol Evol 28:58–66. https://doi.org/10.1016/j.tree.2012.07.013
Smith JN, Kelly N, Renner IW (2021) Validation of presence-only models for conservation planning and the application to whales in a multiple‐use marine park. Ecol Appl 31(1):pe02214. https://doi.org/10.1002/eap.2214
Soto I, Balzani P, Carneiro L, Cuthbert RN, Macedo R, Tarkan AS, Haubrock PJ (2024) Taming the terminological tempest in invasion science. EcoEvoRxiv. https://doi.org/10.32942/X24C79
Srivastava V, Lafond V, Griess VC (2019) Species distribution models (SDM): applications, benefits and challenges in invasive species management. CABI Rev, (2019), pp1-13. https://doi.org/10.1079/PAVSNNR20191402
Stockwell DR, Peterson AT (2002) Effects of sample size on accuracy of species distribution models. Ecol Modell 148(1):1–13. https://doi.org/10.1016/S0304-3800(01)00388-X
Sutherland WJ, Barnard P, Broad S, Clout M, Connor B, Côté IM, Dicks LV, Doran H, Entwistle AC, Fleishman E, Fox M (2017) A 2017 horizon scan of emerging issues for global conservation and biological diversity. Trends Ecol Evol 32(1):31–40. https://doi.org/10.1016/j.tree.2016.11.005
Syphard AD, Franklin J (2010) Species traits affect the performance of species distribution models for plants in southern California. J Veg Sci 21(1):177–189. https://doi.org/10.1111/j.1654-1103.2009.01133.x
Tyberghein L, Verbruggen H, Pauly K, Troupin C, Mineur F, De Clerck O (2012) Bio-ORACLE: a global environmental dataset for marine species distribution modelling. Glob Ecol Biogeogr 21(2):272–281. https://doi.org/10.1111/j.1466-8238.2011.00656.x
Ulman A, Tunçer S, Kizilkaya IT, Zilifli A, Alford P, Giovos I (2020) The lionfish expansion in the Aegean Sea in Turkey: A looming potential ecological disaster. Reg Stud Mar Sci 36:p101271. https://doi.org/10.1016/j.rsma.2020.101271
Ulman A, Ali FZ, Harris HE, Adel M, Al Mabruk SA, Bariche M, Candelmo AC, Chapman JK, Çiçek BA, Clements KR, Fogg AQ (2022) Lessons from the Western Atlantic lionfish invasion to inform management in the Mediterranean. Front Mar Sci 9:p865162. https://doi.org/10.3389/fmars.2022.865162
Valavi R, Elith J, Lahoz-Monfort JJ, Guillera‐Arroita G (2021) Modelling species presence‐only data with random forests. Ecography 44(12):1731–1742. https://doi.org/10.1111/ecog.05615
Van Vuuren DP, Edmonds J, Kainuma M, Riahi K, Thomson A, Hibbard K, Hurtt GC, Kram T, Krey V, Lamarque JF, Masui T (2011) The representative concentration pathways: an overview. Clim Change 109:5–31. https://doi.org/10.1007/s10584-011-0148-z
Yu D, Zhang Z, Shen Z, Zhang C, Liu H (2018) Regional differences in thermal adaptation of a cold-water fish Rhynchocypris oxycephalus revealed by thermal tolerance and transcriptomic responses. Sci Rep, 8(1), p11703. https://doi.org/10.1038/s41598-018-30074-9 6. Statements & Declarations

SupplementaryInformation.docx

Download PDF

Reviewers agreed at journal
17 Aug, 2024
Reviewers invited by journal
12 Aug, 2024
Editor assigned by journal
02 Aug, 2024
First submitted to journal
31 Jul, 2024

You are reading this latest preprint version

Modelling the rise of invasive lionfish in the Mediterranean

Status:

Version 1

Abstract

Figures

1. Introduction

2. Methods

2.1. Study area

2.2. Species occurrence and environmental data

2.3. Modelling development

2.3.1. Maximum Entropy (MaxEnt)

2.3.2. Generalised linear models (GLMs)

2.3.3. Random Forest (RF)

2.4. Assessing predictions

3. Results

3.1. Predicted suitability under current climatic conditions

3.2. Predicted suitability under future climate scenarios

3.3. Assessing predictions

4. Discussion

4.1. Prediction of current spatial distribution

4.2. Looking into the potential future distribution

4.3. To model or not to model?

4.4. Implications for lionfish eradication, emergent uses and management

Declarations

References

Supplementary Files

Status:

Version 1