Selection of species
The list of species for the vulnerability assessment was based on five different criteria. First, we considered the proportion of each species in the total Portuguese landings between 1989 and 2015, using public landings data from the Direção Geral dos Recursos Marinhos de Portugal (DGRM). The most landed species, accounting for 95% of purse seine, 70% of trawling and 70% of the multigear landings, were included. This selection was carried out separately for each combination of gear and region (Supplementary Table SI1-1). Second, species were chosen in regards of their economic relevance, considering the species representing more than 3% of the total economic revenue of the marine landings within each combination of region and gear (DGRM, Supplementary Table SI1-2). Third, we included the most frequent species in the discards of Portuguese fisheries, according to the work of Leitão et al.42, where the top-ten discarded species per métier are listed (Supplementary Table SI1-3). Fourth, we included the species of importance for the canning industry, obtained by means of a survey covering the main can enterprises of Portugal (Supplementary Table SI1-5). Fifth, a selection of the species of relevance for the Moroccan fisheries sector was carried out, using the reports from the Department of Marine Fisheries of the Kingdom of Morocco43 and the FAO software FishStatJ (most captured species between 2007 and 201744) (Supplementary Table SI1-6). Additionally, due to their importance for specific fleet segments, we included some shark species of interest that were not included by the previous criteria. The selection of shark species was based on reports from the Instituto Português do Mar e as Pescas (IPMA) and included: Galeus melastomus, Prionace glauca, Squalus acanthias, Scyliorhinus canicula, and Hexanchus griseus. Some riverine species were finally removed from the list (Petromyzon marinus, Salmo trutta), as well as cod (Gadus morhua), since it is not captured within the area of study. Finally, some extra species were pointed out by experts during the evaluation process as species with economic interest (Pollicipes pollicipes) or with potential distribution shift into/from the area of study in the context of climate change such as the bivalves Callista chione and Ruditapes philippinarum, and the crabs Callinectes sapidus and Carcinus maenas. The final list of species considered, and their functional group are shown in Table 1.
Environmental change
RCP (representative concentration pathway) scenarios of atmospheric greenhouse gas concentration have been proposed by the IPCC for use in research to project the evolution of environmental variables. Using scenarios RCP 4.5 and RCP 8.5 (predicting a global warming of 1.8 and 3.7 °C respectively by the end of the twenty-first century) as forcing, the POLCOMS-ERSEM model45 forecasted a wide array of physical, chemical and biological variables for the Northeast Atlantic and adjacent seas at a resolution of 0.1 degree (approximately 11 km). For the evaluation of the vulnerability of the species of interest, a selection of the most cited variables with impact on the ecology of marine organisms in the Portuguese marine environment was carried out (e.g. Refs.7,8,9). As a result, these variables were finally considered: sea surface temperature (SST, °C), surface pH, surface salinity (psu), surface zooplankton biomass (mol m−3), surface phytoplankton biomass (mol m−3), surface northward and eastward current velocities (m s−1) and river discharge (m−3 year−1). The zooplankton and phytoplankton biomass were summed to obtain an overall plankton biomass (mol m−3) which was finally used in the assessment of vulnerability. Surface variables were calculated using the top sigma layer of the outputs of the model.
Two time slices of the POLCOMS-ERSEM outputs were used to define two periods for comparison. The first was between 2000 and 2019 setting a reference point for the state of the environment at the beginning of the century (hereafter “reference”), then, the period between 2040 and 2059 served to define the likely state of the environment in the near future (hereafter “future”). Defining the future and reference periods allowed us to compare the expected degree of change of the environmental variables between both periods. To do this on a regional basis, we considered the outputs of the model for each region of Portugal (North, Centre, and South; Fig. 1) and calculated a dimensionless variation index (VI) using the mean of each variable during the reference and future periods, and the standard deviation of the reference period:
$$ {text{VI}} = frac{{left( {mu ,future – mu ,reference} right)}}{sigma ,reference} , $$
(1)
where µ future and µ reference represent the regional average values of the corresponding time slice of the variable, and σ reference is the standard deviation of the regional values in the reference time slice (except for the variable river discharge, for which the average and standard deviation are calculated on a temporal basis) VI takes theoretical values between 0 (when there is no variation between future and reference) and ± infinite (when reference shows no variation all over the region of study). VI was used to weight the influence of each variable in the assessment of the exposure of the species to climate change n Table 2. The idea was to capture the degree of variability of each physical variable, so species exposed to the most variable environmental conditions would be more exposed to the effects of climate change. Then, a weight factor was calculated normalizing between 1 and 2 the absolute values of the VI defined above (“weight factor 1” in Table 2).
Since two versions of the future period were available (climate change scenarios RCP 4.5 and RCP 8.5), the level of exposure to changing environmental variables was calculated separately for both climate change scenarios, making it possible to estimate the overall vulnerability of the species under each scenario separately.
Beyond the degree of variability of each variable, a panel of experts on the ecology of marine organisms of Portugal was asked to rank, according to the likely impact on the physiology of marine organisms, the physical variables under consideration. Each expert was asked to order the variables independently, but a consensus answer was finally asked from them. The ranking of the physical variables was posteriorly transformed numerically between 1 and 2, being 1 the less relevant variable and 2 the most relevant variable. Intermediate variables got a value between 1 and 2 following equally distanced steps (see “weight factor 2” in Table 2). The final weight given to each physical variable during the vulnerability assessment was calculated as the average between weight factors 1 and 2 (“final weight factor” in Table 2). It was possible to estimate this parameter for all the exposure indicators with exception of the extreme events frequency, which was not included in the POLCOMS-ERSEM outputs. The likely evolution of this parameter is controversial and thus, a final weight factor of 1 was assigned by consensus with the panel of experts. In the case of oceanic currents, considered as a proxy for upwelling, we considered eastward currents in the North and Centre regions (North–South oriented coast) and northward currents in the South region (East–West oriented coast).
Vulnerability assessment
Indicators
The vulnerability of the species to climate change was evaluated following the conceptual framework described in the 4th Assessment Report of the IPCC29. This approach assumes that the vulnerability (V) of species to environmental change is a function of: (1) their exposure (E) to the changing environmental variables (defined as the overlap between the expected geographic range of change of the variables and the area/habitats of occurrence of a given species), (2) their sensitivity (S) to environmental change (considered as the degree to a which extent a given species will be affected—in terms of population dynamics or life-history traits—by a change in the environment), and (3) their adaptive capacity (AC) to environmental change (understood as the mechanisms of a given species to resist to a specific change of the environment and recover to the state prior to the perturbation).
For each species, the degree of exposure, sensitivity and adaptive capacity was evaluated considering different aspects (hereafter “indicators”) of its biology, ecology, and exploitation (see Supplementary SI2 for a description of the indicators). The selection of the indicators was made considering the context of climate change in the Portuguese marine environment. Hence, for the level of exposure, the most referenced environmental variables with impact on the ecology of the species of interest were chosen. For the analysis of the sensitivity, a selection of life history traits driving the relationship between the species’ population dynamics and the environment was carried out based on existing literature (e.g. Refs.23,26,28,36). The traits finally considered were: trophic level, fecundity, number of reproductive events in a lifetime, egg spawning strategy, individual growth parameters (growth coefficient, k, in Von Bertalanffy’s growth function), age at maturity, longevity, intrinsic population growth rate (r), sexual strategy (gonochorism, hermaphroditism or protogyny/protandry), length of the spawning seasons, planktonic larval duration (PLD), latitudinal range of distribution, temperature range of distribution, adult mobility, seasonal migrations, sociability, and complexity of the reproductive strategy. The adaptive capacity of the species was analysed considering different aspects related to the degree of conservation or exploitation of the species and the kind of fisheries associated, which give an idea of the capacity of response of the populations to environmental change at a national or regional scale. In this case we considered: the ICES stock status (referred to Portuguese or Iberian stocks when available), the general replenishment potential of the species, related to different life-history parameters such as growth and reproduction, the vulnerability degree assigned by the IUCN, the specific vulnerability to fisheries assessed in Cheung et al.26, and the fishing pressure suffered by each species in Portuguese waters.
Expert’s assessment
To evaluate each species from the point of view of each indicator, a fuzzy logic expert-judgement method was applied26. This method consists of categorizing the range of possible answers or values of each indicator into three levels (bins) corresponding to low, moderate, or high levels of exposure, sensitivity and adaptive capacity, respectively. The number of levels considered (3) has been found to be sufficient for this kind of study28,46, and their ranges were defined for each indicator following the existing literature, adjusting their values to the reality of the Portuguese marine environment. For a description of the levels within each indicator see Supplementary SI2.
Assigning each species to each bin of each indicator was carried by a group of experts in marine biology and ecology with experience in the Portuguese marine environment. A variable number of species was assigned to each expert in regards of their field of knowledge and previous experience. Each species received a minimum of three experts and a maximum of four. The number of tallies assigned to each bin of each indicator (variable between 0 and 5) represented the degree of confidence in the answer. In this way, an absolute confidence in the answer provided was represented by allocating 5 tallies in the corresponding bin, while spreading the five tallies among the three bins meant the highest level of uncertainty. In order to avoid biases in the expert evaluations, each expert was provided with the description of the indicators and their bins found in Supplementary SI2, the maps of climate variability found in Supplementary SI3, and a list of online resources to consult. The experts were allowed to consult any other scientific literature for their evaluations if needed.
After the evaluation of each indicator of exposure, sensitivity and adaptive capacity, each expert was asked to provide a formed opinion on the likely direction of the effects of climate change for each species. This directional effect (DE) evaluation had two steps: (1) the allocation of five tallies among three bins representing negative, neutral, or positive DE, and (2) providing a short rationale text explaining the allocation of tallies among the bins.
Experts were also asked to score the quality of the data used to distribute the tallies among the bins of each indicator following the methodology of Hare et al.23. In this case, the experts should assign a value between 0 and 3 to describe the quality of the information. These values correspond to (0) No Data. No information is available to provide an opinion; (1) Expert Judgement. The distribution of tallies among the bins reflects the expert judgement, based on knowledge of the general ecology of the species and its role on the ecosystem; (2) Limited Data. The data used to distribute the tallies may come from similar species or from other geographic regions out of the Iberian Peninsula; (3) Adequate Data. The score is based on data observed, modelled or directly measured for the species in question and is provided by scientific work carried out in the Iberian Peninsula.
After the individual assessments, a 2-day workshop was carried out where the experts were asked to discuss their evaluations and provide a summarizing text on the likely sign of directional effects of climate change on each species. They were also allowed to modify the distribution of tallies of their votes for the directional effects after the discussion.
Regional evaluation
Each expert was asked to perform the evaluation of each indicator independently for each region of Portugal (North, Centre and South; Fig. 1). This procedure made it possible to obtain, for a given species, region-specific assessments of E, S, AC and DE, which could be finally translated into region-specific overall vulnerability assessments.
Calculation of the overall vulnerability score
For each species, the number of tallies assigned by the experts to each bin of each indicator was averaged. Then, each tally was assigned a different value in regards of the bin where it was assigned: 1-low, 2-moderate, 3-high, making possible to calculate the value of each indicator by summing the value of the tallies. The final score of the indicator (minimum: 5; maximum: 15) was standardized between 0 and 1. To obtain the value of each dimension of the vulnerability (E, S, or AC) the sum of the values of the related indicators standardized between 0 and 1 was computed. All the indicators had the same weight.
Finally, to calculate the overall vulnerability, the value of each dimension was standardized between 0 and 1, being V calculated as:
$$ {text{V}}_{{{text{r}} – {text{cc}}}} = , left( {{text{E}}_{{{text{r}} – {text{cc}}}} + {text{ S}}_{{text{r}}} } right) , {-}{text{ Ac}}_{{text{r}}} , $$
(2)
where subscripts indicate region (r) and climate change (cc) specificity, respectively.
The vulnerability score (Vr-cc) obtained was finally categorized as: “very low vulnerability” (Vr-cc < 0.20, note that negative values could exist), low (0.20 < Vr-cc < 0.40), moderate (0.40 < Vr-cc < 0.60), high (0.60 < Vr-cc < 0.80), and very high (Vr-cc > 0.80, note that values higher than 1 could exist).
Probability of distribution change
The method for assessing the potential for a change in species distribution was adapted from that described in Hare et al.23. These authors consider that species with high adult mobility, broadly dispersing early life stages, low habitat specificity, and high temperature sensitivity would have higher potential to change their area of distribution in the context of climate change. Here, we adapted these criteria to the descriptors of vulnerability considered, and calculated the probability of distribution change (P) as a function of these indicators considering that high P will be characterized by high adult mobility, long PLD, broadcast egg spawning strategy, wide latitudinal range, and narrow temperature tolerance range. These indicators were standardized between 0 and 1 and then considered as:
$$ P = frac{{left( {Ad.mobility + PLD + Eggsp.strategy + Lat.range} right) – temp.range}}{4}. $$
(3)
The range of values of P was categorized as: very high (P > 0.80), high (0.60 < P < 0.80), moderate (0.40 < P < 0.60), low (0.20 < P < 0.40) or very low (P < 0.20).
Variability in experts’ voting
To evaluate the inter-experts’ variability in the allocation of tallies among the bins of each indicator, a bootstrap analysis was carried out. This analysis consisted of a random sampling (10,000 iterations) with replacement of the total number of tallies allocated per indicator (5 tallies × 3 or 4 experts), calculating the overall vulnerability as described before. Then, the proportion of iterations resulting in a vulnerability of the same category (very low, low, moderate, high, or very high) as the original was computed to estimate the variability in the assignment of vulnerability scores by the experts.
The same procedure was carried out considering the indicators needed to compute the probability of distribution change and the directional effects, allowing to evaluate the certainty on these parameters independently of the overall vulnerability.
Vulnerability categories in regards of the relationship between the components of vulnerability
Foden et al.47 described four categories of vulnerability based on the relationship between the vulnerability dimensions E, S, and AC. The first category (“highly climate change vulnerable species”) comprises species with high E and S but low AC, which means that they are at great risk due to climate change. The second group (“potential adapter species”) is formed by species with high E, S, and AC, so they may be at risk due to climate change. The third group (“potential persistent species”) considers species with high E and low S and AC, representing those species that may not be at risk due to climate change. Finally, the “high latent risk species” are those with high S and low E and AC, comprising species that would not be currently at risk. Different management perspectives have been proposed for the species within each category (see Foden et al.47).
To allocate the species to these categories, we considered “highly climate change vulnerable species” those with very high and high exposure and sensitivity, and low or very low adaptive capacity. Potential adapter species were those with very high, high, or moderate exposure, sensitivity, and adaptive capacity. Potential persistent species were defined as those with very high, high, or moderate exposure and very low or low sensitivity and adaptive capacity. Finally, high latent risk species were considered those with very low or low exposure and adaptive capacity, and very high or high sensitivity.
Relationship between overall vulnerability and ecosystem indicators
Aiming at providing a simple but reliable approximation to the assessment of the vulnerability of a given species, we also analysed the relationship between the final vulnerability score and the different indicators used in this work by means of linear regressions.
Source: Ecology - nature.com
