Impact du réchauffement climatique sur la distribution spatiale des ressources halieutiques le long du littoral français: observations et scénarios
par Sylvain Lenoir
Université Lille 1 Science - Doctorat 2011
? Estimation of the realised niche from uncorrected and corrected reference matrices
Data occurrences from the uncorrected training sets were first used to determine the realised niche of each fish (Fig. III.S10). The contours of these niches depended on species. Some species exhibited a unimodal distribution close to a normal distribution. For example, the examination of the thermal tolerance of haddock (Fig. III.S10(f)) showed a preferendum ranging between 5°C and 15°C and an optimum around 9.75°C. Pollack was more stenotherm, found between 9°C and 15°C with a thermal optimum at about 12.75°C (Fig. III.S10(d)). However, for other species and parameters, the distribution could be multimodal (e.g. Atlantic horse mackerel and European anchovy; see Fig. III.S10(a), (b)). For all species, the bathymetric distribution did not follow a normal distribution. The frequencies of occurrence were maximal for the first 200 meters (i.e. continental shelf). The contours of the niche for annual SSS were not as well defined as for the other parameters because the amplitude of variation in the region was not generally important. All species showed mainly a maximum of occurrence between 30 and 40. Pollack was less euryhaline than species such as common sole and turbot (Fig. III.S10(e), (h)). The different salinity profiles exhibited by the species highlighted the importance of the parameter as a predicting variable in NPPEN.
Multiple modes are less conform to our common idea of the ecological niche. They can be related to the absence of a specific habitat or to the presence of oversampled or undersampled regions. They could also be due to seasonal migration such as the one performed by sardine. Indeed most often, data used in ENMs do not originate from rigorous sampling protocol (Legendre & Legendre 1983). To consider this bias, the training set was corrected, which resulted in an improvement of most environmental preferendums (Fig. III.2). However, the procedure of correction did not modify substantially the environmental profiles of species, indicating that the profiles were here not too much influenced by the heterogeneity of the spatial information. Most optimums were refound (Fig. III.2). In the case of poorer occurrence dataset (e.g. turbot; Figs. III.S10(h) and 2(h)), the correction allowed contours of the salinity preferendum to be completed. This homogenisation procedure enabled some response curves to be better balanced. Such a correction was essential as most ENMs, as ours, are highly sensitive to the determination of the niche
? Sensitivity of the model to bimodality and sampling density
Previous results showed that some species exhibited multimodal responses (Figs. III.S10 and 2). To evaluate the influence of such patterns on our model, we simulated training sets characterised by different types of bimodality and sampling density (Fig. III.3). This comparisonwas made by simulating the values of two environmental parameters: SST and bathymetry. When the density of sampling was regular and in case of unimodality, the highest probabilities were observed at the centre of the sampling points and declined progressively towards the edge of the distribution (Fig. III.3(a)). This training set was considered here as a reference to evaluate simulated incomplete, heterogeneous and bimodal training sets. The model was robust to incomplete or heterogeneous training sets (Fig. III.3). The mean difference ranged from 0.0940 (Fig. III.3(b)), 0.0732 (Fig. III.3(c)) to 0.0622 (Fig. III.3(d)). These results suggest that the model was quite robust to altered training set that could be related to incomplete sampling coverage or an unrepresented biotope.
Figure III.4 : Estimated probability of occurrence using the model NPPEN for the decade 1960-1969, the time period 2000-2005 in the North Atlantic Ocean. (a) Atlantic horse mackerel, (b) European anchovy, (c) European sprat, (d) pollack, (e) common sole, (f) haddock, (g) saithe and (h) turbot. The western boundary of the model was fixed to 40°W. This was arbitrary selected for species (but haddock and saithe), which are only found on the eastern side of the Atlantic Ocean.
? Modelled spatial distribution of fish species (contemporaneous spatial patterns)
The examination of the modelled spatial distribution for the 8 species considered in this study indicated the presence of three groups (Fig. III.4, period 1960-1969): (1) species with a spatial distribution mainly centred in the Celtic Sea and/or the North Sea (called temperate species; European sprat, pollack and turbot); (2) species with a spatial distribution from southerly regions to the North Sea (called warm-temperate species; Atlantic horse mackerel, European anchovy and common sole); (3) species with a spatial distribution ranging from the North Sea to the Barents Sea (called subarctic species; haddock and saithe). These results showed that the North Sea and more generally regions close to the British Isles represented a biogeographical crossroad between the Atlantic Arctic and the Atlantic Westerlies Wind (temperate) biomes (Longhurst 1998). The probability of fish occurrence calculated from NPPEN exhibited substantial differences with the maps produced by the FAO (Fig. III.S1).
? Comparison of the spatial distribution inferred from NPPEN and AquaMap
We compared the spatial distribution of fish occurrence modelled by NPPEN for the 1960s and the period 2000-2005 with outputs from AquaMap (Kaschner et al. 2006) by correlation analysis (Table III.1). The strength of the correlations, similar for the two periods (190-1969, mean correlation: 0.5075; 2000-2005, mean correlation: 0.53), depended on species. Although all coefficients of correlation were significant, the amount of variance explained ranged from 1.44% for European anchovy to 56.25% for haddock (Table III.1). The minimum degree of freedom needed to have a significant correlation was calculated to evaluate in which measure the spatial autocorrelation could have influenced the results. For example, the high correlation found with haddock remained significant (p= 0.05) with only 5 degree of freedom (Table III.1). The reduction of the degree of freedom ranged from 5 to 255 (European anchovy). In general, the correlation between the two models tended to increase with the number of occurrence data in the reference matrix (correlation calculated between the decimal logarithm of the number of data after simplification of the training set in Supplementary Table III.1 and correlation coefficients in Table III.1; 1960-1969: rp=0.53, p=0.17; 2000-2005: rp=0.45, p=0.27).
Figure III.5 : Maps showing areas where a substantial increase (in red) or decrease (in blue) in the probability of occurrence (changes>0.2, see `Materials and methods') is expected between the 2090s and the 1960s. Green areas denote no substantial change or changes <0.2. (a) Atlantic horse mackerel, (b) European anchovy, (c) European sprat, (d) pollack, (e) common sole, (f) haddock, (g) saithe and (h) turbot. The western boundary of the model was fixed to 40°W. This was arbitrary selected for species (but haddock and saithe), which are only found on the eastern side of the Atlantic Ocean.
Table III.1 : Correlations between the spatial patterns in the probability of fish occurrence between the model NPPEN and AquaMap (Kaschner et al. 2006) for the period 1960-1969 and 2000-2005. n: degree of freedom; rp: Pearson linear coefficient of correlation; p: probability; n.f.: minimum degree of freedom needed to have a significant correlation at p=0.05.
Table III.2 : Thermal preference and the expected area (in km²) gained or lost by the species between the period 2000-2005 and 1960-1969 and 2090-2099 and 1960-1969 for both scenarios A2 and B2. The balance was calculated (in km²) as the difference between gained and lost area (see Materials and Methods).
? Observed changes in species distribution
All species but haddock exhibited a northward movement or an increase in the probability of occurrence at the northern edge of their spatial distribution between the 1960s and the period 2000-2005 (Fig. III.4). The probability of occurrence of European sprat increased substantially in the North Sea in 2000-2005. The reduction in the probability at the southern edge of all species was not evident with the exception of European sprat, saithe and haddock (e.g. the southern part of the North Sea). Between the 1960s and the period 2000-2005, the potential habitat of warm-temperate species increased (Table III.2). Among them, the potential habitat of European anchovy increased substantially (balance of change: 346,334 km²). Both subarctic and temperate species exhibit a weak or a moderate increase between the 1960s and the period 2000-2005 or a moderate decrease for pollack and turbot (Table III.2).
? Changes in species distribution
Modelled spatial distributions based on projected IPCC changes in SST (scenario B2) suggest that northward movements in fish may accelerate in the future with the exception of pollack in the North Sea (Fig. III.S10). These movements will be generally associated with a reduction located at the southern edge of the species spatial distribution (Fig. III.S10). The projections suggest that the potential habitat of European anchovy will increase (Fig. III.5, Table III.2). Results are less obvious for common sole or Atlantic horse mackerel and suggested either a weak decrease (Scenario A2, Table III.2) or a moderate increase (Scenario B2, Table III.2). Atlantic horse mackerel and European anchovy are expected to move northwards along the European shelf-edge and in the North Sea (Figs. III.S10 and III.5, Table III.2). The potential habitat of temperate species is expected to decrease substantially (Fig. III.5, Table III.2) with the exception of European sprat that is expected to migrate to the Barents Sea. However, this species might eventually disappear from the central part of the North Sea at the end of the century (Fig. III.S10). The model suggests that the potential habitat of subarctic species may diminish considerably (Fig. III.5, Table III.2). The reduction of the potential habitat in the North Sea may not be overcome by the increase in potential habitat over the Barents Sea. Projections based on scenario A2 and B2 for subarctic species gave very similar conclusions (Table III.2).