Limits...
A connectivity-based eco-regionalization method of the Mediterranean Sea.

Berline L, Rammou AM, Doglioli A, Molcard A, Petrenko A - PLoS ONE (2014)

Bottom Line: This dispersal effect can be quantified through connectivity that is the probability, or time of transport between distinct regions.Regions are discussed in the light of existing ecoregionalizations and available knowledge on plankton distributions.This objective method complements static regionalization approaches based on the environmental niche concept and can be applied to any oceanic region at any scale.

View Article: PubMed Central - PubMed

Affiliation: Université du Sud Toulon-Var, Aix-Marseille Université, CNRS/INSU/IRD, Mediterranean Institute of Oceanography (MIO), La Garde, France; CNRS, Laboratoire d'Océanographie de Villefranche, Villefranche-sur-Mer, France; Université Pierre et Marie Curie, Paris 6, Laboratoire d'Océanographie de Villefranche, Villefranche-sur-Mer, France.

ABSTRACT
Ecoregionalization of the ocean is a necessary step for spatial management of marine resources. Previous ecoregionalization efforts were based either on the distribution of species or on the distribution of physical and biogeochemical properties. These approaches ignore the dispersal of species by oceanic circulation that can connect regions and isolates others. This dispersal effect can be quantified through connectivity that is the probability, or time of transport between distinct regions. Here a new regionalization method based on a connectivity approach is described and applied to the Mediterranean Sea. This method is based on an ensemble of Lagrangian particle numerical simulations using ocean model outputs at 1/12° resolution. The domain is divided into square subregions of 50 km size. Then particle trajectories are used to quantify the oceanographic distance between each subregions, here defined as the mean connection time. Finally the oceanographic distance matrix is used as a basis for a hierarchical clustering. 22 regions are retained and discussed together with a quantification of the stability of boundaries between regions. Identified regions are generally consistent with the general circulation with boundaries located along current jets or surrounding gyres patterns. Regions are discussed in the light of existing ecoregionalizations and available knowledge on plankton distributions. This objective method complements static regionalization approaches based on the environmental niche concept and can be applied to any oceanic region at any scale.

Show MeSH
Within cluster mean connection time as a function of the cluster number for MCT3depths.White dots are the mean for each cluster, black dots are the mean over all clusters.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4222956&req=5

pone-0111978-g003: Within cluster mean connection time as a function of the cluster number for MCT3depths.White dots are the mean for each cluster, black dots are the mean over all clusters.

Mentions: Finally hierarchical clustering analysis was applied on each of the oceanographic distance matrix. This method has proved to be robust in the classification of atmospheric wind data (e.g. [24]) and hydrological data (e.g. [25]). Hierarchical clustering assigns grid cells to different clusters in a way that each grid cell belongs to only one cluster [26], and each cluster belongs to a larger cluster (Fig. 2). The grid cells are grouped according to their similarity, which here is the oceanographic distance. Thus there is no distance metric applied as in usual clustering exercises. During each sequence of the clustering algorithm, the distances between the new clusters formed and the other grid cells are computed. This step requires a linkage criterion to be defined. Here we used the flexible [27] and Ward linkages [28]. WPGMA linkage was also tested ([27]) but flexible and Ward best balanced the dendrogram. For a given cut-off level of the dendrogram, we obtained a partition of the grid cells in a certain number of clusters, which is, in the spatial domain, a regionalization. Each cluster corresponded to a region on the connectivity grid whose contours were identified. Finally for each cluster, the within-cluster MCT was computed and plotted as a function of the number of clusters from 2 to 31 (Fig. 3).


A connectivity-based eco-regionalization method of the Mediterranean Sea.

Berline L, Rammou AM, Doglioli A, Molcard A, Petrenko A - PLoS ONE (2014)

Within cluster mean connection time as a function of the cluster number for MCT3depths.White dots are the mean for each cluster, black dots are the mean over all clusters.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4222956&req=5

pone-0111978-g003: Within cluster mean connection time as a function of the cluster number for MCT3depths.White dots are the mean for each cluster, black dots are the mean over all clusters.
Mentions: Finally hierarchical clustering analysis was applied on each of the oceanographic distance matrix. This method has proved to be robust in the classification of atmospheric wind data (e.g. [24]) and hydrological data (e.g. [25]). Hierarchical clustering assigns grid cells to different clusters in a way that each grid cell belongs to only one cluster [26], and each cluster belongs to a larger cluster (Fig. 2). The grid cells are grouped according to their similarity, which here is the oceanographic distance. Thus there is no distance metric applied as in usual clustering exercises. During each sequence of the clustering algorithm, the distances between the new clusters formed and the other grid cells are computed. This step requires a linkage criterion to be defined. Here we used the flexible [27] and Ward linkages [28]. WPGMA linkage was also tested ([27]) but flexible and Ward best balanced the dendrogram. For a given cut-off level of the dendrogram, we obtained a partition of the grid cells in a certain number of clusters, which is, in the spatial domain, a regionalization. Each cluster corresponded to a region on the connectivity grid whose contours were identified. Finally for each cluster, the within-cluster MCT was computed and plotted as a function of the number of clusters from 2 to 31 (Fig. 3).

Bottom Line: This dispersal effect can be quantified through connectivity that is the probability, or time of transport between distinct regions.Regions are discussed in the light of existing ecoregionalizations and available knowledge on plankton distributions.This objective method complements static regionalization approaches based on the environmental niche concept and can be applied to any oceanic region at any scale.

View Article: PubMed Central - PubMed

Affiliation: Université du Sud Toulon-Var, Aix-Marseille Université, CNRS/INSU/IRD, Mediterranean Institute of Oceanography (MIO), La Garde, France; CNRS, Laboratoire d'Océanographie de Villefranche, Villefranche-sur-Mer, France; Université Pierre et Marie Curie, Paris 6, Laboratoire d'Océanographie de Villefranche, Villefranche-sur-Mer, France.

ABSTRACT
Ecoregionalization of the ocean is a necessary step for spatial management of marine resources. Previous ecoregionalization efforts were based either on the distribution of species or on the distribution of physical and biogeochemical properties. These approaches ignore the dispersal of species by oceanic circulation that can connect regions and isolates others. This dispersal effect can be quantified through connectivity that is the probability, or time of transport between distinct regions. Here a new regionalization method based on a connectivity approach is described and applied to the Mediterranean Sea. This method is based on an ensemble of Lagrangian particle numerical simulations using ocean model outputs at 1/12° resolution. The domain is divided into square subregions of 50 km size. Then particle trajectories are used to quantify the oceanographic distance between each subregions, here defined as the mean connection time. Finally the oceanographic distance matrix is used as a basis for a hierarchical clustering. 22 regions are retained and discussed together with a quantification of the stability of boundaries between regions. Identified regions are generally consistent with the general circulation with boundaries located along current jets or surrounding gyres patterns. Regions are discussed in the light of existing ecoregionalizations and available knowledge on plankton distributions. This objective method complements static regionalization approaches based on the environmental niche concept and can be applied to any oceanic region at any scale.

Show MeSH