Limits...
Optimising Regionalisation Techniques: Identifying Centres of Endemism in the Extraordinarily Endemic-Rich Cape Floristic Region.

Bradshaw PL, Colville JF, Linder HP - PLoS ONE (2015)

Bottom Line: We show that weighted data (down-weighting widespread species), similarity calculated using Kulczinsky's second measure, and clustering using UPGMA resulted in the optimal classification.Post-clustering GIS manipulation substantially enhanced the endemic composition and geographic size of candidate CoEs.Although there was broad spatial congruence with previous phytogeographic studies, our techniques allowed for the recovery of additional phytogeographic detail not previously described for the CFR.

View Article: PubMed Central - PubMed

Affiliation: Park Planning and Development Unit, South African National Parks, Port Elizabeth, South Africa; Department of Botany, Nelson Mandela Metropolitan University, Port Elizabeth, South Africa.

ABSTRACT
We used a very large dataset (>40% of all species) from the endemic-rich Cape Floristic Region (CFR) to explore the impact of different weighting techniques, coefficients to calculate similarity among the cells, and clustering approaches on biogeographical regionalisation. The results were used to revise the biogeographical subdivision of the CFR. We show that weighted data (down-weighting widespread species), similarity calculated using Kulczinsky's second measure, and clustering using UPGMA resulted in the optimal classification. This maximized the number of endemic species, the number of centres recognized, and operational geographic units assigned to centres of endemism (CoEs). We developed a dendrogram branch order cut-off (BOC) method to locate the optimal cut-off points on the dendrogram to define candidate clusters. Kulczinsky's second measure dendrograms were combined using consensus, identifying areas of conflict which could be due to biotic element overlap or transitional areas. Post-clustering GIS manipulation substantially enhanced the endemic composition and geographic size of candidate CoEs. Although there was broad spatial congruence with previous phytogeographic studies, our techniques allowed for the recovery of additional phytogeographic detail not previously described for the CFR.

No MeSH data available.


A dendrogram of the correlation between the 12 weighting-dissimilarity matrices.Mantel Tests were undertaken using Pearson correlation and 999 permutations. The actual correlation values are provided in S2 Table. All dissimilarity matrices were significantly correlated with p < 0.001. [Unw = unweighted, Bell = Bell weighting, Int = Integration weighting, Inv = Inverse weighting, K2 = Kulczinsky2; J = Jaccard; S = Simpson].
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4493007&req=5

pone.0132538.g005: A dendrogram of the correlation between the 12 weighting-dissimilarity matrices.Mantel Tests were undertaken using Pearson correlation and 999 permutations. The actual correlation values are provided in S2 Table. All dissimilarity matrices were significantly correlated with p < 0.001. [Unw = unweighted, Bell = Bell weighting, Int = Integration weighting, Inv = Inverse weighting, K2 = Kulczinsky2; J = Jaccard; S = Simpson].

Mentions: The performance criteria of the different analysis permutations (different weightings, similarity coefficients and clustering algorithms) in delimiting CoEs are summarised in Table 2. Of the top five ranked analytical approaches (all UPGMA, see Tables 2 & 3), the top three utilised the Kulczinsky2 similarity coefficient (Bell:K2 performed best, followed by Inv:K2 and Int:K2); and three techniques (1st, 4th & 5th) employed Bell weighting function (Table 3). A summary of the optimality of the weighting techniques employed and the clustering / similarity coefficients (Table 4) indicated that overall Kulczinsky2 performed better than Jaccard, while Simpson performed the least optimally. Pair-wise Mantel Tests of the correlation between the dissimilarity matrices indicated that all dissimilarity matrices were significantly correlated (p = 0.001 for all comparisons; R2 values in S2 Table). Overall, dissimilarity matrices generated using the same coefficient but different weighting approaches were more similar to each other than dissimilarity matrices based on different coefficients but the same weighting approach (Fig 5). Further, the Kulczinsky2 and Simpson dissimilarity matrices are more similar to each other than to Jaccard dissimilarity matrices (Fig 5). Of the weighting techniques, Bell performed most optimally, followed by Inv, then Int weighting (Table 4), although these absolute rankings masked similar results (Tables 2 & 3). PAE ranked seventh, ninth or twelfth depending on the weighting, while the unweighted data performed poorly in almost all CoE performance measures (Tables 2 & 3). Weighted PAE performed better than unweighted matrix UPGMA.


Optimising Regionalisation Techniques: Identifying Centres of Endemism in the Extraordinarily Endemic-Rich Cape Floristic Region.

Bradshaw PL, Colville JF, Linder HP - PLoS ONE (2015)

A dendrogram of the correlation between the 12 weighting-dissimilarity matrices.Mantel Tests were undertaken using Pearson correlation and 999 permutations. The actual correlation values are provided in S2 Table. All dissimilarity matrices were significantly correlated with p < 0.001. [Unw = unweighted, Bell = Bell weighting, Int = Integration weighting, Inv = Inverse weighting, K2 = Kulczinsky2; J = Jaccard; S = Simpson].
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4493007&req=5

pone.0132538.g005: A dendrogram of the correlation between the 12 weighting-dissimilarity matrices.Mantel Tests were undertaken using Pearson correlation and 999 permutations. The actual correlation values are provided in S2 Table. All dissimilarity matrices were significantly correlated with p < 0.001. [Unw = unweighted, Bell = Bell weighting, Int = Integration weighting, Inv = Inverse weighting, K2 = Kulczinsky2; J = Jaccard; S = Simpson].
Mentions: The performance criteria of the different analysis permutations (different weightings, similarity coefficients and clustering algorithms) in delimiting CoEs are summarised in Table 2. Of the top five ranked analytical approaches (all UPGMA, see Tables 2 & 3), the top three utilised the Kulczinsky2 similarity coefficient (Bell:K2 performed best, followed by Inv:K2 and Int:K2); and three techniques (1st, 4th & 5th) employed Bell weighting function (Table 3). A summary of the optimality of the weighting techniques employed and the clustering / similarity coefficients (Table 4) indicated that overall Kulczinsky2 performed better than Jaccard, while Simpson performed the least optimally. Pair-wise Mantel Tests of the correlation between the dissimilarity matrices indicated that all dissimilarity matrices were significantly correlated (p = 0.001 for all comparisons; R2 values in S2 Table). Overall, dissimilarity matrices generated using the same coefficient but different weighting approaches were more similar to each other than dissimilarity matrices based on different coefficients but the same weighting approach (Fig 5). Further, the Kulczinsky2 and Simpson dissimilarity matrices are more similar to each other than to Jaccard dissimilarity matrices (Fig 5). Of the weighting techniques, Bell performed most optimally, followed by Inv, then Int weighting (Table 4), although these absolute rankings masked similar results (Tables 2 & 3). PAE ranked seventh, ninth or twelfth depending on the weighting, while the unweighted data performed poorly in almost all CoE performance measures (Tables 2 & 3). Weighted PAE performed better than unweighted matrix UPGMA.

Bottom Line: We show that weighted data (down-weighting widespread species), similarity calculated using Kulczinsky's second measure, and clustering using UPGMA resulted in the optimal classification.Post-clustering GIS manipulation substantially enhanced the endemic composition and geographic size of candidate CoEs.Although there was broad spatial congruence with previous phytogeographic studies, our techniques allowed for the recovery of additional phytogeographic detail not previously described for the CFR.

View Article: PubMed Central - PubMed

Affiliation: Park Planning and Development Unit, South African National Parks, Port Elizabeth, South Africa; Department of Botany, Nelson Mandela Metropolitan University, Port Elizabeth, South Africa.

ABSTRACT
We used a very large dataset (>40% of all species) from the endemic-rich Cape Floristic Region (CFR) to explore the impact of different weighting techniques, coefficients to calculate similarity among the cells, and clustering approaches on biogeographical regionalisation. The results were used to revise the biogeographical subdivision of the CFR. We show that weighted data (down-weighting widespread species), similarity calculated using Kulczinsky's second measure, and clustering using UPGMA resulted in the optimal classification. This maximized the number of endemic species, the number of centres recognized, and operational geographic units assigned to centres of endemism (CoEs). We developed a dendrogram branch order cut-off (BOC) method to locate the optimal cut-off points on the dendrogram to define candidate clusters. Kulczinsky's second measure dendrograms were combined using consensus, identifying areas of conflict which could be due to biotic element overlap or transitional areas. Post-clustering GIS manipulation substantially enhanced the endemic composition and geographic size of candidate CoEs. Although there was broad spatial congruence with previous phytogeographic studies, our techniques allowed for the recovery of additional phytogeographic detail not previously described for the CFR.

No MeSH data available.