Classification & Ordination of Plant Communities

Flashcards for plant community classification and ordination, covering concepts from community coefficients to ordination techniques.

Ordination and classification

Summarize raw data from quadrats, releve's, line intercepts, etc. to effectively communicate patterns by removing redundancy, ignoring noise and uncovering underlying patterns in data, reduced to easily interpreted tables, trees, or graphs, descriptive methods.

Community Coefficients

For each pair of samples in a data set, an index of similarity (0-100) can be calculated. The higher the index, the more alike. All possible pairs of samples may be compared by constructing a matrix of (dis)similarity values. Ordination & classification calculations often begin such a matrix.

Jaccard Community Coefficient

=(Shared species/ sp A + sp B- total) 100 =C/(A+B-C)100

Sørensen CC weighted by cover

=(2x total lowest cover values)/total cover A+ total cover B)100 =(2MC/MA+MB)100

Classification in community ecology

Grouping samples together into similar community types--or grouping together species that occur on similar sites.

Braun-Blanquet Table Arrangement

Rows & columns are moved so stands with similar composition are next to each other in columns & species that occur together in similar stands are next to each other in rows. Both species and sites classified

Cluster Analysis: Agglomerative Classification

Start at bottom with as many clusters as samples and combine all samples that exceed a given similarity. Resulting clusters are then combined at a lower threshold of similarity. Process repeats for lower thresholds until all sites combine into one cluster.

Divisive Classification

Begins with whole data set as a group & repeatedly divides it into increasingly similar clusters. Example: TWINSPAN -divisive hierarchical classification

Multi-Response Permutation Procedures (MRPP)

Tests whether a priori groups differ in their position in a multidimensional space. Performs a multivariate test of no difference between groups.

Indicator Species Analysis (ISA)

Assigns indicator value for each species (assumed to be columns in your main matrix), based the degree to which they discriminate among groups.

Classify image

Draw / digitize polygons on photos. May used a community-based classification as basis. Computer algorithms to classify a digital image to obtain map polygons.

Ordination

Summarizes redundant structure of data. Reduction of dimensionality.

Ordination Diagram

Samples with similar species composition nearby; Dissimilar ones distant. Interpreted as gradients in community composition.

Polar ordination (Bray & Curtis 1957)

Calculation of percentage dissimilarity matrix for samples. The two most dissimilar samples (poles) are identified. Position of remaining samples relative to poles calculated.

Principal Components Analysis (PCA)

Mathematically complex. Conceptually involves 'rotating the axes of a multivariate cloud of points in multidimensional space'. First axis lies in direction of maximum variability. Second axis extracted perpendicular (orthogonal) to the first.

"Arch" or “Horseshoe” Effect in PCA

Non-linear species responses cause distribution of sample points in ordination space to distort into an arch or a spiral; Second axis (principal component) becomes quadratic distortion of the first.

Weighted Averagings

Weights assigned each species by investigator. Weights usually have some kind of ecological or environmental meaning. Sample ordination score = sum (species abundance x weight).

Reciprocal Averaging

use arbitrary species weights to calculate sample (site) ordination scores. Species scores are then re-calculated using these sample scores as weights.

Detrended Correspondence Analysis, (DCA)

Modification of correspondence analysis (=reciprocal averaging). Removes "arch effect" by detrending--- dividing first axis into short segments and adjusting second axis scores so that they have a mean of zero within each segment.

Nonmetric Multidimensional Scaling (NMDS)

Based on ranked (order of more to less similarity) comparisons among samples. Tries to find the best fit between the ranked dissimilarities and distances in an ordination diagram.

Indirect (unguided) approaches

Analyze species data first, then interpret in light of external factors. Informal relation of external factors to ordination results.

Direct (Guided) Approaches

Species weights represent external factors such as succession stage, soil wetness, fire tolerance, elevation, etc. Directly integrates environmental factors into ordination solution.

CCA: Canonical Correspondence Analysis

Ordination scores are constrained to be best-fit linear combination of environmental variables). Integrates ordination with environmental interpretation.

Direct Gradient Analysis

Identify recognizable environmental gradient (elevation, etc.) & order plots along the gradient before sampling. Observe changes in species abundances & community along the gradient.

The Gaussian Model of Community Structure

Species abundances plotted along gradient show a Bell-shaped response curve.