9b: Detrended Correspondence Analysis (DCA)

Practice Task

Practice Task

Work through these questions after reading the Detrended Correspondence Analysis chapter. The single most useful thing a DCA still tells us is the gradient length in \(\beta\)-diversity standard-deviation units, so most of this task is built around extracting and using that number.

1. Run a DCA on the Doubs fish data with decorana() (remember to drop the empty site first). Extract the axis lengths with summary() or the $evals/axis-length output, and report the length of the first DCA axis in SD units.

2. Use the rule of thumb that a first-axis gradient length below ca. 3 SD favours linear methods (PCA, RDA) while a length above ca. 4 SD favours unimodal methods (CA, CCA). What does the Doubs gradient length recommend? Does that agree with the unimodal species responses you saw in the Correspondence Analysis chapter?

3. Plot the DCA site ordination beside the CA site ordination of the same data. Has detrending removed the arch? Has the spacing of the sites along axis 1 changed (the rescaling step)?

4. Apply DCA to a second dataset with a different gradient length, for example the dune meadow data or the bird communities along the elevation gradient in Yushan Mountain, Taiwan. Report its gradient length and state which family of ordination methods it recommends.

5. Given the well-known criticisms of DCA (instability of the detrending-by-segments algorithm), would you present a DCA ordination diagram in a publication, or use DCA only as a diagnostic for gradient length and then ordinate with another method? Justify your answer.

Assessment Criteria

This Task is not formally assessed.