For an ecology project I am analysing how and if the species composition of a habitat type (heathland for example) changes through time. This was done by measuring fixed plots within a habitat type several times in the last 20 years. The database in my project is similar of structure to the Dune dataset from R-package Vegan but larger with more plots and each plot is 5 times present in the database (representing 5 separate years of measurements). Some plots weren’t sampled in each year, so null values are present. I am looking for a way to explore if the composition of this habitat changes over time by means of performing a DCA or an NMDS. However, which method is best suited for my type of analysis? I have read different forums and internet pages on how these two ordination methods work and what their benefits and disadvantageousness are, however am unsure which methods fits better. The two techniques are both used extensively in ecology and what I found was that the large difference between the two methods is that a NMDS can work with different vegetation distant matrices (such as the Bray-Curtis index) and that the DCA uses a fixed distant matrix with the Euclidean distance. (http://ordination.okstate.edu/overview.htm#Contrast_between_DCA_and_NMDS). Which of the two methods is suitable for my project? And what are the benefits of choosing one method over the other?
Your help will be much appreciated!