It looks that you have an over-fitted model. The number of constraints (environmental variables, vector arrows) is seven, and I only found eight points (that I assume to be your sampling locations). There is nothing "canonical" in such over-fitted model, but they are just ordinary non-canonical correspondence analysis plots with fitted arrows. It is just like an ordinary regression: if you have seven independent (explanatory) variables and eight observations, you can get a perfect fit but nothing real.

Further, you should remember that ordination is nothing but a fancy way of making graphs. Do not look at the numbers: they only confuse you. Only look at the graph.

I think it is best to get an instructor who tells you how to use these methods. The second best method is to get a book.

Good luck.

Anyway, what does the arrow direction mean? i.e pH lies in upper left quadrant while most of enviromental factor in lower right, is it means pH has different trend data with the other?

Thank you, Dr. Dasapta Erwin Irawan, hatur nuhun! :)

hi.. I have some papers on this but in a different context.

get a book by Greenacre. Much of the interpretation depends on the position and angle of the plotted point and distance from the origin of the map. You can also plot in the third dimension.

good luck. Brad

If you are interested into the rules of CCA plot interpretation with regards to aquatic ecology (cf. your study about  phytoplancton abundance), you should read the following paper :

Canonical correspondence analysis and related multivariate methods in aquatic ecology

by Cajo ter Braak and Piet Verdonschot

Aquatic Sciences 57/3, 1995, pp. 255-289.

DOI: 10.1007/BF00877430

available at : http://link.springer.com/article/10.1007%2FBF00877430#page-1

Oldy but goldy... Cajo ter Braak is the father of CCA.

Complementary to Bradley's answer, please find enclosed a short introduction by Michael Greenacre devoted to the application of CCA to social sciences, which provides rules of CCA plot interpretation and some useful references.

Hi Akbar,

the interpretation depends on the distance and angle from the map origin, determined for each explanatory variable, that is, the dimensionality of the variables. Apparently the water temperature, alkalinity and the presence of P. clatratum are the variables that most contribute to differentiate your ST3 sampling points from ST1 and ST2. But you need to look at the numbers and not just graphics, because graphics are nice, but numbers are better. See eigen values, the canonical scores, for example, because is different the distance and angle from an axis or other axis, the axis 1 generally corresponds to the lowest gradient axis in the correspondence analysis.

Good luck!

It looks that you have an over-fitted model. The number of constraints (environmental variables, vector arrows) is seven, and I only found eight points (that I assume to be your sampling locations). There is nothing "canonical" in such over-fitted model, but they are just ordinary non-canonical correspondence analysis plots with fitted arrows. It is just like an ordinary regression: if you have seven independent (explanatory) variables and eight observations, you can get a perfect fit but nothing real.

Further, you should remember that ordination is nothing but a fancy way of making graphs. Do not look at the numbers: they only confuse you. Only look at the graph.

I think it is best to get an instructor who tells you how to use these methods. The second best method is to get a book.

Good luck.

Thank you for all of the answers :)