Comment on Haddad: experimental evidence does not support the habitat amount hypothesis?

Submitted by editor on 6 February 2017.

RE: Haddad et al. 2017. Experimental evidence does not support the Habitat Amount Hypothesis. Ecography 40:48-55.

By Lenore Fahrig


In the Acknowledgements section of Haddad et al. (2017) the authors thank me for my comments on an earlier version of the manuscript. Following are those comments, for the record:


Hi Nick,


My student Margaret, whom you met at the IALE meeting, showed my lab your manuscript on testing the habitat amount hypothesis. We had an interesting lab discussion. Thank you very much for letting us see it.


There is one point I would like to clarify. The habitat amount hypothesis does *not* predict the same intercept for fragments as for sample areas in continuous habitat. It predicts the same slopes but not the same intercepts. In fact, if the habitat amount hypothesis is correct, the intercept has to be lower for the fragments. The hypothesis asserts that species richness declines with habitat amount due to the sample area effect. A fragmented landscape has less habitat than a continuous landscape, so it will have fewer species in total. Over time, following habitat loss, even more species will be lost due to relaxation (see text in Fahrig (2013) on p. 1650-1651), but this will occur in proportion to patch area. This will shift the line down, lowering the intercept. However, the slope remains the same because there is no island effect.


If I understand your tests correctly, it is your interaction terms that test the habitat amount hypothesis, i.e. test for differences in slopes. If the hypothesis is correct, the interaction effect should be zero (more or less). You found one significant negative interaction (SRS test 1), one non-significant negative interaction (moss) and one essentially zero interaction (SRS test 2). So you may have some evidence against the hypothesis (SRS 1 and moss), but not as strong as you intimate. I am curious to hear your explanation for why the same data (SRS) go against the hypothesis in test 1 but are consistent with it in test 2.


I don't know whether this has anything to do with the answer, but one thing about the SRS tests is that, effectively, the continuous landscape is smaller than the fragmented landscape. I'm honestly not sure what this means for your tests, if anything. But it reminds me of when Ilkka Hanski used to illustrate increasing patch isolation - he would show isolated patches distributed over a larger area than non-isolated patches. In other words, the landscape would expand with increasing isolation. I've always found that hard to think about, as it adds another confounded (in both senses of the word) variable - landscape size.


In any case, I hope you will remove the predictions about the intercept as I think that might confuse things in future tests of the hypothesis. Tests should focus on comparing the slopes. Just so you know, I am not a reviewer on your submitted paper.


All the best, Lenore.