An interesting piece of ecological detective work from the shores of New England, which came to my attention via this blog post and this op-ed in the Cape Cod Times. Salt marshes on Cape Cod have been suffering local die-back in recent years, in which, for reasons not immediately clear, the cordgrass or marsh hay holding them together (Spartina spp., for those of you keeping score at home) disappears from the banks of channels, creeks, and guzzles, exposing the soil beneath to erosion. This is worrisome, because of the many wonderful things marshes do for us and other creatures: they are nurseries for juvenile fish, they filter water and absorb pollutants, and they hold the coast together against the ocean’s patient battering of waves and storm surges.
The traditional view of marshes was that they were governed by “bottom-up” dynamics—that is, that physics, nutrients, and other non-biological factors determined where marsh plants grew or did not grow. In recent years, a number of widely-read studies have argued that top-down dynamics can have big effects on entire marine ecosystems, in some cases changing their actual physical structure.
Holdredge, Bertness, and Altieri, in a paper published last year in Conservation Biology, propose that something like this is going on in Cape Cod marshes. The culprit behind the cordgrass die-off is not, they believe, any change in nutrients, pollution, or water flow. Rather, it is the action of hungry squareback marsh crabs (Sesarma reticulatum), emerging from burrows at night and aggressively grazing back the marsh’s vegetation. The reason the crabs are eating the grass with impunity, they conjecture, is that the crabs are not being eaten by predators. Holdredge et al. tested their hypotheses using a classical experimental-ecology approach. If it is true that Sesarma crabs are eating the cordgrass, and that predation determines the abundance of crabs, then we would expect several things to be true:
First, areas with more crabs should have less cordgrass. The researchers walked along marsh channels, recording the extent of cordgrass die-back. They also placed small pitfall traps in the ground and counted how many crabs fell into them. As is turns out, die-back is indeed positively correlated with crab density. 
Second, we would expect grass within the reach of the crabs to be grazed more than grass the crabs cannot get to. This hypothesis was tested by transplanting clumps of grass into die-back areas and fencing some of them off to keep out crabs. Again, the data support the hypothesis: fenced-off grass grew normally, while exposed grass clumps were grazed down and clipped off.
Third, if it is true that predators are controlling the abundance of the crabs (which are in turn controlling the extent of the cordgrass), then crabs in places with more marsh die-back should be eaten less frequently. The researchers tied crabs to stakes with fishing line and left them overnight, both on Cape Cod, where die-offs are ocurring, and in Narragansett Bay, RI, where they are not. The lines were long enough that the crabs could reach existing burrows to hide. Neverthless, tethered crabs in Narragansett Bay suffered a much higher rate of predation than those on Cape Cod.
These three lines of evidence make a strong case for the crabs as culprits in cordgrass die-back on Cape Cod. The authors discuss possible reasons for the decreased predation on crabs, and settle quickly on the removal of their predators by humans. In particular they mention the tautog (Tautoga onitus). This fish eats crabs and experienced a fishing-driven population decline in the mid to late 1990’s, coinciding with accelerating marsh die back. The experiments conducted here don’t directly address the link from humans to crabs via tautog, though, and more work is necessary before pointing the finger at fishing as the definitive cause. I was also left wishing I knew more about the spatial pattern and scale of these die-backs. The researchers used aerial photos of the marsh to measure die-back through time, but didn’t show them in the paper.
On the whole, however, these guys deserve kudos for a straightforward and elegant experimental approach to determining the cause of marsh die-back, logically and clearly testing hypotheses with direct biological interpretations. Though they answer the central question (are crabs eating the cordgrass?) my interest is piqued, I’m left with a bunch of other questions I want answered as well. Are marsh die-backs are occurring all along the East Coast, or just on Cape Cod? Are they contiguous or patchy? Are they present all the time, or do they get grazed down and regrow? What is the distribution of the crab’s predators like? Hopefully, we’ll see answers to some of these soon…
For some more photos and information, check out the National Park Service’s page on crab-driven vegetation losses.
HOLDREDGE, C., BERTNESS, M., & ALTIERI, A. (2009). Role of Crab Herbivory in Die-Off of New England Salt Marshes Conservation Biology, 23 (3), 672-679 DOI: 10.1111/j.1523-1739.2008.01137.x
FOOTNOTE OF STATISTICAL GRIPES
I was going to put these in the main text, but decided that would be a little pedantic and overbearing, especially since they don’t change any of the paper’s conclusions. If you don’t care about GLMs or log-transformed count data, feel free to stop reading.
In the regression reproduced above, the authors model % cordgrass blades grazed as a linear function of the number of crabs per trap, getting a decent fit. The problem is that this model is not the right one for a response variable that is a proportion or percentage (i.e., a probability). To illustrate, plug 9 crabs/trap into the regression equation: it predicts the crabs will eat 110% of the grass blades, an impossibility. A better choice of regression model would be a logistic-link GLM with a binomial error distribution.
Also: for a couple of the ANOVAs, count data were log-transformed to meet assumptions of normality. For a discussion of why this isn’t so great, see O’Hara and Kotze 2010: Do not log-transform count data.