Effects of haying on salt-marsh surface invertebrates.

Abstract

Marsh plants may affect the structure of epibenthic communities in a variety of ways. They might reduce temperature stress on organisms by shading, provide refuge from predators, increase the rate of organic matter accumulation at the sediment surface, and serve as a food source. Thus, salt marsh plants might control the spatial distributions of epibenthic invertebrates within the marsh. An ongoing “experiment” on the impact of the periodic removal of plants has been carried out by New England farmers who have been haying some salt marshes for over 300 years. Earlier research by the Plum Island Estuary Long-Term Ecological Research project (PIE-LTER) has shown that haying causes a short-term increase in the growth of benthic algae (1), causes a shift in the diet of some marsh invertebrates to a greater percentage of algae (2), and results in a change in plant species composition (PIE-LTER, unpubl. results). To determine the shortand long-term effects of haying on the epibenthic community and to identify the principal variables controlling the distribution of epibenthic invertebrates, we carried out two sets of experiments: a marsh-grass-removal experiment to mimic the effects of haying and a survey of invertebrates in areas of the marsh subjected to different haying schedules. Our work was conducted in the Plum Island Sound estuary in Rowley, Massachusetts, from June to August 2002. For our marsh-grass-removal experiment, we clipped and removed the vegetation from three randomly selected circular sites of 3-m radius along a 50-m transect in a marsh that is hayed roughly every other year. Before clipping we counted organisms in four 0.25-m sampling areas, located 1.5 m from the center of the circle and equidistant from each other, within each of the three sites. We sampled the same four areas within each site immediately before clipping, and 1 and 4 days after clipping. Differences in abundances of organisms before and after clipping were evaluated by the Kruskal-Wallis test with time of sampling as treatment. Because differences among circular hayed sites were not significantly different, we pooled replicates from each site before our analysis. For our surveys of invertebrates in areas subjected to different intensities of haying, we examined randomly selected sites in recently hayed and reference marshes along 12 previously established transects for which GPS elevation data were available. We sampled 40 sites between 1 and 3 July, and 40 more between 16 and 18 July. At each site, we visually estimated percent cover of different plant species. A count of organisms was made after clipping all vegetation from 0.25-m plots. A 177-cm plant sample was collected, live biomass was separated from standing dead biomass and weighed separately to determine above-ground biomass. Commonly encountered plants included Spartina patens, Spartina alterniflora, Distichlis spicata, Salicornia europaea, and Juncus gerardi. To evaluate the effect of plant cover on predation rate, we tethered amphipods, Orchestia grillus, to thin-diameter line and deployed 5 per site at most sites for 24 h before clipping. We used stepwise multiple regression to describe the relationship between measured environmental variables and abundance of epifauna for the 16–18 July sampling period (criteria P 0.1 to accept, P 0.25 to reject). Environmental variables used in stepwise multiple regression were haying frequency, elevation, temperature, humidity, dried live plant biomass, dried standing dead, total dried weight, percent live, percent dead, percent bare, percent S. alterniflora, percent S. patens, percent D. spicata, percent S. europaea, percent Atriplex patula, percent Triglochin maritima, percent Iva frutescens, proportion tethered live amphipods, and proportion eaten tethered amphipods. Collinearity among environmental variables was assessed by examining the variance inflation factor for each included variable. All variance inflation factors were less than 2, which indicated that collinearity among independent variables did not significantly affect our results. A few sampling sites were omitted from the regression analysis due to missing data. Analyses were performed using SPSS software.