No abstract

The identification and understanding of shifts in resource availability and community structure caused by a variety of anthropogenic perturbations is essential to future rehabilitation efforts and recovery of essential ecosystem processes. We focus here on the contribution of nutrient enrichment to the overgrazing of macrophyte‐dominated systems, which has often been linked to predator and competition release of urchins due to overfishing. However, the contribution of nutrient loading to the progression and persistence of this phenomenon should also be considered. In an effort to understand the contribution of nutrient loading to the overgrazing phenomenon and associated simplified community in a Caribbean seagrass bed, a detailed isotopic (δ13C, δ15N) assessment of the food web structure was conducted at both a nutrient‐enriched and a control site. The general pattern at the enriched site indicated significantly lighter δ13C signatures (i.e., phytoplankton influenced) for non‐generalist primary consumers (i.e., specialist grazers, subsurface deposit feeders) and for the sediment organic material (SOM) when compared to the control site. The contribution of phytoplankton and associated particulate organic material to the SOM was also different, 7% vs. 44% at the enriched and control site, respectively. The loss of the autochthonous seagrass detritus pool, in the wake of high densities of generalist urchins (>66 000 individuals/ha) and low overall consumer diversity, appears to have been partially replaced by opportunistic “alternatives,” epiphytes but mainly phytoplankton, that benefit directly from elevated input of anthropogenic, allochthonous nutrients. The availability of such alternative, allochthonous resources to generalist urchins could potentially allow for the persistence of simplified seagrass communities. Here, elevated densities of urchins enable a persistent suppression of autochthonous benthic‐macrophyte production through grazing and the consumption of newly recruited competitors and predators.

Restoration has become an integral part of coastal management as a result of seagrass habitat loss. We studied restoration of the seagrass (Halodule wrightii) near Tampa Bay, Florida. Experimental plots were established in June 2002 using four planting methods: three manually planted and one mechanically transplanted by boat. Seagrass cover was recorded at high resolution (meter scale) annually through July 2005. Natural seagrass beds were concurrently examined as reference sites. We also evaluated the suitability of a commonly used protocol (BraunBlanquet scores, BB) for comparing the development of seagrass cover using the planting methods and quantifying spatial patterns of cover over time. Results show that BB scores mirrored conventional measures of seagrass characteristics (i.e., shoot counts and above-and belowground biomass) well when BB scores were either low or very high. However, more caution may be required at intermediate cover scores as judged by comparison of BB scores with direct measurement of seagrass abundance. Significant differences in seagrass cover were detected among planting methods and over time (2002)(2003)(2004)(2005), with manual planting of rubber band units resulting in the highest cover. In contrast, the peat pot and mechanical planting methods developed very low cover. Recovery rates calculated from development of seagrass spatial cover were less than those reported for natural expansion. Importantly, time to baseline recovery may be substantially greater than 3 years and beyond standard monitoring timelines. Prolonged recovery suggests that the rate of service returns, critical for estimating compensatory restoration goals under habitat equivalency analysis, may be severely underestimated.

Table S1. ∆ AIC comparison of models for the Long-term Atoll Monitoring Program (LAMP) density in numbers using the negative binomial. Z and Y denote the model terms zone and year, respectively, with YZ denoting an interaction term. The relationship of the variance to the mean may be modeled either as a linear (σ 2 =kµ, called binomial 1), or quadratic (σ 2 =µ(1+µ/k), called binomial 2), where the dispersion parameter (k) is an estimated parameter in the model. If models including year have substantially lower ∆ AIC values than the models without year, then there is evidence of a trend over time. NA means the model did not converge.