From our results, it is clear that there are two distinct phases in mass loss and nutrient dynamics that are likely to follow the deposition of CWD in mangrove forests. The first 2-month period is characterized by a rapid loss of mass and a net input of the nutrients N and P into the system as newly dead wood lost its most labile components. Mass loss from the downed wood would then slow down as bacteria and fungi decompose the more refractory components of the downed CWD and, during the next few years, the decomposing wood would serve as a minor sink for N and P in the forest. The relative importance of the initial phase would be dependent on the species composition of the forest, as we found that the relative proportion of labile to refractory components varied by species. We also found that the fate of the CWD (whether standing dead, lying on the sediment surface or buried in sediment) influenced the dynamics of wood decomposition and therefore the role that the CWD would play in ecosystem dynamics following a catastrophic deposition of CWD following a hurricane.

Unlike most long-term wood decomposition studies (e.g. Lambert et al. 1980; Spies et al. 1988), we used wood disks cut from living stems, which allowed us to do a more controlled comparison study in a shorter period of time. Thus, we were able to estimate the effects of species, location and condition on wood decomposition, using a double exponential decay model to describe the two separate phases of mass loss. The observed length of the period of rapid mass loss is comparable with reported initial rapid losses of nutrients from wood in deciduous forest trees (France et al. 1997). We suspect that the early loss of mass in wood disks was attributed to abiotic leaching, as shown by others (Steinke et al. 1983; France et al. 1997). It was lowest for disks decomposing in the air and highest for disks decomposing on the soil surface, probably due to differences in water availability, as surface disks were subjected to regular tidal inundation at most of our sites, while disks decomposing in the air were dependent solely on the atmosphere for moisture. Little leaching would occur in air disks during the dry season, consistent with other studies that have suggested that leaching is enhanced in areas where there is strong flushing (Twilley et al. 1986; Currie & Aber 1997).

We suspect that difference in early mass loss between buried disks and those on the soil surface may have been a result of differences in both tidal flushing and oxygen availability for microbial decomposition. In most areas of Shark River, the soil surface is subjected to regular tidal flushing; however, areas beneath the soil surface may not be completely flushed after each tide. Tidal water trapped in the interstitial spaces of the soil can diminish the concentration gradient between water in disks and water of the surrounding environment, decreasing the advection potential of watersoluble components from buried disks. As microbial activity may have facilitated the initial rapid loss, anoxic conditions experienced by the buried disks could have depressed rates of mineralization. Other studies have demonstrated similar effects of oxygen availability on microbial degradation of tissue and leachate from Rhizophora mangle (Benner & Hodson 1985; Benner et al. 1986). However, Kristensen et al. (1995) suggested that fresh organic material is degraded at similar rates in the presence or absence of oxygen and that oxygen becomes more critical in the hydrolysis of some structurally complex and aromatic compounds.

We mimicked the composition of CWD deposited in the forest by extreme events, like hurricanes, by beginning our study with freshly harvested wood. The decay constant for the more refractory component of wood (k₂) in this study is analogous to the decay constants obtained from the single exponential model applied in other studies. Many studies on wood decomposition (Lambert et al. 1980; Lambert & Cromack 1982; Robertson & Daniel 1989; Polit & Brown 1996) use aged instead of freshly dead wood and do not therefore describe initial losses from the most labile components.

Due to the increased surface area: volume ratio of the disks used in this study, we estimated higher decay rates for mangrove wood (k₂) than reported for other types of wood in temperate forests. For instance, annual mass loss calculated for mangrove disks decomposing on the soil surface (0.276 year^-1) occurred at rates three times higher than branches of Quercus prinus (0.092 year^-1; Abbott & Crossley 1982) of similar diameter. However, mangrove wood decomposing in the air in our study, where moisture and presumably microbial activity was very low, had similar rates of mass loss (0.048 year^-1) to Q. prinus CWD decomposing in xeric conditions (0.0377 year^-1; Abbott & Crossley 1982). The decay constant for R. mangle (0.288 year^-1) obtained in this study was also higher than that reported for large Rhizophora boles in Australia (0.083 year^-1), but similar to the decay constant obtained from branches (0.276 year^-1) in an Australian mangrove forest (Robertson & Daniel 1989). Unlike the decomposition of fine litter (e.g. leaves, flowers, stipules), the breakdown of woody tissue is largely influenced by the size of the CWD. Research has shown that, in general, decay coefficients decrease with increasing diameter of the bole (Christensen 1977; Abbott & Crossley 1982; Frangi et al. 1997).

Location of the wood within the forest at each site probably accounted for the overall pattern of decay throughout our study. Disks on the soil surface, which were exposed to tidal activity, precipitation and intermittent submergence in water, appeared to offer the best environment for microbial respiration. Disks in this treatment decomposed almost twice as fast as buried disks and five times as fast as disks decomposing in the air. When moisture in the substrate is below 30%, water is unavailable to microbial decomposers (Harmon et al. 1986). At each collection time, disks hanging in the air were always considerably dryer than disks from the other two conditions. In addition, after 7 months of decomposition, fungal mycelia were only present in the surface disks (L. M. Romero, T. J. Smith III & J. W. Fourqurean, personal observation).

We found that leaching and microbial decomposition prevailed during the initial 13 months for wood decomposing on the soil surface. However, by the second year of decomposition, disks lying on the soil surface in the more marine habitats had been colonized by shipworms. In a study of mangrove wood decomposition in Belize, shipworms were found to be the major decomposers (Kohlmeyer et al. 1995). Robertson & Daniel (1989) also found that shipworms were responsible for most of the weight loss from wood.