The double exponential model was generally more efficient at describing decay patterns. However, for disks decomposing in the air, it was not significantly better than the single exponential model (Table 2). We attribute the comparatively poor fit of the single exponential model to its lack of sensitivity to the initial relatively rapid (0-2 months) losses of wood mass, which is explicitly modelled in the double exponential model. The decay constants from the single exponential model and the decay constant from the slow decay component of the two-component model were comparable, with the former being slightly higher. For surface and buried disks, the decay constants of the labile portion, k₁, ranged from 0.367 to 23.702 month^-1 and the decay constant of the more refractory components of the wood, k₂, varied from 0.001 to 0.033 month^-1. The proportion of labile material, , varied from 1.7 to 28.3% of total mass (Table 3).

Relative differences of the exponential decay rates among conditions are readily apparent (Fig. 2). Our hypotheses that decomposition would be affected by the species and condition of the CWD were consistent with the collected data. Condition and species both had an effect on the decomposition model parameters. There was a species effect for (P < 0.005), and a condition effect for both k₁ (P < 0.005) and k₂ (P < 0.0001). Avicennia germinans had the largest proportion in the labile pool (0.18 ± 0.02) and L. racemosa had the least (0.10 ± 0.01; Fig. 3a). There were no significant differences between L. racemosa and R. mangle (Fig. 3a). Disks on the sediment surface had higher rates for k₁ (8.51 ± 1.74 month^-1) than buried disks (2.29 ± 0.96 month^-1; Fig. 3b). The latter was also true for the mineralization rates of the refractory pool (k₂). Decay constants (k₂) for surface and buried disks were 0.022 ± 0.002 month^-1 and 0.012 ± 0.001 month^-1, respectively (Fig. 3c).

Go to Table 2

---

Go to Table 3

We found no support for our hypothesis that decomposition rates would vary along the freshwater-to-marine gradient. There was no significant location effect on decay model parameters for the 28-month decomposition experiments on surface and buried disks. Neither species nor sites had any effect on the decay parameters for wood disks decomposing in the air.

Fig. 2 (left) Rates of mass loss from wood disks averaged across all sites. Wt = weight of disk (dry weight in g) at time t and W0 = the initial weight of the disk. Data points indicate observed values; error bars are ± 1 SD. [larger image]

---

Fig. 3 (left) Distribution of the parameter values of the decay models. (a) Boxplots of the fast decaying fraction () by species. (b) Boxplots of decay constant for the fast decaying fraction (k1) averaged by condition. (c) Boxplots of the decay constant for the slow decaying fraction (k2) averaged by condition. [larger image]