Results and Discussion
After 60 days on the forest floor, we found that just over half of the bags of both tea types had been chewed through by macrofauna: 31 [52%] of the rooibos tea and 30 bags [50%] of the green tea. In our model the only significant terms were the type of tea (rooibos or green) and the interaction between the state of the bags (opened or unopened) and the tea type (see Table 1 for model outputs). There was no significant effect of time on the decomposition rate. This meant that generally green tea decomposed more rapidly than the rooibos tea. However, where the bags were opened, rooibos tea showed a greater response than in green tea, where the open and closed bags had very similar decay slopes (Table 1, Fig 1). This serves to illustrate that the major problem with the TBI in the tropics is with the rooibos tea bags, which were commonly attacked by macrofauna (probably predominantly termites) and the contents often completely removed. In our study, more than half of the rooibos tea bags were chewed open, and more than 10% were emptied. A similar proportion of green tea bags were chewed open, but none of them were emptied. It seems that macrofauna, at least in this context, prefer rooibos tea to green tea.
It is possible to estimate the two tea bag parameters in a subset of bags without holes (i.e. estimate from the undamaged bags, estimating microbial and mesofaunal decomposition only); the values here were S = 0.242 and k = 0.053. This k value from Bornean rainforst is higher than any values in the original teabag paper, including the tropical rain forest site in Panama (S = 0.06, k = 0.0392; Keuskamp et al. 2013 site 13, Table S1). This is, however, a measure of microbial decay only, as the macrofauna were specifically excluded. Keuskamp et al. 2013 could not calculate the standard deviation for their Panama site, ‘due to overdispersion’ of data. It is possible that holded teabags, empty of tea, may have contributed to this overdispersion.
However, it is difficult to assess projects that have used the TBI in the exact way it was recommended, as the method specifically excludes bags that have been chewed. Therefore they are self-selected for microbial and mesofauna decay and explicitly exlude any macrofauna decay.
If the aim is to quantify microbial decomposition it is therefore possible if only undamaged bags are used. However, sufficient bags need to be buried so that one can afford to ignore damaged ones. We suggest that given half of the bags were damaged at least twice as many are needed (>40 in this case for every period, giving a total of 120). Furthermore, the incubation time (ie the time the tea bags are left out) should be reduced to 60 days, as by 90 days many of the teabags would be likely to be chewed by termites (note we recorded 50%).
The TBI is a valuable idea but our results show it can only give an accurate idea of total decomposition rates in the few locations where invertebrate decomposition is negligible and there is overwhelmingly microbial decomposition (e.g. boreal areas). It works as a measure of microbial decomposition alone in all locations provided the bags are not damaged. Unfortunately, the TBI does not incorporate a measure for including the effects of larger decomposers.
We see no easy way to incorporate the contribution of macrofauna (particularly termites) to decomposition processes using the TBI, without also putting tea bags into macroinvertebrate litter bags made with mesh strong enough to withstand termite attack (such as stainless steel mesh or nylon; (Yates & Grace 1999; Lenz et al. 2012) . Other papers that use small and large mesh sizes to exclude or include macrofauna may suffer the same issue (eg (Handa et al. 2014)) if the small sized mesh material is not strong enough, then some samples may (are likely) to be attacked by termites.
This is because of the very high stochasticity that termites, in particular, contribute to the process. The problem becomes one of modelling two different simultaneous processes: one an exponential decay process and the other potentially a binomial process (either the bags are opened or they are not). There are other more costly methods, in both time and expertise, that produce robust total decomposition results and assess the contribution of all decomposer organisms, such as using leaf litter and wood decomposition assays and mesh to include/exclude invertebrates (e.g. (Smith et al. 2009; Davies et al.2013). However, these methods are not in the spirit of the TBI (see Fig 1), which was designed to produce a quick, simple estimate of two meaningful parameters.
In summary, the TBI can work well to assess the microbial contribution to decomposition where no (or few) bags are opened by macrofauna, such as in boreal locations, but the method risks seriously underestimating total (microbial + macrofauna) decomposition in most habitats. (Wallet al. 2008) present a global map showing the areas where macrofauna (not always termites) are important in dead plant decomposition processes, althoughthis figure significantly downplays the effect of macrofauna (especially termites) in savannas, grasslands and other drier areas (Bignell & Eggleton 2000). Wall et al. 2008 used moth balls (active ingredient naphthalene) to exclude invertebrates. As with any insecticidal active ingredient, it has variable effects on different species of insect…
Termites appear to manufacture naphthalene (possibly secreted by themselves, but more likely via microbes(Chen et al. 1998)). So the treatement would have a reduced effect on termites.
While Keuskamp et al. (2013) do specifically state that macrofauna are excluded they make no other comment about the errors that this might introduce. We believe that any method that ignores the contribution of macrofauna to plant decomposition cannot be recommended as a global method. It may still be useful for estimating purely-microbial effects but it cannot yield measures that include all decay agents.
Data accessibility: the data for this paper has been deposited in DRYAD, doi: TBA
Table 1