, 2011), we found that continental H:DBH models only poorly explained the variance observed at our sites, notably in old-growth secondary forests (Fig. 2). We highlight here that the continental model proposed by Feldpausch et al. (2012) was originally developed for unmanaged forests and should be used with caution in secondary forests. For instance, trees growing in logged forests

in the Amazon were found to be shorter with larger crowns (Nogueira this website et al., 2008). This phenomenon might explain our results in secondary forests, where large trees had much smaller heights than expected. We showed that H:DBH model can be fitted with only a small fraction of the forest stand (Fig. 1), as long as the sample is equally distributed along the actual DBH distribution. In a first attempt, trees were randomly chosen, embedding the model to converge in most cases. This result is encouraging and shows that integrating tree height into carbon stock assessment would not require a lot of additional field work. Using the best predictive model (Chave.H), we found an average value of 378 Mg ha−1 in unmanaged and 316 Mg ha−1 in secondary forests. These values are lower than those previously reported for Dipterocarp forests (Paoli et al., 2008 and Slik et

al., 2010). Both studies used Chave’s equation based on DBH and WSG, with AGB stocks ranging from 457 to 606 Mg ha−1. Our study shows that these BGB324 in vivo figures are likely to be overestimated by at least 10%. Lower AGB stock in secondary forests was mainly explained by the absence of very large trees (DBH > 100 cm) that usually encompass a large fraction of AGB in tropical forests (Paoli et al., 2008 and Rutishauser et al., 2010). However, these figures remained relatively high compared cAMP to forests recovering from conventional logging that range between 150 and 300 Mg ha−1 (Berry et al., 2010 and Saner et al., 2012). This strengthens our initial postulate of considering these plots as mature secondary forests and constitutes

one of the reasons we decided not to use allometric models developed in logged-over forests of Sumatra (Ketterings et al., 2001) or Borneo (Kenzo et al., 2009a). At one site (BT_SF), no logging activity was carried out over the last 40 years, while none was carried at the second site (BM_SF). Such systematic assessment should be performed in other forest types and ecoregions across Indonesia in order to determine the validity and the choice of the appropriate allometric model. The choice of a particular allometric model will remain mainly driven by data availability. Due to time and costs constraints, most forest inventories are restricted to DBH measurements and DBH-models will remain widely used. However, accounting for tree heights can reduce uncertainties surrounding biomass estimates in Dipterocarp forests.