There is an urgent need to estimate the biomass of tropical seasonal forests in central Indochina, which have suffered from deforestation and degradation. However, allometric equations specific to such forests are limited. In this study, we destructively sampled 28 trees in a deciduous forest in Kratie Province, Cambodia, and developed new allometric equations for estimating the tree-level biomass of aboveground woody parts (trunk + branches), leaves, total aboveground parts (trunk + branches + leaves), and that of belowground parts (BGB). The sampling of belowground parts is usually very laborious and time-consuming, and entails inevitable root loss during excavation. Thus, it is difficult to accurately quantify tree-level BGB, especially for large trees. We used a new sampling method (called the mound method) to reduce the sampling loss of BGB. The percentage of BGB sampling loss (ratio of sampling loss estimated by the mound method to total BGB) averaged 28.9%. Our models estimated the stand-level biomass of a deciduous dipterocarp forest in Kratie to be 81.9 ± 45.1 Mg ha^-1. The allometric equations and tree-level biomass data presented here will support activities related to REDD+.