Rainfall interception was quantified and modeled for a Lowland Evergreen Rain Forest (LERF) and two Heath Forests (HF) of contrasting stature during a one-year period at a remote site in Central Kalimantan, Indonesia. Throughfall was measured using a roving gauge approach using 18-20 gauges per forest type. Throughfall was 82.8% of incident rainfall in the LERF vs. 89.1% in tall HF and 76.7% in stunted HF. Corresponding stemflow fractions were 0.8%, 1.3% and 2.0% of rainfall, respectively. Interception losses derived for the LERF (16.4%) and the tall HF (9.6%) were close to findings obtained for similar forest types elsewhere but the high interception loss for the stunted HF (21.3%) was unexpected. On the basis of canopy structural characteristics and wet canopy evaporation modeling it is concluded that throughfall in the stunted HF was underestimated and that the specific nature of the HF required a more intensive sampling arrangement. Throughfall sampling schemes in tropical forests, associated errors and hydrological implications are discussed. © 2007 Elsevier B.V. All rights reserved.