Few results are available about carbon balance of tropical terrestrial ecosystems, either for natural or planted areas. Few experiments use chronosequences (time-series) that allow validation of continuous carbon flow measurements through variations of carbon stocks. Chronosequences also take into account the effect of age age on the productivity of ecosystems. Among 110 million hectares of perennial tropical plantations, Coconut represent 11.5. Coconut is a rapid growing and high-yielding (coprah) monocot. Its growth is continuous, notably in areas where climate conditions are fairly constant during the year. Moreover, coconut oil can be used as fuel in diesel engines. These traits make it a good candidate to the "Clean Development Mechanism" of the "Kyoto Protocol". Our goal is: - Measure carbon stocks (biomass, soil cores) and their variations (aerial growth, rhizotron, soil cores, litter fall and litter decomposition) along a chronosequence of coconut aged till 20 years-old (one cultivar, one density, 5 ages), - Measure fluxes (eddy-correlation), climate, water balance (TDR, sapflow), soil respiration (portable chamber), rhizospheric respiration (trenched-plots), - Compare the 2 previous methods, - Model gas exchanges and growth as a function of climate and water constraints. First results cover 6 months during the transition from dry to rainy season 2001-2002. Coconuts are 20 years old and placed in sub-optimal conditions. LAI is 2.7. They are rooted 2.5 m in a very fertile soil with a high water storage content. H+E account for more than 75% of net radiation (Heat storage not included), indicating that energy balance closure is correct.. Sapflow (T) represents ca. 87% of E after calibration through a 2-level eddy correlation set. Ecosystem Respiration (Re) computed from (Fc) filtered for highly turbulent nights (U*>0.4) reach +12 g C m-2/day and Gross Ecosystem Productivity (GEP) -15 g C m-2/day. Individual carbon fluxes are thus very important. However, da