Land use changes in the savannas of the Orinoco lowlands have resulted in a mosaic of vegetation. To elucidate how these changes have affected carbon exchanges with the atmosphere, we measured CO2 fluxes by eddy covariance and soil CO2 efflux systems along a disturbance gradient beginning with a cultivated tall-grass Andropogon field (S1) and extending over three savanna sites with increasing woody cover growing above native herbaceous vegetation. The savanna sites included a herbaceous savanna (S2), a tree savanna (S3) and a woodland savanna (S4). During the wet season, maximum diurnal net ecosystem exchange (NEE) over the S1-S4 sites was 6.6-9.3, 6.6-7.9, 10.6-11.3 and 9.3-10.6 micromol m(-2) s(-1), respectively. The rate of CO2 uptake over S1 was lower than that for C4 grasses elsewhere because of pasture degradation. Soil respiration and temperature were exponentially related when soil water content (theta) was above 0.083 m(3) m(-3); however, soil respiration declined markedly as theta decreased from 0.083-0.090 to 0.033-0.056 m(3) m(-3). There were bursts of CO2 emission when dry soils were rewetted by rainfall. During the wet season, all sites constituted carbon sinks with maximum net daily ecosystem production (NEP) of 2.1, 1.7, 2.1 and 2.1 g C m(-2) day(-1), respectively. During the dry season, the savanna sites (S2-S4) became carbon sources with maximum emission fluxes of -0.5, -1.4 and -1.6 g C m(-2) day(-1), respectively, whereas the tall-grass field (S1) remained a carbon sink with a maximum NEP of 0.3 g C m(-2) day(-1) at the end of the season. For all measurement periods, annual NEP of sites S1-S4 was 366, 6, 116 and 139 g C m(-2), respectively. Comparisons of carbon source/sink dynamics across a wide range of savannas indicate that savanna carbon budgets can change in sign and magnitude. On an annual basis, gross primary production over the S1-S4 stands was 797, 803, 136 and 1230 g C m(-2), respectively. Net primary productivity (NPP) of the S1-S4 stands, calculated from eddy covariance measurements as the daily sum of NEE and day and night heterotrophic respiration was 498, 169, 181 and 402 g C m-2 year-1, respectively. These values were slightly higher than NPP based on harvest measurements (432, 162, 176 and 386 g C m(-2) year(-1), respectively), presumably because fine roots were incompletely harvested. Soil water content limited carbon uptake at all sites, and water-use efficiency (WUE) was related to rainfall dynamics. During the dry season, all sites except the cultivated tall-grass Andropogon field (S1) had a negative WUE. Although our results are specific to the Orinoco vegetational mosaic, the effects of land-use practices on the controls and physiological functions of the studied ecosystems may be generalized to other savannas.
