Deforestation of tropical forests for the establishment of grass monoculture for livestock production is responsible for about 30 % of CO2 emissions. This issue is partic larly severe in degraded pastures because degraded soils favor CO2 flow to the soil surface. Silvopastoral systems could reduce CO2 emissions, notably by using live fences. Here, we hypothesized that live fences of Gliricidia sepium in livestock systems should reduce variations in environmental relative humidity and soil temperature and, in turn, reduce soil CO2 emissions. Here, we studied two livestock systems: (1) grass monoculture of Brachiaria decumbens with live fences of G. sepium and (2) grass monoculture of B. decumbens without live fences. We measured soil CO2 seasonal emissions at different times of the day, soil temperature, and environmental
relative humidity. Nine 600-m2 plots were established in each system. All variables were measured over four 6-h period during a 24-h period, twice a month from April to September. Our results show that soil CO2 emissions showed less variability with G. septum live fences than without live fences. This lower variability is explained by the creation of a microclimate with a higher and more stable environmental relative humidity, provided by the shade of trees. Results also show, however, that global soil CO2 emissions did not differ between the two systems, with and without live fence. Moreover, soil CO2 emissions varied according to season, as shown by 1.082 g CO2m−2 h−1 in thewet season versus 0.871 g CO2 m−2 h−1 in the dry season. Soil CO2 emissions varied also according to sampling time, as shown by 1.116 g CO2 m−2 h−1 in the night versus 0.960 CO 2 m−2 h−1 in the morning.