Résumé : Inorganic carbon, in the form of allogenic (transported) and pedogenic (soil) carbonates in semi-arid soils, may comprise an important carbon sink. Carbon dioxide, CO2, originating from the atmosphere and exhaled by tree roots into the soil, may be hydrated by soil water within the unsaturated zone (USZ) of semi-arid soils to produce the carbonic acid (H2CO3) solutes HCO3- bicarbonate and H+ Hydrogen ion. This H+ may then dissolve relict soil CaCO3 carbonate (calcite), to release Ca+2 calcium cations and more HCO3- bicarbonate. When conditions allow, one mole of Ca+2 and two moles of HCO3- combine to precipitate one mole of calcite, and to release one mole of CO2: Ca+2 + 2HCO3- --> CaCO3 + CO2 + H2O. However, it has been claimed that such carbonates do not sequester significant amounts of present day atmospheric CO2. The reasons given were that they originate in part from the pre-existing limestone; and that for every mole of calcite precipitated, one mole of CO2 may be liberated to the atmosphere. It was argued that only if the Ca+2 cation is derived from a non-carbonate source can sequestration be assumed. We have tested these assumptions under field conditions at two semi-arid sites in Israel. We found that bicarbonate, originating from root exhalation, is depleted and is incorporated within the USZ as carbonates precipitate. Thus, a net sequestration of atmospheric CO2 does occur under semi-arid forests. Moreover, most of the CO2 liberated in the precipitation reaction may remain in the soil. And Ca+2 in the sediment may also be supplied from sources other than pre-existing calcite. Forestation can therefore augment pedogenic carbonate formation. By extrapolating our data globally, we suggest that worldwide semi-arid forests (existing and to be planted) may sequester 5-20% of the current annual anthropogenic increase of atmospheric carbon dioxide as pedogenic carbonate.