because it seems to depend on how much carbon is sequestered in the soil in either natural or cultivated ground:
Options to mitigate CO2 emissions from agriculture include reducing emissions from present sources, and creating and strengthening carbon sinks. Options for increasing the role of agricultural land as a sink for CO2 include carbon storage in managed soils and carbon sequestration after reversion of surplus farm lands to natural ecosystems. However, soil carbon sequestration has a finite capacity over a period of 50-100 years, as new equilibrium levels of soil organic matter are established. Efforts to increase soil carbon levels have additional benefits in terms of improving the productivity and sustainability of agricultural production systems. Soils of croplands taken out of production in permanent set-asides and allowed to revert to native vegetation eventually could reach carbon levels comparable to their precultivation condition. Considering the 640 Mha of land currently under cultivation in the United States, Canada, the former Soviet Union, Europe, Australia, and Argentina, and assuming recovery of the soil carbon originally lost to cultivation, a permanent set-aside of 15% of the land area could sequester 1.5-3 Gt C (over 50-100 years).
A large-scale reversion or afforestation of agricultural land is only possible if adequate supplies of food, fiber, and energy can be obtained from the remaining area. This is currently possible in the European Union and United States through intensive farming systems. However, if farming intensity changes because of environmental concerns or changes in policy, this mitigation option may no longer be available.
Currently, only half of the conversion of tropical forests to agriculture contributes to an increase in productive cropland. The only way to break out of this cycle is through more sustainable use, improved productivity of existing farmland, and better protection of native ecosystems. These practices could help reduce agricultural expansion (hence deforestation) in humid zones, especially in Latin America and Africa.
Management practices to increase soil carbon stocks include reduced tillage, crop residue return, perennial crops (including agroforestry), and reduced bare fallow frequency. However, there are economic, educational, and sociological constraints to improved soil management in much of the tropics. Many tropical farmers cannot afford or have limited access to purchased inputs such as fertilizer and herbicides. Crop residues are often needed for livestock feed, fuel, or other household uses, which reduces carbon inputs to soil. To the extent that improved management is based on significantly increased fossil fuel consumption, benefits for CO2 mitigation will be decreased.
http://www.gcrio.org/ipcc/techrepI/agriculture.htmlThat was based on figures from the SAR of the IPCC, so there may have been different estimates since then; but I think that shows it's not an open-and-shut case.