Carbon losses and primary productivity decline in savannah soils under cotton-cereal rotations in semiarid Togo.

16 pagesInternational audienceSoil degradation in the savannah-derived agroecosystems of West Africa is often associated with rapid depletion of organic carbon stocks in soils of coarse texture. Field experiments were conducted over a period of more than 30 years at two sites in semiarid Togo to test the impact of agricultural management practices on soil C stocks and crop productivity. The resulting datasets were analysed using dynamic simulation models of varying complexity, to study the impact of crop rotation, fertiliser use and crop residue management on soil C dynamics. The models were then used to calculate the size of the annual C inputs necessary to restore C stocks to thresholds that would allow positive crop responses to fertilisers under continuous cultivation. Yields of all crops declined over the 30 years irrespective of crop rotation, fertiliser use or crop residue management. Both seed-cotton and cereal grain yields with fertiliser fluctuated around 1 t ha−1 after 20 years. Rotations that included early maturing sorghum varieties provided larger C inputs to the soil through residue biomass; around 2.5 t C ha−1year−1. Soil C stocks, originally of 15 t ha−1 after woodland clearance, decreased by around 3 t ha−1 at both sites and for virtually all treatments, reaching lower equilibrium levels after 5–10 years of cultivation. Soil C dynamics were well described with a two-pool SOM model running on an annual time step, with parameter values of 0.25 for the fraction of resistant plant material (K1), 0.15–0.20 for the decomposition rate of labile soil C (K2) and 8–10 t C ha−1 for the fraction of stable C in the soil. Simulated addition of organic matter to the soil 30 years after woodland clearance indicated that additions of 3 t C ha−1year−1 for 15–20 years would be necessary to build ‘threshold' soil C stocks of around 13 t ha−1, compatible with positive crop response to fertiliser. The simulated soil C increases of 0.5 to 1.6% per year are comparable with results from long-term experiments in the region. However, the amounts of organic matter necessary to build these soil C stocks are not readily available to resourcepoor farmers. These experimental results question the assumption that crop residue removal and lack of fertiliser input are responsible for soil C decline in these soils. Even when residues were incorporated and fertilisers used at high rates, crop C inputs were insufficient to compensate for C losses from these sandy soils under continuous cultivation