Soil fertility restoration and crop performance in many developing countries with low input agriculture strongly relies on fallow duration and management. More precise information about the availability of fallow land may provide a way to improve the simulation of yarn (Dioscorea spp.) yields at the regional scale which has hardly been considered in prevailing approaches to model regional crop production. The probable reason behind this is scarce availability of data on fallow duration and variation across the farms in a region. Therefore, this study attempts to estimate effective fallow availability for yam production at the regional scale and to simulate the effect of fallow on regional yam yield. Yam growth and yield were simulated with the EPIC model which was incorporated into a Spatial Decision Support System (SDSS) covering a typical catchment with variable land use intensity within the sub-humid savannah zone of West Africa. Yarn-fallow rotations were simulated within 1120 quasi-homogenous spatial units (LUSAC = Land Use-Soil Association-Climate units) and aggregated to the 121 sub-basins and ten districts within the catchment under three different scenarios of fallow availability: (Si) Total savannah area was available as fallow land, (S2) 50% of the bush savannah was available as fallow land and (S3) 25% of the bush savannah was available as fallow land. The aggregation procedure adopted in this study was based on changes in the frequency of fallow-cropland classes within the sub-basins to render the SDSS sensitive to changes in fallow availability. Comparison of the average simulated tuber yield with the observed mean yield over the entire catchment showed that the simulations slightly overestimated the yields by 0.4% for scenario S1, whereas, underestimated by 14.2% and 36.8% in scenario S2 and S3 respectively. When compared with the effective fallow availability to maize, it was concluded that, (1) due to farmers preferences to plant yam mainly on virgin savanna land and as the first crop in the rotation after fallow, the effectively available fallow area for yam is higher than for maize and (2) the applied approach is suitable to derive effective fallow availability for yarn production at the district scale. (C) 2012 Elsevier B.V. All rights reserved.
