teriorate the soil physico-chemical conditions of the often physically and chemically fragile moist savanna soils Nutrient depletion is a major constraint to crop production for in the long term, especially in the case of ammonium moist savanna soils, and inputs of nutrients are required to overcome sulfate [(NH 4 ) 2 SO 4 ] (Pieri, 1992). This observation, tothis constraint. The impact of sole and combined applications of organic inputs (OIs) [fresh tree prunings, Pueraria phaseoloides (Roxb.) Recherche Agronomique, B.P. 1026, Lomé , Togo; J. Diels, N. Sanginga, and O. Lyasse, RCMD, IITA, Ibadan, Nigeria, c/o L.W. Lam-al., 1997), most trials in the West African moist savanna bourn and Co., 26 Dingwall Rd., Croydon CR9 3EE, UK; R. Merckx, deal with manure or crop residues as a source

Legume-maize rotation and maize nitrogen (N)-response trials were carried out simultaneously from 1998 to 2004 in two distinct agro-ecological environments of West Africa: the humid derived savannah (Ibadan) and the drier northern Guinea savannah (Zaria). In the N-response trial, maize was grown annually receiving urea N at 0, 30, 60, 90 and 120 kg N ha -1. In Ibadan, maize production increased with N fertilization, but mean annual grain yield declined over the course of the trial. In Zaria, no response to N treatments was observed initially, and an increase in the phosphorus (P) and sulphur (S) fertilizer application rate was required to increase yield across treatments and obtain a response to N applications, stressing the importance of non-N fertilizers in the savannah. In the rotation trial, a 2-year natural fallowmaize rotation was compared with maize rotated with different legume types: green manure, forage, dualpurpose, and grain legumes. The cultivation of some legume types resulted in a greater annual maize production relative to the fallow-maize combination and corresponding treatments in the N-response trial, while there was no gain in maize yield with other legume types. Large differences in the residual effects from legumes and fallow were also observed between sites, indicting a need for site-specific land management recommendations. In Ibadan, cultivation of maize after the forage legume (Stylosanthes guianensis) achieved the highest yield. The natural fallow-maize rotation had improved soil characteristics (Bray-I P, exchangeable potassium, calcium and magnesium) at the end of the trial relative to legume-maize rotations, and natural fallow resulted in higher maize yields than the green manure legume (Pueraria phaseoloides). In Zaria, maize following dual-purpose soybean achieved the highest mean yield. At both sites, variation in aboveground N and P dynamics of the legume and fallow vegetation could only partly explain the different residual effects on maize.

This paper focuses on testing the Direct Hypothesis using 15N-labelled fertilizers. This hypothesis formulated by Vanlauwe et al. (2001a) originates from data obtained under controlled laboratory conditions showing interactions between organic matter-C and mineral-N. A series of experiments (lysimeter, nanoplots and microplots) with 15N-labelled fertilizer (urea or ammonium sulfate) and organic matter was carried out to test the Direct Hypothesis at different scales. Details of the trials testing the Direct Hypothesis through the use of 15N-labelled fertilizers and organic matter carried on an Arenosol in Ibadan, southwestern Nigeria are discussed.

The functioning of trees as a safety-net for capturing nutrients leached beyond the reach of crop roots was evaluated by investigating changes in exchangeable cations (Ca, Mg, and K) and pH in a wide range of medium to long term alley cropping trials in the derived savanna of West Africa, compared to no-tree control plots. Topsoil Ca content, effective cation exchange capacity, and pH were substantially higher under Senna siamea than under Leucaena leucocephala, Gliricidia sepium, or the no-tree control plots in sites with a Bt horizon rich in exchangeable Ca. This was shown to be largely related to the recovery of Ca from the subsoil under Senna trees. The increase of the Ca content of the topsoil under Senna relative to the no-tree control treatment was related to the total amount of dry matter applied since trial establishment. The lack of increase in Ca accumulation under the other species was related to potential recovery of Ca from the topsoil itself and/or substantial Ca leaching. The accumulation of Ca in the topsoil under Senna had a marked effect on the topsoil pH, the latter increasing significantly compared with the Leucaena, Gliridia, and no-tree control treatments. In conclusion, the current work shows that the functioning of the often hypothesized 'safety-net' of trees in a cropping system depends on (i) the tree species and on (ii) the presence of a subsoil of suitable quality, i.e., clay enriched and with high Ca saturation.

No abstract