A model (ROTATE) of the nitrogen (N) cycle during the tree and crop ph
ases of fallow systems [Robertson, 1994] was used to determine the pri
mary factors influencing longterm crop yields. The model simulated the
expected patterns of increase in old (recalcitrant) soil organic N du
ring tree rotations and their decrease under continuous cropping. Afte
r 3-4 fallow cycles an equilibrium soil organic N content is reached,
where N losses by crop removal are balanced by N gains by the trees (e
ither by fixation or pumping from depth) plus small inputs in rain. Th
e rotation period has two variable components: the cycle length (tree
plus crop period) and the fraction of years in each cycle occupied by
trees (1 = sole trees, 0 = sole crops). Both components have optima de
termined by the time taken for the trees to increase the old soil orga
nic N pool to an optimal (but not maximal) size. This optimum exists b
ecause the rate of increase in old organic N slows as the tree fallow
progresses and a time is reached (often soon after the trees reach ful
l size) when the benefits of further improvements in soil fertility ar
e outweighed by crop yield foregone. In the example chosen of Acacia/s
orghum in the Sahel, the optimum cycle seemed to be about 50 years wit
h half of the time in trees. The optimum fallow period is shortened by
growing fast-growing trees, and the benefit of fallow periods are gre
atest when (i) a large proportion of the N in litter (above and belowg
round) is transferred to the recalcitrant soil pool, and (ii) the tree
s attain a large size with correspondingly large annual additions of N
to the soil.