Sustaining productivity in short-rotation forest plantations over multiple rotations can be problematic, when nutrient inputs to soil are small compared to exports off-site during harvest. For example, there are concerns about sustained productivity of Eucalyptus globulus plantations growing on ex-pasture land in south-western Australia because nitrogen (N) inputs from fertilisation and leguminous species are low. One way of improving the N balance of the plantation system is through retention of tree residues produced during harvest operations. We used the Generic Decomposition And Yield (G'DAY) model of Carbon (C) and N cycling in plant and soil to investigate the relationship between soil N fertility and stem productivity when E. globulus planted on legume-based pasture land is grown over twenty 8-year rotations under different scenarios of harvest residue and N management. The scenarios included stemwood harvesting with retention of harvest residues, denoted "single slash" (SS), double the amount of harvest residues, denoted "double slash" (DS), no harvest residue (NS), burning of harvest residues and tree litter (B) and addition of fertiliser N equal to N removed in stemwood harvest (F). We parameterised and tested G'DAY using data from a field experiment conducted at two second-rotation E. globulus plantation sites with contrasting productivity. Experimental treatments at these sites included SS, DS, NS and B. With the model we were able to accurately reproduce observed stemwood growth of second-rotation stands for all treatments at both sites (r = 0.99). Model simulations overestimated soil N mineralisation at the more productive site (relative error, E, ranged from -14 to -24), and underestimated it at the less productive site (+5 < E < +25). In the long-term scenario simulations we found that the rate of decline in stemwood productivity is sensitive to rates of N removal in harvested stem and to residue management. Simulations suggest that retention o