Linear programming has been remarkably successful in predicting the diet choice of generalist herbivores. We used this technique to test the diet choice of free-living beavers (Castor fiber in the Biesbosch (The Netherlands) under different foraging goals, i.e. maximization of intake of energy, nitrogen, phosphorus or sodium, or minimization of feeding time. Three food types were distinguished, i.e. woody food, herbs and roots of monocots. We assessed forage quality by measuring the dry matter, energy and mineral contents of the food plants as well as food intake rates, digestibility and metabolisability in captive beavers. Actual diet was in accordance with the predicted food choice in the summer when the beavers were minimizing feeding time by mainly eating woody food. However, in the winter and spring, the beavers were predicted to feed upon non-woody food, whereas they (again) nearly exclusively ate woody food. The major reasons for this discrepancy might be: (1) the foraging constraints were inappropriate, (2) the foraging goals were inadequately defined, or (3) the beavers were not foraging optimally. We suggest that future work should take some additional constraints and foraging goals into account.
