DYNAMIC OPTIMIZATION OF NITROGEN-FERTILIZATION OF CITRUS AND THE VALUE OF INFORMATION FROM LEAF TISSUE-ANALYSIS

The optimal long-run fertilisation and irrigation of a perennial crop is approached via two stochastic dynamic optimisation models. The first model assumes that the ex-post level of the state variable - leaf nitrogen (LN) - is known with certainty. The second model assumes that the level of leaf nitrogen is not observed, but that the farmer has a prior probability distribution on the level of leaf nitrogen, which can be updated to yield a series of posterior probability distributions conditioned on past input decisions. Comparison of the values of the objective functions of both models yields an expected economic value of complete and costless information of leaf tissue analysis. The analysis is illustrated for a mature citrus grove. The impact of improved information on nitrogen use and leaching is also examined empirically.