Quantitative models for teak forest management in Indonesia

This study provides (1) a set of growth and yield models and (2) two versions of harvest scheduling models for management of teak plantations in Indonesia. Growth and yield models were based on data collected from permanent plots distributed in Central and East Java. A forest district in Central Java was selected as a model forest for the development of the harvest scheduling models. The first version of the harvest scheduling model treats yields as known with certainty and was formulated as a linear programming (LP) problem. Model outputs under various sets of constraints indicated that: (1) single-rotation management produces less NPV than multiple-rotation management, (2) NDEF constraints are necessary to regulate harvest flows over time, (3) imposing NDEF constraints reduces NPV, and (4) given alternative rotation ages, models tend to allocate a greater portion of the forest to the shortest rotation. The second version of the harvest scheduling model incorporates risk of not achieving a NDEF condition due to yield variability by treating yield as random variables. This version was formulated using chance-constrained programming (CCP). In this version, a strict NDEF condition was no longer feasible. The model was solved by allowing periodic yield to decrease by a maximum of 10%. This second version resulted in different hectare allocations. The total NPV was by and large increased, but the trajectory of the harvest-flow over time was not as smooth as that given by the deterministic (LP) models. An important advantage of the second version is having some degree of assurance (e.g., 95%) that the projected harvest-flow will materialize if the model activities are implemented. In general, this study has shown that the conventional management approach currently practiced on teak plantations in Java possesses several limitations with respect to attaining maximum profits in a sustainable fashion. The quantitative approach presented in this study considerably mitigates those limitations. Some model refinements (e.g., using finer spatial and temporal aggregations) were suggested prior to the implementation of the harvest scheduling model provided in this study.