Management with sparse data
MetadataShow full item record
Original versionThis report is not to be quoted without prior consultation with the General Secretary.
We consider a management that aims at controlling the removal from the stock, guided by a perception of the state of the stock derived from limited data that does not allow an ordinary analytic assessment. In simulation modeling, there is a 'real' stock and 'real' removals, and the stock develops according to these removals. The management part creates a feed back loop through a noisy link between the state of the true stock and the actual removals. The performance of the management will depend on how the real removal responds to the stock, and how the stock responds to the removals. Structurally, this is a quite simple feed-back system. The interior of the building blocks is complex and diverse, however, and include obstacles like time lags and stochastic terms. When an analytical assessment is not available, the link between the real stock and the management decisions is harder to understand and model, and it may be noisier. The main emphasis in this paper is finding decision rules that rely on sparse and noisy data. A simulation tool runs as a bootstrap and was made to cover a variety of stocks, decision rules and noise in a versatile way, but on a quite generic level. The link between state of the stock and the basis of the decision was modeled as SSB (or alternatively total stock biomass TSB) derived from the real stock numbers at age, but with random noise and a random year factor. Several types of harvest rules were explored, and pros and cons of various types are highlighted.