Prawn Fishery Simulation Yield Model

R. Watson, D. J. Sterling, (1990). “Prawn Fishery Simulation Yield Model.,” Torres Strait prawn project: A review of research 1986-88 (Queensland Fisheries, Brisbane, Australia). Chapter 12.

Managers of fisheries have always sought to predict the consequences of any management measure before it is introduced. Even broad predictions have proven extremely difficult to make and, mistakes in predictions can be costly to the industry and possibly the long term durability of the fishery. The industry expects management measures to be precise. For example the optimum dates for seasonal closures are expected to have an accuracy measured in weeks, if not days. Many of the management measures being employed on prawn fisheries in Northern Australia are novel in their application, for example, seasonal closures to protect small prawns have been in use for less than ten years in northern Australia. Few if any experimental controls have been used to establish the effectiveness of these measures. Large inter-year variations in prawn numbers occur naturally and can complicate long-term assessments. In response to industry concern, several different management regimes have been used in successive years, making the explanation of results extremely difficult because of their possible interactions. To assess the response and elasticity of a fisheries’ potential to withstand fishing pressure it is usually necessary to employ widely varying levels of exploitation and then observe the response. This means some degree of calculated under fishing exploitation and over fishing exploitation within the fishery. To be successful these ‘experiments’ have to be large in scale and extend for several years. The very idea of costly manipulation of a viable important fishery is, at least for the present, politically unacceptable. This means that researchers are asked to help managers decide on strategies without the benefit of the type of data from the system which would allow statistical predictions to be made. They must therefore employ data from other fisheries. Fisheries from which data is available are often from temperate regions, and all too often are failed fisheries and not ideal for comparison with a tropical prawn fishery. There is one other tool researchers and managers can employ. This tool is simulation modelling. A fisheries biologist can estimate a number of important biological and population parameters from direct measurement, experimentation, or from the scientific literature. These can be combined with information gathered from economists, commercial processors and fishermen to produce a series of rules and relationships. Such controls often govern the fishery through various limiting factors such as the number of vessels, the available searching time, the biology of the key species, or the economics of the products. One or more of these aspects can be combined using the computational powers of computers in a simulation of the fishery. These computer simulations often allow for graphical or tabular display of the results so that all potential users can visualise the results and gain some understanding of the interplay between the many complex relationships underlying the model. Several types of computer simulation models exist. As the name implies they are meant to simulate the fishery. Given suitable input such as the numbers and sizes of animals recruiting into the fishery, models can produce estimates of potential outcomes. These may be landings, numbers of animals the following year, or even net profits. Most of these simulation programs are based on dynamic models. Dynamic models attempt to model the passage of time in the life of the fishery. Some programs model time on a continuous basis, while most models use discrete time units. Units of years are applicable only to comparitively long- Iived animals. For prawns, units of months or even weeks are more appropriate. The time scale most appropriate is dependent on the generation time of the key species as well as the detail available in the input data, and the precision required of the predictions. Many models use fixed rules to relate the input data to outcomes and do not allow for chance circumstances, these are deterministic models. Others attempt to simulate the natural uncertainty in the reaction of one factor in the fishery to others, or the certainty or potential error in the input data. These models allow for random or pseudo-random processes to occur. These stochastic models produce different results every time they are used even if the same input data is used. While this produces a more realistic approximation of the natural situation, it requires additional information on the rules which regulate how this variation or randomness occurs. When stochastic models are used, it is common to reuse the computer model a number of times, through the process of Monte Carlo simulations, in order to extract the average result. In addition to the average outcome, stochastic models also allow the range of expected results to be determined. Models can be either extremely complex or ‘reductionist’ in nature and attempt to desc~ibe all knowable aspects of the animals’ biology and the fishery in fine detail, or they can be general or ‘holistic’ in approach and deal with only the generalised net or overall affects. They can attempt to explain the relationship between all parameters that can be measured and attempt to predict all aspects of the fishery, or they can use only some of the available data to predict only one result such as the net profit. Modelling has been employed to great effect by Somers (1985) to predict the optimum opening date for the Penaeus merguiensis, banana prawn fishery in the Gulf of Carpentaria. Somers (1985) used information on prawn prices and prawn growth rates, together with weekly size surveys, to predict when harvest would maximise the gross profits of prawn fishermen. Sluczanowski (1984) used historical catch data and existing fisheries models to optimise (through modelling) population parameters for the Spencer Gulf prawn fishery of P. latisulcatus, the western king prawn. These parameters were then used in a subsequent model, which had an economic framework, to find management measures which would optimise total industry profits. The deterministic model that will be described below is not as complex as that used by Sluczanowski (1984) in the Spencer Gulf, and as yet has not been elaborated to produce economic parameters other than gross catch values. This model was originally developed to simulate the growth, immigration, emigration and mortality processes of juvenile P. esculentus, brown tiger prawns, in nursery areas of Torres Strait. 1t was then extended to include the adult or commercial phase of the life cycle, and to include the other two commercial species in Torres Strait, Metapenaeus endeavouri, endeavour prawn, and P. longistylus, redspot king prawn. It has already been employed to make crude predictions of the effects of differing seasonal closure periods on prawn catch values. With modification and further sophistication this model can be used to meet many of the prawn fishery managers future needs.