Non Technical Summary

This study examined the use of commercial landings as an indicator for the management of fisheries. A common example of an indicator used in the management of commercial fisheries is catch-per-unit-of-effort data, where landings are divided by a measure of commercial fishing effort and used as a surrogate measure of the abundance of the fish. Many fisheries, including some in NSW, do not have credible sources of information about fishing effort. This study thus attempted to quantify the effectiveness of using commercial landings alone as an indicator for impacts in a fishery.

Unfortunately the indicator of commercial landings is particularly 'noisy' and could be confused by several processes occurring simultaneously. That is, any information about changes or impacts on the fishery could be easily masked by changes in the myriad of processes that affect total landings. For example, a reduction in the size of the fish stock could be hidden by an increase in fishing effort. A second issue that requires consideration is 'what is an impact in a fishery'? Quantitative analysis of indicators requires a quantitative definition of impact. This paper studied two types of impacts: chronic impacts which affect recruitment, survival or fishing processes at 5% per year; or acute impacts which affect recruitment, survival or fishing processes at 20% per year.

The statistical methods applied in this study were based upon tools from industrial quality control. These methods were originally designed to identify when manufacturing processes starting producing items that were not of an acceptable quality. Using this analogy, impacts on fishery processes are identified when unexpected changes to commercial landings occur. Detection of such a change would trigger an "alarm" that would then be used to review the fishery in detail.

Evaluation of an impact detection scheme is most readily completed with computer simulation. Three types of fisheries were represented with the standard population models that are used in fisheries research: a prawn fishery, a bream fishery and a shark fishery. Various impacts were imposed on these simulated fisheries and the performance of the detection scheme estimated.

Note that there are two measures of performance to consider when investigating such a scheme. There is the 'sensitivity', which is a measure of the rate of failure to detect an impact when it has occurred. There is also the 'specificity', which is a measure of 'false-alarms', or the tendency of the scheme to detect impact when one has not actually occurred. Clearly, neither type of error is desirable, but impact detection schemes will generate false alarms as well as true signals of actual impact. This scheme was defined to maximise the sensitivity to impacts but at the same time accept that up to 20% of signals will be false alarms.

On the basis of computer simulations, the scheme managed to detect acute changes to recruitment, survival or fishing with a sensitivity of 80-90%. Chronic changes were detected with much lower sensitivity. Landed catch appears to be an effective indicator for acute changes in a fishery if a high frequency of false-alarms can be managed. Due to the lack of sensitivity and specificity this scheme should only be used to identify fish stocks that require additional review and assessment.

This research was funded by NSW Fisheries but was completed while James Scandol was employed by The Centre for Research on Ecological Impacts of Coastal Cities (The University of Sydney).