A review of data-limited assessment methods and 

their applications in fishery resource conservation

Abstract: Scientific stock assessments are the key for fishery management. They support sustainable fisheries by providing fisheries managers with the information necessary to make sound decisions. However, due to the high cost of data collection, the majority of the world’s fisheries are datalimited which lack sufficient data to carry out full stock assessment. In the United States, 59% of stocks are datalimited; 165 out of 262 stocks in Europe have varying degrees of data deficiency; in China, most catches data are counted by the aquatic products categories which are not specific single species. Based on these situations, scientists have been seeking simple and lowcost methods to achieve stock assessment. Therefore, datalimited methods (DLMs) have developed rapidly in the last two decades. Many fisheries that are difficult to carry out traditional stock assessment are now able to be scientifically managed. The domestic researches about DLMs are still finite, and few of them has reviewed how the results of these models can be used to help managers to develop conservation strategies. Model structures of DLMs are generally simple, and their results are therefore different from those obtained by traditional stock assessment methods. Appropriate interpretation and rational use of the results from DLMs are crucial to the determination of management measures. In this study, we divide DLMs into three categories according to data requirements: catchbased models, lengthbased models, and multispecies models. And we briefly summarize and review the model data requirements, model output, and the advantages and disadvantages. Then, we mainly discuss the appropriate application of DLMs into fisheries resource conservation and highlight the issues that need to be focused on in practice. Catch data are the most common fishery data. Therefore, catchbased models are the most explored and developed models. Widely used models are depletioncorrected average catch (DCAC), depletionbased stock reduction analysis (DBSRA), an extension of catchMSY(CMSY), catch only modelsampling importance resampling model (COMSIR), statespace catch only model (SSCOM), simple stock synthesis (SSS). All of these models need catch data and some simple biological information. CMSY model can estimate the maximum sustainable yield (MSY), BMSY, and FMSY. COMSIR and SSCOM have higher ability on forecasting stock status. DCAC and DBSRA include biological parameters such as natural mortality, growth, and recruitment. The results are more reliable, but the models are sensitive to stock depletion. The main outputs of catchbased models are MSY and MSYbased reference points. Therefore, fishery management applies the outputcontrol management to set catch limits. Reference points include overfishing limit (OFL), acceptable biological catch (ABC), annual catch limit (ACL), and annual catch target (ACT). An OFL is an estimate of the catch level above which overfishing is occurring. The ABC is lower than OFL which accounts for scientific uncertainty. The ACL is normally smaller than ABC to ensure that the stock is not overfished. The ACT is a level of catch set to account for management uncertainty which is lower than the ABC. Generally, selecting ACL as the catch limit is relatively safe. Length frequency data have the advantage of being relatively cheap, straightforward, and quick to collect from landing sites and markets. Therefore, lengthbased models have developed rapidly in recent years. The lengthbased models include lengthbased spawning potential ratio (LBSPR), lengthbased integrated mixed effects(LIME), lengthbased risk analysis (LBRA), and lengthbased Bayesian approach (LBB). The LBSPR model is based on the assumption of stock equilibrium and assumes that the length frequency data is representative of the exploited population at a steady state. The LIME model is an extension of the LBSPR that accounts for timevarying recruitment and fishing mortality. The LBRA extends the lengthbased model to a risk analysis context which could define sustainability risks in terms of probability distributions. The LBB model can estimate the currently exploited biomass relative to unexploited biomass (B/B0). Therefore, based on the outputs of LBB, the limit and target reference points are usually set as the values of B/B0 equal to 0.2 and 0.4, respectively. Except for LBB, spawning potential ratio (SPR) is used as the biological reference point for other lengthbased models which is defined as the proportion of the unfished reproductive potential left at any given level of fishing pressure. The values of SPR equal to 30% and 40% are often used as limit and target reference points in practice, respectively. Fisheries resource conservation could also apply the inputcontrol management which constrains the fishing effort (e.g., location, gear type, mesh size). Many species are caught at the same time in some fisheries, which are called multispecies fisheries. In these fisheries, the singlespecies assessment method can not be used in management. There have two general approaches. To reduce research costs, one of the methods is to assign datalimited stocks to assemblages according to similar life history, trophic behavior, home range, etc. The stocks can be managed as units in this way. Ideally, each assemblage would include at least one datarich species as a status indicator for the management unit. Managers then can make conservation measures for the entire unit based on the status of the indicator species. The other category of methods is based on the ecosystem. Mainly models have multispecies virtual population analysis (MSVPA), multispecies statistical catch at age model (MSCAA), impact assessment model (IMA), and productivity and susceptibility analyses (PSA). The MSVPA model can quantify the predatorprey interactions and estimate the rates of predation mortality for exploited fish populations. The MSCAA includes the influence of climate change. The IMA is combined with economic dynamics. The PSA is a method for assessing the vulnerability of a fishery species which is based on the attributes of productivity and susceptibility. Ecosystembased models still require extensive biological data. Finally, in addition to strengthening data collection, we suggest two measures that can be used to improve the robustness and effectiveness of management objectives. The first one is to use the hierarchical Bayesian framework to integrate the collected available information as the information prior. Peerreviewed publications and publicly available databases (e.g., Fishbase) are common sources of information collection. Secondly, we recommend that management strategy evaluation (MSE) and harvest control rules (HCRs) can be added in future work.