Anchovy (Engraulis japonicus) is the most abundant fish in the Yellow Sea large marine ecosystem (YSLME) and is also one of the most prevalent and commercially valuable (for fishmeal production; Du, 2016) pelagic fish species in the western North Pacific (Itoh et al., 2009). This leads to a huge demand for anchovy that is unaffected by commercial demand for other pelagic species. Anchovy plays a pivotal role in the YSLME, being both a prey species for dozens of fish species and an important planktivore (Tang and Ye, 1990; Zhu and Iversen, 1990; Zhao et al., 2003). As a key prey species, over the past 2.5 decades anchovy has had the greatest influence on fish community dispersal capacity, exerting strong upward trophic pressure thereby controlling community predator abundance as well as influencing spatial patterns during all seasons (Su et al., 2021a; Su et al., 2021b; Su et al., 2022). Changes in its biomass can cause biomass shifts among multiple predators throughout the food web, resulting in ecosystem perturbations which compromise the sustainability of fisheries (Pauly et al., 2002). Therefore, maintaining the sustainability of anchovy resources is of considerable ecological importance.

Anchovy of the YSLME spawn in nearshore coastal waters and then gather to overwinter in the Yellow Sea wintering grounds ( Figure 1 ; Liu et al., 1990). This concentration of spawning stock results in the central and southern regions of the Yellow Sea yielding the best catches during December and the following January respectively (Lin and Wang, 1992). China’s anchovy fishery exploration began in the late 1980s, and in 1992 anchovy fishing surveys and fishing technology research won the first prize of China’s National Science and Technology Progress Award, after which the catch increased exponentially, reaching a sub-peak (1037,682 tons) in 1997 in the YSLME, before falling precipitously after 2003 ( Figure 2 ). Since the 1990s, pelagic pair trawling has been the main commercial fishing method in the YSLME as the anchovy fishery developed following depletion of traditional demersal fish species such as yellow croaker and hairtail. The shift from demersal to pelagic fishing has resulted in anchovy becoming the main target of the YSLME fishing industry (Jin et al., 2005; Jin et al., 2015; You et al., 2017).

Acoustic surveys had indicated that anchovy resources were in a rapid decline since 1996 (Jin et al., 2001; Zhao et al., 2003). In the winter of 2001−2002, the acoustically estimated biomass of the anchovy population in the YSLME was less than 10% of its original level (Zhao et al., 2003). Catch limits have been considered since then, and Jin et al. (2001) conducted a study on the quota management of anchovy based on acoustic and trawl survey data from 1986−2001 using virtual population analysis (VPA), recommending a total allowable catch (TAC) of no more than 100,000 tons. However, catches in subsequent years far exceeded this TAC recommendation by a factor of 48−100 ( Figure 2 ). Su et al. (2021b) suggested that the removal of anchovy from the fish community in the YSLME had the greatest impact on the dispersion of the community structure. With more than 60% of fish in the YSLME predating upon anchovy as their primary food source (Tang and Ye, 1990), the significantly sharp decline in anchovy biomass caused noticeable shifts in the YSLME food web structure (Zhao, 2006). This had serious implications for the food supply to other commercially important populations in this ecosystem (Jin et al., 2001; Zhao et al., 2003), such as causing poor growth of Scomberomorus niphonius (Jin, 2001). Therefore, scientific management of anchovy is essential and urgent to maintain the sustainability of many important fisheries in the YSLME.

The basis for evidence-based management of fisheries to achieve sustainability is accurate stock assessment. Effective marine fisheries management is dependent upon information about stock status estimated by stock assessment models (Maunder and Punt, 2013). Time series proportional to stock biomass (biomass indices, e.g., catch-per-unit- effort, CPUE) are often the most critical among the integrated data required by assessment models (Francis, 2011), with their accuracy ultimately affecting the performance of fisheries management strategies. Therefore, biomass indices are often considered as fundamental for empirical assessments of stock status in managing landings worldwide (Thorson et al., 2015).

The survey biomass index produced by statistical processing of data extracted from fishery-independent scientific surveys, was intended to be the proxy for population abundance (Maunder and Punt, 2004; Thorson et al., 2015). The reason for this is that scientific surveys provide good spatial coverage of fish population distributions and have survey protocols (including survey timing and survey norms) that are consistent over time. Moreover, there is a preponderance of research on effective practices for standardizing treatments when estimating CPUE from survey data to improve their biomass representation (Thorson and Barnett, 2017).

Unfortunately, many exploited fish populations are not monitored through fishery-independent surveys due to their high cost and time consuming requirements (Grüss et al., 2019). For most developing countries, survey biomass index is difficult to promote, while nominal fishery CPUE is more accessible. For example, in China, nominal CPUE from the YSLME commercial anchovy fishery can be calculated by dividing the catch by the total power of the vessels, and these two types of data are recorded in the China Fishery Statistical Yearbook (Fisheries Bureau of the Ministry of Agriculture and Rural Affairs, 2022). But nominal CPUE does not necessarily track changes in biomass under all circumstances. Commercial fishers tend to fish in areas with high fish densities, and with increasing experience, advanced instrumentation, and greater efficiency in fishing operations, catch rates can increase or be maintained despite declining stock biomass until the fishery eventually collapses (Harley et al., 2001). This was the case for the Gadus morhua fishery in eastern Canadian waters (Walters and Maguire, 1996). In China, the representativeness of nominal CPUE as a proxy for stock biomass has also been doubtful. This is because total vessel power does not account for technological advances, and so has proven to be a poor indicator of fishing mortality (Han et al., 2021; Han et al., 2022). For the YSLME anchovy fishery, total vessel power existed as a measure of effort long before the present technology was developed and matured, and so does not reflect the true changes in fishing effort and efficiency due to the time course of technology diffusion and adoption of further advances in fishing operations.

Given that scientific surveys are expensive, highly vulnerable to budget cuts, and may be subject to unforeseen events e.g. vessel and instrument failures that make it difficult to obtain consistent long time series, this study attempts to adjust CPUE to reconfigure it to provide a more reliable proxy for biomass, in lieu of a reliable scientific survey biomass index as the main input data for stock assessment modeling. In this study, we used the existing survey biomass index of YSLME anchovy as a reference series to explore this issue by comparing Surplus production model (SPMs) estimates based on survey biomass index with raw commercial CPUE and adjusted CPUE as biomass index inputs. This will help promote the sustainable use of anchovy in China and will assist in assessments of other fisheries around the world that are without access to a reliable survey biomass index.

The biomass index from fishery independent surveys during 1989−1992 had high values, whereas the fishery dependent CPUEs were in a trough ( Figure 3 ). The survey biomass index during 1996−2003 implied a collapse of anchovy resources, where the fishery dependent CPUEs showed high biomass values and a relatively slow rate of decline. Reconfiguration of the fishery CPUEs significantly improved this situation i.e. reduced the difference between fishery independent survey and dependent CPUE trends, relative to the difference from the nominal fishery CPUEs.

The Pearson correlation coefficients of the four forms of fishery CPUE (nominal and adjusted) ranged from 0.78 to 0.99, among which the correlation between FAECcpue and FNcpue was relatively weak. The overlap of the two survey indices (trawl index and acoustic index) had similar trend ( Figure 3 ) and relatively high correlation (0.77). The correlation coefficients between trawl index and four forms of fishery CPUE (FNcpue, FAcpue, FECcpue and FAECcpue) were 0.28, 0.45, 0.34 and 0.46, respectively. The correlation coefficients between acoustic index and four forms of fishery CPUE were -0.57, -0.24, -0.48 and -0.18, respectively. In general, the trend change in fishery CPUE was delayed by about 3 years compared to that of the survey index ( Figure 3 ).

In the present study, two MCMC chains were used for scenario. Each model was run for 30,000 iterations, sampled with a burn-in period of 5,000 for each chain and a thinning rate of five iterations. All scenarios based on these biomass indices satisfied the diagnostic tests of Geweke (1992); Gelman and Rubin (1992) and Heidelberger and Welch (1992) and the visual evaluation of the stationary behavior of MCMC chains, with the model converging in each instance ( Supplementary Figures 1 , 2 ).

The inputs to the surplus production models (SPMs) require only observations of commercial landings and a biomass proxy, such as commercial catch-per-unit- effort (CPUE) or indices educed from scientific surveys (Polacheck et al., 1993). Moreover, since their initial development there have been tremendous advances in SPMs to enhance the accuracy of model estimates, such as, the application of stochastic process and observation errors (Punt, 2003; Thorson and Minto, 2015), techniques for Bayesian state space modeling (Meyer and Millar, 1999; Thorson et al., 2014), and time blocks that allow for significant changes in catchability over time (Carvalho et al., 2014). These factors have allowed SPMs to remain, to date, the primary means for scientific assessing stock status when only limited or moderate data are available.

Conservation management of fisheries based on scientific assessments helps to maintain the stability of ‘blue food’ exports and has promoted the recovery of some depleted stocks (Hilborn, 2007; Worm et al., 2009). Conservation effectiveness largely depends on the quality of fishery reference point values obtained from model assessments. The availability and quality of data largely determine the quality of assessments (Parker et al., 2018), and the suspicion of quality assessment that has led to a debate about the basis of the management of many fisheries throughout the world (Beddington et al., 2007; Hilborn, 2007). The accuracy of the results from SPMs, as widely used empirical models, is strongly influenced by data quality, such as the reliability of the input CPUE data (Parker et al., 2018). Therefore, the SPM JABBA was used as a means of testing the improvement in quality of fishery dependent input rendered by adjustment of the nominal fishery CPUE, using the gap between assessments based on adjusted (reconstructed) fishery dependent CPUE and scientific CPUE. Selection of improved fishery dependent CPUE enables better tracking biomass change trends, thereby improving its reliability in setting reference points for fisheries management decisions.

In this study, we explored this approach to improving fisheries management guidance via the YSLME anchovy fishery, because it has been the most productive fish species in China, occupying an important position in the marine fisheries economy of China. Since 1996, following commencement of the summer fishing suspension system (moratorium), the winter fishing grounds in the central and southern Yellow Sea have become the main fishing grounds for anchovy each year (Jin et al., 2005). Therefore, the CPUE from the winter grounds surveys is the most representative of actual biomass.

The CPUE from the overwintering grounds survey and the SPMs based on this index revealed that the anchovy population had experienced a sharp decline since 1996 and was at its lowest biomass level at the beginning of the 21st century. This change had also been observed by many scholars (e.g., Jin et al., 2001; Zhao et al., 2003; Wang et al., 2006). The anchovy, as a pivotal fish of the YSLME, is a key link in the food web (Zhang, 2018), and this decline subsequently caused a significant negative ecological response. The proportion of anchovy consumed by numerous predatory fish species had drastically reduced and shifted to other prey species (Shan et al., 2011). For example, the dominant food of blue-spotted mackerel Scomberomorus niphonius, a traditionally targeted large commercial fish in the YSLME, shifted from anchovy to other species (Mou et al., 2018), negatively affecting its growth (Jin, 2001). This shift led to a decrease in the overall nutrient level of the YSLME (Zhang and Tang, 2004; Zhang et al., 2007) and a weakening/erosion of the pelagic food chain (Zhang, 2018), which adversely affected marine carbon sequestration, along with material and energy flows within the food web, thereby reducing ecosystem productivity (Zhang et al., 2015). In other words, it is thus clear that the decline of anchovy population had serious impacts on the sustainable use of other commercially important populations in the ecosystem. Rapidly capturing changes in anchovy resource status using stock assessment models and taking timely management actions are important to avoid/lighten negative ecological impacts. However, scenarios #2, #4, #5 and #6 in this study did not reflect the decline in anchovy biomass and failed to guide timely management policy adjustments. Scenarios #3, #7, #8, and #9 significantly improved this situation, with Scenario #7 and #9 tracking stock size decline more effectively. Our reconstruction/adjustment of fishery CPUE has a good effect in redressing the main problem.

In addition, we propose to use the improved fishery CPUE series starting from 1993 or later, because both the fishery CPUE series before 1993 and its improvements differ significantly from the survey biomass index and the theoretical true biomass index and are highly unrealistic. The JABBA scenarios for the fishery dependent CPUE series starting in 1989 (nominal and improved) also portrayed a very low stock size followed a rapidly recovering phase prior to 1993, which is uncharacteristic for a fishery in its nascent exploitation phase. The JABBA scenarios based on the improved fisheries CPUE series starting in 1993 and survey biomass index also detected that anchovy population had been recovering in recent years, but the results were cautiously rather than overly optimistic and more consistent with other SPMS based on fisheries CPUE (Zhai et al., 2020). The result of the JABBA stock assessment based on survey biomass index lay between the estimates for the first two improved fishery dependent CPUEs (FAcpue_y and FECcpue_y ), and in the absence of survey biomass index, we recommend combining the assessment results (average) from these two alternatives to guide the anchovy resource management. However, the third adjusted CPUE (FAECcpue_y ) was overestimating the biomass status and overestimating the fishing mortality in recent years. This is due to the fact that this third reconfiguration is a combination of the first two adjustment methods. The first two adjustment methods were both able to account for the improvement in fishing efficiency, and the combination of the two yielded more conservative results. Therefore, the third adjusted index is not recommended.

The trends in the survey, nominal and both types of adjusted fishery CPUEs strongly support the above conclusions. Throughout most of the period analyzed, the trends of the two types of adjusted CPUE (FAcpue_y and FECcpue_y ), relative to the nominal fishery CPUE, were closer to that from the fishery independent survey. As Palomares and Pauly (2019) put it: “…a trawler plowing the sea in the 21st century looks like a trawler at the beginning of the 20th century.” The reason for this is that the nominal fishery CPUE, which ignores the progress of technology, does not fully reflect the changes in biomass. The new technology affects the environment and ecology much more efficiently and strongly than the previous vessels and equipment. Therefore, if we don’t account for progress in fishing technology, vessel construction, and instrument performance, our understanding about exploitation levels and protection of the anchovy, then stocks will fall into the dilemma of managing based on outdated information.

The anchovy has a short life cycle and a wide range of adaptations to differing environmental conditions, so the resource can be highly resilient to changing circumstances (Tang and Ye, 1990; Jin et al., 2005). Jin et al. (2005) concluded that there is no fish species that poses a threat to the anchovy through interspecific competition in the YSLME, and the cycle of change in status of anchovy populations is quite long in terms of environmental changes. The decline of anchovy resources was mainly caused by fishing pressure (Jin et al., 2001; Zhao, 2006). Therefore, scientific evidence-based adaptive management strategies are essential for the recovery and future sustainable use of anchovy resources. The proportion of older mature fish in the YSLME population has declined in recent years and the nearshore spawning biomass has decreased. 2005; Jin et al. (2001) suggested the introduction of quotas and restrictions on vessels to control the rampant increase in fishing mortality.

A total allowable catch (TAC), as an output control, is an important widely applied method of management among current international fisheries resource management (OECD, 2020). The Fisheries Law of the People’s Republic of China mandated the requirement to implement a universal quota fishing system, but it was not until 2017 that the Ministry of Agriculture and Rural Affairs of the People’s Republic of China (2017) issued the Notice on Further Strengthening the Control of Domestic Fishing Vessels and Implementation of Total Allowable Catch Management in the Ocean, which set clear control objectives and provided for the implementation of vessel and catch limits in the form of a national marine TAC schedule. Anchovy catch was significantly controlled, which has facilitated the stock recovery since 2018.

Scientific estimation of sustainable TAC is a necessary condition for the implementation of any catch quota system. The most conservative scenario in this study (Scenario #9) predicted that a TAC of no more than 600,000 t. Another study recommended a precautionary TAC control of 500,000 t (Iversen et al., 2001), and considering the availability of food for other economically important fish in this ecosystem (Zhao et al., 2003) and the aforementioned negative ecological effects of anchovy decline, as well as the severe fluctuations common to pelagic fish populations, plus the delayed response to management intervention, we preferred conservative estimates and recommend that a precautionary TAC of 500,000 t should be set. The current catch level is within this precautionary TAC.

At present, to promote accurate assessments of fish resources worldwide, collection new data and the rational use of existing data are far more important than the development of new models (Ovando et al., 2021). Our research has clearly demonstrated this conclusion. Improving existing data by adding new information can improve the accuracy and objectivity of stock assessments and sustainable utilization. We strongly encourage researchers to extensively review and improve existing fisheries statistical datasets to produce more reliable stock assessments (Rousseau et, 2019; Choi et al., 2021). This will promote the implementation of the United Nations (UN) Sustainable Development Goal 14 (SDG14) in the scientific understanding of exploited populations and fisheries (Hilborn et al., 2020; Ovando et al., 2021).

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

The animal study was reviewed and approved by Animal Ethics Committee of Yellow Sea Fisheries Research Institute.

QH and XS conceived this study. QH conducted to methodology, software, analysis, writing – original draft, visualization. XJ made substantial contribution to data curation, writing – review & editing. HG made substantial contribution to conceptualization, writing – review & editing. All authors contributed to the article and approved the submitted version.