Cataloguing environmental influences on the spatiotemporal variability of Adriatic anchovy early life stages in the eastern Adriatic Sea using an artificial neural network

Abstract

The anchovy (Engraulis encrasicolus, Linnaeus, 1758), one of the most important small pelagic fish species in the Adriatic, is currently described as a species that can be considered overfished. From 2013 to 2020, samples of anchovy eggs and larvae were collected through scientific surveys during the summer months. The collected ichthyoplankton data were combined with environmental data (measured satellite sea surface temperature and chlorophyll data, numerically simulated salinity, maps of primary production) to identify anchovy spawning habitats and environmental conditions affecting the anchovy early life stages. For this large dataset, a nonlinear method called Growing Neural Gas Network analysis was used to explain the multiple dependencies between anchovy and the explanatory environmental variables and represent them in 9 patterns called Best Matching Unit (BMU). Obtained values of anchovy early life stages abundances (eggs/m2; larvae/m2) showed a clear negative trend, which was easily observed both in the time series and in the annual spatial distributions. Among all measured environmental parameters that were previously mentioned, salinity showed a significant increase, which can be attributed to the cyclonic phase of the bimodal oscillatory system of the Adriatic and Ionian Seas. The calculated BMUs showed several interesting results that shed new light on previous findings: (a) there is a split between the richer northern and poorer southern parts of the Adriatic in terms of anchovy eggs and larvae abundances, (b) the Kvarner Bay, the west coast of Istria and the area around Dugi otok are consistently rich spawning grounds, (c) decreased abundance in the southern areas is a result of the influence of salinity, (d) an increase in chlorophyll can lead to an increase in egg count, (e) the positive effects of upwelling can be negated by an increase in salinity, (f) increased primary production is followed by increased egg count. Upwelling, as one of the factors that can influence larval and egg abundance by bringing nutrients up from the seafloor, showed increased spatial and temporal variability during the investigated period, which depended on the wind regime. Our analysis showed that neural network analysis can successfully describe the effects and interplay of environmental factors on the abundance of anchovy early life stages.