Brazilian sardine (Sardinella brasiliensis Steindachner, 1879) spawning and nursery habitats: spatial-scale partitioning and multiscale relationships with thermohaline descriptors

Luiz Eduardo de Souza Moraes 2 Douglas Francisco Marcolino Gherardi 1 2 Mario Katsuragawa 0 Eduardo Tavares Paes 1 3 0 Oceanographic Institute, Department of Biological Oceanography; University of Sa o Paulo , Pra ca do Oceanogra fico 191, Sa o Paulo, 05.508-120 , Brazil 1 Center for Environmental Complexity Synthesis, CENOSYS , Av. Ita lia, km 8, Rio Grande, Rio Grande do Sul, 96.201-900 , Brazil 2 Remote Sensing Division, National Institute for Space Research, Avenida dos Astronautas 1758, Sa o Jose dos Campos, Sa o Paulo , 12.227-010 , Brazil 3 Instituto Socioambiental e dos Recursos H dricos (ISARH), Universidade Federal Rural da Amazonia, Avenida Presidente Tancredo Neves 2501 , Bel em, Para , 66.077-901 Brazil We provide a detailed account of the spatial structure of the Brazilian sardine (Sardinella brasiliensis) spawning and nursery habitats, using ichthyoplankton data from nine surveys (1976 - 1993) covering the Southeastern Brazilian Bight (SBB). The spatial variability of sardine eggs and larvae was partitioned into predefined spatial-scale classes (broad scale, 200 - 500 km; medium scale, 50 - 100 km; and local scale, ,50 km). The relationship between density distributions at both developmental stages and environmental descriptors (temperature and salinity) was also explored within these spatial scales. Spatial distributions of sardine eggs were mostly structured on medium and local scales, while larvae were characterized by broad- and medium-scale distributions. Broad- and medium-scale surface temperatures were positively correlated with sardine densities, for both developmental stages. Correlations with salinity were predominantly negative and concentrated on a medium scale. Broad-scale structuring might be explained by mesoscale processes, such as pulsing upwelling events and Brazil Current meandering at the northern portion of the SBB, while medium-scale relationships may be associated with local estuarine outflows. The results indicate that processes favouring vertical stability might regulate the spatial extensions of suitable spawning and nursery habitats for the Brazilian sardine. - Introduction Pelagic fish spawning and larval distributions usually occur within moderately well-defined geographic limits (Lluch-Belda et al., 1989), often displaying some degree of spatial organization in the form of patches, aggregations or density gradients (Folt and Burns, 1999). Spatial structuring mechanisms result from induced spatial dependency (Legendre, 1993), or the combined effects of species behavioural traits (Folt and Burns, 1999) and spatially structured environmental variables (Dray et al., 2006; Jombart et al., 2009). In the specific case of fish reproduction, the influence of behavioural factors is a consequence of a fundamental requirement for suitable habitats that may provide maximum offspring survival. Classical theories developed over the last decades suggest that the survival of eggs and recently hatched larvae may be maximized by a combination of relative water column stability (Lasker, 1981), enrichment and retention mechanisms (Bakun, 1996), as well as by prey availability (van der Lingen et al., 2006). Notwithstanding, ecological interactions, such as predation (Bakun and Broad, 2003) and interspecies competition, and intrinsic factors, such as species physiology (Takasuka et al., 2007), spawning intensity, population size and age structure, also exert some influence on ichthyoplankton spatial arrangements (Curtis, 2004; Bellier et al., 2007a). In an analytical framework, spatial structures can be understood as the result of the spatial autocorrelation emerging from the processes cited above. The identification of these spatial structures, along # 2012 International Council for the Exploration of the Sea. Published by Oxford University Press. All rights reserved. For Permissions, please email: with the mechanisms responsible for their generation and maintenance, has been regarded as a major ecological concern over the last decades (Levin, 1992; Bellier et al., 2007b; Guenard et al., 2010). The physical and biological mechanisms that induce spatial structuring operate within their own range of temporal and spatial scales (Legendre et al., 1986), and may contribute to the generation of multiscale spatial patterns (Bellier et al., 2007b). On the other hand, interactions between organisms, such as spawning fish and fish larvae, and the environment may occur within particular and limited ranges of time and space scales (Levin, 1992). Furthermore, the effects of these structuring mechanisms on fish survival may vary, as the response of organisms to environmental forcing tends to be scale-dependent, and a single mechanism may induce distinctive responses, according to the spatial scales being considered (Bellier et al., 2007b; McClatchie et al., 2007). Hence, the identification of the relevant scales of variability for species distributions, biological interactions and physical processes, and their proper representation on statistical models, are essential steps for a better understanding of the ecological processes controlling species distribution and abundance (Haury et al., 1978; Legendre and Fortin, 1989; Levin, 1992; Cushman and McGarigal, 2002). In the present study, the terms scale and spatial scale are employed interchangeably throughout the text, as synonyms. Modern techniques, such as Morans eigenvector maps and principal coordinate analysis of neighbour matrices (PCNM) (Dray et al., 2006), provide a mathematical representation of spatial structures on multiple scales through orthogonal sine-like functions, defined by intersample neighbourhood relationships (Borcard et al., 2004; Jombart et al., 2009). Spatial scales are hierarchically represented by decreasing eigenvalues, which account for progressively decaying levels of spatial autocorrelation (Griffith, 2003; Dray et al., 2006). In ecological studies, eigenvector maps or PCNM functions can be employed as spatial descriptors, which are either incorporated into quantitative models, or used to isolate environmental effects on the organisms spatial distribution (Dray et al., 2006; Griffith and Peres-Neto, 2006). Also, orthogonal properties provide the dissection of the spatial variance into independent components, favouring their incorporation in statistical models, and allowing the study of processes on any predefined range of spatial scales (McClatchie et al., 2007; BrindAmour et al., 2005). The Brazilian sardine (Sardinella brasiliensis) plays an important historical role in Brazilian marine fisheries, attested by shares of up to 47% of total annual marine catches (Paiva, 1997). As for other small pelagic stocks, the demographic variability is high (Cergole et al., 2002), and despite the limited distribution of S. brasiliensis within the Southeastern Brazilian Bight (SBB; 228S 418W 298S 498W) (Figure 1), knowledge concerning the mechanisms governing such variability is stil (...truncated)