Selecting statistical models and variable combinations for optimal classification using otolith microchemistry

Mercier, Lény; Darnaude, Audrey M.; Bruguier, Olivier; Vasconcelos, Rita P.; Cabral, Henrique N.; Costa, M. J.; Lara, Monica R.; Jones, David L.; Mouillot, David

doi:10.1890/09-1887.1

Cited by 102 publications

(104 citation statements)

References 64 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Habitat discrimination based on otolith fingerprints from LA-ICPMS measurements was achieved using a random forest (RF) algorithm (Breiman 2001), recently shown to be highly accurate for Sparus aurata habitat differentiation in the study area (Mercier et al 2011b). Although the method is still new to ecologists, it has been widely used in bioinformatics (e.g.…”

Section: Discussionmentioning

confidence: 99%

“…Among the chemical elements measurable in the otoliths, some can be less informative than others, and bring more noise than signal in habitat classification (Mercier et al 2011b). In order to identify the optimal RF classifier for reconstruction of fish movement from otolith signatures, i.e.…”

Section: Discussionmentioning

confidence: 99%

“…In order to identify the optimal RF classifier for reconstruction of fish movement from otolith signatures, i.e. the element combination allowing the best accuracy in discrimination to be reached between all habitats tested, the following cross-validation procedure (Mercier et al 2011b) was applied to all possible combinations of the 10 elements measured (B, Mg, Cr, Mn, Cu, Zn, Sr, Y, Ba, and Pb). To avoid circular reasoning when the same data are used to both build and test a classifier, 75% of the spots from the training dataset were randomly selected and only their signatures were used to build the RF classifiers.…”

Section: Discussionmentioning

confidence: 99%

“…More specifically, the packages FactoMineR, class, gtools, klaR, MASS, mda, and randomForest were used. The R scripts created for element selection (Mercier et al 2011b) are available at www.ecosym. univ-montp2.fr.…”

Section: Discussionmentioning

confidence: 99%

“…Secor et al 2001, Jessop et al 2008) have been based on lifetime fluctuations in otolith strontium (Sr) concentrations, which are a proxy for environmental salinity (Campana 1999). However, in order to confidently discriminate between coastal habi tats with similar salinities, multi-elemental otolith fingerprints are needed (Forrester & Swearer 2002, Mercier et al 2011b. Furthermore, in the coastal zone, temporal variations in otolith elemental composition at a given location can be greater than variation among sites, so the collection of multiannual signatures for each site is necessary to obtain temporally stable signatures that can be exploited to reconstruct lifetime habitat use by coastal fishes (Elsdon et al 2008).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Multi-element otolith fingerprints unravel sea−lagoon lifetime migrations of gilthead sea bream Sparus aurata

Mercier¹,

Mouillot²,

Bruguier³

et al. 2012

Mar. Ecol. Prog. Ser.

View full text Add to dashboard Cite

Precise knowledge of lifetime migrations is vital in exploited fish species, since all essential habitats must be protected to maintain sustainable stock levels. The present study used multi-element otolith fingerprints of the gilthead sea bream Sparus aurata (L.) to discriminate its main juvenile and adult habitats in the Languedoc-Roussillon region (Gulf of Lions, northwest Mediterranean) and infer the lifetime migrations of 12 individuals from the area (11 from the present day and 1 from the Roman era). This allowed for the first time the identification of key habitats for the successful completion of the species' life cycle in the Gulf of Lions, and the connectivity between them. Our results revealed that lagoon use by S. aurata is probably ancient (> 2500 yr) and confirmed its current commonness. Yet, although most observed migration patterns were in accordance with the migratory behavior previously described for the species, strong inter-individual variations and new patterns in habitat use were detected. At the juvenile stage, a preference for shallow lagoons with low salinities was evidenced. Nevertheless, the first year of life can also be successfully completed in marine conditions. At the adult stage, lagoon use was shown to occur until at least age 4 yr, with periods of lagoon residency of up to 11 mo in a year, often including winter months. Because overwintering in the lagoons was previously thought to be impossible for S. aurata due to low temperatures, this finding has important implications for future stock management, especially since the species breeds in winter. KEY WORDS: Otolith microchemistry · Trace elements · Random forest · Key habitats · Gulf of LionsResale or republication not permitted without written consent of the publisher

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Multi-element otolith fingerprints unravel sea−lagoon lifetime migrations of gilthead sea bream Sparus aurata

Mercier¹,

Mouillot²,

Bruguier³

et al. 2012

Mar. Ecol. Prog. Ser.

View full text Add to dashboard Cite

show abstract

Linking otolith microchemistry and dendritic isoscapes to map heterogeneous production of fish across river basins

Brennan

Schindler

2017

Ecological Applications

View full text Add to dashboard Cite

Production patterns of highly mobile species, such as anadromous fish, often exhibit high spatial and temporal heterogeneity across landscapes. Such variability is often asynchronous in time among habitats, which stabilizes production at aggregate scales of complexity. Reconstructing production patterns explicitly in space and time across multiple scales, however, remains difficult but is important for prioritizing habitat conservation. This is especially true for fishes inhabiting river basins due to long-range dispersal, high mortality at early life stages, complex population structure and elusive life history variation. We develop a new approach for mapping production patterns of Pacific salmon across a large river basin by integrating otolith microchemistry and dendritic isoscape models. The geographically continuous Bayesian assignment framework presented here yielded high accuracies (>90%) and relatively high precisions (precisions <4%; i.e., assignment areas of <530 river km of the 13 100 km total river length) when used to determine the natal source of known-origin juvenile Chinook salmon captured throughout the study region. Integrating these methods enabled us to base estimates of provenance and habitat use of individuals on a per location basis using strontium isotopic data throughout the continuous spatial domain of a river network. Such a framework provides substantial advantages over the more common nominal approach to employing otolith microchemistry to reconstruct movement patterns of fish. In doing so, we reconstructed the spatial production patterns of adult Chinook salmon returning to a large watershed in Bristol Bay, Alaska and illustrate the power of such an approach to conservation efforts.

show abstract

Lagoon nurseries make a major contribution to adult populations of a highly prized coastal fish

Tournois

Darnaude

Ferraton

et al. 2017

Limnology & Oceanography

Self Cite

View full text Add to dashboard Cite

This study investigated the contributions of five putative nursery areas, namely four coastal lagoons and the marine coastal zone, to the local offshore fishery of the gilthead sea bream Sparus aurata, in the Gulf of Lion (NW Mediterranean). A detailed database of multi-elemental lagoon and marine otolith signatures was developed, based upon analysis of multiple sequential spots across growth bands of the otoliths of juveniles (n 5 142) sampled from the habitats over a 5-yr interval. Adults of 10 annual cohorts (n 5 114), fished throughout the Gulf, were then submitted to a similar microchemical analysis of the growth band of their otoliths that was formed after nursery settlement in their first year. Using the juvenile signature database and Random Forest classification algorithms, adults were re-assigned to the lagoons or marine coastal zone as a nursery. More than 80% of adults derived from lagoons, being 85.0% for five consecutive cohorts. Therefore, the marine coastal zone apparently contributed few adults, especially as a function of surface area. The four lagoons differed, however, in their contribution to the S. aurata fishery. Two lagoons qualified as "Effective Juvenile Habitats" that contributed more than 20% of the adults fished. Two shallow brackish lagoons were "Nursery-Role Habitats" that contributed almost twice as many adults, per unit surface area, as the overall average. Such information can support strategies for sustainable management of multiple coastal lagoon habitats, and of valuable fisheries that depend upon them.

show abstract

Selecting statistical models and variable combinations for optimal classification using otolith microchemistry

Cited by 102 publications

References 64 publications

Multi-element otolith fingerprints unravel sea−lagoon lifetime migrations of gilthead sea bream Sparus aurata

Multi-element otolith fingerprints unravel sea−lagoon lifetime migrations of gilthead sea bream Sparus aurata

Linking otolith microchemistry and dendritic isoscapes to map heterogeneous production of fish across river basins

Lagoon nurseries make a major contribution to adult populations of a highly prized coastal fish

Contact Info

Product

Resources

About