Predish: a decision support tool for estimating risk of fish predation in aquaculture systems
2025
Lévy, S.A. | Thomas, M. | Pétronin, F. | Lecomte, M. | Mortillaro, J.M. | Carval, D. | Lecocq, T. | Institut des Sciences de l'Evolution de Montpellier (UMR ISEM) ; Centre de Coopération Internationale en Recherche Agronomique pour le Développement (Cirad)-École Pratique des Hautes Études (EPHE) ; Université Paris Sciences et Lettres (PSL)-Université Paris Sciences et Lettres (PSL)-Institut de recherche pour le développement [IRD] : UR226-Centre National de la Recherche Scientifique (CNRS)-Université de Montpellier (UM) | Fonctionnement écologique et gestion durable des agrosystèmes bananiers et ananas (UR GECO) ; Centre de Coopération Internationale en Recherche Agronomique pour le Développement (Cirad) | Laboratoire Animal et agroécosystèmes (L2A) ; Université de Lorraine (UL)-Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE)
International audience
Mostrar más [+] Menos [-]Inglés. The practice of rearing multiple species within the same environment, known as polyculture, is an age-old approach that offers numerous benefits for enhancing the sustainability of aquaculture, provided the species can coexist without detrimental interactions. Predation is one of the most detrimental interactions in polyculture systems and must be avoided to ensure compatibility among species. Since many piscivorous species are commonly reared in aquaculture, it is crucial to avoid species combinations where predation might occur. To address this, we developed a model that estimates predation risk between fish based on their total length in order to avoid empirical trials and thereby uphold ethical standards. This model was calibrated based on niche allometric framework with an extensive dataset on fish predation (4,207 instances) and validated using the specificity, the sensitivity, the True Skill Statistic (TSS) and the area under the curve (AUC) method with a comprehensive dataset of fish predation and non-predation events (13,707 events). The model demonstrated high accuracy with a sensitivity (predation prediction accuracy) of 0.88, specificity (non-predation prediction accuracy) of 0.96, a TSS of 0.84, and an AUC of 0.92. As a practical application, we developed a decision-support tool, Predish, in R software to allow users to estimate predation risk between fish. Predish also includes a database to fill any missing data, making it a valuable resource for designing new polyculture systems that incorporate piscivorous species with minimal predation risk, ultimately supporting aquaculture sustainability.
Mostrar más [+] Menos [-]Palabras clave de AGROVOC
Información bibliográfica
Este registro bibliográfico ha sido proporcionado por Institut national de la recherche agronomique