Search

Over the past 4 decades there has been a growing concern for the conservation status of elasmobranchs (sharks and rays). In 2002, the first elasmobranch species were added to Appendix II of the Convention on International Trade in Endangered Species of Wild Fauna and Flora (CITES). Less than 20 yr later, there were 39 species on Appendix II and 5 on Appendix I. Despite growing concern, effective conservation and management remain challenged by a lack of data on population status for many species, human−wildlife interactions, threats to population viability, and the efficacy of conservation approaches. We surveyed 100 of the most frequently published and cited experts on elasmobranchs and, based on ranked responses, prioritized 20 research questions on elasmobranch conservation. To address these questions, we then convened a group of 47 experts from 35 institutions and 12 countries. The 20 questions were organized into the following broad categories: (1) status and threats, (2) population and ecology, and (3) conservation and management. For each section, we sought to synthesize existing knowledge, describe consensus or diverging views, identify gaps, and suggest promising future directions and research priorities. The resulting synthesis aggregates an array of perspectives on emergent research and priority directions for elasmobranch conservation.

Accurately characterizing the biology of a pelagic shark species is critical when assessing its status and resilience to fishing pressure. Natural mortality (M) is well known to be a key parameter determining productivity and resilience, but also one for which estimates are most uncertain. While M can be inferred from life history, validated direct estimates are extremely rare for sharks. Porbeagle (Lamna nasus) and shortfin mako (Isurus oxyrinchus) are presently overfished in the North Atlantic, but there are no directed fisheries and successful live release of bycatch is believed to have increased. Understanding M, post-release mortality (PRM), and variables that affect mortality are necessary for management and effective bycatch mitigation. From 177 deployments of archival satellite tags, we inferred mortality events, characterized physiological recovery periods following release, and applied survival mixture models to assess M and PRM. We also evaluated covariate effects on the duration of any recovery period and PRM to inform mitigation. Although large sample sizes involving extended monitoring periods (>90 days) would be optimal to directly estimate M from survival data, it was possible to constrain estimates and infer probable values for both species. Furthermore, the consistency of M estimates with values derived from longevity information suggests that age determination is relatively accurate for these species. Regarding bycatch mitigation, our analyses suggest that juvenile porbeagle are more susceptible to harm during capture and handling, that keeping lamnid sharks in the water during release is optimal, and that circle hooks are associated with longer recovery periods for shortfin mako.

Accurately characterizing the biology of a pelagic shark species is critical when assessing its status and resilience to fishing pressure. Natural mortality (M) is well known to be a key parameter determining productivity and resilience, but also one for which estimates are most uncertain. While M can be inferred from life history, validated direct estimates are extremely rare for sharks. Porbeagle (Lamna nasus) and shortfin mako (Isurus oxyrinchus) are presently overfished in the North Atlantic, but there are no directed fisheries and successful live release of bycatch is believed to have increased. Understanding M, post-release mortality (PRM), and variables that affect mortality are necessary for management and effective bycatch mitigation. From 177 deployments of archival satellite tags, we inferred mortality events, characterized physiological recovery periods following release, and applied survival mixture models to assess M and PRM. We also evaluated covariate effects on the duration of any recovery period and PRM to inform mitigation. Although large sample sizes involving extended monitoring periods (>90 days) would be optimal to directly estimate M from survival data, it was possible to constrain estimates and infer probable values for both species. Furthermore, the consistency of M estimates with values derived from longevity information suggests that age determination is relatively accurate for these species. Regarding bycatch mitigation, our analyses suggest that juvenile porbeagle are more susceptible to harm during capture and handling, that keeping lamnid sharks in the water during release is optimal, and that circle hooks are associated with longer recovery periods for shortfin mako.

[EN] Nowadays, the video-game industry has grown so much that it has transformed this sector into one of the most productive and important in the world. Such is its weight in society, that even the field of Artificial Intelligence has turned with it, giving fruit to various investigations aimed at studying the behavior of AI in video games. The seemingly trivial idea that ¿a computer can defeat a human¿ is not as far away as it seems, but thanks to the highly controlled environments that video-games offer us, computer science has found a safe box of experimentation thus demonstrating that a video-game environment can simulate real situations, beyond entertainment. Reinforcement Learning has proven to be a general and flexible approach to the development of agents who play video-games. However, developing efficient reinforcement agents, especially for end-reward video games, is often very expensive and the simulation requires an enormous number of iterations to obtain an effective policy. The goal of this project is to study the representation and use of domain-specific knowledge from a video-game to guide the exploration of a reinforcement learning agent and reduce in this way the episodes necessary to master the game. The work is based on the reward machine¿s concept, a type of finite state machine that takes an abstract description of the environment as input and returns different reward functions at different instants of time depending on the reward machine¿s transitions. In particular, we apply the concept of the reward machine to the Google Research Football environment using deep reinforcement learning techniques., [ES] Hoy por hoy, la industria de los videojuegos ha crecido tanto que ha trasformado este sector en uno de los más productivos e importantes del mundo. Tal es su peso en la sociedad, que incluso el campo de la Inteligencia Artificial se ha volcado con él, dando fruto a diversas investigaciones orientadas a estudiar el comportamiento de la IA en los videojuegos. La idea, aparentemente trivial, de que ¿un ordenador pueda derrotar a un humano¿ no está realmente tan lejos, si no que gracias a los entornos tan controlados que nos ofrecen los videojuegos la ciencia de la computación ha encontrado un cuadro seguro de experimentación, demostrando así que, mas allá del entretenimiento, un entorno de videojuegos puede simular situaciones absolutamente reales. El Aprendizaje por Refuerzo ha demostrado ser una aproximación general y flexible para el desarrollo de agentes que juegan a videojuegos. Sin embargo, desarrollar agentes de refuerzo eficientes, especialmente para videojuegos de recompensa final, suele ser muy costoso y la simulación requiere un número enorme de iteraciones hasta obtener una política efectiva. El objetivo de este proyecto es estudiar la representación y utilización de conocimiento específico del dominio de un videjuego para guiar la exploración de un agente de aprendizaje por refuerzo y reducir así los episodios necesarios para dominar el juego. El trabajo se sustenta en el concepto de máquina de recompensa, un tipo de máquina de estados finitos que toma una descripción abstracta del entorno como entrada y devuelve diferentes funciones de recompensa en diferentes instantes de tiempo dependiendo de las transiciones de la máquina de recompensa. Particularmente, aplicamos el concepto de máquina de recompensa al Google Research Football environment utilizando para ello técnicas de aprendizaje por refuerzo profundo., [CA] Ara com ara, la indústria dels videojocs ha crescut tant que ha transformat aquest sector en uno dels més productius e importants del món. Tal és el seu pes en la societat, que fins i tot el camp de la Inteligencia Artificial s’ha volcat amb ell, donant frut a diverses investigacions orientades a estudiar el comportament de la IA en els videojocs. La idea, aparenment trivial, de que “un ordinador puga derrotar a un humà” no està realment tan lluny, si no que gràcies als entorns tan controlats que ens ofereixen els videojocs, la ciència de la computació ha trobat un cuadre segur d’experimentació, demostrant aixina que, més enllà de l’entreteniment, un entorn de videojocs pot simular situacions absolutament reals. L’Aprenentatge per Reforç ha demostrat ser una aproximació general i flexible per al desenvolupament d’agents que juguen a videojocs. No obstant això, desenvolupar agents per reforç eficients, especialment per videojocs de recompensa final, sol ser molt costós i la simulació requereix un número enorme de iteracions fins a obtindre una política efectiva. L’objectiu d’aquest projecte és estudiar la representació i utilització de coneixement específic del domini d’un videojoc per guiar l’exploració d’un agent d’aprenentatge per reforç i reduir així els episodis necesaris per a dominar el joc. El treball es sustenta en el concepte de màquines de recompensa, un tipus de màquina d’estats finits que pren una descripció abstracta de l’entorn com entrada i torna diferents funcions de recompensa en diferents instants de temps depenent de les transaccions de la màquina de recompensa. Particularment, apliquem el concepte de màquina de recompensa al Google Research Football environment utilitzant per a ell tècniques d’aprenentatge per reforç profund.

[ES] Actualmente, la industria de los videojuegos, se ha convertido en una explotación cuya continua expansión es innegable. Como fruto de ello, han llegado al mercado diferentes herramientas que facilitan el desarrollo de estos. A lo largo del tiempo, ha ido evolucionando no solo el concepto de videojuego en sí, sino también el de videoconsola, hasta el punto de encontrar en el mercado, una gran variedad de hardware que compite con una serie de especificaciones realmente avanzadas a los origines de estas, las máquinas recreativas. Así, este proyecto consiste en el desarrollo de un videojuego ambientado en el entorno de las máquinas recreativas, la llamada temática “arcade”, que comprende un estilo de lucha clásico y que se ha desarrollado con la herramienta de desarrollo de videojuegos Unity., [CA] Actualment, l’industria dels videojocs, s’ha convertit en una explotació de la qual es innegable parlar de las seua continua expansió. Com fruït d’això, han aplegat al mercat diferents ferramentes que faciliten el desenvolupament d’aquests. Al llarg del temps, ha anat evolucionant no sols el concepte de videojoc en sí, si no també el de videoconsola, fins al punt de trobar al mercat, una gran varietat de hardware que competeix amb una sèrie d’especificacions realment avançades als orígens d’aquestes, les màquines recreatives. Així, aquest projecte consisteix en el desenvolupament d’un videojoc ambientat en el entorn de les màquines recreatives, l’anomenada temàtica “arcade”, que compren un estil de lluita clàssic i que s’ha desenvolupat amb la ferramenta de desenvolupament de videojocs Unity., [EN] Nowadays, the industry of video games has become a one whose continuous expansion is undeniable. As a result of this, many different tools have arrived at market that allow us to develop video games. Over time, the concept of video games has evolved and, most important, the concept of video game consoles too. Now, you can see in the market a lot of hardware that compete with specifications that are too much far of his origins, the arcade machines. This project consists in a development of a video game with the environment of arcade machines, that it has a classic fighting style, and it has been developed with Unity, a huge tool to develop video games.

The blue shark (Prionace glauca) is the most frequently captured shark in pelagic oceanic fisheries, especially pelagic longlines targeting swordfish and/or tunas. As part of cooperative scientific efforts for fisheries and biological data collection, information from fishery observers, scientific projects and surveys, and from recreational fisheries from several nations in the Atlantic and Indian Oceans was compiled. Data sets included information on location, size and sex, in a total of 478,220 blue shark records collected between 1966 and 2014. Sizes ranged from 36 to 394 cm fork length. Considerable variability was observed in the size distribution by region and season in both oceans. Larger blue sharks tend to occur in equatorial and tropical regions, and smaller specimens in higher latitudes in temperate waters. Differences in sex ratios were also detected spatially and seasonally. Nursery areas in the Atlantic seem to occur in the temperate south-east off South Africa and Namibia, in the south-west off southern Brazil and Uruguay, and in the north-east off the Iberian Peninsula and the Azores. Parturition may occur in the tropical north-east off West Africa. In the Indian Ocean, nursery areas also seem to occur in temperate waters, especially in the southwest Indian Ocean off South Africa, and in the south-east off south-western Australia. The distributional patterns presented in this study provide a better understanding of how blue sharks segregate by size and sex, spatially and temporally, and improve the scientific advice to help adopt more informed and efficient management and conservation measures for this cosmopolitan species.

The blue shark is the most captured shark in pelagic longline fisheries targeting tunas and swordfish. As part of an ongoing cooperative program for fisheries and biological data collection , information collected by fishery observers and scientific Project from several fishing nations in the Atlantic (EU.Spain, EU.Portugal, Uruguay, Taiwán, USA, Japan, Brasil, Venezuela and South Africa) were analyzed. Datasets include information on geographic location, size and sex. A total of 414428 blue sharks records collected between 1992 and 2014 were compiled, with the sizes ranging from 36 to 394 cm FL (fork length). Considerable variability was observed in the size distribution by región and season, with larger sizes tending to ocsur in equatorial and tropical regions and smaller sizes in higher latitudes. The expected distribution of juvenile and adult specimens also showed considerable variability, and the sex ratios varied between regions and sizes classes. The distributional patterns presented in this study provide a better understanding of different aspects of this species in the Atlantic that can help to promote more informed managemente and conservation measures.