Search for: All records

Abstract BackgroundAnimal movement data are increasingly used to make ecological inferences, as well as to inform conservation and management actions. While advanced statistical methods to estimate behavioral states from these datasets have become widely available, the large number to choose from may make it difficult for practitioners to decide which method best addresses their needs. To guide decisions, we compared the behavioral state estimates and inferences from three methods (movement persistence models [MPM], hidden Markov models [HMM], and mixed-membership method for movement [M4]) when analyzing animal telemetry data. Tracks of post-breeding adult male green sea turtles (Chelonia mydas) were treated as an empirical example for this method comparison. The effect of temporal scale on behavioral state estimates was also investigated (at 1, 4, and 8 h time steps). ResultsThe HMM and M4 models produced relatively similar behavioral state estimates (compared to the MPM) and estimated anywhere from three to five states depending on the time interval of the tracks and the method used. Likewise, for all three methods, sampling movement at coarser time scales smoothed estimates of behavioral transitions. Additionally, the selection of movement metrics for analysis by the HMM and M4 also appeared to be a critical decision regarding state estimation and interpretation. At the longest time step (8 h), all three models were able to distinguish area-restricted search (ARS) behavior from migratory behavior, with greater nuance estimated by the HMM and M4 methods. By comparison, the MPM was the only model that was able to identify fine-scale behavioral patterns when analyzing the shortest time step (1 h). Moreover, the analysis of tracks with short time steps via MPM identified likely periods of resting during long-distance migration, which had only previously been hypothesized in green turtles. ConclusionsWhile there is no single best method to estimate behavioral states, our findings demonstrate that results can vary widely among different statistical methods and that model assumptions should be thoroughly checked during the model fitting process to reduce any potential biases. Thus, practitioners should carefully consider which methods best address their needs while also accounting for the inherent properties of their telemetry dataset. 

In many species, demographic assessments of population viability require an estimate of the number or proportion of breeding adults in a population that are male (the breeding sex ratio). However, this estimate is often difficult to obtain directly in species with multiple paternity when males are difficult to sample. Parentage analysis of breeding females and offspring can produce this estimate by identifying the number of unique males that contribute genetic information to (i.e., sired) a given cohort. There is an added challenge of choosing a sample design with the desired level of confidence to identify all the fathers contributing to a cohort, either at the scale of individual clutches or an entire nesting season, given limited resources. Sampling effort can be defined as the number of offspring sampled per clutch, or the number of clutches sampled per breeding season, depending on the analysis. The minimum number of samples required may depend on the proportions of eggs that different fathers fertilize in a clutch (the paternal contribution mode), the total number of fathers fertilizing a clutch, the proportion of adults available for breeding that are male (the operational sex ratio), and population size. We conducted power analyses to quantify the confidence in identifying all fathers in animal populations with multiple paternity. We simulated sampling a theoretical sea turtle population with a range of population demographics, mating systems, and sampling effort, and used the proportion of 10,000 simulations in which all fathers were identified as a proxy for confidence. At the clutch level, confidence was strongly dependent on the paternal contribution mode, and when it was skewed, it also depended on the total number of fathers contributing and the number of offspring sampled. However, sampling about one third of a clutch was sufficient to identify all fathers with high confidence for most scenarios, unless the paternal contribution mode was extremely skewed and there were many contributing fathers, such that some fathers fertilized very few eggs and were difficult to detect. At the scale of an entire nesting season, confidence was more strongly affected by the operational sex ratio, the proportion of clutches sampled, and the presence or absence of polygyny than by the lesser effects of paternal contribution mode and within-clutch sample size. Sampling fewer offspring from more clutches increased confidence compared to sampling more offspring from fewer clutches. Relaxing the minimum required proportion of fathers identified from 100% to 90% led to high confidence while sampling 50% to a maximum of 75% of clutches, depending on the mating system, even as the population size increased by an order of magnitude. Our approach and results can be widely informative for sample design as well as quantifying uncertainty in existing and future estimates of the number of breeding males in populations with multiple paternity. 

Abstract Entender las respuestas de la población a perturbaciones ambientales, específicamente a pulsadas individuales, es esencial para la conservación y la gestión adaptativa. Las poblaciones de interés pueden reducirse a niveles bajas debido a la perturbación, y es necesario entender las diferencias interespecíficas en las trayectorias de recuperación para evaluar las opciones de gestión. Analizamos modelos para especies individuales para investigar los factores demográficos y de gestión que determinan los dos componentes de la ‘resiliencia’ de la población: la magnitud del impacto inicial sobre la abundancia de la población y la duración del tiempo de recuperación. Simulamos poblaciones estructuradas por edad con reclutamiento que depende de la densidad, las sometimos a una perturbación pulsada que consiste en un período de mayor mortalidad del grupo etário juvenil o de todos los grupos etários, y calculamos tanto el impacto como el tiempo de retorno. A modo de ilustración, utilizamos parámetros demográficos de un conjunto de 16 especies de peces. Formulamos el modelo como una ecuación de renovación, lo que nos permite describir matemáticamente los impactos de las perturbaciones como una convolución. También incluimos dinámicas no lineales que representan poblaciones que se recuperan hacia un estado estable; esto es más realista (en la mayoría de los casos) que los análisis previos de resiliencia en modelos lineales sin la dependencia de la densidad. Cuando la perturbación ha afectado a uno o a algunos pocos grupos etários jóvenes, la longevidad fue el principal determinante de la historia de vida del impacto y el tiempo de recuperación. Las especies de vida más corta sufrieron mayores impactos cuando fueron perturbadas porque cada grupo etáreo representa una mayor proporción de la población. Sin embargo, las especies con vidas más cortas también tuvieron tiempos de recuperación más rápidos, por la misma razón. Cuando la perturbación afectó a los grupos etários adultos, el impacto fue más inmediato y ya no se vio afectado por la longevidad de las especies, aunque se mantuvo el efecto de la longevidad sobre el tiempo de recuperación. Estos resultados mejoran nuestra comprensión de las diferencias interespecíficas de la resiliencia y aumentan nuestra capacidad para hacer predicciones con fin a la gestión adaptativa. Además, formular el problema como una ecuación de renovación y usar convoluciones matemáticas nos permite cuantificar cómo las perturbaciones con distintos lapsos de tiempo (no solo un nivel de perturbación constante e inmediato, sino niveles de perturbación que aumentan o disminuyen gradualmente) tendrían diferentes efectos sobre la resiliencia de la población: respuestas tardías para especies en las que la biomasa se concentra en grupos etários de mayor edad y para perturbaciones que se vuelven progresivamente más severas. 

Abstract Adaptive management of marine protected areas (MPAs) to determine whether they are meeting their intended goals requires predicting how soon those goals will be realized. Such predictions have been made for increases in fish abundance and biomass inside MPAs. However, projecting increases in fishery yield (“fishery spillover”) is more complex because it involves both how the fishery is managed and uncertainty in larval connectivity. We developed a two-patch, age-structured population model, based on a renewal equation approach, to project the initial timing of increase in fishery yield from larvae exported from a no-take MPA. Our results link our understanding of the predicted timing of increases in biomass (and thus reproduction) in MPAs with the time-lags associated with new recruits entering the fishery. We show that the time-lag between biomass peaking within the MPA and the increased fishery yield outside the MPA reaching its maximum depends, in a predictable way, on the age-dependent patterns of growth, natural mortality, and fishing mortality. We apply this analysis to 16 fishery species from the US Pacific coast; this difference ranged from 7 to 18 years. This model provides broadly applicable guidance for this important emerging aspect of fisheries management.