Skip to main content
Dryad logo

Spatial covariation of fish population vital rates in a stream network

Citation

Tsuboi, Jun-ichi et al. (2020), Spatial covariation of fish population vital rates in a stream network, Dryad, Dataset, https://doi.org/10.5061/dryad.jm63xsj6t

Abstract

Animal populations are spatially structured in heterogeneous landscapes, in which local patches with differing vital rates are connected by dispersal of individuals to varying degrees. Although there is evidence that vital rates differ among local populations, much less is understood about how vital rates covary among local patches in spatially heterogeneous landscapes. In this study, we conducted a 9-year annual mark-recapture survey to characterize spatial covariation of survival and growth for two Japanese native salmonids, white-spotted charr (Salvelinus leucomaenis japonicus) and red-spotted masu salmon (Oncorhynchus masou ishikawae), in a headwater stream network composed of distinctly different tributary and mainstem habitats. Spatial structure of survival and growth differed by species and age class, but results provided support for negative covariation between vital rates, where survival was higher in the tributary habitat but growth was higher in the mainstem habitat. Thus, neither habitat was apparently more important than the other, and local habitats with complementary vital rates may make this spatially structured population less vulnerable to environmental change (i.e., portfolio effect). Despite the spatial structure of vital rates and possibilities that fish can exploit spatially distributed resources, movement of fish was limited due partly to a series of low-head dams that prevented upstream movement of fish in the study area. This study shows that spatial structure of vital rates can be complex and depend on species and age class, and this knowledge is likely paramount to elucidating dynamics of spatially structured populations.

Methods

Field surveys were conducted annually (the third weekend of October) in 2009–2017. Fish were captured using a backpack electrofishing unit (300–400 V DC, model 12B or LR20, Smith-Root, Inc., Vancouver, WA, USA) and 3-mm mesh dip nets. Two passes of electrofishing were conducted for fish density estimates of each section with a depletion method (Zippen 1958). Captured fish were anaesthetised with phenoxyethanol (ca. 0.5 ml/L water), measured for fork length (FL: nearest 1 mm), and were marked individually with visible implant elastomer tags (Northwest Marine Technology Inc., WA, USA) or their individual code was recorded if recaptured. A unique combination of four elastomer colours were subcutaneously administered to the forehead of each individual. All captured fish with FL > 43 mm were marked. During each year of the study, juveniles (young-of-the-year, YOY) and adults (age 1+ and older) were distinguished based on length-frequency histograms. For individuals which cannot be assigned to an age class due to their intermediate body size, a few scales were taken using a scalpel and the annuli were counted. All fish were returned alive to the capture site (< 20 m for mainstem, < 40 m for tributaries) after recovering from anaesthesia. All captured individuals retained at least two of the four elastomer colours, and were marked again with the lost colour(s). We identified all individuals uniquely based on species, sex, body size, and study section at mark (i.e., asymmetrical movement at dams).

Usage Notes

A multi-state Cormack-Jolly-Seber (CJS) model was developed to infer state-specific annual survival, capture and transition probabilities for charr and salmon. States were defined based on age (juveniles and adults) and location (tributary, mainstem connected, and mainstem fragmented). Because locations were considered states, the state-transition probability referred to the movement probability among locations, which varied by age. To account for imperfect recapture of individuals, a state-space model was developed, composed of an ecological model following the state transitions over time and an observation model linking the true states to observed data.

Data were formatted in a two-dimensional array (yi,t), where rows indicated individuals (i) and columns indicated annual sampling occasions (t). Elements of the array were observed states defined by age and location. A unique combination of ages (juvenile and adult) and locations (tributary, mainstem connected and mainstem fragmented) resulted in six states (1 = juveniles in tributary; 2 = juveniles in mainstem connected; 3 = juveniles in mainstem fragmented; 4 = adults in tributary; 5 = adults in mainstem connected; and 6 = adults in mainstem fragmented). In addition, sampling occasions on which individuals were not detected were coded as 7. Prior to fitting multi-state CJS models to data, we used the R2ucare package (Gimenez et al. 2018) to confirm that there was no evidence for trap-dependence (p > 0.27), transience (p = 1.00), and memory of past states (p > 0.35) in either charr or salmon.