Data and scripts used in "Age-dependent patterns of spatial autocorrelation in fish populations"

dataset

posted on 2021-06-23, 08:14 authored by Jonatan MarquezJonatan Marquez, Sondre Aanes, Are Salthaug

The population spatial autocorrelations were estimated using data from scientific bottom trawl surveys performed annually by the Norwegian Institute for Marine Research and the Polar Research Institute of Marine Fisheries and Oceanography from January to March, from 1985 to 2016 (Jakobsen et al. 1997; Aanes & Vølstad 2015). The trawl survey was spatially stratified and sampled locations were approximately uniformly distributed in space. The survey has been mostly standardized with respect to sampling gear and performance, except for a reduction in the mesh size of the codend from 35-40 mm to 22 mm in 1994 to prevent potential sampling size bias among 1-year-old cod and haddock. For more details on sampling protocols see Jakobsen et al. (1997), Johannesen et al. (2009), Fall et al. (2020). The fish were sampled onboard, following the instructions given in Mjanger et al. (2020) and otoliths were collected to determine the age of the individuals (Johannesen et al. 2009, Mehl et al. 2016). When the catch so big that length-measuring the entire catch was unfeasible, a representative random subsample was measured. From this subsample, otoliths to age the fish were collected for an extra subsample of the fish, following a length stratified sampling design. Before 1993, 5 individuals per 5 cm length group were aged for a spatially stratified subset of trawls, from 1993 to 1995 only 2 individuals per 5 cm length group were aged, but for a larger subset of trawls. Since 1996, 1 individual per 5 cm length group has been aged in all trawls. Lastly, the collected data were then used to make age-length keys to raise or extrapolate the age distribution of each catch. In total 8288 trawls were performed, where 7037 contained haddock, 8145 contained cod and 5153 contained beaked redfish.

The study region was subdivided using a grid with hexagonal cells because this shape homogenises the distances between centroids of neighbouring cells. The data presented here correspond to a grid cell size resolution of 6400 km2. In addition, the analysis was repeated after shifting the hexagonal grid along the latitudinal and longitudinal gradients 15 times. This resulted in slight differences in how samples were grouped to average cell densities, preventing grid cells with fewer samples from generating outliers that could cause biases in the results.

Files:

Data:

ylu_6400_list.rda

yu_6400_list.rda

zlu_6400_list.rda

zu_6400_list.rda

Scripts:

SI_Functions.R

SI_Hexagon.R

SI_DensityMaps.v1.R

SI_Synchrony.v1.R

SI_Synchrony_Ages.v1.R

Description

Data:

ylu_6400_list.rda – List containing arrays with average density estimates by grid cell for each species and age class. Each array is named as: “species name” (i.e. Cod, Haddock or B_redfish), “_spatial offset of the grid in degrees” (e.g. -1). The dimensions of the arrays correspond to [Years, Grid id, Age class]

yu_6400_list.rda - List containing matrix with total average density estimates by grid cell for each species. Each matrix is named as: “species name” (i.e. Cod, Haddock or B_redfish), “_spatial offset of the grid in degrees” (e.g. -1). The dimensions of the matrix correspond to [Years, Grid id]

zlu_6400_list.rda - List containing arrays with the distances in km between the grids corresponding to the data in “ylu_6400_list.rda”. Each array is named as: “species name” (i.e. Cod, Haddock or B_redfish), “_spatial offset of the grid in degrees” (e.g. -1). The dimensions of the arrays correspond to [Grid id, Grid id, Age class]

zu_6400_list.rda - List containing matrix with the distances in km between the grids corresponding to the data in “yu_6400_list.rda”. Each array is named as: “species name” (i.e. Cod, Haddock or B_redfish), “_spatial offset of the grid in degrees” (e.g. -1). The dimensions of the matrix correspond to [Grid id, Grid id]

Scripts:

SI_Functions.R – Compilation of functions used in the other of the scripts.

SI_Hexagon.R – R script to create the hexagonal grid overlaid across the Barents Sea (i.e. hex_grid_LL_M). Specify the resolution of the grids. The default in the uploaded data and scripts is a<-6400

SI_DensityMaps.v1.R – R script to map the density by grid cells of each study species. Corresponds to Figure 1.

SI_Synchrony.v1.R – R scripts to estimate and plot the spatial autocorrelation of the population and the two life stages (i.e. juveniles or immature, and adults or mature). The spatial autocorrelation estimates (i.e. SynchResults) are represented by three parameters:

logitrhoinf = degree of autocorrelation at infinity

logitrhoinf = degree of autocorrelation as distance approaches 0

logl = spatial scaling (i.e. standard deviation of the gaussian function optimized over the spatial autocorrelation data

The output data is then plotted to:

Synchronies.6400.tiff - Spatial autocorrelation plot with each species at the population level and life stage level (juveniles and adults). Corresponds to Figure 2.

SI_Synchrony_Ages.v1.R – R scripts to estimate and plot the spatial autocorrelation of each age class. The spatial autocorrelation estimates (i.e. SynchResults) are structured in the same way as in “SI_Synchrony.v1.R”. The spatial autocorrelation estimates are then plotted in the following ways:

1. SynchronyDifferences_6400.tiff = Plots the change in the spatial autocorrelation parameters after accounting for the effect of local cohort dynamics. i.e. Figure 4.

2. RidgesSynch_6400.tiff = Ridges plots showing the density distribution of the bootstrapped spatial autocorrelation parameters of density and cohort-independent density for cod and haddock, and ages 1 to 8. i.e. Figure 3.

3. SynchAgesRibbons_Den6400.tiff = Spatial autocorrelation plots corresponding to age classes 1 to 8 of cod and haddock. i.e. Figure S2 in Appendix.