Applications of the flapper family of algorithms

Edward Lavender^1,2*

¹ Scottish Oceans Institute, University of St Andrews, Scotland
² Centre for Research into Ecological and Environmental Modelling, University of St Andrews, Scotland

^* This repository is maintained by Edward Lavender (el72@st-andrews.ac.uk).

Figure 1. A flapper skate (Dipturus intermedius). Photograph courtesy of the Movement Ecology of Flapper Skate project.

Introduction

The flapper family of algorithms is a suite of mechanistic approaches designed to reconstruct fine-scale movement paths and emergent patterns of space use from discrete detections in passive acoustic telemetry arrays. This repository illustrates applications of these algorithms to real-world movement (acoustic and archival) data collected from flapper skate (Dipturus intermedius) tagged in the Loch Sunart to the Sound of Jura Marine Protected Area (West Scotland) by the Movement Ecology of Flapper Skate (MEFS) project in 2016–17. Four analyses are implemented:

• A1: Depth use. The depth-contour (DC) algorithm is used to examine the depth use of a selected individual over a one-month period in the MPA.
• A2: Space use. The mean-position, acoustic-container particle filtering (ACPF) and acoustic-container depth-contour particle filtering (ACDCPF) algorithms are used to reconstruct patterns of space use for a selected individual over a one-month period in the MPA.
• A3: Post-release paths. The depth-contour particle filtering (DCPF) algorithm is used to reconstruct fine-scale post-release movement paths of two individuals suggested to exhibit irregular post-release behaviour following catch-and-release angling in the MPA.
• A4: Coocccurrences. The ACDCPF algorithm is used to reconstruct fine-scale movement paths of two individuals during a period of cooccurring detections to examine evidence for close-knit interactions versus fine-scale spatial partitioning.

Figure 2. Example outputs from the flapper_appl project showing reconstructed movement paths of a selected flapper skate following catch-and-release angling in the Loch Sunart to the Sound of Jura Marine Protected Area. The background shows the bathymetry (in blue), the individual’s release location (in black) and selected paths over an 80-minute period. The reconstructed paths all indicate that the individual descended into a deep-water channel following angling, before rapidly re-ascending via one of two routes into the shallow water around a small island.

Prerequisites

The analyses are written in R and organised as an R Project. For data processing and analysis, the flapper R package is required. For visualisation, prettyGraphics is used, which is a dependency in flapper. For quick data summaries, the utils.add package is used on a few occasions.

Structure

1. data-raw/ contains raw data for the project.

   • movement/ contains raw movement data from the MEFS project:
     • skateids.rds is a dataframe that records tagged individuals and their characteristics;
     • moorings.rds is a dataframe that records acoustic receiver deployments;
     • acoustics.rds is a dataframe that records acoustic detections;
     • archival.rds is a dataframe that records depth observations;
     • dat_iprb.rds is a dataframe that records depth observations around catch-and-release angling events;
   • spatial/ contains spatial data for the study area:
     • bathy/ contains a 5 x 5 m bathymetry raster (named bathy_res_full_ext_full_abs.tif), sourced from Howe et al. (2014);
     • coastline/ contains a 1:10,000 coastline layer (named westminster_const_region.shp) from Digimap;
     • sediments/ contains a map of sediment types (named HI1354_Sediment_Map_v2_WGS84.shp), sourced from Howe et al. (2014) and Boswarva et al. (2018);
   • process_data_raw.R processes raw data as required for each analysis.

2. data/ contains data for the project.

   • movement/contains processed movement time series (from process_data_raw.R) and analysis-specific algorithm outputs;
   • skate/ contains skate datasets, copied from movement/ for publication in this repository:
     • A1-2 contains skate datasets required for A1 and A2:
       • moorings.rds contains passive acoustic telemetry deployment information (copied from movement/generic/);
       • moorings_xy.rds contains receiver deployment locations (copied from spatial/);
       • acoustics_eg.rds is the example acoustic time series (copied from movement/tag/);
       • archival_eg.rds is the example archival time series (copied from movement/tag/);
     • A3 contains skate datasets required for A3 (copied from movement/post_release_paths/):
       • 1507/ contains the data for individual 1507, including the release location (xy_release.rds) and the post-release time series (archival_pr.rds);
       • 1558/ contains the same datasets for individual 1558;
     • A4 contains skate datasets required for A4 (copied from movement/cooccurrences/):
       • acc_1.rds and arc_2.rds contain the acoustic time series for individuals 542 and 560 respectively;
       • arc_1.rds and arc_2.rds contain the archival time series for the same individuals;
   • spatial/ contains processed spatial data (from process_data_raw.R);
   • tmp/ stores temporary files;

3. R/ contains R scripts that implement analyses.

   • define_global_param.R defines global parameters, such as projections, detection and movement parameters;
   • define_study_area_fields.R defines spatial fields for mapping the study area;
   • examine_depth_use.R implements A1;
   • examine_space_use.R implements A2, supported by examine_space_use_time_trials.R, examine_lcps.R and examine_habitat_preferences.R;
   • examine_post_release_paths.R implements A3;
   • examine_coocccurrences.R implements A4;

4. fig/ contains figures.

Note that data-raw/, data/* (except data/skate/) and fig/ are not included in the online version of this repository.

Workflow

1. Set up. Install project dependences (such as flapper) and set up the R Project, including the directory system (as outlined above and in the R scripts). It is desirable to initiate the project on a system with a capacity of at least 4 TB (e.g., an external hard drive) as some routines generate large numbers of files.

2. Data availability. Spatial and movement data need to be obtained and processed. Unfortunately, this repository cannot be published with all the spatial and movement data required to implement the project due to third party restrictions. However, the spatial datasets can be accessed via the references provided above and processed via process_data_raw.R. The raw movement data were collected by NatureScot and Marine Scotland Science and made available for this study by these organisations. Requests to access these data from NatureScot and Marine Scotland Science can be facilitated. In the meantime, the processed skate data (from process_data_raw.R) required to run R scripts are archived in the data/skate/ directory. These can be manually copied into the directory system defined above to run R scripts.

3. Define global parameters. Define global parameters via define_global_param.R and study area fields via define_study_area_fields.R.

4. Implement algorithms. Implement A1–4 via examine_depth_use.R, examine_space_use.R (together with examine_space_use_time_trials.R, examine_lcps.R and examine_habitat_preferences.R), examine_post_release_paths.R and examine_cooccurrences.R respectively.

5. Examine results. Examine reconstructed patterns of depth and space use and their implications in analyses of habitat preferences; post-release movement paths; and fine-scale spatial partitioning during periods of cooccurring detections for the selected individuals.

Figure 3. Example outputs of the flapper_appl project showing the most likely reconstructed path (from Figure 2) over an 80-minute period (purple–yellow).

References

Boswarva et al. (2018). Improving marine habitat mapping using high-resolution acoustic data; a predictive habitat map for the Firth of Lorn, Scotland. Continental Shelf Research, 168, 39–47. https://doi.org/10.1016/j.csr.2018.09.005

Howe et al. (2014). The seabed geomorphology and geological structure of the Firth of Lorn, western Scotland, UK, as revealed by multibeam echo-sounder survey. Earth and Environmental Science Transactions of the Royal Society of Edinburgh, 105(4), 273–284. https://doi.org/10.1017/S1755691015000146

Citation

Lavender et al. (in press). A semi-stochastic modelling framework for passive acoustic telemetry. Methods in Ecology and Evolution.