Changelog • positionR

positionR 1.5.0

Effort-balanced non-detection weighting: New scale_non_detections parameter (default TRUE) in calculate_fish_positions() scales non-detection evidence by total detections per fish/time period, balancing the detection-weighted mean that favours high-count stations.

data.table conversion: All internal WADE helper functions converted from dplyr to data.table for ~10x speedup on large datasets. Affects aggregate_detections_for_prediction(), create_non_detections(), aggregate_non_detections(), aggregate_probability(), normalize_DE_by_station(), and information weighting helpers.
Vectorized barrier detection: New check_line_crosses_barrier_vectorized() calls raster::extract() once per station instead of once per cell (~20-100x faster for barrier checking in calculate_station_distances()).
Optimized reshaping: tidyr::pivot_longer() replaced with data.table::melt() in calculate_station_distances().

Reduced console output verbosity: calculate_station_distances() progress line only in interactive sessions; removed per-station detection rate printout from analyze_detection_performance().
New dependency: data.table added to Imports.

New default integration method: calculate_fish_positions() now defaults to integration_method = "subtractive", which produces tighter position estimates anchored to actual detection zones. The previous additive method inflated spatial footprint by assigning positive probability everywhere non-detecting stations had low detection efficiency. To replicate previous behaviour, set integration_method = "additive".

Integration method parameter: New integration_method parameter in calculate_fish_positions() with three options:
- "subtractive" (new default): Detection field minus non-detection penalty, clamped to 0. Produces detection-anchored positions.
- "multiplicative": Detection field scaled down proportionally by non-detection evidence. Smoother penalty than subtractive.
- "additive": Original WADE formula preserved for backward compatibility.
For subtractive and multiplicative methods, detection_weight is no longer used (detection is always the base); only non_detection_weight controls non-detection penalty strength (0-1).
integrated_prob values are now rescaled to [0, 1] per fish/time period after integration, ensuring consistent probability ranges across all integration methods.

Depth-dependent behavioural state bias: New depth_state_bias parameter in simulate_fish_tracks() biases state transition probabilities based on fish depth relative to species-specific depth thresholds, producing more ecologically realistic movement patterns.
Goal-directed movement (BCRW): New goal_locations, goal_bias, and goal_tolerance parameters enable biased correlated random walks where fish navigate toward specified destinations with tunable directedness.
path_color parameter: New option in plot_fish_tracks() to override automatic colouring with a single colour for all paths.
New vignette: Simulation_Goal_Directed.R demonstrates goal-directed movement across a range of bias strengths.

Added barrier masking functionality to prevent unrealistic detections through land obstacles (islands, peninsulas, shorelines)
New include_barriers parameter available in four core functions:
- calculate_station_distances(): Automatically detects line-of-sight barriers using ray-tracing algorithm
- calculate_detection_system(): Incorporates barriers into system-wide coverage calculations
- simulate_fish_tracks(): Prevents simulated detections through land obstacles
- calculate_fish_positions(): Accounts for barriers in both detection and non-detection probabilities

Line-of-sight analysis identifies when direct path between receiver and location crosses land
Detection efficiency automatically set to 0 where barriers present
Works with both static DE mode (pre-computed values) and temporal DE mode (on-the-fly calculation)
Generates crosses_barrier column in station distances data for downstream use

Enhanced realism in detection simulations by accounting for physical geography
More accurate position estimates that respect landscape features
Updated vignettes with barrier masking examples and visualizations:
- Array Design & Simulation vignette: barrier field visualization and detection masking
- WADE Positioning vignette: barrier-aware positioning workflow

Added run_array_design() function to launch interactive Shiny app for acoustic receiver array design and optimization
App includes four main modules:
- Array Design: Interactive map for manual receiver placement and regular array generation
- Range Modeling: Depth-dependent detection efficiency model fitting
- Array Performance: Quantitative coverage assessment and sampling bias analysis
- Instructions: Comprehensive methodology guide and best practices

Integrated array design app with existing package functions (generate_regular_points(), create_logistic_curve_depth())
Enhanced namespace handling to avoid conflicts between raster and dplyr packages
Added comprehensive documentation for array design workflow

Initial CRAN-ready release
Comprehensive tools for acoustic telemetry array design, fish movement simulation, and WADE positioning methodology