Package 'oystermapR'

Description

Classifies each survey location as intertidal, subtidal, or supratidal based on its depth below Chart Datum (CD) and the tidal range at the nearest harmonic reference port. Chart Datum is approximately Lowest Astronomical Tide (LAT), so:

• Subtidal: depth > 0 (always submerged)

• Intertidal: ⁠-MHWS_above_CD <= depth <= 0⁠ (exposed at some states of the tide; depth is at or above CD)

• Supratidal: depth < -MHWS_above_CD (above the highest astronomical tide; excluded from scoring)

MHWS above CD is approximated from the harmonic constituents of the nearest port as: Z0 + 1.1 * (M2_H + S2_H) – the 10% factor accounts for minor constituents not included in the 5-constituent model.

This flag is used internally by score_locations() to give Magallana gigas (Pacific oyster) full depth scores for intertidal cells, reflecting its strong intertidal ecology in NW European waters.

Usage

add_intertidal_flag(
  df,
  max_port_dist_km = 75,
  depth_col = "depth",
  verbose = TRUE
)

Arguments

Value

The input dataframe with an added intertidal_zone column:

• "subtidal": depth > 0 m CD

• "intertidal": 0 m CD >= depth >= -MHWS_above_CD

• "supratidal": depth < -MHWS_above_CD

Examples

survey <- auto_tidal_correct(survey, datetime_col = "date")
survey <- add_intertidal_flag(survey)

# Check intertidal coverage
table(survey$intertidal_zone)

Description

Vectorised wrapper around detect_season() for use on dataframes.

Usage

add_season_column(df, date_col = NULL, lat_col = "lat")

Arguments

Value

The input dataframe with an additional column season.

Description

Attaches a regulatory shellfish harvesting classification (A/B/C/Prohibited) to each row of a scored result dataframe. This classification affects the economic viability score for aquaculture applications: Prohibited sites receive viability = 0; Class C sites receive a 0.60 multiplier; Class B a 0.80 multiplier; Class A is unpenalised.

Classification can be supplied three ways (in priority order):

1. class_col – name of an existing column in result already containing classification values ("A", "B", "C", "Prohibited", or NA).

2. classified_areas – a dataframe with lat, lon, and shellfish_class columns, spatially matched to result rows within match_radius_deg degrees.

3. fetch_live = TRUE – queries the FSA England/Wales open data API (requires internet). Requires httr package. Works only for sites within the FSA/DAERA coverage area (England, Wales, Northern Ireland).

Unclassified or unmatched sites receive shellfish_class = "Unclassified" and a penalty of 0.70 (precautionary principle).

Usage

add_shellfish_classification(
  result,
  class_col = NULL,
  classified_areas = NULL,
  fetch_live = FALSE,
  match_radius_deg = 0.05,
  verbose = TRUE
)

Arguments

Value

result with two additional columns:

• shellfish_class: "A", "B", "C", "Prohibited", or "Unclassified"

• shellfish_class_penalty: multiplier applied to economic viability (1.0 for A, 0.80 for B, 0.60 for C, 0.0 for Prohibited, 0.70 for Unclassified)

Examples

# Option 1: manual column already in data
result <- add_shellfish_classification(result, class_col = "water_class")

# Option 2: separate classified areas dataframe
areas <- data.frame(lat = c(51.5), lon = c(-4.2), shellfish_class = c("B"))
result <- add_shellfish_classification(result, classified_areas = areas)

# Option 3: live fetch (internet required)
result <- add_shellfish_classification(result, fetch_live = TRUE)

Description

Re-scores each row of a survey dataframe n_boot times, each time adding Gaussian noise to every numeric variable proportional to its measurement uncertainty, and returns the 5th and 95th percentile suitability scores as suit_ci_lower and suit_ci_upper columns (90% interval).

Measurement uncertainty is specified via the uncertainty argument: a named list where each name matches a variable name in tolerances$scored_factors and the value is the 1-sigma absolute error (same units as the variable). Any variable not listed defaults to 2% of its observed value.

Usage

add_suitability_ci(
  result,
  tolerances,
  n_boot = 200L,
  uncertainty = NULL,
  seed = 42L,
  verbose = FALSE
)

Arguments

Value

The input dataframe with additional columns: suit_ci_lower, suit_ci_upper, suit_ci_width.

Examples

result <- predict_oyster(survey, "ostrea_edulis")
tol    <- get_species_tolerances("ostrea_edulis")
result <- add_suitability_ci(result, tol,
                             uncertainty = c(temperature = 0.3,
                                             salinity    = 0.5))
# Columns suit_ci_lower, suit_ci_upper, suit_ci_width now present
generate_report(result, "report.html")   # map rings scale with CI width

Description

Builds a spatial graph of cells above a suitability threshold and identifies connected components – contiguous patches of suitable habitat. Isolated cells or small patches are flagged because they are at higher risk of:

• Recruitment failure – larvae from isolated populations may be flushed away before settling back in the patch.

• Genetic bottlenecks – small isolated populations have lower allelic diversity and reduced adaptive capacity.

• Local extinction without recolonisation – if a patch is lost, there are no connected source populations to recolonise.

Two cells are considered connected if they are within gap_m metres of each other (default 500 m – approximately the scale of larval dispersal during a tidal cycle at moderate current speeds). Connectivity is computed using a fast union-find (disjoint set) algorithm – no external graph packages required.

Usage

analyse_connectivity(
  result,
  min_suitability = 0.5,
  gap_m = 500,
  min_patch_cells = 3L,
  verbose = TRUE
)

Arguments

Value

Input dataframe with additional columns: patch_id (integer – unique patch identifier; NA for non-suitable cells), patch_size (number of cells in the patch), patch_area_km2 (approximate area based on cell density), connectivity_class ("isolated", "small", "moderate", "large"), is_hub (logical – TRUE for the highest-scoring cell in each patch, useful as candidate introduction points).

Examples

result <- predict_oyster(survey, "ostrea_edulis")
result <- analyse_connectivity(result, gap_m = 500)

# Large well-connected patches are the best restoration targets
good_patches <- subset(result,
  connectivity_class == "large" & suitability_class == "High")

# Count patches
table(result$connectivity_class)

# Visualise -- patch_id maps to distinct colours in QGIS
export_geotiff(result, "connectivity.tif")

Description

Evaluates which aquaculture and restoration gear types are physically viable at each survey location based on depth, current velocity, substrate, and (optionally) wave height. Returns a feasibility score and binary flag for each gear type.

This function is part of the planning module aimed at commercial operators. For pure restoration work, the restoration_reef gear type is always included and is the most relevant output.

Usage

assess_gear_feasibility(
  result,
  gear_types = names(.gear_specs),
  depth_col = "depth",
  current_col = "current_velocity",
  substrate_col = "substrate_hardness_class",
  wave_col = NULL,
  verbose = TRUE
)

Arguments

Value

Input dataframe with added columns per gear type: ⁠gear_<type>_feasible⁠ (logical), ⁠gear_<type>_score⁠ [0-1], plus best_gear (most feasible option) and n_gear_options (count).

Examples

result <- predict_oyster(survey, "ostrea_edulis")
result <- assess_gear_feasibility(result)

# Sites where longline is feasible AND suitability is High
longline_sites <- subset(result,
  gear_longline_feasible & suitability_class == "High")

# Restoration reef sites (no farming licence needed)
reef_sites <- subset(result, gear_restoration_reef_feasible)

Description

Finds the nearest standard port to the survey area, checks whether it is within a configurable distance threshold, predicts tidal heights for every survey timestamp using embedded 5-constituent harmonic models (M2, S2, N2, K1, O1), and applies correct_to_chart_datum() automatically.

No external data or internet connection is required. Harmonic constituent data (amplitude and Greenwich phase lag) for 31 UK and European ports are embedded in the package.

Accuracy: 5-constituent predictions are typically accurate to +/-0.15–0.35 m at the standard port. Accuracy degrades with distance from the port and in areas with complex tidal dynamics (e.g. the Bristol Channel, inner fjords). For safety-critical work, use official UKHO EasyTide predictions and supply them manually via correct_to_chart_datum().

Requirements: The dataframe must contain a datetime column. Values must be parseable by as.POSIXct() (e.g. "2024-06-15 10:30:00" or "2024-06-15"). Date-only strings default to noon UTC.

Usage

auto_tidal_correct(
  df,
  datetime_col = "date",
  depth_col = "depth",
  max_port_dist_km = 75,
  keep_raw = TRUE,
  verbose = TRUE
)

Arguments

Value

The input dataframe with depths corrected to Chart Datum, plus a tidal_height_pred_m column containing the predicted tidal height used for each row.

See Also

correct_to_chart_datum() for manual correction, tidal_port_info() for port datum reference data.

Examples

# Automatic correction -- finds nearest port, predicts heights, corrects depths
survey_corrected <- auto_tidal_correct(survey, datetime_col = "date")

# Tighten the distance threshold (only trust ports within 40 km)
survey_corrected <- auto_tidal_correct(survey, max_port_dist_km = 40)

# See which port was selected and inspect predicted heights
survey_corrected$tidal_height_pred_m

Description

Applies hard-stop exclusion criteria for a given species tolerance profile. Any location that violates one or more exclusion rules receives a flag and is removed from the suitability scoring pipeline.

Exclusion factors checked (where data columns are present):

• General temperature range

• Season-specific temperature floors/ceilings (requires season column)

• Salinity (with temperature-dependent minimum)

• Dissolved oxygen (hypoxia threshold)

Usage

check_exclusions(df, tolerances)

Arguments

Value

The input dataframe with two additional columns:

• excluded: logical – TRUE if the location fails any exclusion criterion.

• exclusion_reason: character – semicolon-separated list of failed criteria, or NA if not excluded.

Column Naming

The function recognises these column names (case-insensitive):

Examples

tol <- get_species_tolerances("ostrea_edulis")
df <- data.frame(
  lat = 51.5, lon = -2.5,
  temperature = 8, salinity = 28, dissolved_oxygen = 7
)
check_exclusions(df, tol)

Description

Uses the intensity of acoustic backscatter from the near-seabed bin(s) to classify substrate type. Hard substrates (gravel, rock, shell hash) scatter more sound energy than soft substrates (mud, fine sand). The function converts raw backscatter (dB re 1 m^-1 or instrument counts) to a categorical substrate class and a continuous hardness index [0, 1].

Classification thresholds (default, adjustable):

The hardness index is a min-max normalisation of the mean near-seabed backscatter across the bottom n_bottom_bins cells. If the column contains calibrated volume backscatter (Sv in dB), set is_sv = TRUE and a sign-flip is applied (less negative Sv = stronger return = harder).

Usage

classify_substrate_from_backscatter(
  df,
  backscatter_col,
  is_sv = FALSE,
  bs_min = NULL,
  bs_max = NULL,
  hard_thresh = 0.8,
  mixed_thresh = 0.5,
  soft_thresh = 0.2,
  output_col = "substrate_hardness",
  verbose = TRUE
)

Arguments

Value

Input dataframe with substrate_hardness_index and substrate_hardness_class columns appended. These map directly to the substrate_hardness scored variable in the tolerance spec.

Examples

# ADCP data with near-seabed backscatter column
survey <- classify_substrate_from_backscatter(survey,
            backscatter_col = "bottom_backscatter_db",
            is_sv = TRUE)

# The substrate_hardness_index column feeds directly into predict_oyster()
# as the 'substrate_hardness' variable
names(survey)[names(survey) == "substrate_hardness_index"] <- "substrate_hardness"
result <- predict_oyster(survey, "ostrea_edulis")

Description

Runs predict_oyster() for every specified species against the same survey dataset and returns a combined comparison table – one row per survey location, one suitability column per species. An optional summary prints which species is best-suited to each location.

This is the tool to use when you haven't yet decided which species to stock, or when you want to identify areas where multiple species overlap in suitability (robust site selection).

Usage

compare_species(
  data,
  species = NULL,
  output_dir = NULL,
  min_data_quality = "low",
  verbose = TRUE
)

Arguments

Value

A dataframe with all original survey columns plus:

• ⁠suit_<species_key>⁠: suitability score [0, 1] per species

• ⁠class_<species_key>⁠: suitability class per species

• best_species: key of the species with the highest suitability at each location (NA if all excluded)

• best_suitability: the highest suitability score across all species

• n_species_high: number of species scoring "High" at each location (useful for identifying robust multi-species sites)

Examples

# Compare all supported species
comparison <- compare_species(survey)

# Compare only well-characterised species
comparison <- compare_species(survey, min_data_quality = "high")

# Compare specific species and export per-species heatmaps
comparison <- compare_species(
  survey,
  species    = c("ostrea_edulis", "magallana_gigas"),
  output_dir = "comparison_maps/"
)

# Find sites suitable for both O. edulis AND M. gigas
both_high <- subset(comparison, n_species_high >= 2)

Description

Takes a named list of predict_oyster() result dataframes and produces a single comparative summary suitable for trend monitoring, multi-site selection, or regulatory reporting. Common spatial cells are matched across surveys and a change analysis is run between consecutive surveys if ordered chronologically.

The function returns a list with three elements:

• summary – one row per survey with mean suitability, class breakdown, and data completeness stats.

• spatial – all surveys joined on shared lat/lon cells, with suitability columns per survey and trend_slope (linear regression of suitability over survey index for each cell).

• change – pairwise change table (only if surveys are in order; each consecutive pair shows mean score difference and fraction of cells that improved, degraded, or were stable).

Usage

compare_surveys(
  surveys,
  cell_deg = 0.001,
  stable_threshold = 0.05,
  verbose = TRUE
)

Arguments

Value

Named list: summary (dataframe), spatial (dataframe), change (dataframe or NULL if only one survey).

Examples

r2022 <- predict_oyster(survey_2022, "ostrea_edulis")
r2023 <- predict_oyster(survey_2023, "ostrea_edulis")
r2024 <- predict_oyster(survey_2024, "ostrea_edulis")

comp <- compare_surveys(
  surveys = list("2022" = r2022, "2023" = r2023, "2024" = r2024)
)

comp$summary        # mean scores by year
comp$change         # year-on-year change

# Pass to generate_report() for a full comparative HTML report
generate_report(r2024, "monitoring_report.html",
                title = "Kames Bay 3-Year Monitoring")

Description

Oysters are permanent residents; a location is only truly suitable if conditions are adequate year-round. composite_seasonal() takes a named list of predict_oyster() result dataframes (one per season or survey visit) and produces a single composite dataframe with one row per spatial cell, summarising suitability across all input surveys.

Composite methods:

• "min" (default) – most conservative; a location must be suitable in every season to score well. Best for regulatory site selection.

• "mean" – average across seasons. Good for monitoring trend summaries.

• "prob" – fraction of seasons with suitability >= prob_threshold. Useful when surveys are irregularly spaced or some seasons are missing.

Spatial matching uses a grid cell tolerance of cell_deg decimal degrees (default 0.001 approx. 111 m). Cells not present in all seasons are flagged in n_seasons_present.

Usage

composite_seasonal(
  surveys,
  method = c("min", "mean", "prob"),
  prob_threshold = 0.5,
  cell_deg = 0.001,
  verbose = TRUE
)

Arguments

Value

A dataframe with columns: lat, lon, suitability_composite, suitability_class, n_seasons_present, one ⁠suit_<season>⁠ column per input survey, and suit_range (max - min across seasons).

Examples

spring <- predict_oyster(survey_spring, "ostrea_edulis")
summer <- predict_oyster(survey_summer, "ostrea_edulis")
autumn <- predict_oyster(survey_autumn, "ostrea_edulis")

composite <- composite_seasonal(
  surveys = list(spring = spring, summer = summer, autumn = autumn),
  method  = "min"
)

generate_report(composite, "composite_report.html",
                title = "Year-round Suitability Assessment")

Description

Converts water depths recorded during a survey to depths below Chart Datum (approximately Lowest Astronomical Tide, LAT) by adding the tidal height above Chart Datum at the time of survey. This removes bias introduced by surveying at different states of the tide.

Formula: depth_CD = depth_surveyed + tidal_height_m

Finding your tidal height:

• UK: UKHO EasyTide (easytide.ukho.gov.uk) – free 7-day predictions for any standard port. Download the predicted heights for your survey period.

• Ireland: Marine Institute tide gauges (data.marine.ie)

• France/EU: SHOM (shom.fr) or Copernicus Marine (marine.copernicus.eu)

• Any port: apps such as PocketTides, Tides Near Me, or TidePod give height-above-CD predictions for named ports.

For a typical loch or sheltered bay survey, reading the tidal height from a tide table for the nearest standard port at the mid-point of the survey is sufficient. For multi-day surveys or those spanning a full tidal cycle, supply per-row heights matched to the survey timestamp.

Usage

correct_to_chart_datum(
  df,
  tidal_height_m = NULL,
  depth_col = "depth",
  datetime_col = NULL,
  keep_raw = TRUE,
  verbose = TRUE
)

Arguments

Value

The input dataframe with the depth column corrected to Chart Datum. If keep_raw = TRUE, the original values are preserved as depth_raw_m.

Examples

# Survey conducted around high water at Oban -- tidal height ~3.1 m above CD
survey_corrected <- correct_to_chart_datum(survey, tidal_height_m = 3.1)

# Per-row heights from a tide gauge CSV matched to survey timestamps
tide_series <- read.csv("oban_tide_gauge.csv")
# (match to survey rows by timestamp -- user responsibility)
survey_corrected <- correct_to_chart_datum(
  survey,
  tidal_height_m = tide_series$height_m
)

Description

Returns the meteorological season ("winter", "spring", "summer", "autumn") for a given date and latitude. Hemisphere is inferred from the sign of latitude (positive = Northern Hemisphere, negative = Southern Hemisphere).

Usage

detect_season(date, lat)

Arguments

Value

A character string: one of "winter", "spring", "summer", "autumn".

Examples

detect_season("2024-01-15", lat = 51.5)   # "winter" (UK)
detect_season("2024-07-15", lat = 51.5)   # "summer"
detect_season("2024-01-15", lat = -33.9)  # "summer" (Sydney)

Description

Uses the empirical log-linear relationship between volume backscatter strength (Sv) and chlorophyll-a concentration to add an estimated chlorophyll_a column to a survey dataframe. Intended for surveys where a fluorometer was not deployed and the ADCP backscatter intensity is available.

The conversion is: Chl_a (ugg/L) = 10 ^ (a * Sv + b)

Default a and b constants are derived from Deines (1999) for three common ADCP frequencies (300, 600, 1200 kHz). For best results, calibrate against concurrent water samples from your survey using the calibration argument.

Usage

estimate_chlorophyll_from_backscatter(
  df,
  backscatter_col,
  frequency_khz = 600,
  calibration = NULL,
  min_chl = 0.05,
  max_chl = 50,
  keep_raw = TRUE,
  verbose = TRUE
)

Arguments

Value

The input dataframe with an additional chlorophyll_a column (ugg/L). If a chlorophyll_a column already exists, a warning is issued and the existing column is overwritten.

Improving accuracy with water samples

Collect 5–15 grab/Niskin water samples distributed across your survey area and depth range, then:

# Fit calibration from samples
fit  <- lm(log10(sample_chl) ~ backscatter, data = samples_df)
a    <- coef(fit)[["backscatter"]]
b    <- coef(fit)[["(Intercept)"]]
df   <- estimate_chlorophyll_from_backscatter(
          df, "amp_mean_dB", frequency_khz = 600,
          calibration = c(a, b))

Examples

# Default calibration (300 kHz Nortek Signature)
adcp <- read_nortek_adcp("survey.csv")
adcp <- estimate_chlorophyll_from_backscatter(
          adcp, backscatter_col = "amp_mean_dB",
          frequency_khz = 300)

# Custom calibration from water samples
adcp <- estimate_chlorophyll_from_backscatter(
          adcp, "amp_mean_dB", frequency_khz = 300,
          calibration = c(a = 0.041, b = -0.52))

Description

Simple great-circle ray-tracing to estimate how far open water extends from a point in a given direction, up to max_fetch_km. This is a lightweight GIS-free approximation – for precise fetch, use QGIS or the fetchR R package.

Usage

estimate_fetch(
  lat,
  lon,
  bearing_deg,
  land_polygons = NULL,
  max_fetch_km = 200,
  step_km = 1
)

Arguments

Value

Numeric. Estimated fetch in km.

Description

Derives contour lines from the IDW suitability raster and saves them as a vector GeoPackage (.gpkg). Load alongside the heatmap in QGIS for the depth-contour look shown in BioBase outputs.

Usage

export_contours(tif_path, interval = 0.1, gpkg_path = NULL)

Arguments

Value

The .gpkg file path (invisibly).

Examples

export_geotiff(result, "oyster_heatmap.tif")
export_contours("oyster_heatmap.tif")

Description

Takes the scored point dataset from predict_oyster() and produces a smooth, continuous raster heatmap using Inverse Distance Weighting (IDW) interpolation – the same technique used by systems like Lowrance BioBase. The result looks like a fluid heat surface over your survey area rather than isolated point squares.

The output GeoTIFF contains three bands:

1. suitability – IDW-interpolated score [0, 1]. Use this for the heatmap.

2. excluded_mask – 1 where a survey point was hard-excluded, 0 otherwise.

3. n_observations – how many survey points fell within each raster cell (data density diagnostic).

Usage

export_geotiff(
  df,
  path,
  resolution = 2e-04,
  idw_power = 2,
  idw_max_dist = NULL,
  buffer = 0.001,
  contours = TRUE,
  crs = "EPSG:4326",
  overwrite = TRUE
)

Arguments

Value

The output file path (invisibly). Also writes a .qml QGIS style file alongside the .tif automatically.

Examples

result <- predict_oyster("survey.csv", "ostrea_edulis")

# Standard export -- smooth heatmap with auto radius and contours
export_geotiff(result, "oyster_heatmap.tif")

# Finer resolution for a small harbour survey
export_geotiff(result, "oyster_heatmap.tif", resolution = 0.0001)

# Override IDW radius explicitly (e.g. sparse transect data)
export_geotiff(result, "oyster_heatmap.tif", idw_max_dist = 0.005)

Description

Writes a QGIS Layer Style file (.qml) using a yellow -> orange -> red -> dark red colour ramp, closely matching the BioBase/Lowrance heatmap style. The file is automatically applied when its name matches the .tif – no manual styling needed in QGIS.

Usage

export_qml_style(tif_path)

Arguments

Value

The .qml file path (invisibly).

Description

Convenience wrapper that fetches data from one or more live sources and returns a named list of dataframes that can be passed to the relevant scoring functions. Each source can be fetched independently or all at once.

This function requires internet access and registered API credentials where applicable (see oystermapR_live_config()). If a source fails, it is silently skipped and not included in the output - the function never aborts due to a single failed source.

Usage

fetch_live_environmental_data(
  survey,
  sources = "all",
  date_range = NULL,
  verbose = TRUE
)

Arguments

Value

Named list. Each successfully fetched source produces one element:

• ⁠$cmems⁠: SST, salinity, chlorophyll grid

• ⁠$ices_hab⁠: HAB event records within extent

• ⁠$ices_vms⁠: Swept-area ratio per ICES c-square

• ⁠$emodnet_predators⁠: Predator occurrence records

• ⁠$fsa_shellfish⁠: Shellfish classification polygons / area descriptions

Examples

## Not run: 
oystermapR_live_config(cmems_user = "u", cmems_password = "p")
live <- fetch_live_environmental_data(survey, sources = c("ices_hab", "ices_vms"))
result <- score_hab_risk(result, hab_data = live$ices_hab, species = "ostrea_edulis")
result <- score_anthropogenic_disturbance(result, trawling_data = live$ices_vms)

## End(Not run)

Description

Renders a structured assessment report from the output of predict_oyster(). The report includes an executive summary, suitability class breakdown, per-variable scoring table, most common limiting factors, top introduction sites, and a methodology section.

Output format is PDF by default (requires a LaTeX installation – see Note below). HTML is also supported and requires no additional system software.

Usage

generate_report(
  result,
  output = "oyster_report.html",
  title = "Oyster Suitability Assessment",
  author = "",
  species = "ostrea_edulis",
  top_n = 5L,
  validation = NULL,
  comparison = NULL,
  open = TRUE
)

Arguments

Value

The output file path (invisibly).

Note

HTML reports (recommended) require plotly and scales: install.packages(c("plotly","scales")). PDF reports additionally require a LaTeX installation; see the Note section below.

PDF output requires LaTeX. If you don't have LaTeX installed, either use output = "report.html" (no extra software needed), or install a minimal LaTeX distribution with:

install.packages("tinytex")
tinytex::install_tinytex()

Examples

result <- predict_oyster(survey, "ostrea_edulis",
                         output_geotiff = "kames_bay.tif")

# Standard HTML report (recommended -- Plotly charts, no LaTeX needed)
generate_report(result, "kames_bay_report.html",
                title  = "Kames Bay Oyster Habitat Assessment",
                author = "T. Tucker")

# With validation results embedded
val  <- validate_against_records(result, nbn_records)
generate_report(result, "kames_bay_validated.html",
                validation = val)

# With competition adjustment (when comparing species)
comp <- compare_species(survey,
          species = c("ostrea_edulis", "magallana_gigas"))
generate_report(result, "kames_bay_competition.html",
                comparison = comp)

# PDF report (requires LaTeX / tinytex)
generate_report(result, "kames_bay_report.pdf",
                title = "Kames Bay Oyster Habitat Assessment")

Description

Produces a self-contained multi-page A4 PDF directly from the output of predict_oyster() without requiring LaTeX, pandoc, or Rmd rendering. The PDF is designed for printing and contains four pages:

• Page 1 – Title header, executive summary table, class breakdown bar

• Page 2 – Suitability heatmap (survey points, continuous colour ramp)

• Page 3 – Per-variable mean score chart (horizontal bar)

• Page 4 – Score distribution histogram + class count bar chart

Usage

generate_summary_pdf(
  result,
  output = "oyster_summary.pdf",
  species = "ostrea_edulis",
  title = "Oyster Suitability Summary",
  author = "",
  open = TRUE,
  verbose = TRUE
)

Arguments

Value

The output file path (invisibly).

Examples

result <- predict_oyster(survey_df, "ostrea_edulis")
generate_summary_pdf(result, "kames_bay_summary.pdf",
                     title  = "Kames Bay -- Spring Survey",
                     author = "T. Tucker")

Description

Get species tolerance parameters

Usage

get_species_tolerances(species)

Arguments

Value

Named list of tolerance parameters for the species.

Examples

tol <- get_species_tolerances("ostrea_edulis")
tol$exclusions$temperature

Description

Returns the tolerance parameter list currently held in the session cache, after any update_species_tolerances() calls. If no update has been run, returns the built-in prior parameters unchanged.

Usage

get_tolerance_posteriors(species, var = NULL)

Arguments

Value

Named list of tolerance parameters (same structure as get_species_tolerances()⁠$scored⁠).

Examples

# After calling update_species_tolerances()
post <- get_tolerance_posteriors("ostrea_edulis")
post$temperature  # updated optimal_min, optimal_max with CIs

Description

From a project_suitability() output, identifies locations that maintain at least min_class suitability across all projected scenarios. These are the most robust sites for long-term restoration or aquaculture investment.

Usage

identify_resilient_sites(projections, baseline, min_class = "Moderate")

Arguments

Value

Subset of baseline for locations that meet min_class in every scenario, with an added n_scenarios_qualifying column.

Examples

proj      <- project_suitability(result, tol)
resilient <- identify_resilient_sites(proj, result, min_class = "Moderate")
nrow(resilient)  # sites that stay Moderate+ even under RCP8.5

Description

Converts sparse point observations (CTD casts, ADCP profiles) to a regular grid by spatial interpolation. The interpolated grid is returned in the same dataframe format as the raw survey, and can be passed directly to predict_oyster().

Method selection (applied per variable independently):

• Ordinary kriging is used when >= 50 non-NA observations are available for a variable, the gstat and sp packages are installed, and the variogram can be fitted without error. Kriging is the best linear unbiased estimator (BLUE) under the assumption of spatial stationarity – it minimises mean squared prediction error and returns a kriging variance for each cell.

• IDW (Inverse Distance Weighting, p = 2) is used as a fallback when kriging cannot be applied. It is fast, deterministic, and requires no model fitting, but produces no uncertainty estimate and can create bull's-eye artefacts around isolated observations.

Kriging variograms are fitted automatically using gstat::fit.variogram() with a Matern model (default) – the most flexible and statistically justified model for environmental data. If automatic fitting fails, a spherical model is tried; if that also fails, IDW is used.

Usage

interpolate_survey(
  survey,
  vars = NULL,
  resolution_m = 100,
  method = c("auto", "kriging", "idw"),
  kriging_min_n = 50L,
  idw_power = 2,
  idw_max_radius_m = NULL,
  verbose = TRUE
)

Arguments

Value

A dataframe on a regular grid with interpolated values, a method column per variable (⁠interp_method_<var>⁠), and kriging variance columns (⁠krige_var_<var>⁠) where kriging was used.

Examples

# Auto method -- kriging for dense variables, IDW for sparse
grid <- interpolate_survey(survey, resolution_m = 100)
result <- predict_oyster(grid, "ostrea_edulis")

# Force IDW (fast, no gstat needed)
grid <- interpolate_survey(survey, resolution_m = 200, method = "idw")

# Inspect which method was used per variable
unique(grid$interp_method_temperature)

# Kriging variance -- high values = uncertain interpolation
hist(grid$krige_var_temperature)

Description

List all supported species

Usage

list_species()

Value

Named character vector: names = latin names, values = species keys.

Examples

list_species()

Description

Reads a previously saved .rds file (from save_tolerance_update()) and restores it into the session cache. Subsequent calls to predict_oyster() and score_locations() will automatically use the loaded parameters.

Usage

load_tolerance_update(
  species = "all",
  path = tools::R_user_dir("oystermapR", which = "data"),
  verbose = TRUE
)

Arguments

Value

Invisibly: the loaded tolerance list.

Examples

load_tolerance_update("ostrea_edulis")
# Now predict_oyster() uses the updated tolerances automatically
result <- predict_oyster(survey, "ostrea_edulis")

Description

Takes any number of sensor dataframes (from read_nortek_adcp(), read_generic_csv(), or custom sources) and merges them spatially into a single table ready for predict_oyster().

Each dataset is first rounded to a common spatial resolution grid. Datasets are then joined using the grid cell key – cells that exist in multiple datasets are merged; cells that exist in only one dataset carry NA for variables from other sensors.

Usage

merge_sensor_data(..., spatial_res = 4L, date_source = NULL, verbose = TRUE)

Arguments

merge_sensor_data(
  adcp    = list(adcp_run1, adcp_run2, adcp_run3),
  bathy   = bathy_df,
  biobase = biobase_df
)

Value

A single dataframe with all oystermapR-compatible columns populated from their respective sensor sources. Columns that weren't available from any sensor will be absent from the result – predict_oyster() handles missing columns gracefully.

Examples

# Single run per sensor
adcp    <- read_nortek_adcp("adcp.csv")
biobase <- read_generic_csv("biobase.csv")
survey  <- merge_sensor_data(adcp = adcp, biobase = biobase)

# Multiple ADCP runs (different survey days) -- automatically stacked
adcp1  <- read_nortek_adcp("survey_jan.csv")
adcp2  <- read_nortek_adcp("survey_feb.csv")
survey <- merge_sensor_data(adcp = list(adcp1, adcp2), biobase = biobase)

result <- predict_oyster(survey, "ostrea_edulis", output_geotiff = "map.tif")

Description

Prints a step-by-step workflow guide to the console, covering sensor data ingestion, spatial merging, suitability prediction, and QGIS export. Run this any time you need a reminder of the available functions and typical usage patterns.

Usage

oystermapR_help(topic = NULL)

Arguments

Value

Called for its side-effect (console output). Returns invisible(NULL).

Examples

oystermapR_help()                  # full guide
oystermapR_help("sensors")         # sensor ingestion only
oystermapR_help("qgis")            # QGIS output only

Description

Stores API credentials in R session options so live data fetching can authenticate with external services. Credentials are not written to disk and are lost when the R session ends.

All parameters are optional - only set credentials for the services you intend to use. Functions that require a missing credential will warn and fall back to manual data rather than aborting.

Usage

oystermapR_live_config(
  cmems_user = NULL,
  cmems_password = NULL,
  ices_key = NULL,
  show_config = FALSE
)

Arguments

Value

Invisibly returns a named list of the options set.

Free registrations

• CMEMS: Register at marine.copernicus.eu

• ICES: API access is open (no key required for HAB/VMS endpoints)

• EMODnet Biology: Open WFS, no key required

• FSA/DAERA: Open REST API, no key required

Examples

## Not run: 
oystermapR_live_config(
  cmems_user     = "myusername",
  cmems_password = "mypassword"
)
# Then use fetch_live = TRUE in any supporting function
score_hab_risk(result, species = "ostrea_edulis", fetch_live = TRUE)

## End(Not run)

Description

Converts the output of an OpenDrift settlement simulation (a CSV file with particle origin and final position) into the connectivity matrix format expected by score_larval_connectivity(connectivity_matrix = ...).

OpenDrift setup: Run OpenDrift (OceanDrift or OpenOil adapted for larvae) seeding particles at each source reef location. Export a CSV with at minimum:

• origin_lat, origin_lon – seed position

• final_lat, final_lon – settlement position (stranded or settled)

• status – filter to "stranded" or "active" as appropriate

The function bins particles into a grid and computes the fraction of particles from each source bin that arrive in each destination bin.

Usage

parse_opendrift_connectivity(
  opendrift_csv,
  bin_deg = 0.05,
  status_col = NULL,
  status_keep = c("stranded", "settled"),
  min_weight = 0.005,
  verbose = TRUE
)

Arguments

Value

A dataframe with columns source_lat, source_lon, dest_lat, dest_lon, weight – ready for passing to score_larval_connectivity().

Examples

## Not run: 
# After running OpenDrift simulation and exporting CSV:
cm <- parse_opendrift_connectivity("opendrift_output.csv", bin_deg = 0.05)

result <- predict_oyster(survey, "ostrea_edulis")
result <- score_larval_connectivity(result,
            species             = "ostrea_edulis",
            connectivity_matrix = cm)

## End(Not run)

Description

Estimates the importance of each scored environmental variable by randomly permuting its values and measuring the resulting drop in AUC against known presence/absence records. A large drop indicates the model strongly depends on that variable; a small drop indicates it could be removed with minimal impact.

This approach (Breiman 2001) is model-agnostic, requires no model refit, and is directly interpretable: "how much does AUC fall if I destroy the information in variable X?"

Usage

permutation_importance(
  predicted,
  records,
  presence_col = "presence",
  n_permutations = 50L,
  match_radius_deg = 0.002,
  seed = 42L,
  verbose = TRUE
)

Arguments

Value

A dataframe (sorted by importance, descending) with columns:

• variable: scored variable name

• baseline_auc: AUC of unperturbed model

• mean_permuted_auc: mean AUC after permuting this variable

• importance: baseline_auc - mean_permuted_auc (drop in AUC)

• importance_sd: SD of AUC drop across permutation replicates

• importance_pct: importance as percentage of baseline AUC

• rank: rank by importance (1 = most important)

Interpretation

• importance > 0.10 (> 10% AUC drop): high importance – variable is load- bearing for model discrimination.

• importance 0.02–0.10: moderate importance.

• importance < 0.02: low importance – consider whether this variable adds value relative to its data collection cost.

References

Breiman (2001) Machine Learning 45:5-32. Zurell et al. (2020) Ecography 43:1261-1277.

Examples

records <- read.csv("nbn_ostrea_edulis.csv")
result  <- predict_oyster(survey, "ostrea_edulis")

imp <- permutation_importance(result, records, n_permutations = 100)
print(imp)
# Variable        Importance  Rank
# temperature     0.183       1
# salinity        0.091       2
# depth           0.045       3
# ...

Description

The primary user-facing function of oystermapR. Accepts a tabular dataset (CSV file path or dataframe) containing spatial coordinates and environmental measurements, applies species-specific exclusion and weighted scoring rules, and returns a scored dataframe alongside an optional GeoTIFF heatmap for QGIS visualisation.

Usage

predict_oyster(
  data,
  species,
  output_geotiff = FALSE,
  resolution = 2e-04,
  contours = TRUE,
  verbose = FALSE
)

Arguments

Value

A dataframe (invisibly) with all original columns plus:

• season: detected season at each location/date.

• excluded: logical, TRUE if the location fails a hard exclusion.

• exclusion_reason: character, reason(s) for exclusion or NA.

• suitability: numeric [0, 1], overall weighted suitability score.

• suitability_class: factor, one of "High", "Moderate", "Low", "Very Low", "Excluded".

• ⁠score_<variable>⁠: per-variable component scores.

If output_geotiff is set, a GeoTIFF is written to disk and its path is printed to the console.

Column Naming

Column names are matched case-insensitively. Recognised synonyms:

Examples

# Minimal runnable example using the bundled Example Bay survey data

adcp_f  <- system.file("extdata", "example_bay_adcp.csv",      package = "oystermapR")
bathy_f <- system.file("extdata", "example_bay_soundings.xyz", package = "oystermapR")
ctd_f   <- system.file("extdata", "example_bay_ctd.csv",       package = "oystermapR")
adcp   <- read_nortek_adcp(adcp_f,    verbose = FALSE)
bathy  <- read_soundings_xyz(bathy_f, verbose = FALSE)
ctd    <- read_generic_csv(ctd_f,     verbose = FALSE)
survey <- merge_sensor_data(adcp = adcp, bathy = bathy, ctd = ctd)
result <- predict_oyster(survey, "ostrea_edulis", verbose = FALSE)
table(result$suitability_class)



# Using your own CSV file
result <- predict_oyster(
  data    = "my_survey.csv",
  species = "ostrea_edulis",
  output_geotiff = "oyster_suitability.tif"
)

Description

Re-scores survey locations under user-specified climate perturbations to assess how habitat suitability may change under warming and related oceanographic shifts. Perturbations are applied as additive deltas to temperature and salinity, and multiplicative factors to other variables.

Built-in scenarios align with UKCP18 marine projections for NW European shelf seas:

The function runs score_locations() for each scenario and returns a list with one result dataframe per scenario, plus a comparison summary and a suitability_delta column (projected minus baseline) for each.

Usage

project_suitability(
  result,
  tolerances,
  scenarios = c("rcp26", "rcp45", "rcp85"),
  custom_deltas = NULL,
  horizon = "2050s",
  verbose = TRUE
)

Arguments

Value

Named list:

• One dataframe per scenario (named by scenario key) with suitability, suitability_class, and suitability_delta columns.

• "summary" – dataframe comparing mean suitability, class breakdown, and net change vs baseline across all scenarios.

Examples

result <- predict_oyster(survey, "ostrea_edulis")
tol    <- get_species_tolerances("ostrea_edulis")

# Project under all three UKCP18-aligned scenarios
proj <- project_suitability(result, tol)

# Mean suitability change by scenario
proj$summary

# High-risk cells: currently High but drops to Low under RCP8.5
vulnerable <- subset(proj$rcp85,
  result$suitability_class == "High" & suitability_class == "Low")

# Custom scenario (e.g. local downscaled projection)
proj2 <- project_suitability(result, tol, scenarios = NULL,
  custom_deltas = list(
    ukcp18_p95 = c(temperature = 3.2, salinity = -0.8)))

Description

Detects and flags suspect sensor readings before they enter the suitability model. Three complementary checks are applied to each numeric column:

1. Range check – values outside the physically plausible range for that variable (e.g. temperature above 35degrees C in temperate coastal water, salinity above 42 PSU) are flagged as "range_fail".

2. Statistical outlier – values beyond iqr_k x IQR from the median are flagged as "outlier" (default k = 3, Tukey's outer fence).

3. Temporal gradient – when a datetime column is present and data is sorted chronologically, sequential differences exceeding the max_gradient threshold are flagged as "gradient_fail" (instrument spike or data entry error).

Additionally, cross-variable sanity checks flag physically implausible combinations:

• Dissolved oxygen > 20 mg/L with temperature < 5degrees C (likely sensor error)

• Salinity < 5 PSU with temperature > 25degrees C (likely freshwater intrusion instrument error, not a real coastal reading)

• Chlorophyll_a > 50 ugg/L (extreme bloom or sensor fouling)

Flagged values are not removed – they are marked in a ⁠qc_flag_<col>⁠ column so the user can decide whether to replace with NA, correct, or accept them. A summary qc_n_flags column counts total flags per row.

Usage

qc_survey_data(
  df,
  datetime_col = "datetime",
  iqr_k = 3,
  max_gradients = NULL,
  apply_flags = FALSE,
  verbose = TRUE
)

Arguments

Value

Input dataframe with additional columns: ⁠qc_flag_<varname>⁠ for each checked variable ("ok", "range_fail", "outlier", "gradient_fail", "cross_fail"), qc_n_flags (integer count of flags per row), qc_status ("pass", "review", "fail" – fail = 3+ flags on one row).

Examples

# Run QC before modelling
survey_qc <- qc_survey_data(survey, datetime_col = "timestamp")

# Inspect flagged rows
flagged <- subset(survey_qc, qc_status %in% c("review","fail"))
View(flagged)

# Apply flags (replace flagged values with NA) before scoring
survey_clean <- qc_survey_data(survey, apply_flags = TRUE)
result <- predict_oyster(survey_clean, "ostrea_edulis")

Description

Reads the CSV file exported by Aanderaa Data Studio from an Aanderaa Recording Current Meter (RCM Blue 5450, SeaGuard RCM, or compatible model). The export typically contains time, current speed, current direction, and optionally temperature, pressure/depth, salinity, and dissolved oxygen.

Because Aanderaa instruments are moored at a fixed location, latitude and longitude must be supplied by the user via lat and lon.

Current speed is returned as current_speed (m/s). If the export uses cm/s (common in older instrument firmware), set speed_unit = "cm/s" or leave as "auto" and the function will detect units from the column header or by range-checking the data.

Usage

read_aanderaa_csv(
  file,
  lat = NULL,
  lon = NULL,
  speed_unit = c("auto", "m/s", "cm/s"),
  depth_from_pressure = FALSE,
  verbose = TRUE
)

Arguments

Value

A one-row dataframe (site-averaged) with columns lat, lon, current_speed, and any additional variables present in the file (temperature, salinity, dissolved_oxygen, pressure_dbar, depth) – ready for merge_sensor_data().

See Also

read_rdi_adcp(), read_nortek_aquadopp(), merge_sensor_data()

Examples

# Moored RCM Blue at Strangford Lough narrows
rcm <- read_aanderaa_csv(
  "strangford_rcm_2024.csv",
  lat = 54.398,
  lon = -5.558
)
survey <- merge_sensor_data(adcp = rcm, bathy = bathy)

Description

Reads any sensor CSV (Lowrance BioBase, CTD/probe logger, bathymetric sonar, sidescan export, etc.) and maps its columns to the standard oystermapR variable names. Performs spatial averaging at a specified resolution.

Usage

read_generic_csv(
  file,
  col_map = NULL,
  spatial_res = 4L,
  min_obs = 3L,
  categorical_cols = c("sediment_type", "benthic_communities", "biotope",
    "fishing_intensity"),
  verbose = TRUE
)

Arguments

col_map = c(
  lat                 = "Latitude",
  lon                 = "Longitude",
  depth               = "Depth_m",
  substrate_hardness  = "Hardness",
  temperature         = "Temp_C"
)

Value

A spatially averaged dataframe using oystermapR standard column names.

Examples

# Lowrance BioBase export (automatic column matching)
biobase <- read_generic_csv("biobase_export.csv")

# With explicit column mapping
probe <- read_generic_csv(
  "ctd_data.csv",
  col_map = c(lat="Lat", lon="Lon", temperature="Temp", salinity="Sal",
              dissolved_oxygen="DO_mgl")
)

Description

Reads a raw merged Nortek Signature 500 ADCP CSV, automatically detects and excludes depth bins contaminated by acoustic sidelobe interference, computes current velocity and estimated bed shear stress from the deepest clean bin, and returns a spatially averaged dataframe ready to merge with other sensor data via merge_sensor_data().

Sidelobe detection: Each bin is tested independently. A bin is flagged as contaminated if more than sidelobe_threshold fraction of its speed readings exceed max_plausible_speed. Once a bin is flagged, all deeper bins are also flagged (contamination propagates downward).

Shear stress accuracy: Bed shear stress (tau = rho * Cd * U_bed^2) is most accurate when computed from near-bed velocity. If near-bed bins are contaminated and only near-surface bins remain, a warning is issued and the shear_stress_quality column is set to "low". If two or more clean bins exist, the deepest clean bin is used and quality is "moderate" or "good".

Usage

read_nortek_adcp(
  file,
  spatial_res = 4L,
  min_obs = 5L,
  max_plausible_speed = 1.5,
  sidelobe_threshold = 0.1,
  rho = 1025,
  Cd = 0.002,
  verbose = TRUE
)

Arguments

Value

A dataframe with one row per spatial cell containing:

• lat, lon – cell centroid coordinates

• date – earliest observation date in cell (character, "YYYY-MM-DD")

• current_velocity – mean speed from deepest clean bin (m/s)

• current_velocity_sd – standard deviation within cell (m/s)

• current_velocity_p95 – 95th percentile speed (m/s)

• shear_stress – estimated bed shear stress (N/m^2)

• shear_stress_quality – "good", "moderate", or "low"

• n_ensembles – raw ensembles averaged into this cell

• bins_used – which velocity bins contributed (e.g. "bin1")

• bins_excluded – which bins were removed as contaminated

Examples

adcp <- read_nortek_adcp("S104456A008_AWE_Melfort_merged.csv")
print(adcp)

Description

Reads the ASCII velocity export produced by Nortek's AquaPro software from an Aquadopp Profiler deployment. The function accepts two common export layouts:

• ENU CSV – a single file with named columns for east, north, and up velocity per depth bin (e.g. East_1, North_1, East_2, ...) plus optional time, temperature, and pressure columns.

• Separate beam files – the companion .sen (sensor) file providing time, temperature, and pressure, and the .v1/.v2 files providing beam velocities. Supply the .sen path to file and set beam_files = c("survey.v1", "survey.v2", "survey.v3").

Because Aquadopp instruments are often moored at a single location rather than vessel-mounted, GPS coordinates may not be embedded in the export. In that case supply lat and lon directly; all output rows receive that fixed position.

Usage

read_nortek_aquadopp(
  file,
  beam_files = NULL,
  lat = NULL,
  lon = NULL,
  spatial_res = 4L,
  min_obs = 5L,
  max_plausible_speed = 2.5,
  rho = 1025,
  Cd = 0.002,
  verbose = TRUE
)

Arguments

Value

A dataframe with columns lat, lon, current_speed, bed_shear_stress, shear_stress_quality, and optionally temperature, pressure_dbar, and date – ready for merge_sensor_data().

Examples

# Moored deployment (fixed position)
adcp <- read_nortek_aquadopp(
  "harbour_mouth_aqd.csv",
  lat = 56.512, lon = -5.403
)

# Vessel-mounted with GPS columns in the file
adcp <- read_nortek_aquadopp("transect_enu.csv")

survey <- merge_sensor_data(adcp = adcp, bathy = bathy)

Description

Reads the binary PD0 output file produced by Teledyne RDI instruments (WorkHorse II, Sentinel V, StreamPro, Ocean Surveyor, Long Ranger) via the the oce package's read.adp.rdi() parser. The function extracts east/north velocity profiles, applies sidelobe contamination detection, and derives current speed and bed shear stress – producing output in the same format as read_nortek_adcp().

Optional dependency: reading RDI binary (PD0) files requires the oce package (install.packages("oce")). If oce is absent the function will stop with an informative message. ASCII WinRiver II exports (.txt/.csv) do not require oce.

GPS: vessel-mounted deployments embed GPS in the binary file and are handled automatically. For moored deployments supply lat and lon.

Usage

read_rdi_adcp(
  file,
  lat = NULL,
  lon = NULL,
  spatial_res = 4L,
  min_obs = 5L,
  max_plausible_speed = 2,
  sidelobe_threshold = 0.1,
  rho = 1025,
  Cd = 0.002,
  verbose = TRUE
)

Arguments

Value

A dataframe with columns lat, lon, current_speed, bed_shear_stress, shear_stress_quality, and n_ensembles – ready for merge_sensor_data().

See Also

read_nortek_adcp(), read_nortek_aquadopp(), merge_sensor_data()

Examples

# Vessel-mounted WorkHorse II (GPS embedded in binary)
adcp <- read_rdi_adcp("survey_2024.000")

# Moored Sentinel V (fixed position)
adcp <- read_rdi_adcp("mooring_2024.000", lat = 52.41, lon = -9.20)

survey <- merge_sensor_data(adcp = adcp, bathy = bathy)

Description

Reads a single-band GeoTIFF raster using terra and returns a spatially averaged dataframe suitable for merge_sensor_data().

Two modes are supported, selected via the type argument:

"bathy" – Bathymetric DEM (e.g. from Ping 3DSS or post-processed soundings). Extracts:

• depth – raster cell value (metres)

• slope – computed using terra::terrain()

• roughness – Terrain Ruggedness Index via terra::terrain()

"sidescan" – Sidescan backscatter mosaic (normalised 0–1 or raw DN). Extracts:

• substrate_hardness – backscatter intensity normalised to [0, 1]. High backscatter -> hard/coarse substrate (rock, gravel, shell); low backscatter -> soft substrate (mud, fine sand).

Usage

read_sonar_tif(
  file,
  type = c("bathy", "sidescan"),
  spatial_res = 4L,
  band = 1L,
  nodata_val = NA,
  depth_positive = TRUE,
  verbose = TRUE
)

Arguments

Value

A dataframe with lat, lon, and the derived variable columns ready for merge_sensor_data().

Examples

bathy_df   <- read_sonar_tif("kames_bay_bathy.tif",    type = "bathy")
sidescan_df <- read_sonar_tif("kames_bay_sidescan.tif", type = "sidescan")
survey <- merge_sensor_data(bathy = bathy_df, sidescan = sidescan_df, adcp = adcp_df)

Description

Reads a plain-text XYZ soundings file (comma or space delimited, optional ⁠#⁠-prefixed header), spatially averages onto a regular grid, and derives:

• depth – mean depth per cell (metres, positive downward)

• roughness – rugosity index per cell, derived from intra-cell depth variance using the surface-area approximation rugosity = sqrt(1 + (depth_sd / cell_size_m)^2). Values near 1.0 are flat; higher values indicate more complex relief.

• slope – maximum downslope gradient in degrees, computed by finite differences between neighbouring grid cells (Horn 1981 method).

Usage

read_soundings_xyz(
  file,
  lon_col = NULL,
  lat_col = NULL,
  depth_col = NULL,
  spatial_res = 4L,
  min_soundings = 10L,
  depth_positive = TRUE,
  verbose = TRUE
)

Arguments

Value

A dataframe with columns lat, lon, depth, slope, roughness, and n_soundings – ready for merge_sensor_data().

Examples

bathy <- read_soundings_xyz("kames_bay_soundings.xyz")
survey <- merge_sensor_data(adcp = adcp_data, bathy = bathy)

Description

Clears the session cache for a species, reverting to the built-in prior parameters. Does not affect saved files (use file deletion for those).

Usage

reset_tolerance_update(species = "all", verbose = TRUE)

Arguments

Value

Invisibly NULL.

Examples

reset_tolerance_update("ostrea_edulis")
reset_tolerance_update("all")

Description

Serialises the current session cache for a species to an .rds file so that updated parameters persist across R sessions. Load again with load_tolerance_update().

Usage

save_tolerance_update(
  species = "all",
  path = tools::R_user_dir("oystermapR", which = "data"),
  verbose = TRUE
)

Arguments

Value

Invisibly: path(s) to saved file(s).

Examples

save_tolerance_update("ostrea_edulis")
save_tolerance_update("all", path = "data/bayes_updates/")

Description

Calculates a disturbance index [0,1] from bottom trawling intensity, dredging activity, and anchor damage. The primary data input is the ICES swept-area ratio (SAR), which can be supplied manually or fetched live from the ICES VMS data portal.

This output is primarily relevant to the gear feasibility and economic viability modules. Very high trawling intensity (SAR > 0.5/yr) constitutes a hard gate for bottom culture and restoration reef deployment – physical gear deployment is not viable on actively trawled ground regardless of ecological suitability.

Input options:

1. trawling_data – ICES VMS SAR data, manually downloaded as CSV/dataframe.

2. fetch_live = TRUE – queries ICES VMS GeoServer endpoint.

3. Neither – returns zero disturbance with a warning.

Usage

score_anthropogenic_disturbance(
  result,
  trawling_data = NULL,
  anchor_data = NULL,
  dredge_areas = NULL,
  fetch_live = FALSE,
  match_radius_m = 5000,
  verbose = TRUE
)

Arguments

Value

result with additional columns:

• disturbance_sar: swept-area ratio at or near each site (NA if no data)

• disturbance_score [0,1]: composite anthropogenic disturbance index

• disturbance_class: "Negligible" / "Low" / "Moderate" / "High" / "Active"

• disturbance_gear_gate: logical – TRUE if SAR exceeds gear deployment threshold (used by assess_gear_feasibility())

• disturbance_note: management flag

trawling_data format

A dataframe with columns:

• lat, lon – centroid of c-square or fishing location

• sar – swept-area ratio (numeric; dimensionless ratio per year)

• gear_type (optional) – "otter", "beam", "dredge", "seine" etc. Dredge gear has higher impact per SAR unit.

Additional disturbance inputs

• anchor_data: dataframe with lat, lon, density (vessel anchoring density index) for recreational/commercial anchorage areas.

• dredge_areas: dataframe with lat, lon identifying licensed aggregate extraction areas (all sites within the polygon receive score 1.0).

Examples

# Manual ICES VMS data (downloaded from ICES data portal)
vms <- read.csv("ices_vms_sar_2022.csv")  # columns: lat, lon, sar, gear_type
result <- score_anthropogenic_disturbance(result, trawling_data = vms)

# Live ICES VMS fetch
result <- score_anthropogenic_disturbance(result, fetch_live = TRUE)

Description

Assesses the environmental risk of two key bivalve pathogens at each survey location:

• Bonamia ostreae (Ostrea edulis only) – an intracellular haplosporidian parasite that is the primary constraint on flat oyster restoration in northern Europe. Risk is temperature-driven: transmission peaks 15-20degrees C and is negligible below 8degrees C. The function also accepts a known_sites dataframe of confirmed infected locations to apply a proximity multiplier.

• OsHV-1 microvariant (Magallana gigas) – a herpesvirus causing Pacific Oyster Mortality Syndrome (POMS). Mortality events occur when seawater exceeds 16degrees C for sustained periods. High turbidity amplifies risk by increasing larval stress.

Risk scores are returned on a 0-1 scale (0 = negligible, 1 = highest environmental risk) and classified as Low / Moderate / High / Critical. They are deliberately kept separate from the suitability score – a High suitability site with High disease risk is a meaningful regulatory red flag.

Usage

score_disease_risk(
  result,
  species = "ostrea_edulis",
  known_sites = NULL,
  temp_col = "temperature",
  salinity_col = "salinity",
  turbidity_col = "turbidity",
  verbose = TRUE
)

Arguments

Value

Input dataframe with additional columns: disease_risk_score [0-1], disease_risk_class (Low/Moderate/High/Critical), disease_agent (pathogen name), and disease_risk_note (plain-language interpretation).

Examples

result <- predict_oyster(survey, "ostrea_edulis")

# Basic risk scoring (temperature-driven only)
result <- score_disease_risk(result, "ostrea_edulis")

# With known Bonamia-positive site locations
infected <- data.frame(lat = c(55.8, 56.1), lon = c(-5.2, -5.4))
result <- score_disease_risk(result, "ostrea_edulis",
                              known_sites = infected)

# Sites with high ecological suitability but also high disease risk
flagged <- subset(result,
  suitability_class == "High" & disease_risk_class %in% c("High","Critical"))

Description

Combines ecological suitability, gear feasibility, site accessibility, and patch size into a composite economic viability index [0-1]. This is a heuristic scoring – not a financial model – but it ranks sites consistently and can be used to shortlist candidates for detailed business case development.

The index weighs four components:

1. Ecological suitability (40%) – from predict_oyster()

2. Gear feasibility (25%) – best gear score from assess_gear_feasibility()

3. Accessible patch area (20%) – log-scaled area of connected suitable habitat (from analyse_connectivity() if available)

4. Access score (15%) – distance to nearest harbour/port (if harbours table supplied) or flat 0.5 if unknown

Usage

score_economic_viability(
  result,
  harbours = NULL,
  target = c("restoration", "aquaculture"),
  verbose = TRUE
)

Arguments

Value

Input dataframe with viability_score [0-1], viability_class (Poor/Fair/Good/Excellent), and viability_notes.

Examples

result <- predict_oyster(survey, "ostrea_edulis")
result <- assess_gear_feasibility(result)
result <- analyse_connectivity(result)

harbours <- data.frame(
  name = c("Tarbert","Portavadie"),
  lat  = c(55.865, 55.875),
  lon  = c(-5.425, -5.300)
)

result <- score_economic_viability(result, harbours = harbours,
                                    target = "aquaculture")

# Best aquaculture candidates
subset(result, viability_class %in% c("Good","Excellent"))

Description

Produces a HAB risk index (0 = negligible, 1 = severe) based on historical bloom event frequency and, where available, nutrient and stratification data.

Input options (in priority order):

1. hab_data – manually supplied dataframe of historical HAB events (see Details).

2. fetch_live = TRUE – queries the ICES HAB database for event records within the survey extent and a specified date range. Requires internet access and the httr package.

3. Neither – returns a fixed background-level risk (0.1) with a warning.

Usage

score_hab_risk(
  result,
  hab_data = NULL,
  fetch_live = FALSE,
  date_range = NULL,
  match_radius_m = 5000,
  species = "ostrea_edulis",
  verbose = TRUE
)

Arguments

Value

result with additional columns:

• hab_risk [0,1]: composite risk index

• hab_risk_class: "Low" / "Moderate" / "High" / "Critical"

• hab_closure_days_per_year: estimated days/year with shellfish closures

• hab_dominant_toxin: most frequently recorded toxin class at the site

• hab_risk_note: management flag

hab_data format

A dataframe with columns:

• lat, lon – event coordinates (decimal degrees)

• date – event date (Date or character "YYYY-MM-DD")

• genus (optional) – bloom genus (e.g. "Alexandrium", "Pseudo-nitzschia")

• toxin (optional) – toxin class (PSP, ASP, DSP, AZP) – affects severity weight

• closure_days (optional) – number of days the area was closed; default 1 per event if absent

Examples

# Manual data
hab <- data.frame(lat=52.1, lon=-4.5, date="2022-07-15",
                  genus="Alexandrium", toxin="PSP", closure_days=14)
result <- score_hab_risk(result, hab_data=hab, species="ostrea_edulis")

# Live ICES fetch
result <- score_hab_risk(result, fetch_live=TRUE,
                         date_range=c("2015-01-01","2024-12-31"))

Description

Estimates the larval connectivity of each survey location using one or both of two approaches:

Route 1 – Built-in gap-threshold union-find (no external data needed): Patches above min_suitability that lie within the species' effective dispersal kernel distance are grouped into dispersal clusters using union-find. Within each cluster, a connectivity score is derived from the number, quality, and distance-weighted contribution of potential larval sources (Gaussian decay kernel, sigma = dispersal_km / 2).

Route 2 – External connectivity matrix (particle tracking / literature): When connectivity_matrix is supplied (a dataframe of source -> destination pairs with weights), those weights replace the Gaussian kernel for matched pairs. Rows not covered by the matrix fall back to Route 1. The matrix can come from:

• OpenDrift particle tracking simulations (export as source/destination settlement density CSV)

• FVCOM / ROMS model connectivity matrices

• Published empirical matrices (e.g. Robins et al. 2017)

Usage

score_larval_connectivity(
  result,
  species = NULL,
  dispersal_km = NULL,
  pld_days = NULL,
  tidal_excursion_km = 5,
  min_suitability = 0.4,
  connectivity_matrix = NULL,
  matrix_match_radius_deg = 0.1,
  verbose = TRUE
)

Arguments

Value

result with additional columns:

• larval_dispersal_km: effective kernel distance used

• larval_pld_days: PLD used (species default or override)

• larval_type: "lecithotrophic" or "planktotrophic"

• n_larval_sources: suitable patches within dispersal range

• source_quality_score [0,1]: Gaussian-weighted quality of nearby sources

• larval_cluster_id: dispersal network cluster ID (integer; NA if below min_suitability threshold and no sources nearby)

• larval_cluster_size: number of suitable patches in dispersal network

• nearest_source_km: km to nearest suitable source patch

• larval_connectivity_score [0,1]: composite dispersal connectivity score

• larval_connectivity_class: "Highly connected" / "Connected" / "Low connectivity" / "Isolated"

• larval_connectivity_note: plain-language ecological interpretation

• larval_source: "union_find", "matrix", or "matrix+union_find"

Dispersal distance

If dispersal_km is not supplied, the effective kernel distance is derived from the species' mean pelagic larval duration (PLD) multiplied by tidal_excursion_km (mean daily tidal dispersal distance):

dispersal_km = PLD_mean_days x tidal_excursion_km

PLD values are literature-derived:

References

Cowen R.K. & Sponaugle S. (2009) Annual Review of Marine Science 1:443–466. Robins P.E. et al. (2017) J Applied Ecology 54:1699–1710. doi:10.1111/1365-2664.12854 Siegel D.A. et al. (2008) PNAS 105:8974–8979. Woolmer A.P. et al. (2009) J Shellfish Research 28:107–116. Trimble A.C. et al. (2009) J Shellfish Research 28:43–53.

Examples

result <- predict_oyster(survey, "ostrea_edulis")

# Route 1 only -- built-in dispersal kernel
result <- score_larval_connectivity(result, species = "ostrea_edulis")

# Route 1 with custom dispersal distance (e.g. strong tidal excursion)
result <- score_larval_connectivity(result, species = "ostrea_edulis",
                                    tidal_excursion_km = 10)

# Route 1 with manual dispersal override
result <- score_larval_connectivity(result, dispersal_km = 8)

# Route 2 -- external connectivity matrix from OpenDrift
cm <- read.csv("opendrift_connectivity.csv")
# Required columns: source_lat, source_lon, dest_lat, dest_lon, weight
result <- score_larval_connectivity(result,
           species             = "ostrea_edulis",
           connectivity_matrix = cm)

# Inspect isolated patches -- need artificial seeding
isolated <- subset(result,
  larval_connectivity_class == "Isolated" & suitability_class == "High")

# Strong candidates: high suitability AND high connectivity
targets <- subset(result,
  suitability_class == "High" & larval_connectivity_class == "Highly connected")

Description

Applies .score_row() across every non-excluded row. Excluded rows receive suitability = 0. Adds per-variable score columns and a limiting_factors column identifying the variables most constraining the score at each location.

Usage

score_locations(df, tolerances, verbose = FALSE)

Arguments

Value

The input dataframe with additional columns:

• suitability: weighted score [0, 1]

• suitability_class: "High" / "Moderate" / "Low" / "Very Low" / "Excluded"

• limiting_factors: comma-separated names of the variables most limiting the score (NA if all variables score >= 0.65)

• ⁠score_<variable>⁠: one column per scored variable present in the data

Description

Produces a predation risk index (0 = negligible, 1 = severe) based on predator occurrence data, substrate rugosity, and depth. The output is intended as an overlay on ecological suitability – high predation risk does not reduce the habitat suitability score directly but is flagged in a separate column for management planning (e.g. cage exclusion, deep subtidal placement to avoid intertidal drills).

Input options (in priority order):

1. predator_data – manually supplied dataframe with predator records (see Details).

2. fetch_live = TRUE – queries EMODnet Biology for Asterias rubens and Carcinus maenas occurrence records within the survey extent. Requires internet access and the httr package.

3. Neither – returns a depth-proxy-only risk score with a warning.

Usage

score_predation_risk(
  result,
  predator_data = NULL,
  fetch_live = FALSE,
  match_radius_m = 1000,
  species = "ostrea_edulis",
  verbose = TRUE
)

Arguments

Value

result with additional columns:

• predation_risk [0,1]: composite risk index

• predation_risk_class: "Low" / "Moderate" / "High" / "Severe"

• predation_risk_note: text flag for management planning

• n_predator_records: number of predator records within match_radius_m

predator_data format

A dataframe with columns:

• lat, lon – coordinates

• species – species name or AphiaID (character)

• density (optional) – relative abundance / density index; if absent, each record is treated as presence-only (density = 1)

• date (optional) – survey date (used to filter to relevant season)

Examples

# Manual data
pred <- data.frame(lat=51.5, lon=-4.1, species="Asterias rubens", density=3)
result <- score_predation_risk(result, predator_data=pred, species="ostrea_edulis")

# Live EMODnet fetch
result <- score_predation_risk(result, fetch_live=TRUE)

# Depth-proxy only (no predator data)
result <- score_predation_risk(result)

Description

Calculates the Shields parameter and a sediment stability score [0,1] for each survey location. High sediment mobility (theta >> theta_cr) indicates that the seabed is frequently in motion, which is unfavourable for oyster settlement, juvenile survival, and restoration reef persistence.

Usage

score_sediment_stability(
  result,
  current_col = "current_ms",
  wave_hs_col = "wave_hs_m",
  depth_col = "depth_m",
  substrate_col = "substrate",
  d50_col = "d50_mm",
  wave_period_s = 8,
  drag_coef = 0.003,
  verbose = TRUE
)

Arguments

Value

result with additional columns:

• shields_parameter: Shields number theta at each site

• shields_critical: critical Shields number theta_cr for the substrate d50

• mobility_ratio: theta / theta_cr (> 1.0 = mobile; < 1.0 = stable)

• d50_mm_estimated: grain size used (mm) – measured or substrate-derived

• sediment_stability_score [0,1]: 1 = fully stable, 0 = highly mobile

• sediment_mobility_class: "Stable" / "Marginally stable" / "Mobile" / "Highly mobile"

• sediment_stability_note: management flag

Inputs required

At minimum, the function needs current velocity to estimate bed shear stress. Grain size improves accuracy but can be estimated from the substrate type column if not directly measured.

Examples

# With current velocity and substrate type columns
result <- score_sediment_stability(result, current_col="current_ms", substrate_col="substrate")

# After score_wave_exposure() -- wave_hs_m column already present
result <- score_wave_exposure(result, fetch_km=15)
result <- score_sediment_stability(result)

# With measured grain size
result$d50_mm <- c(0.25, 2.5, 15.0, 0.08)
result <- score_sediment_stability(result)

Description

Assesses whether survey locations meet the conditions required for bivalve larval settlement and metamorphosis. Settlement scoring uses a separate, tighter tolerance specification than adult habitat scoring – larvae are more sensitive to turbidity, require minimum current flow for delivery, and need hard substrate for attachment.

The output can be compared directly against the adult suitability score from predict_oyster() to distinguish "good adult habitat" from "likely natural recruitment site."

Usage

score_settlement(survey, species = "ostrea_edulis", verbose = TRUE)

Arguments

Value

Dataframe with settlement_suitability, settlement_class, and per-variable ⁠settle_score_*⁠ columns. Combine with adult result via cbind() or merge on lat/lon.

Examples

adult_result     <- predict_oyster(survey, "ostrea_edulis")
settlement_result <- score_settlement(survey, "ostrea_edulis")

# Identify cells suitable for both adult survival AND natural recruitment
combined <- merge(adult_result, settlement_result[, c("lat","lon",
              "settlement_suitability","settlement_class")],
              by = c("lat","lon"))
combined$dual_suitable <- combined$suitability >= 0.6 &
                          combined$settlement_suitability >= 0.6

Description

Produces a wave exposure score [0,1] and significant wave height estimate for each survey location. High wave exposure reduces gear feasibility and can destabilise restoration reef structures and spat bags.

Input options:

• fetch_km: effective fetch in kilometres (distance to nearest land or obstruction in the prevailing wind direction). Most useful for estuarine, sea loch, or enclosed bay sites. Can be measured from GIS (e.g. in QGIS using the Fetch tool) or estimated from chart inspection.

• wave_height_m: directly measured or modelled significant wave height (Hs). If supplied, fetch_km is ignored for Hs computation.

• Both absent: wave exposure is estimated from depth alone (coarse proxy; a warning is issued).

Usage

score_wave_exposure(
  result,
  fetch_col = NULL,
  wave_height_col = NULL,
  fetch_km = NULL,
  wave_height_m = NULL,
  wind_speed_ms = 12,
  depth_limit = TRUE,
  verbose = TRUE
)

Arguments

Value

result with additional columns:

• wave_hs_m: significant wave height (m) – computed or supplied

• wave_exposure_score [0,1]: 0 = fully sheltered, 1 = severely exposed

• wave_exposure_class: "Sheltered" / "Moderate" / "Exposed" / "Severe"

• wave_exposure_note: gear implication flag

• wave_source: method used ("measured", "jonswap_fetch", "depth_proxy")

Examples

# Fetch column in result
result$fetch_km <- c(2.5, 8.0, 25.0, 45.0)
result <- score_wave_exposure(result, fetch_col = "fetch_km")

# Uniform fetch for a sheltered sea loch
result <- score_wave_exposure(result, fetch_km = 3.5)

# Measured wave height column
result <- score_wave_exposure(result, wave_height_col = "hs_m")

# Depth proxy only (coarse)
result <- score_wave_exposure(result)

Description

Computes the marginal response curve for a single environmental variable: suitability is predicted at a grid of values for the focal variable, while all other variables are held at their observed median (or a specified background value). This reveals the shape of the scoring function as actually applied to a real dataset.

Partial dependence is a standard diagnostic for SDM publication (Elith et al. 2008; Zurell et al. 2020 ODMAP protocol). Use it to verify that predicted responses match ecological expectations before submission.

Usage

sensitivity_analysis(
  predicted,
  species,
  variable,
  n_steps = 100L,
  background = NULL,
  season = NULL,
  verbose = TRUE
)

Arguments

Value

A dataframe with columns:

• x: focal variable value

• suitability: predicted suitability at that value

• variable: focal variable name

• species: species key

Suitable for plotting with ggplot2 or base R plot().

References

Elith et al. (2008) J Animal Ecology 77:802-813. Zurell et al. (2020) Ecography 43:1261-1277.

Examples

result <- predict_oyster(survey, "ostrea_edulis")

# Temperature response curve
temp_pd <- sensitivity_analysis(result, "ostrea_edulis", "temperature")
plot(temp_pd$x, temp_pd$suitability, type="l",
     xlab="Temperature (degrees C)", ylab="Suitability",
     main="Partial dependence: temperature")

# Salinity response (summer)
sal_pd <- sensitivity_analysis(result, "ostrea_edulis", "salinity",
                               season = "summer")

# ggplot2 version
library(ggplot2)
ggplot(temp_pd, aes(x, suitability)) +
  geom_line(colour="steelblue", linewidth=1.2) +
  geom_ribbon(aes(ymin=0, ymax=suitability), alpha=0.2, fill="steelblue") +
  labs(x="Temperature (degrees C)", y="Suitability [0-1]") +
  theme_minimal()

Description

Individual CTD or ADCP casts are point measurements subject to instrument noise, micro-scale patchiness, and tidal state variability. smooth_suitability() applies a Gaussian distance-weighted kernel to the suitability scores, replacing each cell's score with a weighted mean of all cells within bandwidth_m metres. The result is a spatially coherent surface that better reflects broad habitat quality rather than single-cast noise.

The smoothed score is computed as:

$\hat{s}_i = \frac{\sum_j w_{ij} \cdot s_j}{\sum_j w_{ij}}$

where $w_{ij} = \exp(-d_{ij}^2 / (2\sigma^2))$ and $\sigma$ is the Gaussian bandwidth in metres. Excluded cells (suitability = 0, excluded flag) are down-weighted but included so that unsuitable barriers are preserved.

Usage

smooth_suitability(
  result,
  bandwidth_m = 500,
  max_radius_m = NULL,
  smooth_excluded = FALSE,
  verbose = TRUE
)

Arguments

Value

Input dataframe with suitability_raw (original) and suitability (smoothed, overwrites original) columns, plus a suitability_class column reclassified from the smoothed score.

Examples

result <- predict_oyster(survey, "ostrea_edulis")

# Smooth with 300 m bandwidth (good for dense ADCP surveys)
result_smooth <- smooth_suitability(result, bandwidth_m = 300)

# Compare raw vs smoothed
plot(result$suitability_raw, result$suitability,
     xlab = "Raw", ylab = "Smoothed", pch = 20)

Description

Evaluates model performance using spatial block cross-validation (Roberts et al. 2017), which avoids inflated performance estimates that arise when nearby train and test records share similar environments.

Records are partitioned into n_blocks contiguous geographic blocks. In each fold, one block is held out as test data and the remaining blocks provide training data to re-score the suitability model. AUC, TSS, and Brier score are computed per fold and summarised across folds.

This function is purely evaluative – it does not refit a new model. Instead, it uses the spatial heterogeneity in the existing predicted suitability surface to estimate out-of-block performance.

Usage

spatial_block_cv(
  predicted,
  records,
  presence_col = "presence",
  n_blocks = 5L,
  match_radius_deg = 0.002,
  seed = 42L,
  plot = TRUE,
  verbose = TRUE
)

Arguments

Value

A named list:

• mean_auc: mean AUC across blocks (primary reporting metric)

• auc_sd: standard deviation of per-block AUC

• mean_tss, tss_sd: mean / SD of True Skill Statistic

• brier_mean, brier_sd: mean / SD of Brier score

• n_blocks_actual: number of blocks with >= 5 records (used in CV)

• fold_results: dataframe with per-fold metrics

• spatial_autocorr_range_deg: estimated autocorrelation range (degrees)

• spatial_bias_warning: logical, TRUE if random CV would be misleading

Interpreting results

Compare mean_auc from spatial CV with standard AUC from validate_against_records(). A large drop (> 0.10) indicates strong spatial autocorrelation and means the model is less transferable than the standard validation suggested. This is common for coastal surveys with spatially clustered records.

References

Roberts et al. (2017) Ecography 40:913-929. doi:10.1111/ecog.02881

Examples

records <- read.csv("nbn_ostrea_edulis.csv")
result  <- predict_oyster(survey, "ostrea_edulis")

# Standard validation
std_val <- validate_against_records(result, records)
std_val$auc  # may be optimistically high

# Spatial block CV
sp_cv <- spatial_block_cv(result, records, n_blocks = 5)
sp_cv$mean_auc  # more honest estimate of transferability

# Compare -- if sp_cv$mean_auc << std_val$auc, model is spatially
# autocorrelated and transfers less well than initially apparent.

Description

Combines two or more dataframes from the same sensor (e.g. an ADCP run on Monday and again on Thursday) by row-binding them and re-averaging overlapping cells onto a common spatial grid. Non-overlapping cells are kept as-is.

This is the right tool when you have repeated surveys of the same area – rather than picking one run, overlapping cells are averaged across all runs which improves noise reduction. Non-overlapping cells (different survey extents) are preserved.

After stacking, pass the result into merge_sensor_data() alongside your other sensor datasets as usual.

Usage

stack_surveys(..., spatial_res = 4L, verbose = TRUE)

Arguments

Value

A single spatially averaged dataframe combining all input surveys.

Examples

# Stack two ADCP runs before merging with other sensors
adcp_mon <- read_nortek_adcp("survey_monday.csv")
adcp_thu <- read_nortek_adcp("survey_thursday.csv")
adcp_all <- stack_surveys(adcp_mon, adcp_thu)

survey <- merge_sensor_data(adcp = adcp_all, bathy = bathy_df)

Description

Returns the mean spring tidal range (MHWS - MLWS) and typical Chart Datum offset from Mean Sea Level for a named port. Useful for a quick sanity check on tidal height values before applying correct_to_chart_datum().

This is a reference table only – always use official tide table predictions for actual corrections.

Usage

tidal_port_info(port)

Arguments

Value

A one-row dataframe with columns port, country, mhws_m, mlws_m, spring_range_m, cd_below_msl_m.

Examples

tidal_port_info("oban")
tidal_port_info("brest")

Description

Fits a Bayesian logistic regression linking AHP suitability scores to observed presence/absence, then updates the optimal-range tolerance parameters for one or more scored variables.

Default method (MAP + Laplace): finds the Maximum A Posteriori estimate via optim() (L-BFGS-B) and approximates the posterior as a multivariate normal using the numerical Hessian. Fast; suitable for routine use.

MCMC method: Random-Walk Metropolis-Hastings sampler for full posterior samples. Recommended when n < 50 or when credible intervals are needed for publication. May take several minutes for large parameter sets.

Posteriors are stored in a package-level cache so that repeated calls accumulate evidence (sequential Bayesian updating). Use save_tolerance_update() to persist between sessions.

Usage

update_species_tolerances(
  records,
  species,
  update_vars = NULL,
  presence_col = "presence",
  method = c("map", "mcmc"),
  n_iter = 5000L,
  n_warmup = 1000L,
  prior_sd_fraction = 0.2,
  min_records = 20L,
  verbose = TRUE
)

Arguments

Value

Invisibly: a named list with elements:

• species: species key

• method: "map" or "mcmc"

• n_records: number of matched records used

• loglik_null: log-likelihood of intercept-only model

• loglik_fit: log-likelihood at MAP/posterior mean

• mcfadden_r2: 1 - loglik_fit / loglik_null (pseudo-R^2)

• updated_params: named list of updated parameter values per variable

• posterior_sd: named list of posterior SD per parameter (from Laplace/MCMC)

• prior_params: the parameter values before updating (for comparison)

• convergence: optim() convergence code (0 = success; MAP only)

Side effect: updates the in-session cache accessed by score_locations().

What gets updated

For each variable in update_vars, the function estimates shifts in optimal_min and optimal_max (the sweet-spot bounds), plus acceptable_min / poor_min and acceptable_max / poor_max (shoulder bounds, updated proportionally to maintain curve shape). Parameters are constrained to biologically plausible ranges via box constraints in the optimiser.

See Also

get_tolerance_posteriors(), reset_tolerance_update(), save_tolerance_update(), load_tolerance_update()

Examples

# Field records with presence/absence + environmental measurements
records <- data.frame(
  lat      = c(51.5, 51.6, 51.7, 51.8, 51.4),
  lon      = c(-4.1, -4.2, -4.0, -4.3, -4.0),
  presence = c(1, 1, 0, 0, 1),
  temperature = c(16.2, 17.1, 12.3, 11.0, 15.8),
  salinity    = c(32, 33, 31, 30, 34),
  depth       = c(5, 8, 15, 20, 3)
)

fit <- update_species_tolerances(
  records = records,
  species = "ostrea_edulis",
  update_vars = c("temperature", "salinity", "depth")
)

# See updated parameters
get_tolerance_posteriors("ostrea_edulis")

# Re-run predict_oyster -- it will automatically use updated parameters
result <- predict_oyster(survey, "ostrea_edulis")

# Save to disk for next session
save_tolerance_update("ostrea_edulis")

Description

Compares the suitability scores from predict_oyster() against a presence/absence dataset to quantify how well the model discriminates between occupied and unoccupied habitat. Useful for assessing whether the AHP-weighted scoring produces ecologically defensible outputs before using them for site selection.

Metrics computed:

• AUC (Area Under the ROC Curve): probability that a random presence location scores higher than a random absence location. 0.5 = random; 1.0 = perfect discrimination.

• Optimal threshold: Youden's J statistic (maximises sensitivity + specificity - 1). Used for the confusion-matrix-derived metrics below.

• Sensitivity (true positive rate): fraction of presences correctly predicted above threshold.

• Specificity (true negative rate): fraction of absences correctly predicted below threshold.

• F1 score: harmonic mean of precision and sensitivity.

• Brier score: mean squared error between predicted probability and observed presence/absence; 0 = perfect, 0.25 = uninformative.

• TSS (True Skill Statistic = sensitivity + specificity - 1): values above 0.4 generally indicate useful predictive skill.

Usage

validate_against_records(
  predicted,
  records,
  presence_col = "presence",
  match_radius_deg = 0.002,
  plot = TRUE,
  verbose = TRUE
)

Arguments

Value

A named list:

• auc: numeric [0,1]

• optimal_threshold: numeric [0,1] – Youden's J

• sensitivity: numeric [0,1] at optimal threshold

• specificity: numeric [0,1] at optimal threshold

• f1: numeric [0,1] at optimal threshold

• tss: numeric [-1,1] at optimal threshold

• brier_score: numeric [0,1]

• n_presences: integer

• n_absences: integer

• n_unmatched: integer – records dropped due to no nearby prediction

• roc_df: dataframe with columns threshold, sensitivity, specificity (for custom plotting)

Examples

# Load known flat oyster records (e.g. from NBN Atlas CSV export)
records <- read.csv("nbn_ostrea_edulis.csv")

# Run prediction on same area
result <- predict_oyster(survey, "ostrea_edulis")

# Validate
val <- validate_against_records(result, records,
                                 presence_col = "presence")
val$auc           # overall discrimination power
val$tss           # skill score
val$roc_df        # data for custom ggplot