SeaSCAPE Toolbox: Seabed Shipwreck Classification Assessment and Preservation Evaluation

An open-source ArcGIS Pro toolbox for the automated detection and structural preservation assessment of shipwreck sites from high-resolution bathymetric survey data.

---

Overview

SeaSCAPE implements a two-stage ModelBuilder workflow that takes a single bathymetric raster and produces classified shipwreck detections with preservation scores:

Stage 1 - Derivative Generation Generates six terrain derivatives from the input bathymetry: slope, curvature, TPI (Topographic Position Index), roughness, resampled bathymetry, and Broad-Scale BPI (Bathymetric Position Index).

Stage 2 - Detection & Evaluation Reclassifies the derivatives using Jenks Natural Breaks, applies a variable-threshold detection system that adjusts sensitivity by seabed complexity zone, converts detections to polygons, scores each on five confidence criteria, determines the orientation of each detection, performs cluster analysis to characterise structural fragmentation, and assigns a four-class preservation classification:

Class	Label	Score Range	Description
1	Intact	10–8	Structurally continuous wreck signal
2	Broken	7–5	Distinct sections separated but dominant structure remains
3	Fragmented	4–2	Many small clusters with no dominant structural core
4	Indeterminate	1	Below resolution gate — insufficient spatial information

Outputs

The toolbox produces classified detection polygons in two output feature classes: a full version retaining all attribute fields from the workflow, and a simplified version reduced to key output fields only (Confidence, Degrees, BearingTxt, Pres_Class, Pres_Index). The Confidence field can be used to filter and prioritise detections according to survey objectives. If the objective is to find and analyse extant wrecks, moderate to high confidence levels can be assessed first (e.g. ≥8). If smaller, possible wreck sites are being sought, then all detections can be retained for manual assessment.

Requirements

• ArcGIS Pro 3.7.0
• Extensions: Spatial Analyst, 3D Analyst, Image Analyst
• Python packages: scipy (for MST cluster analysis — included with ArcGIS Pro's conda environment)
• Input data: Bathymetric raster in a projected coordinate system with metres in X, Y, and Z at 1 m resolution

Installation

1. Download or clone this repository
2. In ArcGIS Pro, open the Catalog Pane
3. Right-click Toolboxes → Add Toolbox
4. Navigate to the downloaded folder and select SeaSCAPE.atbx

Once opened as tools in ArcGIS, both workflow stages will provide additional information about each parameter and default settings.

Repository Structure

SeaSCAPE/
├── SeaSCAPE.zip                       # ArcGIS Pro toolbox (unzip this and add the .abtx file to your project)
├── scripts/                           # Python script tools
│   ├── CalculateBroadBPI.py           # Broad-Scale Bathymetric Position Index
│   ├── CalculateTPI.py                # Topographic Position Index
│   ├── ClusterAnalysisTool.py         # MST-based cluster analysis and gap measurement
│   ├── DetectionExport.py             # Export detections to full and simplified outputs
│   ├── NormaliseMetrics.py            # Min-max normalisation of preservation metrics
│   ├── PreservationClassification.py  # Four-class preservation classification
│   └── VariableThresholdDetection.py  # Zone-based adaptive threshold detection
├── README.md
├── LICENSE
├── CITATION.cff
└── .gitignore

Usage

Quick Start

1. Add the toolbox to your ArcGIS Pro project
2. Run Stage 1 — Derivative Generation with your bathymetric raster
3. Run Stage 2 — Detection & Evaluation using the outputs from Stage 1
4. The final output is a polygon feature class with confidence scores (Confidence), orientation (Bearing), and preservation classifications (Pres_Class, Pres_Index)

Custom Script Tools

The seven Python scripts in scripts/ can also be run independently from the ArcGIS Pro geoprocessing pane:

Script	Purpose
CalculateBroadBPI.py	Broad-Scale BPI using annular focal neighbourhood (Lundblad et al., 2006; Walbridge et al., 2018)
CalculateTPI.py	TPI using circular focal neighbourhood (Weiss, 2001; Jenness, 2006; Dilts, 2015)
VariableThresholdDetection.py	Three-zone adaptive detection with configurable thresholds
ClusterAnalysisTool.py	Cluster merging, area statistics, dominance, and MST-based MaxGap
NormaliseMetrics.py	Min-max normalisation of R_mean, R_std, CV_wreck
PreservationClassification.py	Four-class preservation scheme with composite scoring
DetectionExport.py	Export to full and simplified feature classes

Default Parameters

Detection and classification thresholds have sensible defaults calibrated to 1 m bathymetric data across 12 survey regions in the English Channel. All thresholds are exposed as adjustable parameters for calibration to different environments or resolutions.

Key defaults:

• Detection thresholds: High zone = 15, Medium zone = 13, Low zone = 12
• Buffer distances: Positive = 30 m, Negative = −15 m
• Gap threshold: 5 m (structural separation distance)
• Dominance threshold: 0.3 (ratio of largest cluster area to total)
• Resolution gate: 50 m (minimum polygon perimeter length)
• BPI annulus: Inner = 50 cells, Outer = 500 cells (at 5 m resampled resolution)
• TPI radius: 30 cells (at 1 m resolution)

Tutorial

A full tutorial of the toolbox can be found in the Releases section of this page. It contains a .pdf workflow with screenshots based on a provided sample bathymetric survey as well as the expected output detections using default settings.

Citation

If you use SeaSCAPE in your research, please cite:

Pols, C.T. (2026). SeaSCAPE: Seabed Shipwreck Classification Assessment and Preservation Evaluation [Software]. https://github.com/ctpols/SeaSCAPE

A CITATION.cff file is included for automated citation tools.

Acknowledgements

This research is funded by the Leverhulme Trust through the Southampton Marine and Maritime Institute (SMMI), Intelligent Oceans programme, University of Southampton.

License

This project is licensed under the MIT License — see LICENSE for details.