Horizon

Horizon is an uncertainty analysis tool for Navigate, the open-source maritime transition model. It samples uncertain input parameters (Monte Carlo or Latin Hypercube), generates one Navigate simulation per sample across scenario combinations, queues the runs, and aggregates the results — turning a single Navigate scenario into an uncertainty study.

Horizon is a companion to Navigate, not a standalone tool: it prepares, schedules, and post-processes Navigate simulations, and a working Navigate installation is required for any actual simulation run.

Disclaimer

Horizon is an open-source analytical tool intended for research and scenario analysis of maritime decarbonisation pathways. Its outputs depend on the model, assumptions, and parameter ranges selected by the user and are provided for illustrative and analytical purposes only. They should not be interpreted as forecasts, benchmarks, recommendations or commercially optimal outcomes, nor as legal, financial or investment advice. Users are responsible for selecting appropriate assumptions and for exercising their own independent judgement when interpreting any outputs.

Requirements

• Python >= 3.12
• Navigate installed in the same environment (Horizon invokes the navigate command)
• pueue for queuing simulation runs — the daemon must be running (pueued -d)

Installation

Install Horizon into the same environment as Navigate:

conda activate nav       # or whatever environment Navigate lives in
pip install .

For development with lint and test tools:

pip install -e ".[dev]"

Quick start

The bundled example runs without Navigate or pueue — it stops after sampling (SampleOnly = TRUE) and writes sampled values plus sampling diagnostics:

cd examples/uncertainty
horizon fuel_cost_study.hor

See examples/uncertainty/README.md for a walkthrough, and docs/configuration.md for the full configuration reference.

Workflow

A full uncertainty study runs in three steps:

# 1. Sample parameters, generate .nav files, queue Navigate runs via pueue
horizon my_study.hor

# 2. Watch progress (pueue) or check for failures
pueue status
horizon --status path/to/output_dir

# 3. Collect the Excel reports of completed runs into one file
horizon -c path/to/output_dir results.xlsx

The .hor file defines the study: which parameters vary, over what ranges and distributions, across which scenario combinations. The .unc file it points at is an ordinary Navigate simulation file with %TOKEN% placeholders where sampled values are substituted.

Sensitivity analysis

After a study completes, Horizon can compute PRCC (partial rank correlation coefficient) sensitivities linking sampled parameters to output metrics:

horizon --sensitivity-analysis my_study.sen

Command-line options

See horizon --help for the full list. The most used flags:

Editor support

The syntax/ folder contains Notepad++ user-defined-language files with syntax highlighting for .hor, .sen, and Navigate files.

Documentation

• docs/configuration.md — .hor and .sen format reference
• docs/ARCHITECTURE.md — how the code is structured, for contributors
• Navigate documentation — the simulation model itself, including the .nav file format

Testing

pip install -e ".[dev]"
pytest

See CONTRIBUTING.md for contribution guidelines and CODESTYLE.md for coding conventions.

License

Copyright 2026 Fonden Mærsk Mc-Kinney Møller Center for Zero Carbon Shipping.

Horizon is licensed under two licenses, depending on the type of content:

• The software — the horizon package, tests, and all build, tooling, and editor-support files — is licensed under the Apache License 2.0 (see also the root LICENSE file).
• The documentation, examples, and figures are licensed under Creative Commons Attribution 4.0 International (CC-BY-4.0).

Every file declares its license through an SPDX-License-Identifier header or through the metadata in REUSE.toml, following the REUSE specification. The full license texts are in the LICENSES directory.