Gains:
- Ability to understand the four stages of the itinerary and use artificial intelligence safely in appraisal, route alternative and air routing scenarios
- Ability to implement the route produced by artificial intelligence, which is always a no-go in ECDIS and must be subject to UKC control, and to take into account the uncertainty of the weather forecast.
- Ability to understand that the 'optimal' route will be misleading without all constraints such as port window, draft and ETA and that the final approval lies with the captain.
A "straight line" from one point to another at sea is rarely the best route. Wind, waves, currents, shoals, traffic separation schemes (TSS), glaciers, piracy zones and the harbor window (the window of time in which the port will accept the ship) all shape the route. Navigation optimization is planning a safe, economical and timely passage under these constraints. AI is a powerful aid in this work; But the person who approves the plan and is responsible for maritime safety is the navigation officer and the captain.
In this unit, you will learn how to use AI safely in passage planning and weather routing (setting the route according to the weather/sea forecast); You will learn where to stop and turn to ECDIS and official sources.
Four phases of the cruise plan and AI
The international standard navigation plan consists of four stages: appraisal (assessment - collection of all information), planning (planning - drawing the route), execution (execution - execution of the course), monitoring (monitoring - tracking deviations from the plan). AI particularly shines in the first phase: gathering information and generating scenarios.
- Appraisal: AI distills known risks in the transition zone (bottlenecks, TSSs, seasonal weather patterns, piracy zones) into a checklist. But this list is a draft; It is confirmed by official navigation publications (Sailing Directions, Notices to Mariners).
- Planning: AI generates several route alternatives and a time/fuel/air exposure estimate of each. Entering the route into ECDIS, checking no-go areas and UKC (sub-spine space) is human work.
- Execution / Monitoring: AI says, "You may consider adjusting your route as follows" according to the new predictions; The captain approves or rejects.
Attention: The route generated by the AI is a draft and is not a verified route drawn on the official ENC (electronic navigation chart). No route is executed without passing a no-go field in ECDIS and a UKC check.
Air routing: where is the gain, where is the trap
Air routing aims to increase safety and save fuel/time by shifting the route according to the wave and wind forecast. In heavy seas, the ship slows down (added resistance), lurches, and the cargo and boat are stressed. AI quickly compares multiple forecast scenarios and provides predictions such as “this route keeps the highest significant wave height (Hs — statistical height of waves) at 3.1 m instead of 4.5 m.”
The trap is this: the weather forecast is uncertain and becomes less reliable as time progresses. While the 24-hour forecast is quite reliable, the 120-hour forecast carries a wide margin of error. AI can sometimes present this uncertainty as an “exact number.” Therefore:
- Compare multiple forecast sources (different meteorological models); If they all point in the same direction, trust increases.
- Take the long-term recommendation as the "direction", not the "pinpoint"; Update the route with new forecasts.
- Extend the margin of safety according to weather uncertainty: do not approach the storm front for a margin gain.
three mini cases
Case 1 — 14 hours gain, verified. For the North Atlantic crossing, YZ proposes a route around the south of the low pressure center: 14 hours longer distance but much calmer seas, resulting in a 14 hours earlier arrival due to lower added resistance and a fuel saving of 22 tonnes. The captain confirms with two different weather models, draws the route in ECDIS, sees that there is no UKC problem and implements it. The gain is real.
Case 2 — Tempting but dangerous shortcut. On a ferry route, YZ recommends a 40-minute shortcut through a narrow inter-island channel. When the officer checks the ECDIS, he sees that the tidal current of the channel has reached 5 knots at that hour and the UKC is critical. The shortcut is rejected. The AI had calculated the distance correctly and the risk incompletely.
Case 3 — Port window escape. A container ship follows the AI's recommendation to "slow down, save fuel", but the AI is not given a port window constraint. The ship reaches the harbor outside the tide window, cannot enter due to draft, and suffers an expensive anchorage wait. Lesson: "optimal" is misleading if you don't give the optimization all constraints (ETA commitment, port window, tide).
Four copyable templates
1) Itinerary appraisal checklist:
Your role: cruise planning consultant. Transition: [port of departure] ->[port of destination], ship type [x], draft [y]. Outline me an appraisal checklist: bottlenecks/TSSs, seasonal weather risks, piracy/restriction zones, possible UKC points, required navigational broadcasts. Put a note "to be confirmed by official source" on each item. Mark where you are not sure.
2) Route alternative comparison:
I'll give you a 72 hour wind/wave forecast. Ship service speed [x] knots. Generate 3 draft route alternatives; for each: estimated duration, estimated fuel, highest Hs exposure, and most risky area. Constraint: port window [date/time]. I will confirm the shallows and tides; Make a note "Confirmation in ECDIS".
3) Rerouting while tracking:
My current location [latitude/longitude] and planned route are attached. According to the new 24-hour forecast, there is an area of Hs [x] m in front of me. Suggest me 2 options to shift the route to the safe side; Write down the additional time/fuel cost of each. I will make the final decision and UKC confirmation.
4) Captain briefing summary:
Turn this sailing plan draft into a 10-line briefing to the captain: route summary, 3 most critical risks, weather window, decision points. Exaggeration; Write the unclear areas as "unclear/confirmation required".
Weak prompt / Strong prompt
Weak prompt:
Find the shortest route, save fuel.
No restrictions, no air, no security. "Shortest" is often the most dangerous or slowest route.
Powerful prompt:
Your role: air routing consultant. Ship: 180 m bulk cargo, 13 knots service, draft 10.5 m. Route: Rotterdam -> New York, departure 12 March. Attached is a 96-hour wave/wind forecast. Task: 2 safe and economical route suggestions; for each the time, fuel, highest Hs, and the riskiest 6 hours of that route. Restriction: Hs Do not go above 5 m; port window 20 March 06:00-18:00. Mark all depth/UKC and no-gotehid; State the margin of uncertainty in the weather forecast.
Ship, constraint, safety threshold and uncertainty demand make the output usable at sea.
Route decision: who verifies what
input
Contribution of AI
human verification
Weather/wave forecast
Scenario comparison
Multiple model confirmation
Route alternatives
Time/fuel estimation
Drawing in ECDIS
Shallow / UKC
"Confirmation required" warning
Tide + depth calculation
No-go areas
Candidate marking
Official ENC control
port window
ETA account
Agency + tide confirmation
TSS / traffic
General reminder
Live observation with Radar/AIS
Common mistakes
- Executing the route without verifying it in ECDIS. The AI route is drafty; UKC and no-go control on the official map is a must.
- Assuming the long-term forecast is certain. A 96+ hour forecast carries a wide margin of error; take the suggestion as "direction", update it.
- Missing the constraints. "Optimum" is misleading if no port window, draft, ETA commitment is given.
- Relying on a single weather model. If different models diverge, uncertainty is high; expand the margin.
- Putting profit before safety. Do not enter a storm or narrow current channel for fuel savings at the limit.
In summary
Navigation optimization is balancing the safe-economic-on-time triad under constraints. AI saves a lot of time in appraisal checklist, route alternatives and air routing scenarios. But the route is always no-go and UKC checked in ECDIS; the uncertainty of the weather forecast is taken into account; and "optimum" is misleading if all constraints (port window, draft, ETA) are not given to the AI. It is the captain who approves the route and is responsible for maritime safety.
Application task
Select a pass (e.g. Izmir - Barcelona). Have the AI draft two routes with the "Route alternative comparison" template; Give all constraints (draft, harbor window, wave threshold). Then pour the output into a control chart: the source of each number, which fact-checking tool it will verify, and the points where the AI says “confirmation is required.”
checklist
- [ ] I created the appraisal list with AI, but confirmed it with official navigation broadcasts.
- [ ] I drew the route alternatives in ECDIS and performed UKC and no-go checks.
- [ ] I compared the weather forecast with at least two models and took into account the uncertainty.
- [ ] I included all constraints (draft, port window, ETA) in the optimization.
- [ ] As the captain/navigational officer, I gave the final route approval.