# What is a Probabilistic Forecast?

Arguably one of the most valuable types of forecasts for a sailor, but also one of the hardest to find.

“U.S. Navy Ship Routing Officers typically avoid the area with more than a 5% probability of tropical storm force winds [≥ 34 knots], which serves as a reasonable threshold for most large ships.”

Mariner’s Tropical Cyclone Guide

You read that right: Even the U.S. Navy strives to avoid areas with a better than 5% chance of 34-knot winds.

Wouldn’t you too?

But what forecast gives you a percentage chance of winds?

What you need is a probabilistic forecast!

Before getting into the details, we need a bit of background.

## A Brief Intro to Numerical Weather Prediction

Many of today’s forecasts are generated using a method known as numerical weather prediction (NWP).

This method involves three pieces: initial conditions, a model, and outputs.

In simplistic terms, the model consists of a collection of mathematical formulae which are used to compute approximations of the physical processes involved in the weather.

The initial conditions are generally derived from observations (measurements) of the atmosphere. These observations can come from weather balloons, satellites, terrestrial weather stations, buoys, and so on.

The model is then run on a supercomputer, starting with the initial conditions, then using the model’s mathematical formulae to simulate how the weather evolves over time.

The outputs of this simulation are weather forecasts.

This means that the accuracy of the forecasts are dependent on two things: (1) how well the initial conditions reflect the actual state of the physical systems involved (i.e., atmosphere, oceans, etc.) at a point in time; and, (2) how accurately the model represents natural reality.

A perfect glimpse into the exact state of the physical systems is impossible, so there will always be some forecast error due to imperfect knowledge of the system.

And, because global weather is a chaotic system, small variations or imprecisions in the initial conditions can lead to wildly different outcomes in the forecasts, especially the more they extend out into the future. This is a result of the butterfly effect: variations compound at each iteration of the simulation.

## Wind Forecast: An Example

Say you want to generate a forecast of wind speed. You set up the initial conditions and run the model simulation on your supercomputer. A bit later it spits out a forecast showing a wind speed of 15 knots for location 37°N, 122°W at noon.

One might be tempted to think, “Great! I’ll sail out there and at noon will have 15 knots of wind!”

Of course, you know from experience that wind forecasts often don’t pan out. During the author’s Pacific crossings, the wind speed was frequently ± 10 knots from what had been forecast.

The problem is that the model computes only a single wind speed value for each forecast point—there isn’t a mechanism to convey the inherent uncertainty implied in that value.

This is where a probabilistic forecast comes in!

## The Probabilistic Forecast

When running your NWP forecast model, you may not know exactly which of the initial conditions are imprecise and by how much.

However, with a lot of experience and some study, you will likely be able to settle on a list of “usual suspects” which tend to introduce the most critical uncertainties into the forecast.

Then, you can run your forecast model multiple times in various “what if” scenarios (known technically as perturbations), to account for the possibility of these uncertainties, to get a sense of how many scenarios result in a certain outcome.

This, in a nutshell, is probabilistic forecasting!

## Are Probabilistic Forecasts More Accurate?

Actually, it’s not about being more or less “accurate” of a forecast at all.

The probabilistic forecast helps you make more informed decisions in the face of uncertainty.

As a sailor, if I download a GRiB file and see an area of 34-knot winds seven days in the future, what are the chances it’ll actually be 34 knots when I get there?

(The flip side of that—and potentially the more crucial—is, if I see an area of winds in the 20s, what are the chances it’ll actually be in the 30s when I get there?)

With a regular forecast, you have no way of evaluating these questions.

Given the choice of looking at a forecast for areas with 34-knot or higher winds, and looking at a forecast for areas with a 5% chance of having 34-knot or higher winds, I’ll choose to look at the latter.

(Refer back to the U.S. Navy quote at the beginning of this article!)

## Where Can Sailors Get Probabilistic Forecasts?

Weathermuffin has two WxHeadsUp™ alerts based on probabilistic forecasts: sustained winds forecast to have a 5% or greater chance of occurring, and significant wave heights forecast to have a 5% or greater chance of occurring:

The Wind ≥ 5% Chance alert:

The Waves ≥ 5% Chance alert:

You can also retrieve probabilistic wind speed guidance directly from NOAA here, though obtaining these over a low-bandwidth connection will require some effort.

## Conclusion

Probabilistic forecasts are a valuable component of informed decision-making on the water.