Getting accurate weather for open ocean areas

In order to compute the best route for selected objectives, and to correctly optimise voyage for a ship, Syroco EfficientShip uses a high-fidelity digital twin of the vessel that accurately simulates the behaviour of the ship in given operating conditions. For this simulation to be meaningful, the platform requires access to accurate weather in areas that are relevant for users. 

Syroco has chosen to partner with Spire, a global provider of space-based data, analytics and services. Spire uses space-powered weather forecasting technologies to deliver current, forecasted and historical metocean data.  

The reason Syroco has chosen the Spire Weather service is that the metocean data it provides is simply the most extensive and accurate on the market. Spire has an unfair advantage: their proprietary constellation of 100+ multipurpose satellites that use radio occultation technology. This unique asset allows Spire to capture precise atmospheric measurements and therefore to deliver high-quality weather data to enhance forecasting accuracy. 

Measuring initial atmospheric conditions

Initial conditions define the atmosphere’s current state: they are the starting point of any numerical prediction model. Usually, atmospheric conditions are measured through radiosondes, pilot-balloon stations and buoys. Since only buoys can be installed at sea and they collect surface-level data, measures collected through these networks of sensors are less dense and less accurate for open ocean areas than for land. 

On their side, Spire uses satellites to measure temperature, humidity and pressure at different levels of the atmosphere through the use of Radio Occultation (RO) technology.

Clearly, Spire is not the first or only operator to use RO measurements in their models. For example, the COSMIC-2 satellites (a US/Taiwan joint project) have been analysing the atmosphere in a similar way since June 2019, leading to more accurate models than were available before. Yet, COSMIC-2 only uses 6 geosynchronous satellites, whereas the Spire constellation of 100 low earth orbit satellites provides a much better latitudinal distribution leading to measurements even close to the poles. The charts below illustrate this difference (source: ECMWF).

Numerical prediction model and ML-based refinement

The measurements collected by Spire produce broad and comprehensive sets of data that are used to feed the FV3 numerical model from US-government-run NOAA. That model is the same one that is at the core of the Global Forecast System (GFS), but it is fed with more accurate initial conditions - especially for open ocean areas. 

From the model output, Spire runs a set of in-house machine learning algorithms with observational data to adjust the outputs and produce a trustworthy 16 days forecast.

It is the combination of these pillars (atmospheric conditions, well-proven numerical model, ML-based algorithm) that form a virtuous cycle as the model continuously trains on observed conditions to adjust its output.

Comparing accuracy of forecasts

Website PredictWind has conducted a comparison between several forecast providers and models, in two different situations: land areas and open ocean. They compared the accuracy of forecasts from the following organisation:

• ECMWF - the European Centre for Medium-Range Weather Forecasts
• GFS - the US-operated Global Forecast System
• UKMO - the UK Met Office
• Spire 

For all open ocean forecasts studied, Spire was ranked number 1 and was consistently the best on each day of the forecast. For land forecasts, Spire came out number 2, slightly behind the ECMWF. Since Syroco EfficientShip users primarily use the solution on open oceans, choosing Spire Weather was an obvious decision.