Stands for the PredictWind proprietary weather model that uses the NCEP global initial conditions, processed through the CSIRO CCAM model to generate the PWG forecast.

PredictWind's proprietary weather model that uses the ECMWF global initial conditions, processed through the CSIRO CCAM model to generate the PWE forecast.

PredictWind is the only private company in the world with this proprietary technology that generates forecasts down to 1km resolution for popular areas around the world. The 1km resolution forecasts go out to 36 hours, and rate highly in the accuracy scores using the validation tool.

Our proprietary AI-powered model represents the most important forecasting breakthrough ever at PredictWind. PWAi combines the strengths of ECMWF (hi-res), AIFS, Fengwu, and GraphCast to generate a single, optimized forecast covering the globe. It provides 1-hour time steps for the first 3 days, offering greater temporal precision than other AI models. PWAi performs very well in validation studies over short to medium-term periods in larger-scale areas.

Stands for European Center for Medium-Range Weather Forecasts and is highly regarded by Meteorologists and top Navigators around the world. The ECMWF High RES model consistently rates as the top global weather model from a national weather service with the highest rating scores. In March 2016 ECMWF increased the resolution of their model to a record-breaking 9 km resolution, which is currently the highest resolution global model available. ECMWF data has a very high acquisition cost, and this is why the data is not widely used by many weather websites, and has been traditionally used only by top yacht racing teams and meteorologists.

Developed by ECMWF, AIFS is their next-generation AI-driven model. ECMWF AIFS model uses advanced AI to deliver timely, accurate global forecasts. It performs especially well offshore, capturing large-scale weather patterns in the medium to longer term.

Global Forecast System from NCEP. This is used by most other weather websites/apps. We now display the GFS-FV3 model when you see the GFS label in PredictWind. It’s the first significant upgrade to GFS in about 40 years. Unlike the previous GFS model, GFS-FV3 is able to simulate vertical movements such as updrafts, a key component of severe weather, at very high resolution. So far, tests suggest that the FV3 model has more accurate five-day forecasts, as well as better predictions of hurricane tracks and intensification. Although the new FV3 core has shown improvements over GFS it remains ranked 3rd for accuracy behind ECMWF(1st) and UKMO(2nd).

Otherwise known as the “Unified Model” by the UK Meteorological Office has a long reputation as a market leader in forecast modelling. UKMO has very similar accuracy to the ECMWF model offshore, and is slightly behind the ECMWF models for the land-based weather stations.

High-Resolution Rapid Refresh and is an NOAA real-time 3-km resolution, hourly updated, cloud-resolving, convection-allowing atmospheric model, initialized by 3 km grids with 3 km radar assimilation. Radar data is assimilated in the HRRR every 15 min over a 1-h period adding further detail to that provided by the hourly data assimilation from the 13 km radar-enhanced Rapid Refresh. To learn more see the video.

North American Mesoscale Forecast System and is one of NOAA’s major weather models, which in this case covers most of North America. NAM is a mesoscale model, which means that the numerical analysis is able to model land, and other features, at a higher resolution than in a global model, leading to improved forecast accuracy.

A small-scale numerical prediction model, operational at Meteo-France since December 2008. It was designed to improve short-range forecasts of severe events such as intense Mediterranean precipitations (Cévenole events), severe storms, fog, urban heat during heat waves. This model is highly regarded by top racing navigators and beats the ECMWF forecast.

Germany’s global numerical weather prediction (NWP) model, developed by the German Weather Service (DWD) in collaboration with the Max Planck Institute for Meteorology (MPI-M). It uses an icosahedral grid: the Earth’s surface is divided in a grid system based on subdivided icosahedra, giving roughly even coverage globally. This avoids the distortion and computational inefficiencies encountered near the poles in traditional latitude-longitude grids.

The SPIRE forecast is being removed from PredictWind as of October 2025, as the model used, SOF-D, is being deprecated by SPIRE.

Depending on the nature and volume of data required to populate a given map or table, this varies by weather model, parameter and App/Forecast Website.

The table below summarizes the weather models available by parameter by app:

Website - PredictWind Forecast Website

PWA - PredictWind App

OA - Offshore App

Please note that the exact update times may vary slightly due to variations in the source data delivery and the ingestion process.