Like all other weather models available, accuracy relies on a weather input file which can be comparable to a photographic "snapshot" of the earth's atmosphere. If these inputs are inaccurate, the model forecast is likely to be inaccurate. Fortunately, we provide you with six forecasts twice daily created by independently generated weather input files. If there is a discrepancy between the model forecast and real-time observations of more than 30 degrees and/or 5 knots, the forecast should be discarded. This will happen in approximately 15% of cases.

Unfortunately, there is no "golden rule" for which model is more accurate. Often there are periods where one model is better; then it swaps around. If all models are similar, you can have more confidence in the forecast. If they are very different it just shows the uncertainty in the situation.

Make sure you are reviewing the highest-resolution maps available.

The higher the resolution (1km is higher resolution than 50km), the more accurate the forecast. Also, it is worth checking local observation stations and your visual/ instrument observations for the area and matching those to the forecast.

Which model is closest to the actual observation? If, for example, PWG was forecasting 25 knots from the south with rain and the actual weather is 25 knots from the South with rain, then the PWG model is doing a 'good' job of the forecast, and you should put more weight on that forecast for the next 24 hours.

Please see the accuracy ratings of the models from a national weather centre at https://www.predictwind.com/model-validation-technical-report/. With 1-2 models, it can be difficult to decide on which forecast to select, but with up to 9 models, the decision is easier to see a consensus.

To clarify the terminology for the different models that PredictWind offers, see the Help article at http://help.predictwind.com/en/articles/2884560-what-does-pwg-pwe-gfs-ecmwf-spire-ukmo-hrrr-nam-arome-stand-for.