ForecastWatch Provides The Highest Quality Data

ForecastWatch is a name you can trust. Before either forecasts or actual weather measurements are accepted for accuracy calculations, they must pass through several layers of validation. We would rather reject a questionable forecast or actual measurement than use it with uncertainty. When we present accuracy calculations, we want to be sure that the data presented reflects actual forecasts, and not errors or incorrect data.

ForecastWatch Validates All Forecasts

ForecastWatch checks forecasts before using them in accuracy calculations. This validation is also used as the basis for the ForecastWatch Quality Assurance product, which continually monitors forecasts for errors or problems. Thankfully, less than 100 out of every million temperature forecasts retrieved by ForecastWatch end up permanently invalid. All invalid forecasts are reviewed by ForecastWatch staff, and if the problem is correctable, it is corrected and then marked as valid.

ForecastWatch Validates All Actuals

Actuals go through a similar process. Initial checks are performed to ensure that the high and low are reasonable. Additionally, the daily high and low are verified against hourly data, and declared invalid if there is not a substantial match. Approximately one percent of all actual observations are determined to be of questionable validity and are not used.

ForecastWatch Team Hand Audits Accuracy Measurements

After both forecasts and actuals go through the automated validity checks, there is yet another layer of checks. The ForecastWatch team uses web-based tools to hand-validate the remaining forecasts and actuals. The team can review all forecasts and actuals, but the tool is most useful by flagging outliers that deserve additional inspection. For example, actuals that forecasts all missed, or large forecast errors by a single provider.

The screenshots above show some of the online audit tools used to help ensure that the data ForecastWatch presents is valid, accurate, and of the highest quality.