Please my belated reply on this topic. Yes this year we applied a different definition of the weather data category. Last year we could not find many short-term forecasts that would span more than three days (starting with the current day). So we changed this and shortened the time span of weather forecasts. On the other hand we were more vigilant to make sure that temperature, wind, and precipitation values are included in our reference dataset.
More details can be found in my review diary.