User Feedback for App Weather Predictions
-
Subscription user here. One thing I learnt about probabilistic models, machine learning and A.I is the "feedback loop" to refine and increase accuracy of the model.
When Windy App throws a 1pm to 4pm rain prediction at me (for my GPS location), there is no way for me to feedback to the system whether this prediction was 0%, 50% or 100% accurate. Basically, your A.I. model only dishes out forecasts and has no real world feedback. Isn't that a recipe for zero self-improvement and a way to repeat mistakes?
Please give paid users the ability to rate poor forecasts that made them stop an outdoor job and made lots of people really angry.
-
@asheem01
Windy shows several deterministic weather models. The probability of a forecast is not displayed as this is deduced from ensemble weather models.
But in Windy the user can compare several models. By using the “Compare” table the user can feel the probability of the forecast.
Then you can rely on Meteoblue which itself is a Multimodel with A.I processing, as Meteoblue explain.
Looking ahead, there is no doubt that artificial intelligence (AI) and machine learning (ML) will become increasingly important in weather forecasting. The ECMWF, whose IFS model is the one used by default in Windy, is already carrying out parallel work with its new AI model:
https://www.ecmwf.int/en/about/media-centre/aifs-blog/2023/ECMWF-unveils-alpha-version-of-new-ML-model
