Pirate-Weather / pirateweather

Code and documentation for the Pirate Weather API
Apache License 2.0
649 stars 29 forks source link

Data accurancy #207

Open lollita opened 5 months ago

lollita commented 5 months ago

Describe the issue

I compare: https://merrysky.net/forecast/roma/si https://www.ilmeteo.it/meteo/Roma How do you explain a 5° difference? How reliable are the data? What is the source for Italian data?

Acknowledgements

cloneofghosts commented 5 months ago

Thanks for opening this issue. The reason for the discrepancy is Pirate Weather uses the global GFS model which is fairly low resolution and only updates every 6h. I'm having trouble finding the current conditions on the other site you linked. From what I see it looks like they're using ECMWF data which is higher resolution than the data Pirate Weather uses.

cloneofghosts commented 1 month ago

@alexander0042 It seems like this issue was missed and I know the solution here is to integrate more models (or use observation data). I know you have some on the roadmap but these may also be helpful?

I saw this posted in the Open-Meteo Repo but it looks like this model covers most of Europe and is updated hourly. https://dataplatform.knmi.nl/group/weather-forecast?q=UWC&sort=metadata_modified+desc

I know DWD Icon is another popular source of model data that could be integrated https://www.dwd.de/EN/ourservices/nwp_forecast_data/nwp_forecast_data.html

I know you're busy with open-sourcing the code but for 2.3 a new model could be integrated? You said the Canadian models have a fire index which would close out the Fire Index issue.

EDIT: Also found this on AWS which may be useful?

https://registry.opendata.aws/met-office-uk-deterministic/ (global and UK) https://registry.opendata.aws/meteo-france-models/ https://registry.opendata.aws/hirlam/ (finland only)