Climate data from High-Resolution Rapid Refresh (HRRR) is only available starting from september 2014 while SWE labels are available starting from january 2013. Which climate data should we use between january 2013 and september 2014 ?
Would be also nice to confirm we can use the archived HRRR data for training from other sources since the allowed data access location only has information for the last two days.
It’s not an issue for inference but makes the location not as useful for training.
Citation from main website:
All data sources and access locations detailed below are pre-approved for both model training and inference** . These data have undergone a careful selection process and are freely and publicly available. Note that only the listed data products and access locations are pre-approved for use. You must go through an approved access location when using these sources for inference.
You must go through an approved access location when using these sources for inference. - It could mean, that for training, it is allowed to use any data sources we want, and restrictions are only imposed during inference time. That would mean, that it is allowed to supplement training data with another climatic data source.
Hi all - Thanks for reaching out about this. @rszostak is correct - approved sources and access locations are the only ones allowed for inference, but you may explore and incorporate additional sources of historical data during training. Hope that helps!
Hi @glipstein . Since we can only use approved data products for inference, and HRRR data through the allowed access only supports data within the most recent 5 days of data retrieval, I want to confirm that we will not be using any direct climate data for inference (no HRRR data available through this source during 2020).
If I am mistaken and there is a way to access historical HRRR data through the allowable data repositories, can you please let us know?
@whitelightning450 Through the approved HRRR data access link
https://noaahrrr.blob.core.windows.net/hrrr, you can access HRRR data that dates back to March 21, 2021 during inference. You are permitted to explore and incorporate additional historical HRRR archives during model training. Please see this discussion for more information.
@tglazer The inference period begins 1/7/2020. So to be clear, we will not be able to use HRRR climate information for this year (2020), and then only half of 2021 (3/21/2021 onward).
@whitelightning450 Thank you for clarifying. Because HRRR data is unavailable for the first half of the inference period in the Development Phase, you are permitted to use HRRR from the AWS access point for inference in the Development Phase only. During the real-time Evaluation Phase, you should continue to use the Planetary Computer access point.