I'm working with Landsat-8 imagery for sea ice research. The attached image shows a case where the metadata reports only 7% cloud cover, but when I apply the CFMask, a much larger area is clearly masked as cloud (as seen on the right). This discrepancy seems too large to be a coincidence.

I thought the cloud cover percentage in the metadata was calculated based on CFMask results. If that's true, shouldn't it match the masked area more closely?

• How exactly is the "cloud cover" value computed in Landsat-8 metadata?
• Is CFMask used directly for that value?
• Why might this large difference happen?

Any insight would be really helpful. Thank you!