Training precipitation models with Tweedie deviance instead of root-mean-square error improves extreme-rainfall recall in downscaling and improves nowcasting skill, with gains compounding over autoregressive lead times.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
physics.ao-ph 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Stop using root-mean-square error as a precipitation target!
Training precipitation models with Tweedie deviance instead of root-mean-square error improves extreme-rainfall recall in downscaling and improves nowcasting skill, with gains compounding over autoregressive lead times.