Temperature bias correction #65
Replies: 1 comment 2 replies
-
Hey Simon, I think there is another complementary field related to this known as empirical-statistical downscaling. The guys at metno, especially Rasmus Benestad (this review-like paper somehow got reject but it looks interesting despite the strange name: I know some people at NERSC in Bergen (Richard Davy, see his KrigR tool) that use kriging methods for downscaling, and I think they would be interested in developing such hybrid tools. You could consider reaching out to them. Still, with sparse observations I would be quite curious about what is validating what when comparing a kriged field to TopoSCALE. I think TopoSCALE would be better at capturing local variability that kriging can never really resolve (since it is limited by the density of the observing network). Still, kriging could be used to do a larger scale bias correction which would probably make a lot of sense, while keeping the spatial patterns from TopoSCALE. I think you have outlined most of the methods quite nicely, but you are maybe missing (multiple) linear regression which is usually the first thing people try for a simple bias correction that allows for (somewhat risky) extrapolation of the bias to unobserved areas. Note that kriging is actually a form of linear regression too, but it allows you to include more covariates and is based on more explicit modeling of spatial variability. Happy to discuss more. |
Beta Was this translation helpful? Give feedback.
-
Hey @joelfiddes @yelizy @krisaalstad ,
Do you know what would be the best strategy to bias correct the downscaled temperatures when using
toposub
given there is handful of weather stations in the domain?Some ideas:
Beta Was this translation helpful? Give feedback.
All reactions