Is it a good practice to bias-correct a projected climate model with the exact same biases obtained from the historical climate model against historical observations?
As an example, we have a historical and projected climate model for temperature (tas) along with gridded observations for a local area.
Firstly we modify the historical climate model data by shaping its data distribution and making it closer to the observation’s distribution.
We do the quantile-mapping for each pixel by considering all data that is falling in its closest 15days of the year along all years.
Afterwards we will use the exact same biases to shape the climate projection data distribution.
By doing that, we would expect to have a similar distribution to the historical climate model plus a warm bias. Does that make sense?
Fulco Ludwig can comment on that in more detail. In general, it depends on the variable you would like to correct. For more information so far – please have a look in the tutorials for this specific topic: http://climateservice-global.eu/tutorials/#close
As far as I understand it makes sense. However, I do not understand what you mean with: “Firstly we modify the historical climate model data by shaping its data distribution and making it closer to the observation’s distribution”. How have you modified the data?