search

ISI-MIP / attrici

Produce counterfactual climates for ISIMIP.
GNU General Public License v3.0
9 stars 2 forks source link

Implement tasrange #60

Closed SimonTreu closed 4 years ago

matthiasmengel commented 4 years ago

Implemented via a88befa0d8fa79f1fff778124b5d7dcce80a2cab.

matthiasmengel commented 4 years ago

This is the current performance by 503a41ef2b2ad93221ba9a9d530a8f6264532bd0 via isicf022_gswp3_tasrange_sub20. This can be improved yet.

image

matthiasmengel commented 4 years ago

Improved via narrower priors via ac4e877984592cf6, see isicf022_gswp3_tasrange_sub20_sharperslopepriors image

SimonTreu commented 4 years ago

The worst cell in the run on all cells: Looks like an issue with the logistic function ( intercept can not be set as needed) Screenshot from 2020-01-30 09-46-08

SimonTreu commented 4 years ago

Relu model fixes this cell

Screenshot from 2020-01-30 19-49-06

SimonTreu commented 4 years ago

with relu. Works at least as good as the logistic model. Screenshot from 2020-01-31 09-38-54

matthiasmengel commented 4 years ago

On the full dataset, using RELU via isicf023_gswp3_tasrange_sub01_relu. We need to check why some grid cells fail. @SimonTreu any ideas how to deal with the regions that still show significant trend?

image

SimonTreu commented 4 years ago

Fixed when no trend in sigma is modeled