Replication data for "Observations are the dominant climate uncertainty in impact projections"

This dataset contains the replication data for "Observations are the dominant climate uncertainty in impact projections" by Schwarzwald et al., currently in review. 

It includes, in subdirectories: 

`reg_inputs`: Input datasets used to train climate impact models

`reg_outputs`: Regression coefficients of climate impact models

`aggs`: nonlinear climate metrics from global climate models (GCMs) used as inputs to climate impact models, aggregated to relevant shapefiles for each climate impact model

`uncert_parts`: Partitioning of uncertainty in damage projections into contributions from regression uncertainty, (climate) model uncertainty, internal variabiltiy, and reanalysis uncertainty

To use the replication code, set the path (in the dir_list.csv file) of the `proc` row to the parent directory of this data. 

Note that not all files necessary to reproduce the manuscript figures are included here due to space limitations. They can instead be regenerated from the replication code:

Damage projections can be regenerated from the files in `aggs` using the notebook `3_project_impacts.ipynb`

Intermediate files required for particular figures can be regenerated in the relevant figure creation notebooks