Climate shift uncertainty and economic damages

Replication for CESifo paper by Romain Fillon, Manuel Linsenmeier, Gernot Wagner

Most data are not available here but can be downloaded from the following sources:
DOSE v2 (econ data + shapefiles) is here: doi:10.5281/zenodo.7573249
ERA 5 on EU climate data store: doi:10.24381/cds.adbb2d47
CMIP6 projections on ISIMIP Protocol 3B
Köppen-Geiger regions in doi.org/10.1038/s41597-023-02549-6
Because climate data is too large, I upload a subsample to see how the code runs (only illustrative)
Furthermore, I provide intermediate outputs once the distribution (control, synthetic, projections) are built (intermediate outputs are random sample from real outputs).

Step 0: dezip all files :
/data_additional/DOSE/geometries
/climate_weighting/

Before the pre-analysis step (which processes the raw data to build the temperature distribution), the datasets are very large. Therefore, I provide a first subsample in /subsample_reproducibility/, which can be used to run and understand preanalysis.r.
After the pre-analysis, the data remain heavy due to the fine binning resolution. For this reason, a second subsample is provided in /climate_outputs_pretreated_subsample/, which can be used to run and understand analysis.r.
/climate_outputs_pretreated/ is empty because complete preprocessed data is heavy.

Step 1: compute warming patterns

[might need climate data] File: preanalysis.r
CMIP6 outputs (2041-2060, 3 SSP, 5 ESM) from ISIMIP and data in /data_additional/
As CMIP6 data is not directly available here, and for illustration only, we leave a subsample of our climate data in /subsample_reproducibility/. The idea is to avoid downloading the whole climate data. Subsample is based on first weeks of January 2010 (picontrol) and January 2050 (each SSP*ESM).

3 possibilities:

1. [need climate data] Run complete preanalysis with complete climate data (subsample_creation=0, subsample_run=0). Output of this run is in /climate_outputs_pretreated/ and used for final results
2. [need climate data] Run complete preanalysis and create subsample data with complete climate data (subsample_creation=1, subsample_run=0). Output of this run is in /subsample_reproducibility/
3. [no need climate data] Run complete preanalysis with subsample climate data already created in /subsample_reproducibility/ (subsample_creation=0,subsample_run=1), for illustration of how preanalysis.r works

Step 2: compute damage patterns

[need climate data] File: p_estimate_model.py
DOSE and ERA5 data
Save damage to /damage_frompython/
Damage functions can be (i) standard GDP growth (ii) indexed fd if in GDP levels.

Step 3: compute omitted damages and plot results

File: analysis.r
2 possibilities:

1. [need climate data] Run analysis with complete results from preanalysis, either from subsample (step 1, subsample_run=1), or complete data (step 1, subsample_run=0). Data to be created in /climate_outputs_pretreated/ before this step is feasible
2. [no need climate data] Run analysis from subsample from complete data stored in /climate_outputs_pretreated_subsample/ (real data but aggregated to 1°C)

Additional data in /data_additional/
Dose-response functions in /damage_frompython/

Additional file

File: /Figures/illustrative_graph.py
Illustrative first graph.

Contact for questions: rfillon@protonmail.com
Distributed under the GNU GENERAL PUBLIC LICENSE. See LICENSE for more information.