This project is licensed under the MIT License - see the LICENSE.md file for details.
This repository contains the R code used for simulation, data processing, and figure generation in Jacobson et al. (2020).
Important note: Throughout the simulation code, correlation length is represented by the variable s, whereas the paper denotes correlation length as a.
The corresponding data repository containing simulation results and forecast data can be found here. The repository contains the following
- Simulation results: Each .RData file contains eleven sets of 5000 samples of fraction of threshold exceedance (FTE) values for a simulated verification field and corresponding 11-member ensemble. The verification field's correlation length
s_1is fixed according to the file name (e.g._s2.5indicates correlation length 2.5), and the correlation length of the ensemble means_2ranges from0.5*s_1to1.5*s_1by0.1*s_1, yielding the eleven simulation sets per file. refcstv2_precip_ccpav3_subset_066_to_072.nc: GEFS coarse resolution reforecast data and corresponding CCPA analyses.GSDM_downscaled_*_066_to_072.nc: GEFS forecast data from January, April, July, and October, downscaled to CCPA resolution according to a simplified version of the Gibbs sampling disaggregation model (GSDM).
Code used to process simulation results and forecast-analyses data can be found in process_results.R and process_gefs.R, respectively.
Directory containing simulation code and shell scripts used to carry out the simulation experiment.
Code used to produce all figures.
Jacobson, J., Kleiber, W., Scheuerer, M., and Bellier, J.: Beyond univariate calibration: verifying spatial structure in ensembles of forecast fields, Nonlin. Processes Geophys., 27, 411–427, https://doi.org/10.5194/npg-27-411-2020, 2020.