The dgevbhm package is an R and Stan-based package for Bayesian hierarchical regional modeling of extreme precipitation. It has the ability to provide robust intensity-duration-frequency (IDF) curve estimations in data-scarce scenarios (e.g. records under 50 years).

The requirements are R (>=4.4.0), rstan and torch.

To install please run: pak::pak("AlexLeeR/dgevbhm")

Basic workflow:

1. Data pre-processing: Use data_read() to format the sub-daily precipitation TXTs
2. Model Fitting: Use dgev_bhm() to fit the Bayesian hierarchical model with processed data
3. Model evaluation and application tools: Utilize diagnostics tools such as pp_plot_torch() and waic, and application tools like idf_plot_torch().

library(dgevbhm)
data('UK_data')

# Fit a model with a shape parameter hierarchical over durations
fit <- dgev_bhm(UK_data, iter=2000, cores=8, shp_d='d')

# Generate an IDF plot for the second locality
idf_plot_torch(fit, j=2, rp=c(2, 10, 100))

The data_read() function...

data_processed <- data_read(
  dir_path = 'path/to/directory',  # path to the directory containing TXTs
  missing_data = 5,                # minimum missing data per year (in percent)
  durs = c(1,3,6,12,24,24,48),     # durations e.g. in hours
  cores = 4,                       # cores for parallel processing
  min_year = 25                    # minimum years for station to be selected
)

This project is licensed under the MIT License.

For questions or support, please contact: Alexander Lee Rischmuller - alex.rischmuller@gmail.com