Extreme Precipitation

This is the source code that accompanies the paper [1].

Setup

General

Some R-packages (terra, sf) have a couple dependencies. Install: bash sudo add-apt-repository ppa:ubuntugis/ubuntugis-unstable sudo apt-get update sudo apt-get install libudunits2-dev libgdal-dev libgeos-dev libproj-dev libsqlite0-dev

R

rig

We use rig to manage the R version.

To install rig on Ubuntu: bash curl -Ls https://github.com/r-lib/rig/releases/download/latest/rig-linux-latest.tar.gz | sudo tar xz -C /usr/local

We are working with 4.3.0, so we add this version and set it as the default: bash rig add 4.3.0 rig default 4.3.0

renv

Use renv to install all required R modules from renv.lock: r renv::restore()

After manually adding a package (by updating DESCRIPTION), update the renv.lock using this command: r renv::init() and choosing 2: Discard the lockfile and re-initialize the project.

Usage

The provided code is quite memory-consuming and was executed on a workstation with 32GB RAM.

Data

The data is provided by the Austrian Zentralanstalt für Meteorologie und Geodynamik.

Method

The methods used are summarized in the submodule mevr. This file contains all functions related to the MEV, SMEV and TMEV. Note: It is planned to release the submodule as a separate R-package.

Runnable files (Spatiotemporal model)

All the files concering the spatio-temporal model are contained in the folder spatio_temporal_model. The order of the following list defines the order in which the scripts should be run. * spattempmodel_etl.R: Transforms the raw data into the format that is required for training, testing and analysing with the following files. * spattempmodel_train.R: Trains a spatio-temporal model using the bamlss framework. * spattempmodel_predict.R: Computes a prediction of the underlying Weibull parameters over the whole domain of Austria. * spattempmodel_returnlevels.R: Evaluates the return levels for daily precipitation sums based on the TMEV approach. Either the return levels are based on the whole year or separatly computed for each month. * plot_spattempmodel.R: Plots the return levels. In case of monthly return levels a figure is created for each month and in addition a map of the month with the highest return is plotted.

Runnable files (Individual stations)

• 10_years_window.R: Median 10-, 50- and 100-year daily rainfall return levels of Austrian stations with more than 50 years series length are computed using the TMEV.
• Poinestimates_error.R: Compares the performance of the TMEV with the SMEV as pointestimates.
• monthly_returns.R: For each available Austrian weather station the month with the highest 50-year daily return level is computed.
• motivation.R: Illustrates the advantage of the TMEV comparing two stations with different seasonal behaviour.
• parameterevolution.R: Shows the annual trend of the underlying Weibull parameters from 1900 to 2016.
• data_mayrhofen_plots.R: Creates the data for station mayrhofen, that is used for plotting.
• plotting_mayrhofen.R: Creates several plots for the station of Mayrhofen illustrating the change in seasonality.

Helper files

• ccc.R: Defining some global constants.
• plotting_insert.R: Plotting an insert with a stations location.

References

[1] M.-A. Falkensteiner, H. Schellander, G. Ehrensperger, and T. Hell (2023). Accounting for seasonality in the metastatistical extreme value distribution. Weather and Climate Extremes, vol. 42, p. 100601, Dec. 2023 (https://doi.org/10.1016/j.wace.2023.100601)