Data Analysis to Protect Against Climate-Driven Extremes

Elias Theis, University of Wuppertal
Institute of Technologies and Management of the Digital Transformation
Sumbitted for the seminar "Selected Topics in Data-Science" 2025

Paper

Here is the final paper.

Recreating Results

The following paragraphs briefly describe how to reconstruct the results and graphics of the paper.

Preparing the Dataset

In order to reproduce any analysis, the dataset has to be prepared by decompressing and homogenizing the resolution. This can be done by executing prepare_dataset.py. You just have to change the paths and the bottom of the file before. The data set is thereby inflated from around 10GB to around 400GB.

Settings the path

After the data set has been prepared, all data accesses are carried out via a central method (util.read_radar_data). The correct path to the unpacked data can be set at the top of the util.py file. This variable is called PATH_TO_UNCOMPRESSED_DATA and has to be a Path-object (from pathlib).

Recreating the basic figures

First, uncomment the create_basic_figures() call at the bottom of the main.py and then run it. This could take a relatively long time.

Creating a GIF animation of raw data

First, uncomment the create_gif() call at the bottom of the main.py and then run it.

Create a CPD animation

First, uncomment the create_daily_precipitation_file() at the bottom of main.py, set a range of years and run main.py. Afterwards, set the correct starting date and run the cumulative_precipitation_with_decay.py file. This could take a relatively long time. Note: dont interact with the pyplot window.

Hardware Requirements

• At least 32GB of RAM
• About 400GB of free disk space
• A CUDA capable GPU