Assessing the performance of regional landslide early warning models: the EDuMaP method.

This repository contains a python implementation of the EDuMaP ("event, duration matrix, performance") framework for assessing the performance of regional landslide early warning models presented in Assessing the performance of regional landslide early warning models: the EDuMaP method.

Implementation.

Assessments are carried out in two main steps:

1. Data collection and preparation. Data consisist of a list of issued warnings and landslide occurences along with the time of the occurence. Example data has been downloaded from Rio Alert System and is stored in data/landslides_warnings.xlsx. The preprocessing of this data is executed in the notebook preprocess.ipynb and generates two csv files: landslides.csv and warnings.csv, stored in the generated folder.
2. Calculation of duration matrix and metrics. Duration matrix and associated metrics are calculated in the notebook edumap.ipynb using the landslides.csv and the warnings.csv as input.

Requirements.

The analysis has been implemented using python 3.10 and is based on numpy, pandas and matplotlib for visualization.