This repository contains scripts that together perform an analysis of GloFAS's and Google FloodHub's predictive performance of river discharge data. For both, the data is downloaded, processed to a uniform type, and, subsequently, their accuracies (e.g. by probability of detection (POD) and false alarm ratio (FAR)) are assessed using impact data.
As of now, late 2024, the scripts focus on floods in Mali, but later, the goal is to generalize to more countries. For questions, contact elskuipers@rodekruis.nl and toldenburg@rodekruis.nl.
An analysis can be divided into three main parts: (1) preparing the GloFAS data; (2) preparing the FloodHub data; and (3) comparing them. The directories are set-up comparably:
- Preparing GloFAS data: GloFAS
- Preparing FloodHub data: the GoogleFloodHub directory contains scripts to query the (beta) FloodHub API and transform the downloaded data into a uniform format suitable for comparison with GloFAS forecasts. A set-up and manual for the downloading part can be found here.
- Comparison: ...
...