Skip to main navigation Skip to search Skip to main content

DBSCAN 3D Clusters of SPEI-90 days – Linguere, Senegal, 1981-2023

Dataset

Description

Science Case Name: Multi-Hazards in Senegal.

Dataset Name/Title: DBSCAN 3D Clusters of SPEI-90 days – Linguere, Senegal, 1981-2023

Dataset Description: The dataset contains gridded data on SPEI-90 days over Linguere area of Senegal.

Key Methodologies: Droughts were computed with SPEI-90, with daily precipitation from Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS) (1981-2023) and daily maximum, average and minimum air temperature from ERA5-Land. Potential evapotranspiration (PET) was computed with the Hargreaves equation from the SPEI R package. Water balance, the difference between precipitation and PET, was aggregated to 90-day rolling sums and Z-scores were computed from distributions of values from day of year (43 points).  Days with Z-scores below or equal to –1 were marked as droughts.

Spatio-temporal DBSCAN was conducted with Python packages st_dbscan https://github.com/eren-ck/st_dbscan (Cakmak et al., 2021). Spatial proximity (epsilon 1) was set to 0.5 (0.5 degree), temporal proximity (epsilon 2) was set to 1.5 (1.5 days); min number of samples was set to 30.

The values in the NetCDF files represent cluster numbers (from 0 to 11), with values -1 (negative 1) representing outliers.

A summary table in CSV represents rows for each cluster with start and end dates, average severity and intensity, and maximum number of affected cells.

Temporal Domain: 1981–2023, daily

Spatial Domain: Linguere, Senegal, West Africa, Spatial resolution ca 0.1°x0.1° (EPGS:4326)

Key Variables/Indicators: Spatio-temporal clusters of dry/drought events

Data Format: netCDF, CSV 

Source Data: ERA5-Land daily min, max and average air temperature and CHIRPS daily precipitation

Accessibility: Zenodo, https://doi.org/10.5281/zenodo.15212446

Stakeholder Relevance: Identifying and assessing past drought events for multi-hazard events monitoring, prediction and preparedness.

Limitations/Assumptions: The clustering was done over the specified region only. Each calendar year was clustered independetly.

Additional Outputs/information: The dataset access is currently restricted due to pending related publication.

Contact Information: Egor Prikaziuk (UT-ITC, Faculty of Geo-Information Science and Earth Observation, ITC, University of Twente, the Netherlands)
Date made available14 Apr 2025
PublisherZenodo

Cite this