DBSCAN 3D Clusters of SPEI-90 days Values – Italian NUTS3 (ITH31, 32, 34, 35, 36, 37), 1950–2023

Science Case Name 

Multi-Hazards in the Downstream Area of the Adige River Basin. 

Dataset Name/Title 

DBSCAN 3D Clusters of SPEI-90 days Values – Italian NUTS3 (ITH31, 32, 34, 35, 36, 37), 1950–2023

Dataset Description 

Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm output based on the daily Standardized Precipitation Evapotranspiration Index (SPEI) with a 90 days timescale for elevation lower than 1500 m a.s.l. (https://doi.org/10.5281/zenodo.13778103) applying the threshold SPEI-90 days  ≤ -1.

Key Methodologies 

The DBSCAN algorithm included in the scikit-learn package in Python environment (https://scikit-learn.org/stable/modules/generated/sklearn.cluster.DBSCAN.html) was used to detect the spatio-temporal clusters of droughts, taking the SPEI index as input. Three parameters guide the DBSCAN clustering procedure: the neighbourhood parameter (ε), which defines the search radius around a point (a SPEI index value); the spatio-temporal ratio (r), which controls the importance of spatial distance relative to temporal lag when computing the Euclidean distance between data points; the density threshold parameter (μ), representing the minimum number of neighbouring SPEI pixels required for a point to be considered as a core point (a point representing a suitable point to generate a new drought cluster).

The selected parameter values are: neighbourhood parameter (ε) = 30, spatio-temporal ratio (r) = 3 and density threshold (μ) = 10. These parameters were selected based on their physical significane and then refined through the comparison with the drought historical events retrieved from newspapers, official technical reports and regional council resolutions.

Temporal Domain 

1950–2023

Spatial Domain 

The spatial domain of the dataset is represented by the Italian Provinces identified by the NUTS3 codes ITH31, ITH32, ITH34, ITH35, ITH36, ITH37 considering elevation lower than 1500 m a.s.l..

Key Variables/Indicators 

Spatio-temporal clusters of dry/drought events

Data Format 

Comma Separated Values (CSV) 

Source Data 

E-OBS dataset (v29.0e), https://zenodo.org/records/13778103

Accessibility 

https://zenodo.org/records/13785996

Stakeholder Relevance 

Both the SPEI index and its adoption as an input in the DBSCAN algorithm used to identify drought spatio-temporal clusters represent a key step to identify the spatio-temporal footprints of hazard events. The identification of spatial patterns enables a greater understanding of hazard dynamics, it can be coupled with other hazard footprints (e.g., heatwaves) and fosters the use of EO data by providing a robust meteorological-based event identification to be then further refined by the use of higher spatial resolution EO data capable of capturing subtle spatial patterns (e.g., soil moisture as influenced by droughts, or land surface temperature as a response to different land uses during hot events).

Limitations/Assumptions 

Excludes areas above 1500 m a.s.l.

Additional Outputs/information

Daily SPEI-90 Days Values - Elevation Below 1500 m a.s.l. (https://zenodo.org/records/13778103)The dataset access is currently restricted due to pending related publication.

Contact Information 

Maraschini, Margherita (CMCC Foundation - Euro-Mediterranean Center on Climate Change) - Data managerMasina, Marinella (CMCC Foundation - Euro-Mediterranean Center on Climate Change) - Data curatorFurlanetto, Jacopo (CMCC Foundation - Euro-Mediterranean Center on Climate Change, National Biodiversity Future Center) - Data curatorFerrario, Davide Mauro (CMCC Foundation - Euro-Mediterranean Center on Climate Change) - Data curatorTorresan, Silvia (CMCC Foundation - Euro-Mediterranean Center on Climate Change, National Biodiversity Future Center) - Data manager