HDBSCAN Clusters Rice Crop Stress - Senegal River Delta, 2019-2020

Demonstration Case Name 

Multi-hazards in Senegal (transferability case)

Dataset Name/Title 

HDBSCAN Clusters Rice Crop Stress - Senegal River Delta, 2019-2020

Dataset Description 

The dataset contains HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise) clusters based on a synthetic stress indicator obtained through a PCA (Principal Component Analysis) on NDVI (Normalized Difference Vegetation Index) and NDMI (Normalized Difference Moisture Index) Sentinel-2 based indices. The dataset contains identified vegetation stress clusters on the following dates: wet year: '2019-06-02', '2019-06-07', '2019-06-17', '2019-06-22', '2019-06-27', dry year: '2020-05-27', '2020-06-11', '2020-06-21', '2020-06-26',  '2020-07-01' , covering the end portion of the rice cropping season in the area. All observations are located in a plain area that relies on the SenegalRiver for cropland irrigation. For each row, the dataset contains the following columns: 

date: date of the clustering, in YYYY-mm-dd format

cluster: cluster number, -1 identifies the noise (unclassified) cluster

ndvi_mean: mean NDVI across field polygon

ndmi_mean: mean NDMI across field polygon

bsi_mean: mean BSI across field polygon

distance_t: average distance (m) of a rice field from the closest irrigation channel

geometry: polygon geometry of the cluster in EPSG:32628

Key Methodologies 

Sentinel-2 Level-2 NDVI, NDMI and BSI were computed, averaged per field and exported for 2019-2020. The rice growing period is April to July; the end of season is May-June.  Bare fields were filtered out by BSI >= 0.08. For the remaining fields, NDVI and NDMI  were combined in a principal component (PC) that underwent HDBSCAN clustering. Clusters with average NDVI < 0.45 and NDMI < 0.15 were considered stressed crops.

Temporal Domain 

5 days in 2019, 5 days in 2020

Spatial Domain 

Senegal River Delta, West Africa, EPSG:32628 (UTM Zone 28N)

Key Variables/Indicators 

HDBSCAN clusters

Data Format 

ESRI Shapefile

Source Data 

Sentinel-2 Level-2

Accessibility 

Zenodo, https://doi.org/10.5281/zenodo.17116834

Stakeholder Relevance 

The dataset provides valuable post-disaster information on crop vegetation dynamics during hot and dry events.

Limitations/Assumptions 

The HDBSCAN algorithm was not optimized for the dataset; a fixed version of hyperparameters was directly taken from https://doi.org/10.5281/zenodo.15301314 

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)