HLS Burn Scars: Satellite Imagery of U.S. Wildfires, 2018-2021

Harmonized Landsat and Sentinel-2 satellite imagery of burn scars across the contiguous United States from 2018 to 2021. The collection contains 804 scenes, each 512x512 pixels, with associated segmentation masks. It was created by ibm-nasa-geospatial and uploaded to Hugging Face in September 2023 for training geospatial machine learning models.

Use Cases

• Train semantic segmentation models to identify burn scars based on the provided imagery and masks.
• Develop change detection algorithms to assess wildfire damage over the 2018-2021 period.
• Benchmark geospatial AI models on a standardized dataset of harmonized Landsat and Sentinel-2 imagery.

Strengths

• Contains 804 distinct image scenes for model training.
• Covers a four-year time range (2018-2021) across the contiguous United States.
• Provides pre-processed, harmonized imagery from two major satellite sources (Landsat and Sentinel-2).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the associated masks is unknown, which may limit suitability assessment.
• Last updated 2023-09-26; freshness should be verified for applications requiring recent data.

Provenance

Source

ibm-nasa-geospatial

Collection Method

Likely compiled from Harmonized Landsat Sentinel-2 (HLS) satellite data products.

Time Range

2018-2021

Freshness

Last updated 2023-09-26.

Geography

Contiguous United States