Datasets:

Modalities:
Image
Languages:
English
DOI:
Libraries:
Datasets
License:
hls_burn_scars / README.md
CEPhillips's picture
Create README.md
a8b42ec
|
raw
history blame
1.74 kB
---
size_categories:
- n<1K
---
# Dataset Card for HLS Irrigation Scenes
## Dataset Description
- **Homepage: https://huggingface.co/datasets/nasa-impact/hls_irrigation_scenes**
- **Point of Contact: Dr. Christopher Phillips ([email protected])**
### Dataset Summary
This dataset contains Harmonized Landsat-Sentinel imagery of burn scars and the associated masks for the years 2018-2021 over the contiguous United States. There are 512x512 scenes. It's primary purpose is for training geospatial machine learning models.
## Dataset Structure
## TIFF Metadata
Each tiff file contains a 512x512 pixel tiff file. Scenes have 6 bands, and masks have 1 band. For satellite scenes, each band has already been converted to reflectance.
## Band Order
For scenes:
Channel, Name, HLS S30 Band number
1, Blue, B02
2, Green, B03
3, Red, B04
4, NIR, B8A
5, SW 1, B11
6, SW 2, B12
Masks are a single band with values:
1 = Burn scar
0 = Not-burned
-1 = Missing data
## Class Distribution
Burn Scar - 11%
Not-burned - 88%
No Data - 1%
## Data Splits
The 805 files have been randomly split into training (2/3) and validation (1/3) directories, each containing the masks, scenes, and index files.
## Dataset Creation
The data were from shapefiles maintained by the Monitoring Trends in Burn Severity (MTBS) group. The original data may be found at:
https://mtbs.gov/
After co-locating the shapefile and HLS scene, the 512x512 chip was formed by taking a window with the burn scar in the center. Burn scars near the edges of HLS tiles are offset from center.
Images were manually filtered for cloud cover and missing data to provide as clean a scene as possible, and burn scar presence was also manually verified.