GeoAI Arctic Challenge Dataset

The GeoAI Arctic Challenge dataset is an instance segmentation benchmark for detecting and delineating retrogressive thaw slumps (RTS), landscape disturbances caused by permafrost thaw, in Arctic image chips. The dataset builds on Yang et al. (2023), which provided semantic segmentation masks that label each pixel as RTS or non-RTS. For this challenge, those labels have been extended and reformatted so each RTS feature is represented as an individual instance.

Participants receive multimodal image data and train models to predict one mask for each RTS instance in the hidden test set. This instance-level formulation supports evaluation at the feature level, including how well models separate and delineate individual RTS boundaries, rather than only measuring pixel-wise foreground and background accuracy.

Training labels are provided in COCO instance segmentation format. Test labels remain hidden and are used by the official scorer.

Why it matters: RTS are sensitive indicators of permafrost thaw, which releases greenhouse gases and alters Arctic landscapes. By leveraging AI, we aim to accelerate RTS detection and improve understanding of climate-driven change.

Satellite Image (RGB)	Semantic Mask (Original)	Instance Mask (This Challenge)

Figure 1. Conversion from semantic RTS labels into instance-level masks. The challenge dataset uses connected-component instance labels so models can be evaluated at the feature level rather than only pixel-wise.

Geographic Coverage & Study Sites

The source data spans 7 Arctic subregions, including:

Canada: Herschel Island, Horton Delta, Tuktoyaktuk peninsulas, Banks Island
Russia: Yamal and Gydan peninsulas, Lena River, Kolguev Island

Figure 2. Spatial coverage of the source Arctic RTS dataset. The competition release removes geospatial metadata from distributed image chips while preserving multimodal image information for modeling (Li et al., 2025).

Public Release Contents

The public package contains training images and labels, hidden-label test images, metadata, and starter tools:

competition_release/
  README.md
  metadata/
    band_names.json
    sample_submission.json
    train_manifest.csv
    test_manifest.csv
  tools/
    coco_utils.py
    validate_submission.py
    evaluate_coco.py
    inspect_dataset.py
  examples/
    load_image_and_label.py
    make_sample_submission.py
    encode_predictions.py
  train/
    images/*.npz
    annotations/instances_train.json
  test/
    images/*.npz

Each .npz file contains one array named image with shape H x W x 8 in HWC order.

Bands

Each image chip contains eight co-registered channels. The bands combine optical imagery, spectral features, and topographic context so models can learn both visual RTS appearance and environmental cues that affect slump boundaries.

Data Layer	Source / Feature Type	Bands	Role in RTS Mapping
RGB imagery	Maxar optical imagery	`red`, `green`, `blue`	Provides high-resolution visual context for identifying exposed soil, vegetation disturbance, and visible RTS morphology.
Spectral features	Vegetation, water, and near-infrared features	`ndvi`, `ndwi`, `nir`	Helps models distinguish thaw-related disturbance from vegetation, water, snow, and other spectrally distinct surface conditions.
Terrain features	ArcticDEM-derived topographic features	`relative_elevation`, `shaded_relief`	Adds terrain structure that can improve boundary delineation and help separate RTS features from surrounding slopes.

The array band order is:

Index	Band Name	Description
0	`red`	Maxar red
1	`green`	Maxar green
2	`blue`	Maxar blue
3	`ndvi`	Normalized Difference Vegetation Index
4	`relative_elevation`	Relative elevation
5	`shaded_relief`	Shaded relief
6	`nir`	Planet near-infrared
7	`ndwi`	Normalized Difference Water Index

The same band list is available in machine-readable form in metadata/band_names.json.

Key Dataset Statistics

Property	Description
Training Images	756 image chips with public labels
Test Images	138 image chips without public labels
Training RTS Instances	1,783
Hidden Test RTS Instances	299
Total RTS Instances	2,082 train + hidden-test labels
Image Array Format	`.npz` files containing `image` arrays with shape `H x W x 8`
Annotation Format	COCO instance segmentation JSON with compressed RLE masks
Task	RTS instance segmentation
Category	`{"id": 1, "name": "rts", "supercategory": "landform"}`
Original Source	*Yang et al.* (2023)*, Remote Sensing of Environment*

Label Conversion

Training annotations are stored in train/annotations/instances_train.json using COCO instance segmentation format. Instance masks are encoded as compressed COCO run-length encoding (RLE).

The label conversion rule is: