Burned-Area Map Accuracy Metrics for Indonesian Peatlands

1,042 validation points assess the accuracy of three burned-area maps across seven Indonesian provinces covering 87.60 million hectares. The dataset reports overall, user, and producer accuracy metrics with 95% confidence intervals. David L. A. Gaveau published this assessment on figshare in April 2026.

Use Cases

• Compare the overall accuracy (OA) of three different burned-area mapping methods for the same region.
• Analyze user accuracy (UA) and producer accuracy (PA) to understand commission and omission errors in fire detection.
• Use the 95% confidence intervals for each accuracy metric to statistically compare map performance.
• Validate new burned-area detection algorithms against the reported baseline metrics for Indonesian peatlands.

Strengths

• Validation based on 1,042 ground reference points from stratified random sampling.
• Reports three standard accuracy metrics (OA, UA, PA) with 95% confidence intervals for each of three maps.
• Covers a large geographic area of 87.60 million hectares targeted for peatland restoration.

Limitations

• Small dataset size of 5.5 KB indicates limited scope, likely a summary table rather than raw point data.
• Unknown number of rows and specific column structure limits detailed analysis.
• Temporal coverage of the underlying maps and validation data is not specified.

Provenance

Source

David L. A. Gaveau via figshare.

Collection Method

Accuracy metrics estimated from 1,042 points distributed using stratified random sampling.

Freshness

Last updated in April 2026, indicating recent assessment.

Geography

Seven Indonesian provinces (87.60 Mha) targeted for peatland restoration.