Can Single-Date Post-Fire Algorithms Replace Bi-Temporal ΔNBR? A Multi-Algorithm Assessment Across Mediterranean Fires

Bi-temporal change detection using the Differenced Normalized Burn Ratio (ΔNBR) remains the operational standard for satellite-based burned area mapping, yet cloud-free pre-fire imagery is frequently unavailable. While single-date methods bypass this temporal constraint, conventional index thresholding suffers from high false-positive rates and substantial area estimation errors, often exceeding a MdAPE of 63.7%. To address this limitation, we assess single-image approaches against EFFIS reference perimeters across 30 Euro-Mediterranean wildfire sites from 2021 to 2026, and introduce EMBER, an unsupervised framework designed to delineate burned perimeters from a single Sentinel-2 scene. The methodology expands surface reflectance bands into a 20-dimensional feature space using 3×3 local spatial means, applying principal component analysis and a Gaussian Mixture Model governed by a near-infrared maximum-gap rule and spatial connectivity constraints. Equivalence testing via Two One-Sided Tests (TOST) confirmed formal statistical equivalence between EMBER and bi-temporal ΔNBR within a δ=±0.10 tolerance bound (pTOST<0.01) across the pixel-level overlap metrics (F1-score: 83.9% vs. 85.6%; IoU: 72.3% vs. 74.9%; Precision: 94.7% vs. 94.6%; Recall: 78.5% vs. 80.5%). For boundary displacement (HD95=792.5m vs. 758.5m), no statistically significant difference was detected relative to ΔNBR (padj=0.393). Processing each scene in an average of 10.2s, EMBER does not aim to replace bi-temporal analysis, but demonstrates that local spatial-spectral clustering can serve as a reliable alternative when pre-fire baseline data is missing, bypassing data retrieval delays without the computational overhead of deep learning architectures.

Authors

Institutions

Publication Details

Journal
Journal of Imaging
Published
2026-11-16
DOI
https://doi.org/10.3390/jimaging12110580
Primary Topic
Fire effects on ecosystems
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Can Single-Date Post-Fire Algorithms Replace Bi-Temporal ΔNBR? A Multi-Algorithm Assessment Across Mediterranean Fires

Nazario Tartaglione, Alessandro Mercatini
Journal of Imaging
Fire effects on ecosystems
article

Can Single-Date Post-Fire Algorithms Replace Bi-Temporal ΔNBR? A Multi-Algorithm Assessment Across Mediterranean Fires

Nazario Tartaglione, Alessandro Mercatini
article en

Abstract

Bi-temporal change detection using the Differenced Normalized Burn Ratio (ΔNBR) remains the operational standard for satellite-based burned area mapping, yet cloud-free pre-fire imagery is frequently unavailable. While single-date methods bypass this temporal constraint, conventional index thresholding suffers from high false-positive rates and substantial area estimation errors, often exceeding a MdAPE of 63.7%. To address this limitation, we assess single-image approaches against EFFIS reference perimeters across 30 Euro-Mediterranean wildfire sites from 2021 to 2026, and introduce EMBER, an unsupervised framework designed to delineate burned perimeters from a single Sentinel-2 scene. The methodology expands surface reflectance bands into a 20-dimensional feature space using 3×3 local spatial means, applying principal component analysis and a Gaussian Mixture Model governed by a near-infrared maximum-gap rule and spatial connectivity constraints. Equivalence testing via Two One-Sided Tests (TOST) confirmed formal statistical equivalence between EMBER and bi-temporal ΔNBR within a δ=±0.10 tolerance bound (pTOST<0.01) across the pixel-level overlap metrics (F1-score: 83.9% vs. 85.6%; IoU: 72.3% vs. 74.9%; Precision: 94.7% vs. 94.6%; Recall: 78.5% vs. 80.5%). For boundary displacement (HD95=792.5m vs. 758.5m), no statistically significant difference was detected relative to ΔNBR (padj=0.393). Processing each scene in an average of 10.2s, EMBER does not aim to replace bi-temporal analysis, but demonstrates that local spatial-spectral clustering can serve as a reliable alternative when pre-fire baseline data is missing, bypassing data retrieval delays without the computational overhead of deep learning architectures.

Journal of ImagingVol. 12(11)
Istituto Superiore per la Protezione e la Ricerca Ambientale (IT)
Fire effects on ecosystems
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon