Table 1 Comparison of our flood detection method with existing datasets

From: Mapping global floods with 10 years of satellite radar data

Region

Additional Flood Area Identified

Overlap With Existing Flood Detections

  

GSW

MODIS

Global

+71%

35% (48%†)

28% (33%†)

Africa

+90%

44% (57%†)

33% (37%†)

Ethiopia

+194%

51% (78%†)

21% (44%†)

Semera

+96%

59% (68%†)

31% (41%†)

Dolo Ado

+1013%

59% (78%†)

72% (76%†)

  1. Our model successfully detects a substantial portion of flood-prone areas identified in existing datasets while revealing significant additional flood extent. The first column shows the percentage of additional flood-prone area identified by our model relative to a combination of GSW and MODIS flood detections. Our results identify considerably more flood extent than previous datasets, particularly in Ethiopia and its sub-regions. The next two columns show the percentage of flood areas from established datasets (GSW and MODIS) that our model detected. Numbers in parentheses (†) indicate detection rates when excluding areas where SAR detection may be unreliable (e.g., urban areas, steep terrain).For context, when comparing the existing datasets to each other, GSW captures 36% of MODIS-identified flood areas, while MODIS captures 23% of GSW-identified areas - similar to the overlap rates we observe with our model.
  2. GSW Global Surface Water dataset, MODIS Moderate Resolution Imaging Spectroradiometer dataset.
Back to article page