Niloofar Azad

November 6th, 2026 – 12:00-1:00 pm, EC4-2101A

Accurate environmental monitoring often requires information from multiple sources, making multi-sensor fusion a promising approach to improving our understanding of the Earth’s surface. By combining observations from different sensors, we expect to capture complementary information and improve prediction accuracy. But does combining more sensors necessarily lead to better predictions? What limits the effectiveness of multi-sensor fusion? Differences in sensor characteristics, redundant information, and the choice of fusion strategy can all influence how effectively observations are combined. This seminar explores the challenges of multi-sensor fusion in the context of Arctic sea-ice mapping, using Sentinel-1 and the RADARSAT Constellation Mission (RCM) as a case study. Through a newly developed cross-sensor SAR dataset and an investigation of different fusion strategies, the seminar examines the complexities of combining observations for predicting sea-ice parameters. It also explores emerging approaches for understanding sensor complementarity, identifying shared and unique information across sensors, and selectively integrating the most relevant features.