Modern radar systems equipped with agile-beam technology support multiple modes of operation, including, for
example, tracking, automated target recognition (ATR), and synthetic aperture radar imaging (SAR). In a multimode
operating environment, the services compete for radar resources and leave gaps in the coherent collection
aperture devoted to SAR imaging. Such gapped collections, referred to as interrupted SAR, typically result in
significant image distortion and can substantially degrade subsequent exploitation tasks, such as change detection.
In this work we present a new form of exploitation that jointly performs imaging and coherent change detection
in interrupted environments. We adopt a Bayesian approach that inherently accommodates different interrupt
patterns and compensates for missing data via exploitation of 1) a partially coherent model for reference-pass to
mission-pass pixel transitions, and 2) the a priori notion that changes between passes are generally sparse and
spatially clustered. We employ approximate message passing for computationally efficient Bayesian inference
and demonstrate performance on measured and synthetic SAR data. The results demonstrate near optimal
(ungapped) performance with pulse loss rates up to ∼ 50% and highlight orders of magnitude reduction in false
alarm rates compared to traditional methods.