Abstract: Global floods, exacerbated by climate change, pose severe threats to human life,infrastructure, and the environment. Recent catastrophic events in Pakistan and NewZealand underscore the urgent need for precise flood mapping to guide restorationefforts, understand vulnerabilities, and prepare for future occurrences. WhileSynthetic Aperture Radar (SAR) remote sensing offers day-and-night, all-weatherimaging capabilities, its application in deep learning for flood segmentation islimited by the lack of large annotated datasets. To address this, we introduceKuro Siwo, a manually annotated multi-temporal dataset, spanning 43 flood eventsglobally. Our dataset maps more than 338 billion $m^2$ of land, with 33 billiondesignated as either flooded areas or permanent water bodies. Kuro Siwo includesa highly processed product optimized for flood mapping based on SAR GroundRange Detected, and a primal SAR Single Look Complex product with minimalpreprocessing, designed to promote research on the exploitation of both the phaseand amplitude information and to offer maximum flexibility for downstream taskpreprocessing. To leverage advances in large scale self-supervised pretrainingmethods for remote sensing data, we augment Kuro Siwo with a large unlabeled setof SAR samples. Finally, we provide an extensive benchmark, namely BlackBench,offering strong baselines for a diverse set of flood events from Europe, America,Africa, Asia and Australia.