Floods in ideal landscapes follow a coherent pattern where single water-covered areas expand and afterwards recede following the inverse sequence but deviate in real landscapes, due to natural or human factors, resulting in flood coverage shifts. Using remote sensing, we introduced two indices to describe the discrepancies between spatially integrated vs. pixel-level frequency distributions under maximum flooded conditions (dext) and throughout all flooding conditions (dtot), expressed as the relative weight of shifts on each landscape’s maximum registered coverage, theoretically ranging between no displacement (<20%) to maximum displacement (< < inf). Globally, over 36 years floods 26 exhibited redistributions representing, on average, 25% and 45% of their peak extents 27 revealing previously unnoticed extra flooded areas and rotational movements within flood28 ing events, rising up to 500% in meandering rivers (South America) and irrigated crop29 lands (Central Asia). We also assessed the influence of natural and human variables and 30 discussed the indices’ potential for advancing flood research.