Lakes are a ubiquitous landscape feature in northern permafrost regions. They have a strong impact on carbon, energy and water fluxes and can be quite responsive to climate change. The monitoring of lake change in northern high latitudes, at a sufficiently accurate spatial and temporal resolution, is crucial for understanding the underlying processes driving lake change. To date, lake change studies in permafrost regions were based on a variety of different sources, image acquisition periods and single snapshots, and localized analysis, which hinders the comparison of different regions. Here, we present a methodology based on machine-learning based classification of robust trends of multi-spectral indices of Landsat data (TM, ETM+, OLI) and object-based lake detection, to analyze and compare the individual, local and regional lake dynamics of four different study sites (Alaska North Slope, Western Alaska, Central Yakutia, Kolyma Lowland) in the northern permafrost zone from 1999 to 2014. Regional patterns of lake area change on the Alaska North Slope (â0.69%), Western Alaska (â2.82%), and Kolyma Lowland (â0.51%) largely include increases due to thermokarst lake expansion, but more dominant lake area losses due to catastrophic lake drainage events. In contrast, Central Yakutia showed a remarkable increase in lake area of 48.48%, likely resulting from warmer and wetter climate conditions over the latter half of the study period. Within all study regions, variability in lake dynamics was associated with differences in permafrost characteristics, landscape position (i.e., upland vs. lowland), and surface geology. With the global availability of Landsat data and a consistent methodology for processing the input data derived from robust trends of multi-spectral indices, we demonstrate a transferability, scalability and consistency of lake change analysis within the northern permafrost region.