ABSTRACT For the past two decades, satellite-derived activef fire data have been used in a multitude of operational applications and in a large and growing body of research on the role of fire within the Earth system. More recent work with satellite-based active fire data has been directed toward estimating what are in effect broad-scale fire spread rates that are in turn used as an important temporal parameter for the extraction of individual-fire boundaries from burned area maps. Here we use data mining to identify active fire clusters that serve as an input to a fire spread reconstruction algorithm to derive optimal global fire spread rates suitable for fire-perimeter extraction. The spread rates calculated for the active fire clusters, which are useful for applications beyond perimeter extraction, correlate with the spread rates based on reference fire boundaries (R 2 = .82, NRMSE = 2.6%) and are generally compatible with other studies, despite key differences in data acquisition methods and quantities measured.