We describe a new global multiyear satellite fire product designed to meet the needs of the global modeling community. We use the new data set to analyze the global distribution of biomass burning using five different temporal metrics derived from 5 years of high-quality satellite data acquired with the Moderate Resolution Imaging Spectroradiometer (MODIS), on board NASA's Terra satellite. The global distributions of fire pixel density, peak month, season length, and annual periodicity are described. As part of our analysis we show, for the first time, the global distribution of the fire radiative power (FRP), a relatively new remotely sensed quantity. We find that low FRP tends to be associated with areas of cropland burning. In the tropics and much of the subtropics, low FRP is also associated with more heavily forested areas, while higher FRP tends to occur in areas of grassland burning. In boreal forests this trend is reversed, with higher FRP occurring in areas of greater tree cover. We next combine 3 years of Terra and Aqua MODIS observations to show that a strong diurnal fire cycle is prevalent at tropical and subtropical latitudes. We also consider the consistency of the fire time series recorded by the two MODIS instruments, and find the month of peak burning and fire season length observed by each to be in good agreement in most areas. However, significant discrepancies with respect to seasonality do occur in some relatively small areas, and are most pronounced in tropical rain forest.