Power law frequency-size distributions of forest fires have been observed in a range of environments. The scaling behaviour of fires, and more generally of landscape patterns related to recurring disturbance and recovery, have previously been explained in the frameworks of self-organized criticality (SOC) and highly optimized tolerance (HOT). In these frameworks the scaling behaviour of the fires is the global structure that either emerges spontaneously from locally operating processes (SOC) or is the product of a tuning process aimed at optimizing the trade-offs between system yield and tolerance to risks (HOT). Here, we argue that the dominant role of self-organized or optimised fuel patterns in constraining unplanned-fire sizes, implicit in the SOC and HOT frameworks, fails to recognise the strong exogenous controls of fire spread (i.e. by weather, terrain, and suppression) observed in many fire-prone landscapes. Using data from southern Australia we demonstrate that forest fire areas and the magnitudes of corresponding weather events have distributions with closely matching scaling exponents. We conclude that the spatial scale invariance of forest fires may also be a mapping of the meteorological forcing pattern.