Odour source localization is crucial in life-saving scenarios like pinpointing gas leaks, detecting explosives, searching for earthquake survivors, or locating fires at their origin. The turbulent character of natural environments makes this task very challenging. The absolute concentration of odour plumes carries little meaning and these plumes are only encountered in an intermittent, transient fashion. However, navigation algorithms that are driven by odour encounter events, can successfully find odour sources by extracting spatiotemporal information. The event driven nature of odour plumes motivates a fully event-driven sensing and processing pipeline for robot navigation. Hence, we developed a spiking neural network, implemented on neuromorphic hardware, that can successfully decode odour-puff direction from a pair of enose-systems. This is to our knowledge the first fully event driven neuromorphic system for odour localization.