Examining Uncertainty in Tornado Intensity Estimates Using Observing System Tests with High-Resolution Simulations

This presentation describes observation experiments based on high-resolution (5 m horizontal and 2.5 m vertical grid spacing) tornado simulations. Each simulation uses the Bryan et al. "eddy injection" technique (see abstract in this conference) to ensure fully developed turbulence in the boundary layer. The simulated vortex is translated across groups of stationary mock "sensors" at 10 m AGL. We repeatedly sample the resulting wind swaths for a range of tornado sizes and intensities, both randomly and systematically; also, some experiments use “perfect” sensors while others limit the maximum measurable wind speed to mimic the characteristics of damage indicators (DIs) specified in the Enhanced Fujita scale. Statistical analysis of the results provides quantitative insight into the limitations of tornado intensity estimates in rural/semi-rural areas due to sparse sensor/DI coverage, structural limitations of the DIs, and spatial and temporal variations in tornadic intensity. Mean errors for these experiments exceed 10 m/s in all cases and approach 40 m/s in some cases.