Limits to the Practical Predictability of Convective‐Scale Forecast Systems With World‐Class Data Assimilation

Abstract The uncertainty of the analysis and the error growth of the forecast from two of the best prediction systems based on radar data assimilation are studied to determine how close we are from accurately forecasting thunderstorms and what additional information would be required to achieve this goal. Even if these two state‐of‐the‐art systems frequently assimilate all available measurements for one, or assimilate radar data every 30 s for the other, considerable uncertainty remains in unobserved atmospheric properties such as temperature or humidity. Worse, limited improvement can be expected as limited direct measurements exist, and essentially all the covariance‐based information is being exploited. Unless a massive investment in new measurement technologies occurs, or radically new ideas are proposed, forecast accuracy will remain insufficient to properly simulate the complete lifetime of a thunderstorm, limiting the accuracy and value of threat forecasting.