An Object-Based Approach to Atmospheric Predictability

Understanding the predictability of the atmosphere is crucial for determining tractable forecast problems. However, the predictability of weather systems like cyclones and convective storms is not fully understood, in part because conventional predictability methods based on error growth in grid-point or spectral space have limitations when dealing with coherent features like weather systems. To address this gap, we introduce an object-based approach to atmospheric predictability. We leverage a global convection-permitting identical twin predictability experiment consisting of a control and perturbed simulation and use the MODE tool, originally developed for verifying high-resolution model forecasts, to quantify the predictability of synoptic-scale and mesoscale weather systems. The MODE tool ''matches'' weather systems in the control and perturbed simulations, and by analyzing how the number of matched objects decreases over time, we can make statements about the predictability of these systems. Our study suggests that mesoscale weather systems have longer predictability than estimated through conventional predictability methods, potentially because mesoscale weather systems such as convective storms are often forced by the more predictable large-scale flow. By introducing a novel object-based approach to quantifying the predictability of weather systems, we provide insights into the behavior of mesoscale and synoptic-scale weather phenomena, which can inform the development of improved weather forecasting strategies.