A New Lens for Examining Storms in Climate Models
In the United States, mesoscale convective systems (MCSs)—assemblies of convective storms on scales of 100 km or larger—are responsible for most warm-season extreme rainfall events. Systematically assessing representations of MCS and non-MCS precipitation characteristics in current climate models has proven challenging because of their low resolutions. Researchers developed a process-oriented approach that explicitly tracks the clouds and precipitation simulated by next-generation, high-resolution climate models to evaluate MCS occurrence frequency and intensity, as well as the large-scale weather patterns that favor MCS formation. This work provides a new methodology for identifying the sources of model biases in simulating MCS precipitation that can guide future model development to improve MCS representations.
Next-generation, high-resolution global climate models may provide improved simulations of MCSs compared to currently used models. Determining model skill at simulating MCSs, beyond inferences based on precipitation characteristics, requires explicit tracking of MCSs via real-world observations and in simulations. This newly developed methodology better equips scientists for benchmarking MCS simulations and relating the model’s skill to the representation of underlying fundamental processes. The latter is critically important for establishing confidence that new models can provide reliable projections of future changes in regional hydrological cycles. The diagnostic approach developed in this work represents an important framework for evaluating the mesoscale processes next-generation climate models can resolve.
Researchers developed an observation-driven methodology to track MCSs at three resolutions (50 km, 25 km, 12 km) likely important to next-generation climate models. The new tracking algorithm can consistently track MCSs across these resolutions to reproduce MCS characteristics from a reference MCS observation database (4 km resolution). To better evaluate large-scale weather patterns (LSWPs) that represent multi-day regional weather conditions and are favorable for simulated MCS formation, researchers developed a diagnostic framework to identify different types of LSWPs associated with observed MCSs.
Researchers applied the tracking algorithm and LSWP diagnostic framework to two variable-resolution global climate simulations that possess regional refinement at 50 km and 25 km resolutions over the United States. Researchers found that the models consistently underestimated the number of warm-season simulated MCSs and the associated precipitation amount and frequency in the central United States compared to observations. Additionally, the simulated MCS precipitation lasts too long, is over too large of an area, and is too low in intensity. While the model can simulate different types of observed LSWPs, their frequencies of occurrence are too low because of identifiable biases in the model. Precipitation simulated under LSWPs also evolves differently compared to observations, indicating further biases in the modeled responses to LSWPs. The diagnostic framework and understanding of model shortfalls gained from this study are currently guiding the development of new climate modeling to improve future MCS simulation.