Does Increased Horizontal Resolution Improve Modeled Precipitation in Hurricane Simulations?

Ryder Fox, University of Miami

Photo of Ryder Fox

At $24 billion in damages, slow-moving Hurricane Florence (2018) became the ninth-most destructive hurricane to impact the United States. Persisting post-landfall, Florence flooded North Carolina with more than 35 inches of record-breaking rain. Numerical weather modeling simulations of three-dimensional severe weather have improved substantially since the 1990s. Despite such advancements, computational resources have been insufficient for optimal modeling of precipitation particles (clouds and raindrops, ice, snow, and hail), which are orders of magnitude smaller than the grid sizes they occupy. This means that many parameterization schemes have been developed over the decades to act as best estimates of precipitation evolution in extreme events. Currently, hurricane models average roughly 3 km of horizontal grid spacing to resolve severe weather scales from 10 to 1,000 km; yet studies have shown that grid spacing on the order of 0.1 to 0.25 km is optimal for representing the turbulent scale mixing of cloud particles. This study is the first of its kind to investigate a large domain (300 km), ultra-fine resolution (0.2 km) nest inside a hurricane model. Specifically, various simulations have embedded one, two, or three nests of increasing resolution in order to test model sensitivity to horizontal grid resolution. Additional simulations are carried out using three different microphysics parameterization schemes to test model sensitivity to differing representations of cloud particles. Questions addressed herein include examining how and where flooding is generated in a hurricane model simulated within ultra-fine resolution. Is rainfall overestimated in potentially overactive cloud systems? Do simulations tend to favor rain or ice precipitation pathways, and how does that compare to observations? By addressing such questions, we hope to offer recommendations on improvements that can be made to microphysics schemes as computational modeling improves to the point of commonly simulating hurricanes within sub-kilometer grid spacing.

Abstract Author(s): K. Ryder Fox, Brian McNoldy, David S. Nolan