Microphysical Impacts to Landfalling Hurricanes
Simulating hurricanes in a regional numerical weather model typically means using grid sizes on the order of 1 kilometer. Getting accurate cloud behavior, then, can be very challenging, especially because particles like ice, snow, hail (called graupel when it is beginning to form), and rain droplets, which make up clouds, are less than 1 millimeter in size. It would take more computer power than is favorable for a computer model to resolve those itty, bitty particles while also resolving phenomenal storms, like hurricanes.
In order to conserve computing energy, we must use what are called “microphysical schemes”. This means that instead of literally simulating every single particle inside our models, then following each one as they evolve in time, we instead represent them mathematically by packing up (or “bulking”) similar particles with each other. However, this method also has limitations, as the scientists who develop mathematical microphysical schemes must make certain assumptions about the average particle’s size, density, and distribution. Those assumptions certainly make our computers run faster, but they also add uncertainties.
For the type of research that I do currently–improving computer models of hurricanes–these uncertainties mean that a modeled hurricane will “rain” in different geographical locations. If I choose a different microphysical scheme for my hurricane, then my model will predict differing total flooding amounts in different towns. And, the only thing that was changed in the model was the cloud particles themselves!
I investigate the dependencies of flooding on ice particles in clouds by focusing my current research on the case study of Hurricane Florence (2018), which had a long-lasting period offshore of North Carolina before making landfall on 14 September. Not eager to leave the Carolinas, Florence caused large flooding, with a record of almost 36 inches in Elizabethtown. In the Figure below, we can see the simulation as Florence is making landfall in the Carolinas. Specifically, the figure is a 2-hour time composite of reflectivity, taken from the 3-km altitude in the model.
To date, my Hurricane Florence simulations have been tested with three different microphysical schemes and produce a range of total simulated rainfall between 40 and 50 inches. Of the three schemes, one distributes particles throughout the simulated hurricane’s outer rainbands, while the other two retain more precipitation within their stormiest inner cores. The two figures below show the differences in total rainfall totals for two of the microphysical schemes (Morrison, top and Thompson, bottom). Not only does one (Morrison, top) produce more rainfall in a broader region, but it also predicts the highest flooding (dark green) in a larger area near the North Carolina / South Carolina border (south of the region where the highest flooding was recorded near Elizabethtown).
I can evaluate the contribution of different microphysical particles in a few different ways. The figure below shows two methods I use to evaluate the Morrison Microphysics scheme. In the first (left hand side), I have vertical profiles of each type of particle, taken from the same landfall period as the reflectivity figure above. In the second (right hand side), I have a CFAD (contour frequency by altitude diagram), also during Florence’s simulated landfall. The vertical profiles show generous contributions from snow and ice (blue), but also strong contributions from graupel (yellow) and rain (green). The CFAD confirms the generous contribution from snow and ice–the bulge to the right near 6 kilometers in height, with higher reflectivities above 50 dBZ, is what is called the “bright band”. This is near the freezing level, where snow melts and turns into larger rain droplets than would form from ice pellets. Below the freezing level, the range of reflectivity below about 45 dBZ are reflective of light (stratiform) rainfall.
This work is ongoing, so stay tuned for updates!