Dynamical and Physical Meteorology

Hurricane Boundary Layer Turbulent Flow

Present-day supercomputers allow for high-resolution, turbulence-resolving simulations of hurricanes with grid-spacing less than 100 m. To evaluate these simulations rigorously, high-frequency observations in the lowest kilometer of the hurricane boundary layer are needed. To accomplish this goal, MMM scientists have teamed up with NOAA’s Hurricane Research Division and the University of Miami to collect measurements with an Unmanned Aerial System (UAS). Six flights from UASs deployed in Hurricane Maria in 2017 sampled winds at up to 10 Hz frequency at altitudes as low as 100 m above sea level. Analysis of turbulence metrics like wind speed variance and turbulence intensity (see figure) have been compared to numerical model output and show encouraging results. Preliminary results suggest that observational turbulence is slightly more intense, by nearly a factor of 2 on average. Further work is underway to determine whether the simulated turbulent properties are biased low, or whether the experimental UAS measurements might have a positive bias.

Figure: Observations and numerical simulations of turbulent flow in the boundary layer of hurricanes
Figure: Observations and numerical simulations of turbulent flow in the boundary layer of hurricanes. (a) An instantaneous snapshot of wind speed (m s-1) at 10 m above sea level from a CM1 simulation using 32 m horizontal grid spacing. (b) A picture of the Raytheon/NOAA Coyote Unmanned Aerial System (UAS) (wingspan approximately 1.5 m). (c) Time series of wind speed (m s-1) from a Coyote UAS flight in Hurricane Maria at 400 m above sea level. (d) Analysis of Turbulence Intensity (T.I., defined as the standard deviation of wind speed divided by average wind speed) from Coyote UAS observations (colored dots, where each color is a different flight) and from numerical simulations (solid lines, where different colors denote different simulations).

Hail and Hailstorms: Now and in a Changing Climate

During July 2018, MMM scientist Andy Heymsfield and Ian Giammanco of the non-profit organization Insurance Institute for Business & Home Safety conducted a three-day hail workshop. This workshop brought together almost 200 people from the academic (measurements, modeling), building (roofing materials and repair), insurance (property damage), and litigation (real vs false damage reports) communities from the U.S. and abroad to discuss common interests in the area of hail and hailstorms. This workshop was funded by a grant from the National Science Foundation and by sponsors from the business community. The workshop was organized in recognition of the considerable monetary damage that results from hailstorms (see figure).

Figure: Estimated Hail Damage
Figure: Estimated Hail Damage. Data courtesy of Steve Bowen, Aon Benfield Inc., Chicago IL.

A few of the key points from the workshop:

  • Research on hailstorms and hail has languished since the mid-1980’s
  • From 1955-2017, the annual number of reported hail days and the number of hail observations with >2 inch hail has increased steadily
  • Sophisticated catastrophe models for hail damage that use a stochastic catalog containing information on location, footprint of storm, type of home and roofing product or automobile, and other vulnerabilities have been developed
  • Companies in the insurance industry have developed and deployed hundreds of weather stations that continuously monitor for hail occurrence
  • In a changing climate, counterintuitively hailfall in general will decrease, with the exception of storms that produce very large hail

Continental Convection in a Changing Climate

Accurate prediction of climate change impacts on clouds and precipitation, moist convection in particular, is difficult because climate models represent these small-scale weather phenomena in a very crude way. We report results of a modeling study that considers effects of climate change on unorganized (“popcorn”) deep convection developing during daytime over summertime continents. The specific modeling case was developed about a decade ago based on observations over the Amazon region in the 1990s (Grabowski et al. 2006). We contrast development and evolution of moist convection in the current climate using the morning sounding from Grabowski et al. 2006 (simulation “CoNTRol” or CNTR) with convection developing in mean conditions predicted by an ensemble of climate models to be representative for the end of the 21st century; simulation “Climate CHange or CLCH. The difference in the initial sounding between CNTR and CLCH is the increase of the temperature (by 4 to 5 K in the lower and middle troposphere and even higher in the upper troposphere) and small (5-10%) reduction of the relative humidity (RH) across the troposphere. The goal is to separate dynamical impacts (e.g., stronger convective updrafts) from effects of thermodynamics, that is, changes in cloudiness and surface precipitation resulting from the increase of water vapor that the warmer atmosphere can hold and convection can work with. We also consider a simulation with similar temperature increases as between CNTR and CLCH, but with an unchanged relative humidity. This simulation features a uniform 4 K increase of the temperature and is referred to as “Control Plus 4K”, or CP4K. The key result is a significant increase of the convective available potential energy (CAPE) in the future and thus significant increase of convective updraft strength. This is illustrated in the figure below. Stronger updrafts lead to increases of the upper-tropospheric cloudiness and the vertical extension of upper-tropospheric anvil clouds. Surprisingly, surface precipitation changes little, and the sign of the change depends on the RH profile: the mean surface rain accumulation slightly increases in the CP4K and it slightly decreases in CLCH.

Grabowski, W. W., and Coauthors, 2006: Daytime convective development over land: A model intercomparison based on LBA observations. Quart. J. Roy. Meteor. Soc., 132, 317–344

Figure:  The left panel shows the spatial variation of vertical temperature flux in the MABL across a sea surface temperature jump of 2 K
Figure: Histograms of the updraft statistics in CNTR (red), CLCH (blue) and CP4K (green) at height of 9/3 km (upper/lower panels) for the deep convection phase in hours 6 and 7 of the simulations. Histogram bin is 2 m/s. Bins between -2 and 2 m/s not shown.

Impact of SST on MABL

Sharp spatial jumps in sea surface temperature (SST) of varying strength and size are ubiquitous features across the oceans. How these temperature fronts and filaments impact the surface fluxes of momentum and scalars and couple with the overlying winds is not well understood. We recently developed a large-eddy simulation (LES) model that allows us to examine the impact of (non-periodic) SST heterogeneity on the marine atmospheric boundary layer (MABL). We find that the response of the MABL depends on the sign (positive, negative) of the SST jump and the magnitude and direction of the surface winds relative to the front, i.e., cross-front and down-front winds generate a different response. An example is depicted in the figure below where we show the vertical temperature flux for winds blowing across a positive SST jump of 2K. First notice the extended distance downstream, nearly 20 kilometers, before the boundary layer reaches an equilibrium state. Surprisingly, because of horizontal advection the scalar flux reaches a maximum in the middle of the transition region. In other words, the approach to the equilibrium state downstream of the SST jump is not monotonic. The work sponsored by ONR is a collaborative effort between NCAR, University of Notre Dame, and UCLA.

Figure:  The left panel shows the spatial variation of vertical temperature flux in the MABL across a sea surface temperature jump of 2 K
Figure: The left panel shows the spatial (x, z) variation of vertical temperature flux in the MABL across a sea surface temperature jump of 2 K: the x location of the jump is indicated by the white line. The large scale geostrophic winds are 10 m s-1 and moving from left to right. The right panel shows individual vertical profiles of scalar flux at selected x-locations. The lines are color coded to match the x intervals in the left panel. Notice the maximum scalar flux occurs in the middle of the transition region.

Wildfire Forecasting Behavior and Prediction

In FY2018 Janice Coen worked with Wilfrid Schroeder (NOAA/NESDIS, formerly University of Maryland) and Brad Quayle (USDA Forest Service) to investigate the weather and fire behavior that shaped past wildfire events and to develop and apply forecasting methodologies to predict how wildfires will grow and behave by integrating satellite active fire detection data and the CAWFE® coupled numerical weather prediction - wildland fire behavior model.

Coen et al. applied CAWFE to investigate the flow regime and underlying mechanisms associated with the Tubbs Fire, which ignited during the North Bay firestorm on October 8-9, 2017, travelled over 19 km in 3.25 h, and burned into the city of Santa Rosa, CA. Figure (a) shows a snapshot of a CAWFE simulation of the Tubbs Fire at 3:09 AM and for comparison, Figure (b) shows the fire's extent as mapped by the Visible and Infrared Imaging Radiometer Suite (VIIRS) at the same time. The simulation recreates the Diablo downslope wind event, localized strong winds over the coastal ranges where winds reached 30-40 m s-1, extreme wind peaks exceeding 40 m s-1 that formed over smaller hills including near the reported ignition site, and rapid fire growth. A 4-domain CAWFE forecast of this event refining to 370 m horizontal grid spacing was performed and showed similar fidelity, forecasting the fire event 4 times faster than real time on a single processor.

Figure: Retrospective simulation of the Tubbs Fire using CAWFE
Figure: (a) Retrospective simulation of the Tubbs Fire at 3:09 AM PDT using CAWFE. The fire's heat flux is show in the upper color bar (in W m-2). Wind speed arrows (in m s-1) point downstream and are colored according to the color bar at right. Vectors are shown every 3 grid points. (b) The first VIIRS active fire detection data of the Tubbs Fire at 3:09 AM PDT.