A LATENT HEAT RETRIEVAL IN A RAPIDLY INTENSIFYING HURRICANE 34th Conference on Radar Meteorology A LATENT HEAT RETRIEVAL IN A RAPIDLY INTENSIFYING HURRICANE Steve Guimond and Paul Reasor Florida State University

Background/Motivation Main driver of hurricane genesis and intensity change is latent heat release Observationally derived 4-D distributions of latent heating in hurricanes are sparse Most estimates are satellite based (i.e. TRMM) Coarse space/time No vertical velocity Few Doppler radar based estimates Water budget (Gamache 1993) Considerable uncertainty in numerical model microphysical schemes McFarquhar et al. (2006) Rogers et al. (2007)

Current Approach Model testing: Examine assumptions Refined latent heating algorithm (Roux and Ju 1990) Model testing: Non-hydrostatic, full-physics, quasi cloud-resolving (2-km) MM5 simulation of Hurricane Bonnie (1998; Braun 2006) Examine assumptions Uncover sensitivities to additional data Uncertainty estimates

Numerical Model Testing

Structure of Latent Heat Goal saturation using production of precipitation (Roux and Ju 1990) Divergence, diffusion and offset are small and can be neglected

Magnitude of Latent Heat Requirements Temperature and pressure (composite eyewall, high-altitude dropsonde) Vertical velocity (radar)

Putting it Together Positives… Uncertainties to consider… Full radar swath of latent heat in various types of clouds (sometimes 4-D) Uncertainties to consider… Estimating tendency term Steady-state ? Thermo based on composite eyewall dropsonde Drop size distribution uncertainty and feedback on derived parameters

Model Heating Budget Results

Examining Assumptions with Doppler radar

Impact of Tendency on Heating Clouds are not steady state Guillermo TA  tendency term with ~34 min delta T Sufficient to approximate derivative? Typical value of tendency term for ∆t  0 ? Should tendency term be the same or similar regardless of delta T? This should depend on the time scale of the feature. Derivatives are scale independent, correct? Does this rule only apply for linear functions?

Impact of Tendency on Heating Should tendency term be the same or similar regardless of delta T? This should depend on the time scale of the feature. Derivatives are scale independent, correct? Does this rule only apply for linear functions?

Impact of Tendency on Heating All heating removed

Impact of Tendency on Heating R2 = 0.714 How to parameterize tendency term? Using 2 minute output from Bonnie simulation (2) Coincident (flight level) 2 RPM LF data

Impact of Tendency on Heating Including parameterization 14

P-3 Doppler Radar Results Rapidly intensifying Hurricane Guillermo (1997) NOAA WP-3D airborne dual Doppler analysis (Reasor et al. 2009) 2 km x 2 km x 1 km x ~34 min 10 composite snapshots

Hurricane Guillermo (1997)

Uncertainty Estimates Mean =117 K/h Bootstrap (Monte Carlo method) Auto-lag correlation  ~30 degrees of freedom 95 % confidence interval on the mean = (101 – 133) K/h Represents ~14% of mean value 27

Conclusions and Ongoing Work New version of latent heat retrieval Identified sensitivities, constrained problem with more data (e.g. numerical model) Developed tendency parameterization Statistics with P-3 LF data Validate saturation with flight level data Ability to accept some errors in water budget Local tendency, radar-derived parameters, etc. Monte Carlo uncertainty estimates (~14 % for w > 5) Goal: Understand impact of retrieved forcings on TC dynamics Simulations with radar derived vortices, heating Smaller errors with retrieved heating vs. simulated heating

Acknowledgments References Scott Braun (MM5 output) Robert Black (particle processing) Paul Reasor and Matt Eastin (Guillermo edits) Gerry Heymsfield (dropsonde data & satellite images) References Roux (1985), Roux and Ju (1990) Braun et al. (2006), Braun (2006) Gamache et al. (1993) Reasor et al. (2009) Black (1990)

Thermodynamic Sensitivity

Testing algorithm in model How is Qnet related to condensation? Only care about condition of saturation for heating Some error OK Tendency, reflectivity-derived parameters Testing algorithm in model How is Qnet related to condensation?

Hurricane Katrina (2005) particle data from P-3 Constructing Z-LWC Relationships Below melting level: Z = 402*LWC1.47 n = 7067 RMSE = 0.212 g m-3 Above melting level (Black 1990): Z = 670*IWC1.79 n = 1609 r = 0.81 Hurricane Katrina (2005) particle data from P-3 August 25, 27, 28 (TS,CAT3,CAT5) Averaged for 6s  ~ 1km along flight path Match probe and radar sampling volumes

Doppler Analysis Quality Comparison to flight-level data at 3 and 6 km height Vertical velocity (eyewall ~1200 grid points) RMSE 1.56 m/s Bias 0.16 m/s

Dropsondes Composite sounding DC8 and ER2 (high-altitude) total of 10 samples Deep convection Sat IR, AMPR, wind and humidity

Testing algorithm in model Non-hydrostatic, full-physics, cloud-resolving (2-km) MM5 simulation of Hurricane Bonnie (1998; Braun 2006)

Testing algorithm in model

Testing algorithm in model

Testing algorithm in model