Tropical Weather

Research focused upon the collection and analysis of observations of hurricanes and other tropical weather systems. Research activities include identifying and validating observational needs, developing instrumentation, obtaining observations, studying the optimum configurations for observation networks, modeling and data assimilation, expediting and facilitating the transition of research to operations, and developing analysis and forecasting applications for operations.

There are two representative projects below:

Representative Projects


Ensemble-Based High-Resolution, Vortex-Scale Data Assimilation for Hurricane Model Initialization

A. Aksoy, K. Sellwood and S. Lorsolo (UM/CIMAS); S. Majumdar (UM/RSMAS), S.D. Aberson (NOAA/AOML); F. Zhang (Pennsylvania State University)

 

Long Term Research Objectives and Strategy to Achieve Them:

Objectives: improve hurricane intensity and track forecasts through improved representation of hurricane vortex structures in the initial conditions of hurricane forecast models.

Strategy: To better utilize high-resolution observations (dropsonde, radar, flight level, surface wind speed, etc.) collected during the Hurricane Field Program run by NOAA/AOML/HRD by taking advantage of flow-dependent covariance structures that can be obtained from an ensemble of model forecasts that will form the basis of an ensemble Kalman filter data assimilation system.

 

An ensemble Kalman filter (EnKF) data assimilation system has been built to assimilate highresolution, vortex-scale observations that are routinely collected and transmitted real-time, during NOAA’s annual Hurricane Field Program and regular reconnaissance flights. Coined Hurricane Ensemble Data Assimilation System (HEDAS), it will be tested in semi-real-time as part of NOAA’s Hurricane Forecast Improvement Project (HFIP) during the 2010 hurricane season. Various observation types that will be assimilated include Doppler radar radial winds, dropwindsonde wind, pressure, temperature, and humidity, flight-level wind, pressure, temperature, and humidity, and stepped-frequency microwave radiometer (SFMR) surface wind speed. The system will also make it possible to evaluate impacts of other potential observation platforms through observing system and observing system simulation experiments (OSE and OSSE, respectively). The EnKF is a state-ofthe-art data assimilation system first proposed for geophysical applications by Evensen (JGR, 1994).
In this specific application, we implement the “ensemble square root” filter of Whitaker and Hamill (MWR, 2002) and covariance localization by Gaspari and Cohn’s (QJRMS, 1999) compactly supported fifth-order correlation function.

HEDAS has been developed within the framework of AOML/HRD’s experimental HurricaneWeather Research and Forecast (HWRFx) model. Data assimilation will be performed on a 3-kmnest, while HWRFx will run in a nested 9/3-km configuration during cycling. Initial ensembleperturbations are obtained from NOAA/ESRL’s EnKF system, which will run with NCEP’s GlobalForecast System (GFS) modeling framework and Gridpoint Statistical Interpolation (GSI) system’s forward operators.

Currently, there are two major research directions with HEDAS. In one approach, a detailed diagnostic of HEDAS performance is being carried out in an OSSE environment. For this purpose, a higher-resolution version of HWRF-x, at 4.5/1.5-km horizontal resolution with explicit convection in all domains, is used to obtain a nature run, which will be the basis for simulating various observation platforms. The case of interest is Hurricane Paloma of 2008. The nature run is initialized from one ensemble member of the GFS-EnKF analysis valid for 7 November 2008 12Z. The availability of a nature run enables detailed diagnostics of a data assimilation system when observations simulated from that nature run are assimilated. The fact that such observations are obtained from a model run also eliminates non-meteorological noise from them so that the performance of a data assimilation system entirely depends on the underlying modeling and data assimilation characteristics. The OSSE aspect of the project is motivated by these factors and will be completed in the near future using the Paloma 2008 nature run presented here.

In Figure 1, some characteristics of the nature run are presented for a 24-h period. In panel (a), a general northeasterly track of the simulated storm is apparent. The 10-m radius of maximum wind speed (10-m RMW, measured as the distance of the azimuthally-averaged 10-m tangential wind speed maximum from the center), the distance of absolute maximum 10-m horizontal wind speed from the center, and the azimuthal standard deviation of
distance of maximum tangential wind to center (coined “10-m RMW standard deviation” hereafter) are shown in panel (b). While the distance of actual 10-m wind maximum is not expected to exactly follow the RMW pattern (because of potential vortex asymmetries), a general agreement within one 10-m
RMW standard deviation is evident beyond 6 h. Meanwhile, a large fluctuation (relative to 10-m RMW standard deviation) between the two quantities within the first 6 h likely points to an imbalance within the vortex as the simulation spins up from the much-coarser-resolution GFS-EnKF initial conditions. It is also evident in this panel that RMW fluctuates at ~30 km 6 h and beyond. Panel (c) shows the RMW slope (as approximated from the linear regression between RMW and height at each time) as a proxy for eyewall slope. During the 6-24 h of simulation, a general decreasing trend is evident. Finally, panel (d) depicts estimations of intensity from minimum central surface pressure and maximum 10-m horizontal wind speed. While minimum pressure appears to be steadily decreasing in the simulation, maximum-wind intensity is steady between 6-18 h and then increases by ~10 m s-1 during the last 4 h of simulation. This intensity increase is preceded with an increase in RMW and decrease in RMW slope.

Figure 1. 24-h nature run simulation for Hurricane Paloma (2008), initialized at 7 November 2008 12Z. (a) Hourly center position as obtained using the 10-m centroid vorticity method, (b) 10-m RMW (solid dark, km), distance of absolute 10-m maximum horizontal wind to storm center (solid gray, km), and 10-m RMW standard deviation (dashed, km), (c) vertical slope of RMW (km RMW km-1 height), and (d) minimum central surface pressure intensity (dark, mb) and maximum 10-m horizontal wind intensity (m s-1).

In parallel work, the performance of HEDAS has been evaluated with real observations for the case of Hurricane Bill (2009), using inner-core Doppler radar, dropwindsonde, SFMR, and flight-level data that were collected by NOAA and Air Force. The HWRFx model is initialized from the same GFS-EnKF ensemble as the OSSE experiment, for 30 ensemble members, at 9/3-km horizontal resolution and the aircraft data are assimilated on the 3-km inner nest. In these experiments, the impact of assimilating the different data types into the model background forecast is assessed by examining the model diagnostic variables and the evolution of their associated error statistics over the 5-h interval in which observations were available. Additional experiments are being performed to study the effect of varying certain HEDAS parameters in order to determine the optimal configuration for real-time data assimilation. This configuration is planned to be utilized for the near-real-time HEDAS runs during the 2010 hurricane season.

In Figure 2, the ensemble-mean horizontal fields of wind and sea-level pressure are shown for the model forecast/analysis with no data assimilation (a, b) and the analysis produced after 5 forecast-assimilation cycles (c, d). The forecast without data assimilation produces a weaker, less symmetric vortex, while the analysis which incorporates the aircraft data is more consistent with the observed structure and intensity of Hurricane Bill.

Figure 2. Mean sea-level pressure (MSLP) and 10-m horizontal wind speed for Hurricane Bill valid at 13Z on August 19, 2009. Top row: Control ensemble mean forecast without data assimilation. Bottom row: Ensemble mean analysis after five data assimilation cycles.

 

Characterization of Turbulent Energy in Hurricanes Using Doppler Measurements

S. Lorsolo and J. Zhang (UM/CIMAS); J. Gamache, P. Dodge and F. Marks (NOAA/AOML)
 

Long Term Research Objectives and Strategy to Achieve Them:

Objectives: To estimate the distribution and evolution of hurricane boundary layers’ turbulent kinetic energy (TKE) for a better understanding of the processes that influence hurricane intensity change and for a more accurate parameterization in numerical weather prediction models.
Strategy: To analyze Doppler radar and in-situ measurements from various instruments from an extensive data set to provide a comprehensive evaluation of the turbulent structure of hurricanes. To develop methods to estimate low-level turbulent parameters that will provide an accurate assessment of hurricane turbulent energy to use in numerical weather prediction parameterization.

 

One of the main challenges of hurricane research is to better understand the processes that influence hurricane intensity change, which could ultimately lead to improved intensity forecasts. Among the various parameters believed to impact hurricane intensity change are air-sea interaction and turbulent
energy transport within the hurricane boundary layer (HBL). An accurate estimate of hurricane turbulent energy is crucial to identify the role of turbulent processes in the energy transport within a storm in general and in the hurricane boundary layer (HBL), in particular.

Figure 1. Lorsolo_T5_ Conceptual model of TKE distribution in hurricanes.

The goal of this research project is to characterize turbulent parameters of hurricanes, especially in the HBL, using remote sensing observations as they are difficult to assess from direct measurements in high wind regions such as the eyewall and the HBL. Turbulent data retrieved from instruments such as Doppler radars are crucial to identify turbulent processes impacting hurricane intensity and to evaluate models.

The activities of the past year have focused on continuing processing and analyzing the data to improve the robustness of the results, completing the manuscript related to the study and developing a new method to retrieve turbulent dissipation rate from Doppler measurements to help model evaluation. A new algorithm is being developed to test the method. The TKE study was refined by providing general results from the combination of data from multiple hurricanes and by the realization of a conceptual model of the TKE distribution with a tropical cyclone.