FIFTH INTERNATIONAL WORKSHOP on TROPICAL CYCLONES
Topic 1 TROPICAL CYCLONE STRUCTURE AND STRUCTURE CHANGE
Topic Chair: Dr. Russell L. Elsberry (USA)
Naval Postgraduate School
589 Dyer Rd., Room 254
Monterey, CA 93943-5114
The purpose of this section is to give an overview of the topic on tropical cyclone (TC) structure and structure change, which includes both the inner-core intensity change and the outer structure changes. As will be evident from the topic Rapporteur reports, this topic is being addressed primarily from a research perspective. While intensity change prediction is considered to be a high priority for improvement at many forecast centers, the discussion in Topic 1.6 indicates little skill is evident. This lack of skill may be related to: lack of observations (Topic 1.1); lack of understanding of environmental effects (Topic 1.2); need for better understanding of convective-scale processes (Topic 1.3); and deficiencies in understanding of the air-ocean interface processes and the feedback from the upper-ocean (Topic 1.4).
The TC structure change problem is complex because it involves interactions of physical processes on a multitude of scales from the sea spray droplets to interactions of the TC circulation with adjacent synoptic systems that may be in the midlatitudes. Whereas improved observations and statistical technique developments are the path for short-term gains in intensity and outer structure forecasting, substantial advances (especially at 36 h and beyond) are most likely to come from numerical model guidance. However, the development of that operational model guidance must be based on understanding of the individual physical processes and their interactions achieved via research models (Topic 1.5). This structure change problem is much more complex than the TC motion prediction problem that does not depend so critically on the convective-scale and air-sea transfer processes. Thus, a multi-scale approach is required and collaboration among scientists from many countries is desired. Advances are most likely if observationalists, modelers, and forecasters work together on this Grand Challenge problem of TC structure change.
The status of the analysis and forecast capability at a number of the forecast centers is addressed in Topic 1.6. For the TC intensity problem, the first step is to obtain an accurate estimate of present intensity. The venerable Dvorak technique is being supplemented with the Objective Dvorak Technique (ODT), and an Advanced ODT is being developed to extend the technique to less intense storms. Advances in intensity prediction are mainly being achieved via multi-spectral satellite approaches, as described in the keynote session. Blending of the microwave sensor estimates with the infrared sensor sensors appears to be the most-promising avenue for improving the initial intensity estimate.
Statistical-dynamical approaches are the mainstay for intensity prediction, with the Statistical Hurricane Intensity Prediction System (SHIPS) at the U. S. National Hurricane Center being an example. The SHIPS is being improved by adding satellite-based predictors and from ocean heat content predictors. The neural net approach may provide an alternative to the multiple regression approach. Thus far, the skill relative to a climatology-persistence approach remains small. None of the dynamical models has skill exceeding the statistical-dynamical approach.
One of the helpful concepts in intensity forecasting is the Maximum Potential Intensity (MPI), e.g., the empirical relationships developed separately by K. Emanuel and by G. Holland. Given the present intensity and the MPI, the forecaster has some bounds on the possible intensity changes. However, improved understanding of the physical factors favoring or inhibiting a TC from reaching its MPI is required. Since these physical factors are likely to be similar in TCs around the globe, collection, analysis, and sharing of information should improve TC intensity forecasting in the future.
Definition of the TC outer wind structure over the ocean has improved from the scatterometer instruments such as Quikscat (see keynote session talks by T. Nakazawa and R. Edson). Although quality control issues need to be resolved, the scatterometer offers hope for defining the outer surface wind structure in those areas that do not have aircraft reconnaissance. In principle, the physics of outer wind structure change is understood and is contained in the high-resolution, full-physics numerical models. Thus, a forecast of the outer wind structure change should be possible given an accurate initial condition specification.
The purpose of the keynote session was to describe the present and future satellite-based observations, and many of these instruments have the potential for improved TC structure analysis. Topic 1.1 emphasizes the additional information that has been gained from radar (especially deployed on research aircraft). These radar observations have documented significant asymmetric motions and clouds (precipitation) in addition to the better-known axisymmetric aspects. The ability to map the complete three-dimensional circulation of the TC from airborne Doppler radar on time scales of an hour has enabled a partitioning of the wind to examine the roles of asymmetries in studies of intensity change.
One of the science issues is the effect of the vertical tilt of the TC vortex in intensity change and motion. The aircraft radar indicates the vertical tilt in the lower troposphere is small (~ 1 km) and may generally be less than 5 km over the deep troposphere in the absence of strong environmental shear. The aircraft radar data also provide an estimate of the vertical wind shear in the inner core. As expected, the vortex tilts are along the direction of the shear vector with increasing height, with larger tilts associated with larger shear magnitudes. However, the cases studied thus far do not have a clear relationship between the storm motion and the vertical wind shear.
A unique airborne radar data set obtained in Hurricane Olivia during a period of rapid increase in environmental vertical wind shear provides insights. The eyewall radar reflectivity became increasingly asymmetric as the maximum vertical wind shear increased from about 5 m s-1 to about 15 m s-1 over a few hours. The outer rings of high reflectivity largely disappeared, which left the eyewall convection on the left side of the shear vector as the dominant reflectivity feature of the storm. The vertical-incidence Doppler radar documented an 8 m s-1 updraft on the downshear side of the eyewall and a 5 m s-1 downdraft on the upshear side. The asymmetric radial flow became more pronounced with inflow at 3 km on the downshear side and outflow on the upshear side as the storm moved relative to the surrounding low-level air. The conceptual model by Black et al. (2002) of the effect of vertical wind shear on the convective asymmetry should be useful to forecasters and assist in interpreting the numerical model simulations.
As indicated above, one of the key forecast issues is what favors or inhibits a TC from reaching its MPI. Environmental effects discussed in Topic 1.2 are certainly one factor. In numerical simulations on a f-plane with no environmental flow (and with no negative feedback allowed from a cooling of the upper ocean beneath the TC), an asymmetric vortex develops that will spinup to its MPI. Introduction of eastward (westward) environmental flow that is uniform in the vertical in a numerical model on a beta-place results in asymmetric convection and different intensities.
Considerable progress has been made in numerical simulations of the effect of vertical wind shear on the vortex structure (see Topic 1.2 report). Many of the structural modifications such as the convective asymmetries observed in the Hurricane Olivia case described in section 1.0.3 above are being reproduced in the models. However, it appears some model vortices are more susceptible to environmental shear magnitudes that are smaller than are observed in nature. For example, the model vortices appear to tilt downshear too strongly. The model simulations suggest the dispersal of the warm core during increasing vertical wind shear begins in the upper troposphere where the inertial stability of the vortex is smaller, and is thus less resistant to radial inflows. As the warm core is diminished, the vortex weakens and the accompanying decreases in inertial stability aloft allow a deeper layer of the upper-troposphere warm core to be advected downshear. Observational support for this negative feedback sequence is being found in the transformation stage of the extratropical transition of a TC (see Topic 4.6).
The improved understanding of vertical wind shear effects may then help understanding how an upper-level trough interaction with a TC may be a good trough (contribute to TC intensification) or a bad trough (inhibit intensification). That is, the vertical wind shear associated with the leading edge of the approaching trough may account for the ventilation of the warm core described above and the development of the asymmetry in the cloud/precipitation distribution that is less favorable for intensification than a symmetric cloud distribution. The good trough scenario involves a positive eddy momentum flux convergence and/or an enhancement of a jet streak between the approaching cold trough and the warm outflow from the TC. The specific conditions in which these intensification-favoring aspects of the trough-TC interaction dominant over the inhibiting vertical shear effects are not known. These conditions will involve the juxtapositioning sequence and the three-dimensional structure of the TC (especially the outflow layer) and thus are complex and nonlinear. Progress in understanding these physical processes and the conditions that apply in specific cases is required if improved structure change guidance is to be provided the forecasters.
The vertical redistribution of the heat/moisture gained from the tropical ocean occurs in convective clouds, which is discussed in Topic 1.3. Dramatically different evolutions of the intensification of a TC are simulated in relatively simple models for different closures of the cloud-base mass flux parameterizations. This sensitivity to the details in the representation of the convection is a sobering consideration in the outlook for accurate numerical guidance for TC structure change prediction. One interesting result from the simplified models is that the presence of the peripheral convection causes changes in the vortex structure that are unfavorable for future inner convection. This result is consistent with the presence of an extensive moat of minimum radar reflectivity outside the small radius eyewall reflectivity of intense hurricanes such as Hurricane Gilbert.
The development of convectively-induced asymmetries in rainfall outside the eyewall leads to coherent structures that undergo a life cycle that may include a mesoscale convective vortex. The evolution of the coherent structures and their effect on formation and intensification of the overall vortex structure is an area of active research. An emerging concept is that structural change (and thus intensity) may occur through convective bursts and their associated circulations in a stochastic process, rather than through a slow, continuous evolution associated with symmetric convection about the vortex circulation center.
Another active research topic is the mechanism of vortex structural modifications by vortex Rossby waves. Two-dimensional simulations with relatively high wave amplitudes suggest a wave-induced eigenmode interacts with the vortex structure, which can sustain the eigenmode and interact with the convection. This feedback loop involving the vortex, vortex Rossby wave, and convection would then need to be correctly modeled to obtain accurate structure modification predictions. Further discussion of the characteristics of vortex Rossby waves and especially their potential role in vortex formation is given in Topic 4.4.
Detailed studies of the eyewall convection and mixing processes are in progress. Trajectories in some numerical simulations suggest the primary source for the buoyancy of these clouds is the surface fluxes of moisture and heat that is added in the lowest part of the boundary layer. Those trajectories that penetrate farther into the eye appear to accelerate outward sharply while rising out of the boundary layer. If air parcels within the nearly saturated lower part of the eye are engulfed into the roots of the eyewall convection, the equivalent potential temperature of the ascending air could be enhanced over that achieved by air-sea fluxes along the inward trajectory.
1.0.6 Air-ocean interface and ocean effects
Just as the numerical model simulations of TCs are sensitive to the convective parameterization scheme, the same is true for the atmospheric boundary layer (ABL) parameterizations. In fact, these sensitivities are linkedwhatever the vertical fluxes of heat and moisture in the ABL, the convective cloud distribution determines where that heat and moisture is pumped out of the ABL, and cooler, dry air is entrained into the ABL, which affects the rate of surface fluxes. The physical processes in the ABL are also linked to the air-sea interface process, and consequently with the upper-ocean processes. These coupled processes are the focus of Topic 1.4.
One of the missing physical processes in traditional ABL parameterizations is the sea spray effects in high wind regions (especially above 35 m s-1). The sea spray size distribution is important because of the different equilibration rates for heat and moisture for different size droplets as they travel from their source in the breaking ocean surface waves until they return to the ocean. Clearly, obtaining measurements of sea spray distributions is difficult in such high wind and wave heights. Obtaining such measurements, and determining their effects on the exchange coefficients for heat, moisture, and momentum, is one of the objectives of the Coupled Boundary Layer Air-Sea Transport (CBLAST) field experiments during 2002-2004.
In addition to the sea spray aspect, the ocean surface wave distribution is important for the surface stress distribution, which affects the frictional drag on the tropical cyclone circulation and generation of ocean currents and mixing effects. Thus, an advanced TC structure prediction model must be coupled to an ocean surface wave prediction model. These wave models include wave-wave interaction terms that are not well-known for the high wind conditions of a tropical cyclone. Furthermore, these models need to be extended to the higher frequencies at which wave breaking occurs. Whether the waves are young or more mature may affect the way in which they interact with the wind, and thus impact the air-sea fluxes. The direction of wave propagation relative to the local wind is hypothesized to affect the wave steepness, and the wave steepness distribution has been related to the magnitude of the entrainment mixing process at the base of the ocean mixed layer (OML).
It is well-known that the TC generates a wake of lower sea-surface temperatures (SST) as it moves over the ocean. The magnitude of the SST decrease is larger for a more intense, slowly-moving cyclone over a shallow OML. If the TC becomes stationary, the wind generates a net ocean current transport away from the center, which induces upwelling of colder water from below and the SST may decrease by 4_C-6_C. The negative feedback to the TC occurs if the lower SSTs decrease the sensible and latent heat fluxes that are the ultimate energy source for driving the TC circulation. Should the high winds of the TC pass over a warm ocean current (such as the Gulf Stream or Kuroshio) or eddy with a much deeper OML, the SST decrease will be smaller, which will diminish the negative feedback mechanism. Such a scenario is believed to have contributed to the rapid deepening of Hurricane Opal when it was over a Loop Current eddy in the Gulf of Mexico.
The effects of the SST value and the OML magnitude can be combined in one measure the Ocean Heat Content (OHC), which is the integrated heat content above the 26_C isotherm. Another reason for the interest in the OHC is that its distribution can be inferred from satellite radar altimeter measurements that detect bulges (depressions) in the ocean surface elevation over warm (cold) pools. Thus, an opportunity exists from satellite observations to characterize the OHC (and SST) conditions that the TC is moving toward, and thus infer whether one of four environmental conditions for TC intensification is going to be more or less favorable. Even higher horizontal resolution OHC distributions can be analyzed where Airborne Expendable Bathythermograph (ACBT) instruments can be deployed from reconnaissance aircraft.
Since the OHC can be changed by horizontal and vertical advection, surface heat fluxes, and vertical mixing through the 26_C isotherm, knowledge of the time-dependent distribution of OHC requires an ocean thermo-haline circulation model. However, some researchers suggest that the primary changes in SST and OML can be approximated by one-dimensional (stick) mixed layer models that do not include horizontal or vertical advection. Comparisons between the full three-dimensional and the stick models in a large number of cases are needed to resolve the conditions in which the simpler models may be adequate.
Regardless of whether three-dimensional or one-dimensional ocean models are employed, the greatest uncertainty will be in the entrainment mixing parameterizations at the base of the OML. The evolution of the SST and OML is quite sensitive to which mixing parameterization is utilized. These differences need to be resolved with new observations and model sensitivity tests, which is expected to occur with the CBLAST data sets.
1.0.7 Numerical modeling for research and operational prediction
It was stated above that advances in prediction of TC structure change beyond 36 h will require numerical model guidance. None of the present operational TC dynamical models has skill in intensity forecasts relative to the statistical-dynamical models. The development of skillful operational models, and the research modeling needed to guide that development, is the focus of Topic 1.5. More detailed discussion of the science requirements for research numerical models for TC intensity prediction can be found in the report of a U. S. Weather Research Program Workshop held in San Diego, California, during 3-4 May 2002. The plans for an operational Hurricane Weather and Research Forecast (HWRF) model at the National Centers for Environmental Prediction (NCEP) will be available in a second workshop report. Many other agencies and national forecast centers are also developing intensity prediction models.
Both idealized simple dynamical models and high-resolution, full-physics numerical models are being used to improve our understanding of all of the physical processes described above that are involved in TC structure change. Here the focus is on full physics dynamical models that will pave the road for development of the next generation of operational forecasting models. The conclusion from the San Diego workshop was that the research models will need to have 1-2 km grid spacing to resolve all of the physical processes. Several research groups have run multiply-nested models with inner grid resolutions with such small grid sizes. Among the realistic features simulated for the first time was the development of concentric eyewalls and a complete eyewall replacement cycle.
As indicated in section 1.0.4, other numerical models are realistically simulating many observed features of the asymmetric convection in response to idealized environmental vertical wind shears. Now real-data predictions are required for vertical shears associated with bad troughs, as well as for good troughs.
A key issue for these high-resolution models is the representation of the microphysical processes. Fortunately, the NOAA Hurricane Research Division and the NASA Convection and Mesoscale Experiments (CAMEX) have collected microphysical observations that can be used to calibrate the models.
Although many critical issues related to air-sea interface are discussed in section 1.0.6, the CBLAST experiments offer considerable hope that these issues will be addressed in the next few years. These research understandings will need to be incorporated in numerical models and demonstrated with real-data experiments. This strategy should also include the turbulent mixing processes in the ocean so that coupled ocean-TC models can be tested.
It would be most helpful if the relative contributions of all the physical processes involved in TC structure change were already known, since this would make the design of the operational model more straight-forward. In lieu of such information, the development of the operational model will have to proceed on various fronts depending on computational resources.
In contrast to the research modeling studies above in which the necessary high resolution and multiple physical process representations can be achieved with integrations that might take days to finish, the operational forecast model must be completed in approximately one hour. Thus, it will not be possible to have 1-2 km horizontal resolution in an operational model within the next five years. Indeed many compromises will also need to be made with regard to the physical processes. Rather than having an explicit representation of the moist processes over nearly all of the major convection area in the TC as planned for the research models above, the operational model will still need to have a cumulus parameterization technique. As indicated in section 1.0.6, the choice of the cumulus parameterization is intimately related to the choice of the frictional parameterization. Thus, many comparisons with different combinations will be necessary. Similarly, the choice of microphysical representation in those inner areas with explicit moist physics will require many tests.
Other choices to be made for an operational model are the benefits to be gained with coupling with a land surface model, an ocean surface wave model, and an ocean circulation model. Each of these options will be costly, and tradeoffs with adding more horizontal or vertical representation or some internal atmospheric physical process representation will need to be considered.
Two other big scientific challenges for the operational TC structure model are the adequacy of the observation database and the data assimilation technique. Clearly, the initial conditions of all model variables must be specified which includes microphysical distributions, plus all of the land surface processes, ocean surface wave variables, and ocean variables if those options are included. Data assimilation techniques are needed for environmental data sources, plus the special vortex data such as radar reflectivity and airborne Doppler radial wind components. All of these atmospheric data assimilation tasks are challenging, and similar challenges exist for ocean data assimilation.
In summary, the numerical modeling required for both research studies and operational forecasting of TC structure change will be difficult. Given the uncertainty in understanding all of the complex physical processes, advances must be made in the science to support the development of an accurate operational model.