热带气旋强度变化机理及关键物理过程研究
Study on the Mechanisms and Key Physical Processes of Tropical Cyclone Intensity Change
台风数值模式关键模式物理构成研究
Study on the Key Physical Processes in Numerical Typhoon Models（2017YFC1501602）
1. Introduction
Tropical cyclones (TCs）are synotpic weather systems that form overthewarm tropical oceans. After their geneses, TCs are often affectd by complex multi-scale processes among convections, storm scale dynamics and the varying environment. Because of their complexity, 3-7 day forecastis o f TC track, structure and intensity are quite chanlenging, and thus how to imrpove the TC track, intensity and structure 3-7 day in advance is one of the difficult issues in the world. On such time scale, the dynmaic and thermodynamic processes and the predictability of TC intensity and structure have been one of the international frontier research. Numerical weather prediciton (NWP) is the basis of weather forecasts nowadays. To imrpove 3-7 forecasts of TC track, intensity and structure, almost all forecast centers in countirs around the world are devoting great efforts to the development and improvement in parameterizations of cumulus convection, planetary boundary layer (PBL) processes, cloud-radiation interactions, and so on, as well as the coupled ocean-wave-atmosphere models.NWP is the main tool and approachto the fine-scale prediction of TCs, however, although theoretically the spacial resolution of an NPW model can be given as high as possible, the TC fine-scale forecasts could not be achieved by increasing model resolution, it is equally important to improve the representation of model physical processes. The ability of the state-of-the-art NWP models to represent the TC fine-scale structure and evolution is still limited. It is urgent to achieve breakthrough advancements in the air-sea coupled strategy and air-sea interaction processes, turbulent exchanges in TC boundary layer, cloud microphysics processes, and other subgrid-scale processes.
This project focues on the technical aspects of key physical processes in NWP models used for the 3-7 forecasts of TC intensity and structure. Based on the multi-scale nature of TC developments and the uncertainty of the model physical processes, we investigate the TC structure characteristics under complex topography and the scale-dependence of the air-sea interfacial exchange, cumulus convection, and etc, and their influences on TC track, intensity and structutre. The main research targets include to develop the dynamical TC-vortex initialization technique, scale-aware cumulus convective parameterization scheme, new PBL parameterization, and air-sea momentum and enthalp exchange algorithms. We consdier the scale adaptivity, nonlocal and non-equilibrium, and random nature of atmospheric physical processes. We attempt to develop scale-aware cumulus onvective parameterization, to optimize and improve the PBL by including the effects of boundary layer rolls and low-level strong vertical wind shear on TC intensity and structure and parameterization, and to develop simplified system to include the coupled ocean-atmospehre-wave processes. Our ultimate goal is to develop scale-aware schemes/algorithm suitable for the use in models with 2-25 km horizntal resolutions, including scale-aware cumulus convective parameterization shceme, improved the representation of PBL scheme, construct new algorithm to perform dynamical TC-vortex initilization scheme under the effec tof meso-scale terrains, and to develop ocean-atmosphere-wave coupled techiques under strong wind conditions.
Through this project, we have achieved almost all our objectives,inclduing an improved dynamical TC-vortex initialization (DVI) scheme including the effect of mesoscale terrains, a parameterized air-sea coupled feedback of the change in SST on surface enthalpy flux, tested the dependence of sea spray effect on suther enthalp flux and TC intensification on the surface drag parameterizations, diagnosed the performance of an earlier PBL scheme, which incudes an explicit prognostic euqation for the turbulent kinetic energy (TKE) dissipation and is called E-ε scheme, implemented and tested a scale-aware cumulus parameterization scheme for TC forecasting.We found that PBL scheme is important for TC intensity and structure changes, in particular the inclusion of dissipative heating due to the TKE dissipation, which affects the intensification of strong TCs. In addition, we also studied the dynamical/physical proceeses to understand the TC intensity change, the weakening of landfalling TCs over China and the effects of land surface process on landfalling TC intensity and precipitation, and also the long-term trends in landfalling TCs over China under the background of global warming.
2. Dynamical TC vortex initialization under the influence of mesoscale terrains
Although observing platforms for TCs have been significantly advanced in the past decades, data obtained from them still have limitations in fully resolving the spatial structure and temporal evolution of TCs for initializing numerical forecast models. Therefore, various initialization methods have been attempted to obtain improved vortex representation in model initial conditions. One of the methods is the so-called dynamical vortex initialization (DVI). Various DVI schemes have led to improvements in TC forecasts by numerical models in recent decades. Although the research community believes that data assimilation can eventually improve the initial TC vortex conditions, the poorly revolved TC vortex intensity and structure in the global analysis often limit the expected improvements from data assimilation because of the strong background dynamical constraints. Previously developed DVI schemes are all applied to TCs over the open ocean without significant effect from any meso-scale topography. When a TC approaches mesoscale terrains (e.g. Taiwan Island and Luzon island in the western North Pacific), the present TC initialization schemes cannot reasonably resolve TC structures, often leading to large forecast errors in intensities, precipitation and wind of TCs during their landfalls. Therefore, this project developed a new DVI scheme for TCs under the influence of mesoscale terrains, which used a terrain-compensation method. A large set of hindcasts were conducted to examine the performance of the new DVI. Results showed that the 72-h position errors and the intensity errors up to the 36-h forecasts using the new DI decreased significantly and were smaller than those from the HWRF forecasts for the same TCs. The new DVI scheme is also shown to produce the TC inner-core structure and rainbands more consistent with satellite and radar observations (see Liu et al. 2018). Figures 1 and 2 show the sea level pressure and the model forecast TC intensity for Typhoon Dujuan (215). In the CTRL simulations, the DVI was skipped, namely the global analysis was interpolated to get the high-resolution model initial condiciton. In the DI simulaitons, the new DVI scheme was used to initialize the high-resolution model. We can see that both the structure and intensity of the storm at the initial time and the forecasts were both improved considerably.
The new TC DVI scheme has been applied in the Typhoon Regional Assimilation and Prediction System (T-RAPS) and tested in terms of 22 TCs in 2016. The TC initial intensity can be improved by the new DVI. During the typhoon seasons since 2019, the T-RAPS is conducted four times per day for real-time forecast of TCs over the western North Pacific. The real-time forecasting products and data are supplied to National Meteorological Center (NMC) and Public Meteorological Service Centre (PMSC) of China Meteorological Administration (CMA). Five clusters of 32 forecasting products in T-RAPS postprocessing module are output in real time during typhoon season based on NCAR Command Language (NCL). The Java-based website of the T-RAPS platform has been updated to display on multi platforms and self-adaption browser, with an optimized webpage layout. These achievements have been included in the real-time forecast system in typhoon seasons since 2019.
Figure 1. Sea level pressure (hPa, contours) and 10-m wind speed (m s–1, shading) for Super typhoon Dujuan (2015) at the initial times 0000 UTC 28 Sep (a, b) and 1200 UTC 28 Sep (c, d). The left panels are for CTRL runs (a, c) and the right panels are for DI runs (b, d).
Figure 2. Temporal evolution of maximum 10-m wind speed (m s–1) of Dujuan (2015) from 0000 UTC 24 to 1800 UTC 29 Sep in the JTWC best-track data (black thick solid) and in the hindcasts from (a) CTRL runs and (b) DVI runs.
3. TC-ocean interaction processes and their parameterization
TCs spend most of their lifetime over the ocean. The energy for TC development and maintenance comes from the underlying ocean through enthalpy flux. In term, the TC may lead to ocean upwelling and strong vertical mixing in the upper ocean mixed layer, leading toe the cooling of the ocean surface, which often reduces the ocean surface enthalpy flux. In addition, wave breaking under strong wind conditions in tropical cyclones (TCs) can generate sea spray droplets. During their suspension in the air spray droplets release sensible heat due to the air-sea temperature difference while absorb sensible heat from the environment when they evaporate and release latent heat to the environment. The interaction between spray and surrounding air may mediated the air-sea enthalpy transfer and affect the TC intensity evolution. Therefore, it is key to the TC simulations and prediction to understand and represent the air-sea interaction processes to TC numerical prediction models. Under the support of this project, we have investigated several important air-sea interaction processes, including the coastal downwelling-induced coastal SST warming and its possible contribution to TC rapid intensification of landfalling TCs, the sensitivity of sea spray induced enthalpy flux to the surface drag parameterization scheme, and the development of parameterization scheme for sea surface temperature (SST) cooling induced by a TC. The major results are briefly introduced below.
Coastal downwelling-induced coastal SST warming and its possible contribution to TC rapid intensification of landfalling TCs. The coastal ocean response and its feedback to Typhoon Hato (2017), a category-3 typhoon, were studied based on both coupled and uncoupled high-resolution simulations. The fully coupled WRF-ROMS model reproduced the TC track and intensity reasonably well, including the rapid intensification prior to landfall and the time of landfall, compared with the best-track data (Figure 3).
Figure 3. Comparison of track and intensity of Typhoon Hato (2017) from the best-track data (black: JTWC, and blue: CMA) and the control simulation (CTL-1, red). (a) 6-hourly 10-m maximum sustained wind speed (solid, m s-1) and (b) central SLP (dash, hPa). (c) The zoomed map of the ocean topography in the northern SCS (shading, m) and the 6-hourly storm centers from the JTWC best-track data (black) CMA (blue), and the CTL-1 simulated track (red) of Hato. The dots with numbers are time in UTC on a specific day (22 and 23 August), and line segments A and B to the right and left of the track denote the cross-sections to be shown in Figs. 6 and 7 and Fig. 10, respectively. HNI, TWI and LZI are short for Hainan, Taiwan and Luzon Island.
Figure 4. Comparison between the satellite retrieved SST and the ROMS SST output. (a) MV_IR_SST at 1200 UTC 22 August, (b) MV_IR_SST at 1200 UTC 23 August with the location of the storm center shown in red cross in the JTWC best track data, and (c) and (d) are the ROMS output SST at the same time as in (a) and (b). The strom track simulated in CTL-1 is shown in black line, and the contours denote the bathymetry of the coastal ocean.
As Hato moved northwestward to approach the coast of South China, warm SST patches developed to the right of the track over the continental inner shelf. This coastal warming was found to be related to the onshore surface currents forced by the onshore winds to the right of the storm track, walled by the inner ocean shelf and then forced downward, returned offshore along the continental shelf at the bottom, and finally upwelled about 120 km offshore in the outer shelf break, leading to a two-layer oceanic circulation across the continental shelf. Offshore upwelling-induced cooling of the ocean surface occurred in the outer shelf break. The observed SST changes in response were well captured by the coupled model simulation (Figure 4). Results from an ocean temperature budget analysis indicate that the onshore advection of warm water was the primary mechanism inhibiting the seasonal coastal upwelling and warming up the inner shelf water column. The wind-driven vertical mixing was significant over the stratified ocean but contributed little to the coastal warming because of the uniform water column in the inner shelf region. The TC-forced ocean response was further demonstrated in a sensitivity experiment using the coupled model, in which the TC vortex was removed from the initial conditions. As expected, with the TC forcing removed, the two-layer circulation across the continental shelf and the warming in the coastal SST were largely suppressed (not shown), further confirming that it was the strong onshore winds to the right of the storm track that forced the onshore surface currents and the SST warming in the inner sea shelf.
Figure 5. Comparison of the hourly TC intensity (solid: maximum 10-m height wind speed in m s-1 and dashed: central SLP in hPa) evolution in coupled simulation (CTL-1, black), atmospheric-alone simulation (CTL-2, red), and the atmospheric-alone simulation with coastal SST warm SST anomalies removed (LT-EXP, green).
The possible influence of the TC-forced coastal warm SST patches on the rapid intensification of Hato was investigated based on two additional experiments using the atmosphere-only WRF_ARW model. Results show that the storm in the experiment with the coastal warm SST anomalies removed intensified slower and reached a peak intensity about 5 m s-1 weaker than that in the experiment with the coastal warm SST anomalies included (Figure 5). The surface enthalpy flux over the coastal warm SST was enhanced, providing surplus energy to the storm and thus leading to more rapid intensification of the storm. Therefore, results from our numerical sensitivity experiments demonstrate that the TC-induced coastal SST warming contributed partly to the rapid intensification of Typhoon Hato prior to its landfall and also slowed down the weakening of Hato at and shortly after its landfall over South China.
Sensitivity of spray-induced surface flux to surface drag parameterization. As the spray generation is closely related with sea surface roughness, which is often parameterized by surface drag coefficient (CD), we investigated how the spray effects on TC intensity evolution depend on the CD scheme in idealized numerical simulations. Two different surface drag (CD) parameterizations (WRF default scheme and Donelan scheme, see Figure 6) are used to perform four numerical experiments, namely the experiments with and without the inclusion of the sea spray effect with the two surface drag coefficient parameterization schemes, respectively.
Figure 6. The dependence of surface drag coefficient (CD) on surface wind speed. Orange represents the default scheme for TC simulations in the WRF model, and blue represents the Donelan scheme. The two schemes were used to simulate TC intensification under idealized conditions.
Results show that the sea spray effect on TC intensity and size evolution depends on the surface drag (CD) scheme used and the stages of the TC lifetime (Figure 7). The dependence results from different wind speed dependence of CD, which affects the radial distribution of the spray generation and thus spray-mediated latent flux. However, although spray can speed up or slow down the intensification rate, it contributes positively to the mature stage TC intensity. The finding demonstrates that caution needs to be given to the surface drag parameterization when the sea spray effects on TC evolution is studied and discussed using numerical sensitivity experiments. It is also suggested that efforts to measure spray properties under TC conditions should be conducted to provide data for validation and improvements of current spray parameterization scheme in future studies.
Figure 7. The time evolution of maximum surface wind speed (m s？1) in experiments with (blue) and without (orange) the sea spray effects (a) with the WRF CD scheme (WRF_CTRL and WRF_SPY) and (b) with the Donelan CD scheme (Donelan_CTRL and Donelan_SPY). The two stages (primary intensification and mature stages) are marked with gray shadings.
Parameterized SST cooling induced by a TC. SST cooling (SSTC) induced by a TC could impose a significant impact on the its intensity. Although a coupled atmosphere–ocean model could provide such SSTC, various challenges associated with the coupled models lead to the continuous use of atmosphere-only models. In such a case, how the SSTC is included in the model simulation and prediction of TCs is an interesting topic to pursue. In this regard, we have tried to fill this gap by constructing a fast, robust, and effective parameterization scheme for TC-induced SSTC that can be used in atmosphere-only TC models without the need to couple with an ocean model.
Figure 8. Flow chart of the parameterization scheme for Sea surface temperature cooling (SSTC). The red variables are the input parameters and the blue variables are output from the parameterization scheme, the green is for explanation of the processes involved in the parameterization scheme, and the purple means the list of symbols.
To achieve this goal, the following three steps are taken (see Figure 8): (i) results from an idealized ocean simulation together with temperature budget analysis is analyzed to isolate each major mechanism causing TC-induced SSTC, which is then used as a basis for the parameterization; (ii) based on the idealized ocean simulation, a new SSTC parameterization scheme, including vertical mixing, advection, and SST recovery processes under the influences of sea surface height anomalies and ocean subsurface temperature, is developed; and (iii) this SSTC parameterization scheme is evaluated through numerical simulations of Typhoon Matsa (2005) and validated against remote sensing data. Results show that with the application of this parameterization scheme, significant improvements in the simulated TC intensity and SST changes can be achieved (Figure 9, see Liu et al. 2018 for more details). The validity of the parameterized SSTC was further evaluated using 57 drifter data under the influence of 9 TCs in 2016 (Figure 10) and promising results. Therefore, the parameterization scheme is indeed compatible with any weather prediction model for TC forecasting.
Figure 9. Comparison of the (left) SST cooling from the parameterized run and (right) the observed from TMI (shaded blue, °C) valid at (top) 0000 UTC 5 August and (bottom) 0000 UTC 7 August. Land masses are shaded in yellow. Black lines in (a) and (c) and (b) and (d) denote the simulated track and Japanese Meteorological Agency track of Matsa (2005), respectively. TMI means the Tropical Rainfall Measuring Mission/Microwave Imager.
Figure 10. Scatterplots of SSTC (°C) from 57 drifters versus SSTC (°C) calculated from the parameterization scheme for 9 TCs in 2016. R is the correlation coefficient between the two variables with the statistical significance at an above 95% confidence level. SSTC = TC‐induced SST cooling.
4. Planetary boundary layer processes and their parameterizations
The planetary boundary layer (PBL) determines momentum, heat and water vapor exchanges and vertical mixing between the underlying surface and the atmosphere above. Surface fluxes and vertical mixing in the boundary layer play a key role in determining TC intensity and its change. Previous studies have illustrated the sensitivity of TC simulation and prediction to PBL parameterization. Therefore, it is key to better representation of PBL processes in numerical models used for TC forecasts. To accurately simulate the atmospheric state in the boundary layer by a PBL parameterization scheme in different regions with their dominant weather/climate regimes is important for global/regional atmospheric models. Under the support of this project, we have performed large-eddy-simulations (LESs) of TCs to understand the turbulent processes in TC boundary layer and their representation in numerical models at different horizontal resolutions. We have also implemented and evaluated an advanced turbulent closure scheme, called the E-ε closure scheme and tested its skill in simulating a record-breaking TC.
Large-eddy-simulation (LES) of TC boundary layer. In view of the increasing interests in the explicit simulation of fine-scale features in TC boundary layer (TCBL), the effects of horizontal grid spacing on the 7~10 hours simulation of an idealized TC were examined using the Weather Research and Forecast (ARW-WRF) mesoscale model with one-way moving nests and nonlinear backscatter with anisotropy (NBA) sub-grid-scale (SGS) scheme. The simulations with the LES resolution (e.g., 166 and 55 m) were compared with the relatively coarser resolutions (500 m and 2 km). In general, reducing the horizontal grid spacing from 2 km to 500 m tends to produce a stronger TC with lower minimum sea level pressure (MSLP), stronger surface winds, and smaller TC inner core size. However, large eddies cannot be resolved at these grid spacings. In contrast, reducing the horizontal grid spacing from 500 m to 166 m and further to 55 m leads to a decrease in TC intensity and an increase in the inner-core TC size (Figure 11). Moreover, although the 166-m grid spacing starts to resolve large eddies in terms of TCBL horizontal rolls and tornado-scale vortex, the use of the finest grid spacing of 55 m tends to produce shorter wavelength of the turbulent motion and stronger multi-scale turbulence interaction (Figure 12).
Figure 11. Time series of the simulated TC (A) minimum sea level pressure (MSLP), (B) maximum 10-m wind speed (Vmax) and (C) the corresponding radius of maximum wind speed (Rmax) in different grid spacing domains. (D) As (B), but after runs have been averaged to grids with a common grid spacing of 2 km.
Figure 12. Horizonal cross-section of instantaneous 10-m wind speed (m s？1) at t = 9 h of the simulation from the grid spacings of (A) 2 km, (B) 500 m, and (C) 166 m with the domain size the same as d05; (D) and (E) are as in (C) but with the domain size the same as d06. Dash and solid circles indicate the radii of 30 and 60 km from the TC center, respectively.
Figure 13. Vertical profiles of (A) SGS, (B) resolved and (C) total turbulent vertical momentum fluxes at t = 9 h of the simulation from the grid spacing of 500, 166, and 55 m, averaged in the radius of 41.25 km from the storm center (namely, in the d06 domain).
There are two components of the vertical momentum fluxes from, respectively, the parameterized sub-grid scale (SGS) and the resolved eddy processes (Figure 13). The resolved fluxes are estimated by the resolved turbulences with scales smaller than 2 km. The SGS fluxes directly output by WRF model. The SGS momentum fluxes in the TCBL decrease as the grid spacing decreases. The SGS momentum fluxes in the 55-m domain is very smaller than those in the 166-m and 500-m domains. In contrast to the SGS vertical momentum fluxes, the vertical momentum fluxes by the resolved eddy motions increase as the grid spacing decreases. Importantly, the total vertical momentum fluxes also decrease as the grid spacing decreases in the boundary layer. For example, the vertical momentum fluxes in the boundary layer from the 55-m grid spacing is much larger than that from the 500-m grid spacing, but show less discrepancies between the heights of 50 and 200 m. Furthermore, although the difference in the total vertical momentum fluxes between the 166-m and 55-m grid spacings are much closer, the 55-m grid spacing still produces slightly larger vertical momentum fluxes in middle and smaller in the lower boundary layer. This means that the relatively coarse horizontal resolution may considerably underestimate (overestimate) the middle (lower) boundary layer vertical mixing in the simulated TCs.
The results from our comparison demonstrate that a grid spacing of sub-100-meters is desirable to produce more detailed and fine-scale structure of TCBL horizontal rolls and tornado-scale vortices, while the relative coarse grid spacing of sub-kilometer (e.g., 500m) is more cost-effective and feasible for research that is not interested in the turbulence processes and for real-time operational TC forecasting in the near future.
Implementation and evaluation of the E-ε turbulent closure scheme. Under the support of this project, we have implemented and tested the turbulence kinetic energy (TKE) and TKE dissipation rate (ε) based 1.5-order closure PBL parameterization (E-ε, EEPS) in the Weather Research and Forecasting (WRF) model. The performances of the newly implemented EEPS scheme and the commonly used Yonsei-University (YSU) scheme, the University of Washington (UW) scheme, and Mellor-Yamada-Nakanishi-Niino (MYNN) scheme were evaluated over the stratocumulus dominated Southeast Pacific (SEP) and over the Southern Great Plains (SGP) where strong PBL diurnal variation is common. The simulations by these PBL parameterizations were compared with various observations from two field campaigns: the VAMOS (the Variability of the American Monsoon Systems Project) Ocean-Cloud-Atmosphere-Land Study (VOCALS) in 2008 over the SEP and the Land-Atmosphere Feedback Experiment (LAFE) in 2017 over the SGP. Results show that the EEPS and YSU schemes perform comparably over both regions, while the MYNN scheme performs differently in many aspects, especially over the SEP. The EEPS (MYNN) scheme slightly (significantly) underestimates liquid water path over the SEP. Compared with observations, the UW scheme produces the best PBL height over the SEP (Figure 14). The MYNN produces too high PBL height over the western part of the SEP while both the YSU and EEPS schemes produce too low PBL and cloud top heights. The differences among the PBL schemes in simulating the PBL features over the SGP are relatively small (Figure 15).
Figure 14. Box-and-whisker plots showing cloud top height (CTH; unit: meters, upper panel) and liquid water path (LWP, only under cloudy conditions larger than 1 g m-2) from observations and simulations with four different PBL schemes. From the bottom to the top, each box-and-whisker plot shows the 5th, 25th, 50th, 75th, and 95th percentiles of CTH. The black dots are mean values. The selection of the simulated grid cells follows the time and locations of observations.
Although the four schemes perform similarly in many aspects over the SGP, the EEPS scheme captures the slightly unstable lower part and weakly stable upper part of the CBL, which is consistent with observations. Therefore, results presented in this study suggest that the EEPS scheme is an alternative option in mesoscale models. It is worthy of noting that the TKE dissipation rate (ε) from the EEPS scheme is an independent variable and is important for many applications, such as comparing the prognostic ε with the diagnostic ε in other PBL schemes and evaluating the effect of dissipative heating on TCs, such as in a case we will show below. In addition to the importance of the TKE dissipation rate for estimating dissipative heating in TCs, the directly simulating ε is also an advantage of the use of the EEPS scheme in high-resolution application models.
Figure 15. The observed TKE from Doppler Lidar (a) and the EEPS scheme simulated TKE (b) and EDR (c) in time and height space. The horizontal axis is local time (LT, hour/day). The red dots represent the time of sunrise and the black dots represent the time of sunset.
We also compared the track and intensity of Hurricane Patrcia (2018) simulated with the EEPS scheme and those with YSU and MYNN PBL schemes in Figure 16. Note that Hurricane Patricia is the most intense TC on record so far on the earth. We can see from Figure 16 (left panel) that the track simulaiton is not too sensitive to the use of the PBL scheme with similar track errors but the bias from EEPS scheme is slightly smaller at the later simulation stage. In sharp contrast, the intensity simulated with the EEPS scheme reproduced the rapid intensification and the final intensity reasobable well in terms of the near-surface wind speed (right panel in Figure 16). Note also that in all three simulations, only the simulaiton with the EEPS scheme captured the timing of peak intensity of the TC, consistent with observations.
Figure 16. The observed and simulated track and intensity for Huricane Patricia (2018). The simulaitons with three PBL schemes are compared as marked in the figure.
Figure 17. The observed and simulated radar reflectivity for Hurricane Patricia (2018), showing the realistic simulation of the secondary eyewall structure with the EEPS PBL scheme only.
The weakening following the peak intensity of the storm in observation resulted from the formation of a secndary eyewall and the associated eyewall replacement and followed by landfall. As we show in Figure 17, the EEPS scheme simulated the observed secondary eyewall processes as in observations although the secondary eyewall had a relatively smaller size than the obserged. However, other two schemes failed to capture such a secondary eyewall formation and thus faield to simulated the timing of the intensifty peak. The better simulation of the storm intensity with the EEPS partially attributed to the inclusion of dissipative heating related to the TKE dissipation rate.
5. Scale-aware convective parameterization and evaluation
In addition to the discontinuity of model resolution between different grids, the impact of different grid spacings is not considered in the model physics, which leads to computational discontinuity and negative effects in simulations and forecasts with multiple nested models. Meanwhile, in the real atmosphere, model physical processes are highly dependent on both the space and time scales, especially convective processes. It is generally considered that scale-aware cumulus convective parameterization should be studied from scale adaptability, non-local, non-equilibrium, and random processes in weather and climate models.
An important aspect of this project is to develop a scale-dependent convective parameterization scheme suitable for TC numerical prediction models. As a first step, we have evaluated and improved the quasi-equilibrium closure assumption based on Meso-SAS convection parameterization scheme, and established a scale-aware closure scheme considering the effect of model resolution. Considering the principle of cumulus convection parameterization, it is used to describe the sub-grid processes which the model cannot be explicitly resolved. As the model resolution increases, its contribution should decrease accordingly. To achieve this goal, a scale-related parameter X is introduced to adjust convention trigger and closure as a function of the mode resolution.
Based on a case study of Typhoon Chan-hom (2015), we compared the effect of X at 60,120 and 250 [km (m/s)] on 24-hour precipitation prediction (Figure 18). In the original Meso-SAS simulation, the typhoon primary rainband, secondary spiral rainband and heavy precipitation center in the inner-core region was simulated, and the spiral rainband in the periphery was mainly from parameterized convective precipitation. Although there is some parameterized convective precipitation in the inner-core area, the heavy precipitation was mainly produced from the resolved microphysical processes.
Figure 15. 24HR forecast of precipitation rate (mm/hr) for Typhoon Chan-hom (2015) Meso-SAS: (a) total precipitation rate, (b) convective parameterized precipitation rate, and (c) microphysical precipitation rate. X=60: (d) total precipitation rate, (e) convective parameterized precipitation rate, (f) microphysical precipitation rate; X=120: (g) total precipitation rate, (h) convective parameterized precipitation rate, (I) microphysical precipitation rate; X=250: (j) total precipitation rate, (k) convective parameterized precipitation rate, and (l) microphysical precipitation rate
When X was taken to be 60, compared with the original Meso-SAS scheme, the distribution of typhoon precipitation is basically similar, but the outer spiral rainband became stronger and narrower, the heavy precipitation center in the inner-core region was significantly enhanced, the parameterized convective precipitation was mainly distributed in the periphery of the typhoon, and the proportion of the resolved microphysical precipitation in the inner-core region was increased significantly. When X was increased to 120, the spiral rainband in the left quadrant of the typhoon became narrower and stronger, while the spiral rainband in the right quadrant basically disappeared, and was replaced by strong and concentrated precipitation in the inner-core region. When X was further increased to 250, the parameterized convective precipitation became extremely weak, and the typhoon precipitation was almost produced by resolved microphysical processes. As a result, the distribution of typhoon precipitation was dominated by heavy precipitation near the inner-core region, with no strong spiral cloud band structure, which is significantly different from the precipitation distribution in mature typhoons.
The above results suggest that the scale-dependent adjustment parameter X can effectively control the contribution of the parameterized convective scheme to the moist process in model prediction and can be used as an effective way to develop a scale-aware parameterized convective scheme. However, the proposed scheme needs to be tested with a large number of simulations with a two-way nested model. In that case, the different values of X in different grid meshes would be used to ensure the continuity of the prediction variables at the grid interface(s). The efforts are being proceeded to further improve the scale-aware convective parametrization scheme and systematically evaluate the scheme based on both idealized and real-case TC simulations/predictions in the subsequent study. Such accomplishment is expected to improve 3-7 day forecasts of TCs with two-way interactive multi-nested models.
6. Summary
The main research objectives of this subproject are to develop the dynamical TC-vortex initialization technique, scale-aware cumulus convective parameterization scheme, new PBL parameterization, and air-sea momentum and enthalp exchange algorithms. The main efforts are to consdier the scale adaptivity, nonlocal and non-equilibrium, and random nature of atmospheric physical processes, such as the turbulence and large eddies in TC boundary layer, and the sub-grid scale cumulus convection. We have attempted to develop and evaluate the scale-aware cumulus onvective parameterization, to optimize and improve the PBL parameterization scheme by including the effects of boundary layer rolls and low-level strong vertical wind shear on TC intensity and structure and parameterization, and to develop simple parameterization to take into account of SST cooling induced by a TC so that the coupled ocean-atmospehre processes can be roughly included in atmospehric-alone numerical models for TC prediction. Our goal is to develop scale-aware schemes/algorithm suitable for the use in multiple nested-mesh models with various resolutions of 2-25 km grid spacings.
As briefly introduced above, through this project, we have developed a dynamical TC-vortex initialization (DVI) scheme that includes the effect of mesoscale terrains, a parameterized air-sea coupled feedback induced SST cooling for improving the surface enthalpy flux under a TC, tested the dependence of sea spray effect on surface enthalp flux and TC intensification on the surface drag parameterization, evaluated the performance of the earlier PBL scheme, which incudes an explicit prognostic euqation for the turbulent kinetic energy (TKE) dissipation, namely the called E-ε scheme, and its application to the simulation of Hurricane Patricia, implemented and tested a scale-aware cumulus parameterization scheme for TC forecasting. We have shown that the PBL E-ε scheme has some strength in simulating TC intensity and structure changes, including the formation of the concetnric eyewall structure. In addition, the inclusion of dissipative heating due to the TKE dissipation contributed to the rapid intensification and the record-breaking intensity of Hurricane Patricia. The all components developed and tested so far are being integrated into the T-RAPS. Systematic evaluation of the developed physical parameterizations will be conducted for TC hindcats/forecasts with more real TC cases and evetnaully used for real-time forecasting of TCs.
References cited and supported by the project
Liu, H.-Y., Y. Wang, J. Xu, and Y.-H. Duan, 2018: A dynamical initialization scheme for tropical cyclones under the influence of terrain. Weather and Forecasting, 33, 641–659, https://doi.org/10.1175/WAF-D-17-0139.1.
Liu, X., D.-L. Zhang, and J. Guan, 2019: Parameterizing sea surface temperature cooling induced by tropical cyclones: 2. Verification by ocean drifters. Journal of Geophysical Research – Oceans, 124. 1215–1231, https://doi.org/10.1029/2018JC014118.
Liu, X., J. Wei, D.-L. Zhang, and W. Miller, 2019: Parameterizing sea surface temperature cooling induced by tropical cyclones: 1. Theory and An application to Typhoon Matsa (2005). Journal of Geophysical Research – Oceans, 124, 1215–1231, https://doi.org/10.1029/2018JC014117..
Xu, H.-X., and Y. Wang, 2021: Sensitivity of fine-scale structure in tropical cyclone boundary layer to model horizontal resolution at sub-kilometer grid spacing. Frontiers in Earth Science, 9:707274, doi:10.3389/feart.2021.707274.
Zhang, C.-X., Y. Wang, and M. Xue, 2020: Evaluation of an E-ε and three other boundary layer parameterization schemes in the WRF model over the Southeast Pacific and the Southern Great Plains, Monthly Weather Review, 148(3), 1121–1145, https://doi.org/10.1175/MWR-D-19-0084.1.
Zhang, Z., Y. Wang, W.-M. Zhang, and J. Xu, 2019: Coastal ocean response and feedback to Typhoon Hato (2017) in the South China Sea: A coupled model study. Journal of Geophysical Research – Atmospheres., 124(24), 13731-13749, https:/doi.org/10.1029/2019JD031377.