TMCnet News
Impacts of 4DVAR Assimilation of Airborne Doppler Radar Observations on Numerical Simulations of the Genesis of Typhoon Nuri (2008) [Journal of Applied Meteorology and Climatology](Journal of Applied Meteorology and Climatology Via Acquire Media NewsEdge) ABSTRACT The Weather Research and Forecasting Model and its four-dimensional variational data assimilation (4DVAR) system are employed to examine the impact of airborne Doppler radar observations on predicting the genesis of Typhoon Nuri (2008). Electra Doppler Radar (ELDORA) airborne radar data, collected during the Office of Naval Research-sponsored Tropical Cyclone Structure 2008 field experiment, are used for data assimilation experiments. Two assimilation methods are evaluated and compared, namely, the direct assimilation of radar-measured radial velocity and the assimilation of three-dimensional wind analysis derived from the radar radial velocity. Results show that direct assimilation of radar radial velocity leads to better intensity forecasts, as this process enhances the development of convective systems and improves the inner-core structure of Nuri, whereas assimilation of the radar-retrieved wind analysis is more beneficial for tracking forecasts, as it results in improved environmental flows. The assimilation of both the radar-retrieved wind and the radial velocity can lead to better forecasts in both intensity and tracking, if the radial velocity observations are assimilated first and the retrieved winds are then assimilated in the same data assimilation window. In addition, experiments with and without radar data assimilation led to developing and nondeveloping disturbances in numerical simulations of Nuri's genesis. The improved initial conditions and forecasts from the data assimilation imply that the enhanced midlevel vortex and moisture conditions are favorable for the development of deep convection in the center of the pouch and eventually contribute to Nuri's genesis. The improved simulations of the convection and associated environmental conditions produce enhanced upper-level warming in the core region and lead to the drop in sea level pressure. (ProQuest: ... denotes formulae omitted.) 1. Introduction Initial conditions have substantial impacts on the numerical prediction of tropical cyclone (TC) genesis. Davis and Bosart (2002) suggested that the numerical simulation of the genesis of Hurricane Diana (1984) is very sensitive to the specification of the upper-level trough and ridge in the initial conditions. Specifically, with a preexisting upper-level disturbance in the initial conditions, the simulation predicts Diana's genesis. In contrast, when the upper-level trough and ridge are removed from the initial conditions, the simulation fails to produce Diana's genesis. Kieu and Zhang (2010) demonstrated the importance of a mesoscale vortex (length scale ;500 km) in the initial conditions for simulating the genesis of Tropical Storm Eugene (2005). They found that Eugene's genesis is successfully simulated through the merging of two preexisting vortices that originate from the initial conditions. However, the simulated Eugene is unable to intensify to tropical storm strength after the removal of one of the vortices associated with the intertropical convergence zone breakdown. Nolan (2007) also noted that the initial vortex structure has a large impact on TC genesis simulation: a deeper vortex from the surface to the middle levels in the initial conditions tends to produce an earlier TC genesis than does a shallower vortex in the initial conditions. In addition to the sensitivity to the flow fields associated with the dynamic conditions (e.g., the upper-level trough and ridge and the mesoscale vortex), the thermodynamic aspects of the initial conditions are also important for TC genesis. Both temperature and moisture profiles in the initial conditions can considerably influence the simulation of TC genesis through changing the vertical distribution of the vertical mass flux in the convection (Raymond and Sessions 2007). Reducing the relative humidity 4 km above the ground level (AGL) in the initial conditions can greatly delay the time of TC genesis (Nolan 2007). In light of the significant influence of the initial conditions on numerical simulations of TC genesis, it is important to improve the initial conditions with data assimilation (DA). However, conventional observations are usually rather sparse in terms of both temporal and spatial resolutions over the ocean, where TC genesis commonly occurs. Remotely sensed observations from satellite and radar become important data sources. Previous studies (Pu and Zhang 2010; Liu et al. 2012) have demonstrated that assimilating satellite observations into numerical weather prediction models can have positive impacts on tropical cyclone genesis forecasts, mainly because of improvements in the TC genesis environmental conditions. An attempt has also been made to assimilate cloud radiances over the core region of a TC (e.g., M. Zhang et al. 2013) in order to improve intensity forecasts, but this method has been tested for mature hurricane only. In past years, airborne Doppler radar has demonstrated the advantage of its mobility in sampling detailed TC structural features (Marks and Houze 1987; Reasor et al. 2000; Houze et al. 2007) near the TC core and associated major convective systems. Previous studies have indicated that assimilating these Doppler radar observations near the TC core region into numerical models could improve the numerical simulation and prediction of mature TCs (Pu et al. 2009; Xiao et al. 2009; Zhang et al. 2009; Zhang et al. 2012). Although in some cases the TC forecasts can be improved through downstream effect of assimilating data outside the core region (Wang and Huang 2012), assimilation of observations over the TC inner-core region can still lead to significant improvements in the forecast skill (Zhang et al. 2011). Despite many previous efforts in data assimilation, research on the assimilation of airborne observations (especially radar data) is rarely conducted for TC genesis cases. Fortunately, during the World Meteorological Organization's The Observing System Research and Predictability Experiment (THORPEX) Pacific Asian Regional Campaign (T-PARC) and the Office of Naval Research- sponsored Tropical Cyclone Structure 2008 (TCS-08) field experiment, airborne radar observations were collected during the genesis phase of typhoons. These airborne Doppler radar data provide important information for understanding TC genesis (e.g., Bell and Montgomery 2010). They also provide a good opportunity to examine the impact of radar data assimilation on the numerical simulation of TC genesis. In this study, we examine the impacts of assimilating radar observations on predicting TC genesis with Typhoon Nuri (2008), one of the major typhoons during T-PARC/TCS-08. Several ways of assimilating the radar radial velocities in the TC genesis environment are evaluated: the assimilation of the retrieved radar wind analysis, the direct assimilation of the radar radial velocity, and the assimilation of both the radar wind analysis and the radial velocity. In addition, the influences of radar data assimilation will be evaluated to reveal the beneficial conditions that are important for the genesis of Typhoon Nuri. The paper is organized as follows. Section 2 introduces the airborne Doppler radar observations taken during the field experiment, the radar data quality control, and the wind analysis algorithm. Section 3 describes the numerical experiments including the numerical model, data assimilation systems, and experimental configurations. Then, the impact of the radar data assimilation on the numerical simulations of Nuri is evaluated through the presentation of the numerical results in section 4. The implications of improved initial conditions and forecasts for the factors that contribute to Nuri's genesis are examined in section 5. The summary and concluding remarks are presented in section 6. 2. Brief overview of Typhoon Nuri and ELDORA observations a. A brief overview of Typhoon Nuri Typhoon Nuri was initiated in the form of an easterly wave over the ocean with high SSTs (above 268C). The Joint Typhoon Warning Center (JTWC) began tracking the pre-Nuri disturbance at 0000 UTC 16 August 2008 with an initial location of 13.28N, 146.88E. At that time, a low-level circulation center formed to the east of Guam as the easterly wave increased its low-level vorticity. With a well-organized convective system and low vertical wind shear, the system reached tropical depression intensity with a maximum surface wind speed (MSW) of 12.8m s21 and a minimum sea level pressure (MSLP) of 1004 hPa at 1800 UTC 16 August. Then, it moved farther westward and formed a tropical storm with an MSW of 18.0 m s21 and an MSLP of 996 hPa at 1200 UTC 17 August 2008. Twenty-four hours later (1200 UTC 18 August), it developed rapidly, achieving typhoon strength and acquiring the name Nuri. The central MSLP reached 974 hPa. In this study, the genesis time of Nuri is defined as the time at which the JTWC designated it as a tropical depression, namely, at 1800 UTC 16 August. Numerical simulations are conducted between 0000 UTC 16 August and 0000 UTC 18 August 2008, which covers the period of Nuri's genesis and rapid intensification. b. ELDORA observations During T-PARC/TCS-08, the Naval Research Laboratory (NRL) P-3 aircraft carried the National Center for Atmospheric Research's (NCAR) Electra Doppler Radar (ELDORA) to obtain the observations, including radar reflectivity and radial velocity. The research flight was operated from 2300 UTC 15 August to 0350 UTC 16 August around Guam (13.458N, 144.788E), targeting the pre-Nuri disturbance. During this period, Nuri was in its pregenesis phase, with 15-18 h remaining until its genesis time (1800 UTC 16 August). The P-3 aircraft flew at a height of about 2.4 km for most of this mission. Two X-band radars with forward- (fore) and backward- (after) detecting beams conically scanned the convective systems at the same time. ELDORA was configured with an unambiguous range of 75 km, meaning that the radar measurements were made in a 150-km swath, centered on the P-3 flight track. More radar configurations and flight mission information can be found in Raymond and López Carrillo (2011) and Raymond et al. (2011). The raw radar data were first processed with a quality control procedure, using the Solo software package (Oye et al. 1995; Bell et al. 2013). For automatic quality control, we followed the method introduced by Wolff et al. (2009) to remove spurious echoes in the ELDORA data. Following quality control, the radial velocity was available to be directly assimilated into the simulations of Nuri's genesis in the following experiments. A radar wind analysis was also conducted to obtain the three-dimensional wind field in order to fulfill diagnosis and analysis purposes. c. Radar wind analysis Since radar-measured radial velocities are not common variables for synoptic analysis, it is very common to derive u and y wind components from radar-measured radial velocities in research and operational practice. The outcomes of the three-dimensional wind retrieval provide three-dimensional wind fields in a conventional format (i.e., u, y, and w wind components), making the radar observations more convenient for diagnosing kinematic characteristics during TC formation and evolution. In this study we will also assimilate radar-derived u and y wind components and compare the results using direct assimilation of radar-measured radial velocities. From the quality-controlled radar radial velocity, the three-dimensional wind field can be retrieved using a variational approach (Reasor et al. 2009). This variational minimization matches the observed Doppler information to the gridded three-dimensional wind field. This scheme enforces the anelastic mass continuity equation, with vanishing second derivatives of the wind field and vertical velocity boundary conditions as weak constraints (Gao et al. 2001; Gamache et al. 2008). During the retrieval process, 4.5-h ELDORA observations as mentioned above are divided into nine legs according to the observing time order: each leg contains a half-hour of airborne radar data. The radar observations in each respective leg are interpolated onto a gridded mesh with a horizontal grid spacing of 2 km and a vertical interval of 250 m. In this case, the lowest level of the gridded domain is at a height of 0.5 km and the highest level extends to 14.5 km. Most of the available radar data are distributed between heights of 0.5 and 10 km. With the interpolated radial velocities, the three-dimensional wind analysis for each half-hour leg is produced. Once the nine sets of half-hour retrieved wind analyses are produced using this variational analysis scheme, the horizontal winds from the retrieved wind analysis are assimilated into the simulations of Nuri's genesis in the following experiments. 3. Numerical experiments a. WRF model An advanced research version of the Weather Research and Forecasting (WRF) Model (Skamarock et al. 2008) is employed for numerical simulation. A total of 30 terrain-following s levels in the vertical coordinate are used, with the top of the model at 5 hPa. A two-way interactive technique is used for three-level nested domains (D01, D02, and D03) with horizontal grid spacings of 36, 12, and 4 km, respectively. D03 is a movable domain that follows the movement of Typhoon Nuri. Figure 1 shows the locations of the model domains. For all simulations, model integration starts at 0000 UTC 16 August 2008 and ends at 0000 UTC 18 August 2008. Unless otherwise noted, all simulation results in this paper are from the D03 domain (4-km grid resolution). The physics scheme options are as follows: the WRF double-moment five-class (WDM5) cloud microphysics scheme (Lim and Hong 2010), the Rapid Radiative Transfer Model for longwave radiation with six molecular species (Mlawer et al. 1997), the Dudhia shortwave radiation scheme (Dudhia 1989), a modified version of the Kain-Fritsch cumulus parameterization scheme (Kain and Fritsch 1993), and the Yonsei University planetary boundary layer parameterization with the Monin-Obukhov surface layer scheme (Hong et al. 2006). b. WRF 4DVAR The four-dimensional variational data assimilation (4DVAR) system in version 3.4.1 of the WRF data assimilation (WRFDA) is used in this study. A prototype version of the WRFDA 4DVAR system was introduced by Huang et al. (2009). More recent upgrades and software improvements in WRFDA are described in Zhang et al. (2012). The 4DVAR system has recently been used for ground-based radar data assimilation by Wang et al. (2013) and Sun and Wang (2013). Generally, the new analysis X is generated in a 4DVAR system through the minimization of a predefined cost function: ... The cost function measures the misfit between the model variables and both the background (first guess) Xb (i.e., Jb term) and the observations y (i.e., Jo term). In the background term (the first term on the rhs of the equation), the analysis vector Xi denotes an intermittent analysis after the ith outer loop. The outer loop index i varies from 1 to n , where n is the desired total number of the outer loop. The final analysis of WRF 4DVAR after the last (nth) outer loop is denoted as Xn or equivalently Xa. The analysis vector from the last second outer loop is Xn21. Normally, Xb is taken as the first guess X0. In the observation term (the second term on the rhs of the equation), the whole assimilation time window is split into K observation windows. Here, dk 5 yk 2 Hk [Mk (Xn21)] is the innovation vector for the observation wind k. The nonlinear and tangent linear observation operators transforming variables from model space to observation space are Hk and Hk. In the cost function, Mk and Mk are the nonlinear and tangent linear models, respectively, propagating the guess vector Xn21 and analysis increments Xn 2 Xn21 from the first to the kth observation time window. After cost function minimization is achieved, the analysis is fitted to the background and the observations according to their weights: the estimated background error covariance matrices B and the estimated observational error covariance matrices R.Inthis study, the background error covariance matrices B use the default setting in WRFDA, which is estimated by the so-called National Meteorological Center [NMC, now known as the National Centers for Environmental Prediction (NCEP)] method (Parrish and Derber 1992). Since the radar wind retrieval process does not provide the wind observational error, the default NCEP errors for the wind profiler observations are given to estimate the observational error covariance matrices R for the assimilation of the retrieved winds (except another assignment of the observation error is given in some experiments). As for the direct assimilation of the radial velocity, the standard deviation of the observation error for the radial velocity is set to 2ms21.Thiserrorvalueissimilartothevaluesusedin Dowell and Wicker (2009), Xiao et al. (2009),andLi et al. (2012). c. Experimental setup and data quality control To examine the impact of radar observations with the 4DVAR method, four experiments are conducted (see Table 1): a control experiment (Ctrl) without radar data assimilation, an experiment that assimilates the radarretrieved wind (UV), an experiment that directly assimilates radial velocity (RV), and an experiment that assimilates both radar-retrieved wind and radial velocity (RV-UV). The simulation without radar data assimilation (Ctrl) is conducted with initial and lateral boundary conditions derived from the NCEP Global Forecast System Final Analysis (FNL) at 18318 horizontal resolution. The airborne Doppler radar data were not assimilated in FNL. As mentioned above, according to data availability, the time window of the 4DVAR is 5 h, from 0000 to 0500 UTC 16 August 2008. The tangent linear model is fairly consistent with the nonlinear model in the short range (e.g., 5-h window). An early experiment shows that the use of the 5-h assimilation window led to better data assimilation and forecast results, compared with the cycled data assimilation experiments with short (e.g., 1 or 2 h) assimilation windows. Figure 2 shows a sample of the radial velocity and horizontal winds in the retrieved wind analysis around 0130 UTC 16 August 2008. Because of concerns about the data quality of the retrieved winds and because fewer samples are available at upper levels, winds above 10 km AGL are not used in experiment UV. In addition, before the data assimilation experiments, a data-thinning process is deployed for the wind observations: the horizontal resolution of the retrieved wind analysis is thinned on the 36-km gridforD01,12-kmgridforD02,and4-km gridfor D03; the vertical levels of the retrieved wind analysis are selected to match the heights of the model vertical levels. The innovation vectors (i.e., observation minus background) are checked for quality control. The examination reveals that radial velocity observations near the flight track are always larger than the radar radial velocities in the background by about 5 m s21. The reason for these large discrepancies may be that precipitation with a large downward motion has a considerable influence on the radial velocity measurements when the radar beam is directed downward vertically along the flight track. Such a large departure between the observations and the background is partly artificial; thus, the data assimilation system may not appropriately handle and assimilate these radial velocities. A test experiment including these radial velocities indicates that data assimilation results in degraded forecasts for both Nuri's intensity and track. Considering the difficulty the data assimilation system has with assimilating these data well, the radial velocity data near the flight track (inside 8 km) are all removed before data assimilation. To make a fair comparison with the assimilation of radar-derived winds, data in the same area are also removed in the UV experiment. 4. The impact of radar data assimilation on intensity and track forecasts a. Assimilation of radial velocity versus retrieved winds The impact of radar data assimilation on the storm track and intensity forecasts is first examined. The simulation results are compared with the JTWC best-track data. In Fig. 3a, all the experiments, including the control experiment and the data assimilation experiments (UV and RV), are initialized at 0000 UTC 16 August 2008 with a similar weak storm intensity (the initial MSLP in all experiments is about 1010 hPa). In the following 48-h forecasts (Fig. 3a), Ctrl does not predict Nuri's genesis, with a slight drop in MSLP (less than 5 hPa). In contrast, with radar data assimilation, UV and RV both predict significant enhancement in Nuri's intensity. Specifically, RV results in a greater degree of improvement in the intensity forecasts, compared with UV. Figure 3b shows Nuri's track forecasts from the control experiment and the data assimilation experiments. Compared with the best-track data, Ctrl produces a southeastern bias for most of the simulation period. At the end of the simulation (48 h), a large track error with an eastern bias appears. Assimilation of the radar wind analysis (UV) leads to a notable improvement in the track forecasts: the track shifts to the northwest from 12 to 36 h and to the west after 36 h. Overall, UV predicts the track better than Ctrl does. The track error is significantly reduced by the assimilation of the radar wind analysis (Fig. 4). In contrast, experiments assimilating radar RV result in less impact on the track forecasts (Fig. 3b) than does assimilating the radar wind analysis in UV. The track forecasts show a northern bias and slower westward movement. Since the results so far indicate that UV predicts more accurate track forecasts, while RV produces better intensity forecasts, we further investigate the different processes in experiments UV and RV that lead to the different results in the track and intensity forecasts. As the track forecasts are mostly controlled by the largescale flow, it is necessary to check whether more improvement in the large-scale flow conditions results in the better track forecasts in UV. Figure 5 compares the surface wind data from data assimilation experiments with the National Aeronautics and Space Administration (NASA) Quick Scatterometer (QuikSCAT) satellite derived wind vectors over the ocean surface around 0830 UTC 16 August 2008. It is apparent that experiment UV reproduces more realistic environmental flow in terms of the large-scale wind speed and circulation, compared with QuikSCAT data and experiment RV. Since westerly flow dominated the lower atmosphere around pre-Nuri during that time, Fig. 6 illustrates the averaged u-component wind at 800 hPa within a radius of 600 km from the center of the simulated Typhoon Nuri. It shows that experiment UV produces a greater speed of the averaged westward wind (negative value of the u-component wind), compared with Ctrl and RV. The stronger large-scale westward flow in UV leads to the faster westward movement of the simulated Typhoon Nuri, and more track errors (mostly because of the eastward bias) are reduced in UV than in RV. The analysis of the large-scale flow conditions indicates that the assimilation of the wind analysis in UV leads to more improvement in the large-scale flow conditions and more accurate track forecasts. This is because the retrieval process can filter small-scale information and provide more large-scale information. Specifically, as mentioned in section 2c, three-dimensional variational analysis of radar wind is constrained by a mass continuity equation. This will help in the retrieval process by constraining the consistency of the vertical motion but also by filtering out some small-scale signals. In addition, the threedimensional wind analysis is down to a 30-min window, implying a temporal average of the wind information, which also filters out some small-scale variations. Section 5b will further illustrate that the assimilation of radial velocity provides more beneficial information at the convective scale and thus leads to improvements in the intensity forecasts. b. Assimilation of both radial velocity and radar wind analysis As discussed above, although retrieved winds are derived from the radar radial velocity, they represent different aspects of information in terms of temporal and spatial scales due to smoothing and constraints used in the retrieval process. In addition, results from RV and UV indicate that the large impacts of radar radial velocity and retrieved winds are from different aspects of TCs (i.e., their track or intensity). Therefore, it should be advantageous to combine retrieved winds and radial velocity measurements in the data assimilation, so additional experiments RV1-UV2, UV1-RV2, and UV-RV were conducted (see Table 1). In RV1-UV2, the radial velocity information is first assimilated into the assimilation window (0000-0500 UTC 16 August 2008), and the outcome at the initial time then provides a first guess for assimilating the retrieved winds in the same window. It is apparent that the result produces a significant enhancement in the intensity and track forecasts (Fig. 7). Specifically, the underestimation in the intensity forecasts is largely removed in RV1-UV2 as compared with UV and RV (Fig. 7a). In Fig. 7b, RV1-UV2 improves the track forecasts (relative to Ctrl and RV) by increasing the speed of the westward movement. The track error is significantly reduced in RV1-UV2 in the late phase of the forecast (after 36h) as compared with Ctrl and RV (Fig. 7b). In contrast to RV1-UV2, UV1-RV2 first assimilates the retrieved winds in the assimilation window (0000- 0500 UTC 16 August 2008), and the outcome at the initial time then provides a first guess for assimilating the radial velocity in the same window. In UV-RV, the retrieved winds and radial velocity are assimilated at the same time in the assimilation window (0000-0500 UTC 16 August 2008). Figure 8 compares the intensity forecasts in terms of minimum sea level pressure from different experiments with the best-track data. It seems that the intensity from experiment UV-RV is similar to that from experiment UV, and the intensity from UV1-RV2 is similar to that from RV only. Meanwhile, the track forecasts from UV-RV and UV1-RV2 are almost the same as those obtained from UV and RV, respectively (figure not shown). UV information seems to dominate the impact in the former (UV-RV) and RV information seems to dominate the latter (UV1-RV2); this has been further confirmed by checking the analysis increments (figures not shown). In RV1-UV2, the data assimilation process improves the forecasts more than it does in any other experiment. 5. Impact of initial conditions on numerical simulations of Nuri's genesis: Implications So far, results from the above discussion clearly indicate that the assimilation of radar observations results in improved forecasts of Typhoon Nuri's genesis. Specifically, without radar data assimilation, the WRF model fails to predict Nuri's genesis, leading to a nondeveloping case. With radar data assimilation, the model successfully forecasts the genesis of Nuri, leading to a developing case. In this section, we further examine how the changes in the initial conditions through radar data assimilation lead to successful forecasts of Nuri's genesis. Since UV-RV (UV1-RV2) produces results similar to those from UV (RV), the analysis in this section will focus on UV, RV, and RV1-UV2. a. Impact of data assimilation on the inner core and surrounding environment 1) VORTICITY Figure 9 compares the vertical structure of the initial (at 0000 UTC 16 August 2008) relative vorticity between Ctrl and the experiments with radar data assimilation (e.g., UV, RV, and RV1-UV2), averaged over an area within a radius of 300 km from the simulated pre-Nuri's circulation center. In Ctrl (Fig. 9a), the positive relative vorticity is concentrated at low levels, with a maximum relative vorticity at about 950 hPa. In UV (Fig. 9b), the changes induced by the assimilation of radar wind analysis are notable: the relative vorticity in UV presents larger positive values at middle levels (800-400 hPa) when compared with those in Ctrl (Fig. 9a). RV also shows (Fig. 9c) that the direct assimilation of radial velocity results in an increase in relative vorticity, especially at middle levels (700- 500 hPa). Similarly, the assimilation of both radial velocity and radar wind analysis (RV1-UV2) produces an enhanced positive vorticity at middle levels in the initial conditions (Fig. 9d). The vertical profiles of the areaaveraged relative vorticity (Fig. 10) demonstrate that radar data assimilation produces an enhanced positive vorticity at middle levels compared with the control experiment. Previous studies (Dunkerton et al. 2009; Montgomery et al. 2010) have indicated the importance of the midlevel vortex for protecting moist air from the intrusion of environmental dry air. As a result, the initial stronger midlevel vortex in UV, RV, and RV1-UV2 maintains the moist environment, which is favorable for the persistent development of deep convection and Nuri's genesis. 2) DIVERGENCE To further illustrate the influence of the assimilation of radar data over the inner core and the surrounding environment, Fig. 11 shows the averaged divergence profiles over the area within a radius of 500 km around pre-Nuri's circulation center at 0500 UTC 16 August 2008 from Ctrl, UV, RV, and RV1-UV2. It is obvious that the assimilation of radar data enhances the lowlevel convergence and upper-level divergence, compared with Ctrl, confirming that radar data assimilation is beneficial for intensifying the vortex inner core. 3) TEMPERATURE AND MOISTURE The impacts of radar data assimilation on the forecast temperature and moisture structure of Nuri are examined. Figure 12 shows an east-west vertical cross section of the temperature anomaly, calculated by the differences between the temperature in horizontal grids over the inner-core region (within a radius of 300 km) and the averaged temperature in the outer rainband region (within a radius of 450-600 km from the simulated Nuri's center) at corresponding pressure levels at Nuri's observed genesis time (1800 UTC 16 August 2008). Since Ctrl and UV did not predict Nuri's genesis at that time, the warm core has not formed in the lower to middle troposphere. Even though there is some development of warm temperature anomalies in UV compared with that in Ctrl, they are not well organized and are mainly at a large scale. In contrast, with the assimilation of radar radial velocity, a strong warm temperature anomaly develops and organizes between 500 and 300 hPa in the inner-core region in RV, showing the development of a tropical cyclone warm core. With the assimilation of both radial velocity and radarderived 3D wind retrievals, RV1-UV2 demonstrates clearly the formation of a strong and well-organized warm core, both in the middle to upper levels (600- 150 hPa) and in the lower level (900-850 hPa). Similar structural variations are found in the moisture field in the different experiments (Fig. 12). Moist air is found to dominate the lower to middle levels over the area around the circulation center in Ctrl (Fig. 12a). With the assimilation of radar-retrieved 3D winds in UV, the moisture field is smoothed over the circulation center (Fig. 12b). With the assimilation of radar radial velocity, the inner core of Nuri forms an ??early eyelike structure'' as the air in the eye region becomes drier and that in the eyewall region becomes moister, compared with the conditions in Ctrl. With the assimilation of both the radial velocity and radar-derived 3D wind retrieval, a mature tropical cyclone eye and an eyewall are formed and are clearly illustrated by both moisture and temperature. In addition, the temperature and moisture structures in Fig. 12 confirm that the assimilation of the radar radial velocity improves the representation of the TC inner core. b. Impact of data assimilation on convective development and precipitation 1) PRECIPITATION Many studies (Montgomery et al. 2006; Wang et al. 2010a,b; Komaromi 2013) have suggested that the development of intense convection is very important for TC genesis because intense convection can transport moisture upward to middle levels and produce warming at upper levels induced by latent heat release. Since rainfall is a good indicator of convection, the impact of the radar data assimilation on precipitation is examined here. Figure 13 illustrates the hourly precipitation from the four simulations (Ctrl, UV, RV, and RV1-UV2) at 1600 UTC 16 August 2008, compared with the satellitederived [NASA Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) on board Aqua] rainfall rate at 1619 UTC 16 August 2008. Although Ctrl produces precipitation associated with Nuri, the rainfall rate is much weaker (Fig. 13b) than the observed rainfall rate (Fig. 13a). As a result of the weak convection associated with the weak precipitation, Ctrl does not predict Nuri's genesis. With the assimilation of the radar wind analysis, UV (Fig. 13c) enhances the rainfall rate in Nuri's inner core. The rainfall rate in UV (Fig. 13c) is also approximately consistent with the observed precipitation (Fig. 13a). In the other two experiments with radar data assimilation, RV (Fig. 13d) and RV1-UV2 (Fig. 13e), an even higher rainfall rate is produced (Fig. 13d), which also corresponds with the stronger intensity of the simulated Typhoon Nuri in the two experiments. As a result, intense convection is necessary for the simulations to be successful in predicting Nuri's genesis. Meanwhile, these results indicate that radar data assimilation significantly improves the convection and precipitation forecasts during Nuri's genesis. 2) DEVELOPING INTENSE CONVECTION INTO THE POUCH In addition, Montgomery and his colleagues (Dunkerton et al. 2009; Montgomery et al. 2009; Wang et al. 2010a,b) suggested that the pouch circulation within the comoving frame is a good way of examining the evolution of convection in the pregenesis disturbance because the center of the pouch circulation is a favorable region for TC genesis. Thus, Fig. 14 presents the closed circulation (the so-called pouch circulation) in the comoving frame, obtained by subtracting the propagation speed of the disturbance wave. The propagation of the disturbance is approximately westward at the following speeds: 3.9 m s21 in Ctrl, 3.1 m s21 in UV, 4.0ms21 in RV, and 5.2ms21 in RV1-UV2. In addition to the wind vector, which represents the relative wind after subtracting the propagation speed, Fig. 14 also shows the divergence at 850 hPa as the low-level convergence (negative values of divergence) indicates convection. The evolution of the convection in the pouch circulation in Ctrl (without radar data assimilation) is shown in Figs. 14a-c. At 3 h, the convergent and ascending motions associated with the initiated convection emerge on the eastern side of the circulation (Fig. 14a). In the following 6 h, the convection is not enhanced significantly. Only weak convection is scattered on the eastern side of the pouch, and no obvious convective system has developed in the pouch's center, which is the intersection of the critical layer (purple line) and the wave trough axis (black line). Therefore, without the strong convective initiation and development in the pouch circulation, Ctrl eventually leads to a nondeveloping system (as shown in Fig. 3a). With radar data assimilation, UV, RV, and RV1-UV2 show different processes of convection initiation and development in the disturbance. In UV, the convection with the low-level convergence is initiated on the eastern side of the pouch circulation at 3 h (Fig. 14d), and then the convection develops noticeably from 6 to 9 h (Figs. 14e,f). By 9 h, UV (Fig. 14f) shows a stronger convergence in a large area in the pouch than does Ctrl (Fig. 14c). Consequently, the stronger convection in UV leads to a developing system with an evident drop in MSLP (as shown in Fig. 3a). In the other two experiments with radar data assimilation, RV (Figs. 14g-i) and RV1-UV2 (Figs. 14j-l), convective initiation on the eastern side of the pouch also appears at 3 h along with strong convective development in the subsequent 6 h. In contrast, strong convection occurs closer to the pouch's center inRVandRV1-UV2thaninUVby9h.It isnoted that in RV (Figs. 14h,i)andRV1-UV2(Figs. 14k,l), the evident southeast flow appears in the eastern region of the pouch and wraps the strong convection into the center of the pouch. This strong convection concentrated in the pouch's center further enhances Nuri's genesis, with a significant drop in MSLP in RV and RV1-UV2 (as shown in Fig. 3a). The above comparison between the control experiment and the radar data assimilation experiments regarding the convection in the pouch demonstrates that radar data assimilation significantly improves the simulated convection initiation and development. Particularly, in RV and RV1-UV2, the southeast flow contributes to the concentration of the initiated convection into the pouch's center. The enhanced convection in the pouch eventually leads to a developing system (Nuri's genesis) in the experiments with radar data assimilation. 3) UPPER-LEVEL WARMING AND THE DROP IN MSLP Although the above results show that strong convective development plays a significant role in Nuri's genesis, the question remains as to what process leads to the evident MSLP drops in the experiments with radar data assimilation. Previous studies have suggested that warming in the TC center accounts for the drop in MSLP (Holland 1997; Zhang and Chen 2012; Zhang and Zhu 2012), as diabatic heating leads to a warming process. In this section, we examine whether radar data assimilation results in an enhanced warming process. Figure 15 shows the time-pressure plot of the temperature perturbation T 0(p, t) in the inner core. We define T 0 (p, t) as the anomaly of the mean inner-core temperature profiles Tin(p, t), within a radius of 50 km from the simulated Nuri's center, with respect to the mean environmental temperature profile Tenv(p), within a radius of 500 km from the center at the model initial time. In Ctrl (Fig. 15a), only weak warm anomalies emerge, mostly at the lower level around 750 hPa, which corresponds to the weak diabatic heating. In contrast, the three experiments with radar data assimilation (Figs. 15b-d) present much warmer anomalies than Ctrl resulting from stronger diabatic heating (figure not shown). Specifically, evident warm anomalies (larger than 3 K) appear at upper levels after 18 h and increase significantly after 36 h. Up to 48 h, inner-core temperature anomalies reach a maximum value of over 5 K, suggesting that radar data assimilation has significant impacts on the warming process in Nuri's simulations. To validate the contribution of the warming process to the drop in MSLP, a simple calculation of the hydrostatic equation, similar to that done by Zhang and Zhu (2012), is applied to quantitatively evaluate the relationship between the warming process and the drop in MSLP. Figure 16 shows the estimated MSLP, which is calculated by hydrostatically integrating the temperature profile in the circulation center from the top model level to the lowest model level. The integration formula originated from the hydrostatic equation is shown as follows: ... The subscripts bottom and top indicate the bottom and top model levels, respectively; R and g are the specific gas of air and standard gravity constants, respectively. It is apparent that the time series of the integrated MSLP in Fig. 16 matches the simulated MSLP in Fig. 7a quite well. Such similarity between the integrated MSLP and the simulated MSLP suggests that the warming temperature in the center column leads to the variation in MSLP in the simulations, because the integrated MSLP is calculated from the integration of the temperature profile. More importantly, the integrated MSLP also shows a significant discrepancy between the simulations with and without radar data assimilation. The integrated MSLP in Ctrl shows only a slight drop (Fig. 16); by contrast, UV and RV both show a more significant drop in the integrated MSLP (Fig. 16), which corresponds with the enhanced warming process (Figs. 15b,c); RV1-UV2 presents the lowest value of the integrated MSLP (Fig. 16) due to the strongest warming process (Fig. 15d). These differences confirm that the stronger warming process in the inner core contributes to the lower MSLP in the experiments with radar data assimilation, compared with the weaker warming process and less drop in MSLP without radar data assimilation (Ctrl). Note that the initial temperature fields are not so different among these experiments. Because of radar data assimilation, the temperature, velocity field, and pressure are all adjusted within the 4DVAR assimilation window. The impacts also extended to the subsequent forecasts. In addition, the adjustments in the velocity field further control the changes in the mass (temperature and pressure) field in the genesis stage (Wu et al. 2006). 6. Summary and concluding remarks In this study, several data assimilation experiments and numerical simulations are conducted using the WRF model and its 4DVAR system to predict the genesis of Typhoon Nuri (2008). The impact of ELDORA airborne radar data, collected during TCS-08/T-PARC, on the accurate prediction of Nuri's genesis is evaluated. Considering the available radar radial velocity direct measurements and common existence of radar-retrieved three-dimensional wind fields, three radar data assimilation methods, including the assimilation of the retrieved 3D wind analysis, the direct assimilation of the radar radial velocities, and the assimilation of both the radar wind analysis and radial velocity, are evaluated. Results from data assimilation and numerical simulations are examined and compared. The main conclusions are summarized as follows. d The radar data assimilation of the 4DVAR system significantly improves the numerical simulations of Nuri's genesis. Simulations with radar data assimilation (in all experiments) predict Nuri's genesis, while the control experiment (Ctrl, without radar data assimilation) fails to predict Nuri's genesis. The simulations with radar data assimilation also produce realistic rainfall structures and reasonable environmental flows before, during, and after Nuri's genesis. d Among various ways to assimilate radar radial velocities, direct assimilation of radar radial velocity leads to better intensity forecasts, as it enhances the development of convective systems and improves the inner-core structure of Nuri, whereas assimilation of the radar-retrieved wind analysis is more beneficial to tracking forecasts, as it results in improved environmental flows. The assimilation of both the radarretrieved wind and the radial velocity can lead to better intensity and track forecasts, if radial velocity observations are assimilated first and retrieved winds are then assimilated in the same data assimilation window. d Experiments with and without radar data assimilation lead to developing and nondeveloping TC genesis in numerical simulations. The improved initial conditions and forecasts from the data assimilation imply that the enhanced midlevel vortex and moisture conditions are favorable for the development of deep convection into the center of the pouch and eventually contribute to Nuri's genesis. The improved simulations of the convection and associated environmental conditions produce enhanced upper-level warming in the core and lead to the drop in MSLP. This study demonstrates the benefits of radar radial velocity data assimilation in the simulation of Nuri's genesis. Because the importance of the mesoscale convective system in the pre-Nuri disturbance was noted by Montgomery et al. (2009), the improvement in intensity forecasts induced by radar data assimilation in this study is attributed to at least two reasons in terms of improving the forecasts of the convective system. First, the airborne ELDORA provides sufficient highresolution observations for representing and predicting the convective system in the pre-Nuri disturbance. Second, because such a convective system shows rapid temporal variations, the 4DVAR system is able to take advantage of assimilating the observations at multiple times in order to accommodate the rapid changes. For these two reasons, detailed conditions associated with the convective system are imposed by 4DVAR into the model to make it produce more accurate predictions of the convective system, which eventually improves the forecasts of Nuri's genesis. In addition, different ways of assimilating radar radial velocity data result in various impacts on numerical simulations. Specifically, direct assimilation of radarmeasured radial velocity resulted in a significant improvement in intensity forecasts, while assimilation of radar wind analysis (UV) led to a notable improvement in the track forecasts. Although retrieved winds derived from the radar radial velocity, they represent different aspects of information in terms of temporal and spatial scales due to smoothing and constraints used in the retrieval process. We combined both retrieved winds and radial velocity measurements in the 4DVAR system to obtain maximum benefits for both track and intensity forecasts. To the best of our knowledge, this paper presents the first numerical experiment that assimilates both radar radial velocity and retrieved wind field together for mesoscale simulations. Future work should be conducted with more case studies using airborne radar observations from different platforms for robust conclusions. In addition, various data assimilation methods (e.g., ensemble Kalman filter) with different cycling periods of radar data assimilation should be examined. Acknowledgments. The authors appreciate the NCAR WRF and WRFDA model development groups for their efforts in developing the community model and data assimilation systems. The computer support from the Center for High Performance Computing (CHPC) at the University of Utah and the S4 computer resource from the NOAA Joint Center for Satellite Data Assimilation (JCSDA) are acknowledged. This study is supported by the Office of Naval Research Award N000141310582, National Science Foundation Award AGS-1243027, and NOAA Grant NA14NWS4680025. Review comments from three anonymous reviewers and Dr. Andrew Jones were helpful for improving the manuscript. REFERENCES Bell, M. M., and M. T. Montgomery, 2010: Sheared deep vortical convection in pre-depression Hagupit during TCS08. Geophys. Res. Lett., 37, L06802, doi:10.1029/2009GL042313. -, W.-C. Lee, A. W. Cory, and H. Cai, 2013: A Solo-based automated quality control algorithm for airborne tail Doppler radar data. J. Appl. Meteor. Climatol., 52, 2509-2528, doi:10.1175/JAMC-D-12-0283.1. Davis, C., and L. F. Bosart, 2002: Numerical simulations of the genesis of Hurricane Diana (1984). Part II: Sensitivity of track and intensity prediction. Mon. Wea. Rev., 130, 1100-1124, doi:10.1175/1520-0493(2002)130,1100:NSOTGO.2.0.CO;2. Dowell, D. C., and L. J. Wicker, 2009: Additive noise for stormscale ensemble data assimilation. J. Atmos. Oceanic Technol., 26, 911-927, doi:10.1175/2008JTECHA1156.1. Dudhia, J., 1989: Numerical study of convection observed during the Winter Monsoon Experiment using a mesoscale twodimensional model. J. Atmos. Sci., 46, 3077-3107, doi:10.1175/1520-0469(1989)046,3077:NSOCOD.2.0.CO;2. Dunkerton, T. J., M. T. Montgomery, and Z. Wang, 2009: Tropical cyclogenesis in a tropical wave critical layer: Easterly waves. Atmos. Chem. Phys. Discuss., 9, 5587-5646. Gamache, J. F., P. P. Dodge, and N. F. Griffin, 2008: Automatic quality control and analysis of airborne Doppler data: Realtime applications, and automatically post-processed analyses for research. Preprints, 28th Conf. on Hurricanes and Tropical Meteorology, Orlando, FL, Amer. Meteor. Soc., P2B.12. [Available online at http://ams.confex.com/ams/pdfpapers/ 137969.pdf.] Gao, J., M. Xue, A. Shapiro, Q. Xu, and K. K. Droegemeier, 2001: Three-dimensional simple adjoint velocity retrievals from single-Doppler radar. J. Atmos. Oceanic Technol., 18, 26-38, doi:10.1175/1520-0426(2001)018,0026:TDSAVR.2.0.CO;2. Holland, G. J., 1997: The maximum potential intensity of tropical cyclones. J. Atmos. Sci., 54, 2519-2541, doi:10.1175/ 1520-0469(1997)054,2519:TMPIOT.2.0.CO;2. Hong, S.-Y., Y. Noh, and J. Dudhia, 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318-2341, doi:10.1175/MWR3199.1. Houze, R. A., Jr., S. S. Chen, B. F. Smull, W.-C. Lee, and M. M. Bell, 2007: Hurricane intensity and eyewall replacement. Science, 315, 1235-1239, doi:10.1126/science.1135650. Huang, X.-Y., and Coauthors, 2009: Four-dimensional variational data assimilation for WRF: Formulation and preliminary results. Mon. Wea. Rev., 137, 299-314, doi:10.1175/ 2008MWR2577.1. Kain, J. S., and J. M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain-Fritsch scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 46, Amer. Meteor. Soc., 165-170. Kieu, C. Q., and D.-L. Zhang, 2010: Genesis of Tropical Storm Eugene (2005) from merging vortices associated with ITCZ breakdowns. Part III: Sensitivity to various genesis parameters.J. Atmos.Sci., 67,1745-1758,doi:10.1175/2010JAS3227.1. Komaromi, W. A., 2013: An investigation of composite dropsonde profiles for developing and nondeveloping tropical waves during the 2010 PREDICT field campaign. J. Atmos. Sci., 70, 542-558, doi:10.1175/JAS-D-12-052.1. Li, Y., X. Wang, and M. Xue, 2012: Assimilation of radar radial velocity data with the WRF hybrid ensemble-3DVAR system for the prediction of Hurricane Ike (2008). Mon. Wea. Rev., 140, 3507-3524, doi:10.1175/MWR-D-12-00043.1. Lim, K.-S. S., and S.-Y. Hong, 2010: Development of an effective double-moment cloud microphysics scheme with prognostic cloud condensation nuclei (CCN) for weather and climate models. Mon. Wea. Rev., 138, 1587-1612, doi:10.1175/ 2009MWR2968.1. Liu, H., J. L. Anderson, and Y.-H. Kuo, 2012: Improved analyses and forecast of Hurricane Ernesto's genesis using radio occultation data in an ensemble filter assimilation system. Mon. Wea. Rev., 140, 151-166, doi:10.1175/MWR-D-11-00024.1. Marks, F. D., and R. A. Houze Jr., 1987: Inner core structure of Hurricane Alicia from airborne Doppler radar observations. J. Atmos. Sci., 44, 1296-1317, doi:10.1175/ 1520-0469(1987)044,1296:ICSOHA.2.0.CO;2. Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663-16 682, doi:10.1029/ 97JD00237. Montgomery,M.T.,M.E.Nicholls,T.A.Cram,andA.B. Saunders, 2006: A vortical hot tower route to tropical cyclogenesis. J. Atmos. Sci., 63, 355-386, doi:10.1175/JAS3604.1. -, L. L. Lussier III, R. W. Moore, and Z. Wang, 2009: The genesis of Typhoon Nuri as observed during the Tropical Cyclone Structure 2008 (TCS-08) field experiment-Part 1: The role of the easterly wave critical layer. Atmos. Chem. Phys. Discuss., 9, 19159-19203, doi:10.5194/acpd-9-19159-2009. -, Z. Wang, and T. J. Dunkerton, 2010: Coarse, intermediate and high resolution numerical simulations of the transition of a tropical wave critical layer to a tropical storm. Atmos. Chem. Phys., 10, 10803-10827, doi:10.5194/acp-10-10803-2010. Nolan, D., 2007: What is the trigger for tropical cyclogenesis? Aust. Meteor. Mag., 56, 241-266. Oye, R., C. Mueller, and S. Smith, 1995: Software for translation, visualization, editing and interpolation. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Soc., 359-361. Parrish, D. F., and J. C. Derber, 1992: The National Meteorological Center's spectral statistical-interpolation analysis system. Mon. Wea. Rev., 120, 1747-1763, doi:10.1175/ 1520-0493(1992)120,1747:TNMCSS.2.0.CO;2. Pu, Z., and L. Zhang, 2010: Validation of AIRS temperature and moisture profiles over tropical oceans and their impact on numerical simulations of tropical cyclones. J. Geophys. Res., 115, D24114, doi:10.1029/2010JD014258. -, X. Li, and J. Sun, 2009: Impact of airborne Doppler radar data assimilation on the numerical simulation of intensity changes of Hurricane Dennis near a landfall. J. Atmos. Sci., 66, 3351-3365, doi:10.1175/2009JAS3121.1. Raymond, D. J., and S. L. Sessions, 2007: Evolution of convection during tropical cyclogenesis. Geophys. Res. Lett., 34, L06811, doi:10.1029/2006GL028607. -, and C. López Carrillo, 2011: The vorticity budget of developing typhoon Nuri (2008). Atmos. Chem. Phys., 11, 147- 163, doi:10.5194/acp-11-147-2011. -, S. L. Sessions, and C. López Carrillo, 2011: Thermodynamics of tropical cyclogenesis in the northwest Pacific. J. Geophys. Res., 116, D18101, doi:10.1029/2011JD015624. Reasor, P. D., M. T. Montgomery, F. D. Marks, and J. F. Gamache, 2000: Low-wavenumber structure and evolution of the hurricane inner core observed by airborne dual-Doppler radar. Mon. Wea. Rev., 128, 1653-1680, doi:10.1175/1520-0493(2000)128,1653: LWSAEO.2.0.CO;2. -, M. Eastin, and J. F. Gamache, 2009: Rapidly intensifying Hurricane Guillermo (1997). Part I: Low-wavenumber structure and evolution. Mon. Wea. Rev., 137, 603-631, doi:10.1175/ 2008MWR2487.1. Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech Note NCAR/ TN-4751STR, 113 pp. [Available online at http://www2. mmm.ucar.edu/wrf/users/docs/arw_v3.pdf.] Sun, J., and H. Wang, 2013: Radar data assimilation with WRF 4DVAR. Part II: Comparison with 3DVAR for a squall line over the U.S. Great Plains. Mon. Wea. Rev., 141, 2245-2264, doi:10.1175/MWR-D-12-00169.1. Wang, H., and X.-Y. Huang, 2012: TAMDAR observation assimilation in WRF 3DVAR and its impact on Hurricane Ike (2008) forecast. Atmos. Oceanic Sci. Lett., 5, 206-211. -, J. Sun, X. Zhang, X.-Y. Huang, and T. Auligné, 2013: Radar data assimilation with WRF 4DVAR. Part I: System development and preliminary testing. Mon. Wea. Rev., 141, 2224-2244, doi:10.1175/MWR-D-12-00168.1. Wang, Z., M. T. Montgomery, and T. J. Dunkerton, 2010a: Genesis of pre-Hurricane Felix (2007). Part I: The role of the easterly wave critical layer. J. Atmos. Sci., 67, 1711-1729, doi:10.1175/ 2009JAS3420.1. -, -, and -, 2010b: Genesis of pre-Hurricane Felix (2007). Part II: Warm core formation, precipitation evolution, and predictability. J. Atmos. Sci., 67, 1730-1744, doi:10.1175/ 2010JAS3435.1. Wolff, C. A., M. M. Bell, and W.-C. Lee, 2009: Automatic quality control of airborne Doppler radar data. Preprints, 34th Conf. on Radar Meteorology, Williamsburg, VA, Amer. Meteor. Soc., P9.7. [Available online at http://ams.confex.com/ams/ pdfpapers/155949.pdf.] Wu, C., K. Chou, Y. Wang, and Y. Kuo, 2006: Tropical cyclone initialization and prediction based on four-dimensional variational data assimilation. J. Atmos. Sci., 63, 2383-2395, doi:10.1175/JAS3743.1. Xiao, Q., X. Zhang, C. Davis, J. Tuttle, G. Holland, and P. J. Fitzpatrick, 2009: Experiments of hurricane initialization with airborne Doppler radar data for the Advanced Research Hurricane WRF (AHW) Model. Mon. Wea. Rev., 137, 2758- 2777, doi:10.1175/2009MWR2828.1. Zhang, D.-L., and H. Chen, 2012: Importance of the upper-level warm core in the rapid intensification of a tropical cyclone. Geophys. Res. Lett., 39, L02806, doi:10.1029/2011GL050578. -, and L. Zhu, 2012: Role of upper-level processes in tropical cyclogenesis. Geophys. Res. Lett., 39, L17804, doi:10.1029/ 2012GL053140. Zhang, F., Y. Weng, J. A. Sippel, Z. Meng, and C. H. Bishop, 2009: Cloud-resolving hurricane initialization and prediction through assimilation of Doppler radar observations with an ensemble Kalman filter. Mon. Wea. Rev., 137, 2105-2125, doi:10.1175/2009MWR2645.1. -, -, J. F. Gamache, and F. D. Marks, 2011: Performance of convection-permitting hurricane initialization and prediction during2008-2010 with ensembledata assimilation of inner-core airborne Doppler radar observations. Geophys. Res. Lett., 38, L15810, doi:10.1029/2011GL048469. Zhang, L., Z. Pu, W.-C. Lee, and Q. Zhao, 2012: The influence of airborne Doppler radar data quality on numerical simulations of a tropical cyclone. Wea. Forecasting, 27, 231-239, doi:10.1175/WAF-D-11-00028.1. Zhang, M., M. Zupanski, M.-J. Kim, and J. A. Knaff, 2013: Assimilating AMSU-A radiances in the TC core area with NOAA operational HWRF (2011) and a hybrid data assimilation system: Danielle (2010). Mon. Wea. Rev., 141, 3889- 3907, doi:10.1175/MWR-D-12-00340.1. Zhang, X., X.-Y. Huang, and N. Pan, 2013: Development of the upgraded tangent linear and adjoint of the Weather Research and Forecasting (WRF) Model. J. Atmos. Oceanic Technol., 30, 1180-1188, doi:10.1175/JTECH-D-12-00213.1. ZHAN LI AND ZHAOXIA PU Department of Atmospheric Sciences, University of Utah, Salt Lake City, Utah JUANZHEN SUN AND WEN-CHAU LEE National Center for Atmospheric Research,* Boulder, Colorado (Manuscript received 12 February 2014, in final form 4 August 2014) * The National Center for Atmospheric Research is sponsored by the National Science Foundation. Corresponding author address: Dr. Zhaoxia Pu, Dept. of Atmospheric Sciences, University of Utah, 135 S 1460 E, Rm. 819, Salt Lake City, UT 84112. E-mail: [email protected] (c) 2014 American Meteorological Society |