• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Multi-scale Incremental Analysis Update Scheme and Its Application to Typhoon Mangkhut (2018) Prediction

    2023-02-08 08:16:36YanGAOJialiFENGXinXIAJianSUNYulongMADongmeiCHENandQilinWAN
    Advances in Atmospheric Sciences 2023年1期

    Yan GAO, Jiali FENG, Xin XIA, Jian SUN, Yulong MA, Dongmei CHEN, and Qilin WAN*

    1Guangdong-Hong Kong-Macao Greater Bay Area Weather Research Center for Monitoring Warning and Forecasting, Shenzhen 518038, China

    2CMA Earth System Modeling and Prediction Centre, China Meteorological Administration, Beijing 100081, China

    ABSTRACT In the traditional incremental analysis update (IAU) process, all analysis increments are treated as constant forcing in a model’s prognostic equations over a certain time window.This approach effectively reduces high-frequency oscillations introduced by data assimilation.However, as different scales of increments have unique evolutionary speeds and life histories in a numerical model, the traditional IAU scheme cannot fully meet the requirements of short-term forecasting for the damping of high-frequency noise and may even cause systematic drifts.Therefore, a multi-scale IAU scheme is proposed in this paper.Analysis increments were divided into different scale parts using a spatial filtering technique.For each scale increment, the optimal relaxation time in the IAU scheme was determined by the skill of the forecasting results.Finally, different scales of analysis increments were added to the model integration during their optimal relaxation time.The multi-scale IAU scheme can effectively reduce the noise and further improve the balance between large-scale and small-scale increments in the model initialization stage.To evaluate its performance, several numerical experiments were conducted to simulate the path and intensity of Typhoon Mangkhut (2018) and showed that: (1) the multi-scale IAU scheme had an obvious effect on noise control at the initial stage of data assimilation; (2) the optimal relaxation time for large-scale and small-scale increments was estimated as 6 h and 3 h, respectively; (3) the forecast performance of the multiscale IAU scheme in the prediction of Typhoon Mangkhut (2018) was better than that of the traditional IAU scheme.The results demonstrate the superiority of the multi-scale IAU scheme.

    Key words: multi-scale incremental analysis updates, optimal relaxation time, 2-D discrete cosine transform,GRAPES_Meso, Typhoon Mangkhut (2018)

    1.Introduction

    In terms of spatial coverage, numerical models can be classified into two major categories: global models and regional models.Both have unique advantages in describing analysis information at different scales.Because global analysis is not affected by biases at lateral boundaries, the largescale aspect of global analysis is superior to the aspect of regional analysis (Peng et al., 2010; Zhuang et al., 2020).However, after assimilating observation data with high spatial density, the regional analysis may yield a more accurate small-scale analysis (Zhuang et al., 2018).To improve the forecasting performance of a regional model, one possible approach is to use a blending technique, which includes an incremental spatial filter to blend large-scale analysis from the global model with small-scale fields from the high-resolu-tion regional model (Denis et al., 2002; Yang, 2005; Hsiao et al., 2015).

    Although the advantages of the blending technique have been collectively presented by some previous studies(Keresturi et al., 2019; Feng et al., 2021), this method still has some limitations in short-term forecasts, such as shortterm precipitation climatology, likely owing to the imbalance between large-scale analysis from the global model and small-scale analysis from the regional model (Polavarapu et al., 2004; Schwartz et al., 2021).To further improve the performance of the blending technique, it is necessary to employ some strategies to combat this imbalance.One effective method to accomplish this task is to apply an incremental analysis update (IAU) technique.Bloom et al.(1996) first proposed the IAU technique, which gradually incorporated the analysis increment.Since this technique emerged, it has been widely used in the atmospheric and oceanic fields and is considered to be successful in keeping the mass and momentum fields in dynamic balance by removing spurious gravity waves (Ourmières et al., 2006; Zhang et al., 2015;Lei and Whitaker, 2016).

    In a traditional IAU scheme, the background system is considered to be linearly evolving over a given time window.Therefore, the traditional IAU scheme has a fixed time window, which is called the relaxation time, τ (Xu et al.,2019; Li et al., 2021).All analysis increments are gradually incorporated into the model integration during the relaxation time (Chen et al., 2020).However, in real weather processes,different scales of incremental analyses are produced by corresponding scales of weather systems, which have unique evolutionary speeds, life histories, and response times.Compared with large-scale information, small-scale information evolves more rapidly and has a shorter life history in the regional model.However, both large- and small-scale increments have the same time window in the traditional IAU method, which cannot entirely represent the propagation characteristics of information having different scales.

    To remedy the shortcomings of the traditional IAU technique and improve the forecast performance of the numerical model, this paper proposes a multi-scale IAU scheme for the first time.Through the blending method, the analysis increments were divided into two categories, large- and small-scale increments.Several IAU experiments were designed to determine the optimal relaxation time (ORT)for each scale increment.Using the multi-scale IAU scheme,different scales of analysis increment were added into model integration during their own ORT, and the corresponding forecast performance was evaluated in the prediction of Typhoon Mangkhut (2018).

    The remainder of this paper is structured as follows.Section 2 describes the methodology used in detail.The model configuration and the numerical experiments conducted in this study are introduced in section 3.The differences in the simulation results between different kinds of IAU schemes in Typhoon Mangkhut (2018) are analyzed in section 4.Finally, a discussion and conclusion are given in section 5.

    2.Methodology

    The typhoon initialization scheme proposed in this paper is based on the multi-scale IAU technique.There are three main steps to achieve the initialization scheme.First,through the 3DVAR system, the background field in the regional model is transformed into a 3DVAR analysis field.Then, the blended analysis field, which consists of largescale analysis information obtained from the global analysis field and small-scale analysis information obtained from the 3DVAR analysis field, is obtained through the blending method, with a spectral transform called the two-dimensional discrete cosine transform (2D-DCT) (Denis et al., 2002).Finally, by using the multi-scale IAU technique, the largescale and small-scale analysis increments are incorporated into the integration of the regional model over their own time windows.The blending method and the IAU scheme are introduced in detail in the following sections.

    2.1.Blending Method

    As mentioned in section 1, the regional model has more accurate small-scale analysis information than the global model.However, the solution obtained from the regional and global models is inconsistent.This inconsistency may produce unexpected noise and cause instabilities owing to wave reflection along the lateral boundaries (Davies, 1983).Therefore, systematic large-scale errors can be made in the limited domain, leading to a distortion of large-scale information in the regional model after a significant period of integration (Peng et al., 2010).A blending method is applied in the regional model to solve this problem.

    In the regional model, the background field and the 3DVAR analysis field are designated asGxbandGxa, respectively.Then, the incremental analysis betweenGxbandGxais expressed asGdxa.Moreover, the global analysis field obtained from NCEP GFS analysis is defined asGglobal.Through the 2D-DCT blending method, the blended analysis fieldGblndcan be calculated as:

    whereFfilter,Lis a low-pass spatial filter with a spectral transform called 2D-DCT.The filter cutoff length adopted in the blending method is chosen as 600 km, which is guided by the GRAPES-RAFS, a rapid analysis and forecast system operated at the Center of Numerical Weather Prediction,CMA (Zhuang et al., 2020).

    As shown in Eq.2, the blended analysis incrementGdblndis obtained asGblnd-Gxb, which includes the largescale increment, Δl=Ffilter,Land the smallscale increment, Δs=Gdxa-Ffilter,L(Gdxa).It follows:

    2.2.Multi-scale IAU Method

    To reduce the high-frequency oscillations induced by the analysis increment, different types of IAU methods are widely applied in atmospheric models.The basic principle of IAU is to gradually incorporate an increment calculated from the analysis during the model integration.In the IAU scheme, analysis increments are treated as constant additional forcing terms in the model’s prognostic equations over a certain time window (relaxation time, τ):

    whereFis the prognostic variable, ··· represents the increment introduced by the model integration, andFaandFbare the analysis and background fields, respectively.

    The IAU scheme called three-dimensional IAU(3DIAU) uses a single increment that is assumed to be constant over an assimilation window during model integration.In contrast to 3DIAU, the propagation of the increment information over an assimilation window is considered in fourdimensional IAU (4DIAU).Time-varying analysis increments are gradually incorporated into the model integration through 4DIAU (Lorenc et al., 2015; Lei and Whitaker,2016).

    Regardless of which type of IAU scheme is adopted, all scales of analysis increments are incorporated in the model integration over the same time window.However, different scales of incremental information have different evolutionary speeds and life histories in the regional model, which cannot be represented by the traditional IAU method.

    In this section, based on the principle of the IAU technique, a first proposal of the multi-scale IAU technique is outlined.In the multi-scale IAU scheme, different scales of analysis increments are added to the model integration over different time windows:

    where Δland τlare the large-scale analysis increment and its relaxation time, respectively, while Δsand τsrepresent the small-scale analysis increment and its relaxation time,respectively.

    Figure 1 shows the schematic illustration of two different categories of IAU schemes.The blue area above the coordinate axis represents the traditional IAU technique, and τ is the relaxation time.The red area below the coordinate axis displays the multi-scale IAU scheme, in which the largescale increment and small-scale increment are incorporated into the model integration over the time windows τland τs,respectively.

    3.Model Configurations and Experimental Design

    3.1.Model Configuration

    The numerical weather prediction model used in this study is the mesoscale version of the Global/Regional Assimilation and PrEdiction System (GRAPES_Meso), which is a new generation of numerical weather forecasting systems for mesoscale weather prediction developed by the China Meteorological Administration (CMA) (Chen and Shen,2006).The model adopts a height-based terrain-following coordinate, a semi-implicit and semi-Lagrangian (SI-SL)time difference scheme, a fully compressible non-hydrostatic balance dynamic framework, and a physical parameterization package (Wu et al., 2005; Chen et al., 2008).The physical parameterization schemes selected in this study include the rapid radiative transfer model (RRTM) for the long-wave scheme (Rosenkranz, 2003), the Dudhia shortwave radiation scheme (Dudhia, 1996), the WRF single-moment six-class(WSM6) microphysics scheme (Hong and Lim, 2006), the Noah land surface scheme, the Monin-Obukhov surface layer scheme (Johansson et al., 2001), and the mediumrange forecast (MRF) planetary boundary layer scheme(Hong and Pan, 1996).

    The model uses a single domain, which covers the area 15°-30°N, 104°-122.9°E, with a horizontal grid spacing of 3 km (6 31×501 grid points) and 51 vertical layers reaching up to 33 km.The time step of the model is 30 s.The initial and lateral boundary conditions are obtained from NCEP GFS analyses at 0.25° resolution.

    Fig.1.Illustration of two IAU schemes.The blue area is the traditional IAU technique, of which the relaxation time is τ.The red area is the multi-scale IAU approach, and the relaxation times of the large-scale and small-scale increments are τl and τs, respectively.The abbreviation for relaxation time is rt in the figure.

    3.2.Experimental design

    Some numerical experiments were designed to evaluate the performance of the multi-scale IAU scheme.The blending experiment, illustrated in red in Fig.2, was first designed to ensure that the regional model derived a physically valid state after initialization (Ulmer and Balss, 2016).The background field was integrated from the initial condition of the regional model and downscaled from the global analysis at(t0-6)h, with a 6 h warm-up time.Then, the background fieldGxbat the analysis timet0was generated.The GRAPES_Meso 3DVAR (Xue et al., 2008) and the blending method were successively applied att0.The Gaussian correlation model was used in the GRAPES_Meso 3DVAR system.The horizontal correlation length of specific humidity was shorter at 200 km, while that of the other variances was 500 km.The 3DVAR analysis fieldGxaatt0was obtained by using the 3DVAR system.After that, the 2-D DCT blending method mentioned in section 2.1 was immediately performed to obtain the blended analysis fieldGblndatt0.Specifically, the large-scale analysis increment from NCEP GFSGglobaland the small-scale analysis increment from the 3DVAR analysis fieldGxawere blended to create the blended analysis increment.After the background fieldGxbwas replaced by the blended analysis fieldGblnd, the model performed a 36-h continuous forecast.

    In addition to the blending experiment, the IAU experiment was also designed and presented in blue in Fig.2.The background field was integrated from the global analysis field at (t0-6)h, and the increment obtained from the blending experiment was treated as constant forcing in a model’s prognostic equation through the IAU scheme over the time window (t0-τ/2,t0+τ/2).Here, the center of the time windows was att0, and the relaxation time was τ.Aftert0, the model performed a 36-h continuous forecast.

    For the IAU scheme, the relaxation time, τ, is a crucial parameter that determines the filtering properties of the analysis increments.IAU experiments with different configurations were designed to test initialization sensitivity to relaxation time and find the optimal relaxation time (ORT) for each scale increment.In the seven large-scale IAU experiments, the large-scale increment was added into the integration at every time step over different relaxation times in Typhoon Mangkhut (2018).The CMA tropical cyclone (TC)database was chosen to verify the TC track and intensity forecasts (Ying et al., 2014; Lu et al., 2021).Based on the verified results, the ORT of the large-scale increment could be determined.Similarly, seven small-scale IAU experiments were also designed to obtain the ORT of the small-scale increment.For the multi-scale IAU experiment, the relaxation times of the large- and small-scale increments were derived from the ORTs of the large-scale and small-scale IAU experiments, respectively.Finally, three all-scale IAU (traditional IAU) experiments and a control experiment (without IAU)were also conducted to evaluate the performance of the multi-scale IAU scheme.Table 1 shows the configuration of all numerical experiments in detail.

    4.Analysis of results

    4.1.Description of case study

    In this paper, Typhoon Mangkhut (2018) was chosen as a case study.Mangkhut (2018) was the strongest typhoon to make landfall in China in 2018.It formed as a tropical depression on 7 September 2018 before moving westward and intensifying into a typhoon.On 15 September 2018, it first battered Cagayan, Philippines, and on 16 September 2018, it made a second landfall in Jiangmen, Guangdong Province, China,as an approximately 900-km-wide super typhoon with a central minimum pressure of 955 hPa and maximum sustained winds of 42 m s-1at 0900 UTC.

    Fig.2.Illustration of two categories of numerical experiments, which included the blending experiment (red lines) and the IAU experiment (blue lines).In the blending experiment, the background field G xb was integrated from the global analysis field at (t0-6) h, and the 3DVAR and blending methods were successively applied at the analysis time t0 to obtain the blended analysis field G blnd.In the IAU experiment, the background field was also integrated at (t0-6) h, and the increment, which was calculated in the blending experiment, was added to the integration at every time step during the time window.Both the blending experiment and the IAU experiment performed a 36-h continuous forecast after t0.

    For the Mangkhut (2018) case, 20 comparison experiments with different initialization schemes were conducted(Table 1).According to the model domain shown in Fig.3,the background field, in this case, was first integrated from the global analysis field at 0000 UTC on 15 September 2018, which is presented as the red dot in Fig.3.After 6-h of spin-up time, the analysis time (t0) was 0600 UTC on 15 September 2018 (blue dot in Fig.3).For the blending experiment, the blended analysis increments were obtained by using the 3DVAR and blending methods.In this work, the atmospheric motion vectors (AMVs) derived from the infrared channel (IR1) and water vapor channel (IR3) of the satellite FY-2G were assimilated by the GRAPES_Meso 3DVAR.The temporal and spatial resolutions of the observation data were 6 h and 5 km × 5 km, respectively.The distribution of AMVs shows that 90% of the observation data were high-level winds (above 399 hPa), while the middlelevel (400-699 hPa) and low-level winds (below 700 hPa)accounted for 8% and 2%, respectively.The arrows corresponding to the wind vectors in Fig.3 represent the distribution of high-level winds.In this case, the small-scale increment was only added to the wind field, while the large-scale increment, obtained from the global model, was applied to wind, pressure, water vapor mixing ratio, and potential temperature.Figure 4a shows the wind in the background field at the 200-hPa layer att0, and Figs.4b-c display the large-and small-scale wind increments at 200 hPa, respectively.Aftert0, the model performed a 36-h continuous forecast with outpout every 6-h and ended its simulation at 1800 UTC on 16 September 2018, a timestamp represented by the black dot in Fig.3.

    Table 1.Configuration of 20 numerical experiments.

    Fig.3.The domain of the GRAPES_Meso adopted in this work and the CMA best track and intensity of Typhoon Mangkhut (2018) from 0000 UTC on 15 September to 1800 UTC on 16 September 2018.The numerical experiments began at the red dot and ended at the black dot.The red arrow of wind vectors represents the distribution of high-level winds (above 399 hPa)derived from FY-2G at the analysis time t0, which is denoted by the blue dot.

    Fig.4.(a) Wind in the background field, (b) large-scale, and (c)small-scale wind increments at the 200-hPa layer.All wind fields were obtained at 0600 UTC on 15 September 2018 (t0).

    4.2.Optimal Relaxation Time for IAU Scheme

    Seven large-scale IAU (LS_IAU) and seven small-scale IAU (SS_IAU) experiments were conducted to find the optimal relaxation time in the IAU scheme.The simulation results were analyzed in detail.

    4.2.1.Large-scale IAU Experiment

    Among all large-scale experiments, there was one experiment that did not adopt the IAU scheme (LS_No_IAU),which means that the large-scale increment was completely incorporated into the model integration att0.The other six LS_IAU experiments used the IAU scheme, with relaxation times of 0.5 h, 1.0 h, 1.5 h, 3.0 h, 6.0 h, and 9.0 h, respectively.For comparison, the simulation result of the control(CTL) experiment is also presented in this section.

    The average tendency of the surface pressure can be used as an indicator of the noise level in the external gravity wave component.Figure 5 displays the average tendency of the surface pressure in the CTL experiment and seven LS_IAU experiments.Note that the presented surface pressure tendency was area-averaged over the whole domain and output every 10 min.The red dashed curve represents the surface pressure of LS_No_IAU, of which the peak value reached over 1 hPa (10 min)-1att0.Throughoutt0to(t0+4)h, the surface pressure of LS_No_IAU was evidently larger than that of the other curves.Excessive noise, presented by the red dashed curve, indicates the obvious presence of gravity waves, and the spin-up time here was estimated to be 4 h.The other six solid lines represent LS_IAUs with different relaxation times.The results show that IAU schemes were important in removing high-frequency noise, and larger relaxation times were associated with lower noise levels.Because there was no analysis increment introduced,the surface pressure tendency of CTL, shown as the black dashed line in Fig.5, was minimal.

    The track errors between the simulated track and the CMA best track versus time are presented in Fig.6a.Overall, the track error between the observation and forecast results increased with time.The mean track error in each experiment is shown in Fig.6b.The mean values ranged from 55.6 km to 69.5 km, and the minimum appeared in the experiment with a relaxation time of 6 h (LS_IAU_6h).Compared with CTL, the mean track error of LS_IAU_6h was reduced by about 21%.

    The minimum sea level pressure (MSLP) at the typhoon center was the simple measure of intensity for Typhoon Mangkhut (2018).Figure 7a displays the intensity errors between the forecast and observation results and shows that the simulated MSLP was larger than observed att0.After a 12-h forecast, the simulated MSLP became smaller than the observation data, and the intensity errors increased with time.Figure 7b shows the mean values of the intensity errors obtained from CTL and seven large-scale IAU experiments.The mean intensity errors ranged from 11 to 12.1 hPa, and the minimum value was also obtained from LS_IAU_6h.Compared with CTL, the intensity error of LS_IAU_6h was reduced by about 9.1%.

    After comprehensively considering the track and intensity errors presented in this section, the optimal relaxationtime for LS_IAU in Typhoon Mangkhut (2018) was 6 h.

    Fig.5.Surface pressure tendency versus time.The black and red dashed lines represent CTL and LS_No_IAU, respectively.The other six solid curves are the simulation results of large-scale IAU experiments with different relaxation times.The surface pressure tendency was area-averaged over the whole domain, with output every 10 min.

    4.2.2.Small-scale IAU Experiment

    Similar to the LS_IAU experiments, seven small-scale IAU (SS_IAU) experiments were also conducted.One of them was without the IAU scheme (SS_No_IAU).While the other six applied the IAU technique, and their relaxation times were 0.5 h, 1.0 h, 1.5 h, 3.0 h, 6.0 h, and 9.0 h, respectively.

    Figure 8 displays the average tendency of the surface pressure in eight experiments.The red dashed curve, representing SS_No_IAU, illustrates the obvious presence of spurious gravity waves introduced by the small-scale analysis increment.The surface pressure caused by the small-scale increment was smaller than that caused by the large-scale increment.Moreover, some high-frequency oscillations occurred in the SS_IAU experiments, and the spin-up period of SS_No_IAU was about 5 h.The solid lines in Fig.8 show that the IAU technique effectively removed these noises.

    The typhoon track error versus time is presented in Fig.9a.The track error in each experiment showed a slight decrease in the first 6 h and then gradually increased with time.Figure 9b shows the mean values of the track errors in CTL and the seven SS_IAU experiments.These ranged from 66.6 km to 70.8 km, and the minimum error was obtained from SS_IAU_3h.Compared to CTL, the mean track error of SS_IAU_3h was reduced by about 4.2%.

    Figure 10a shows the intensity errors versus time in the CTL and SS_IAU experiments.Aftert0, the simulated MSLPs were smaller than observations.Figure 10b presents the mean values of the intensity errors simulated by CTLand seven SS_IAU experiments.The mean intensity errors had a small range, between 12.1 and 12.5 hPa; for the SS_IAU experiments, SS_IAU_3h had a minimal intensity error, essentially the same as the CTL.The mean values of the intensity errors obtained from the other six SS_IAU experiments were slightly larger than CTL.Such results imply that introducing a small-scale analysis increment produces little improvement in typhoon intensity forecasts.In this case,the optimal relaxation time for SS_IAU was determined to be 3 h.

    Fig.8.Surface pressure tendency versus time.The black and red dashed lines represent CTL and SS_No_IAU, respectively.The other six solid curves are the simulation results of SS_IAUs with different relaxation times.

    Fig.9.(a) Track errors of Typhoon Mangkhut (2018) versus time and(b) the mean values of track error in CTL and the seven small-scale increment experiments.

    Fig.10.(a) Intensity errors of Typhoon Mangkhut (2018) versus time and (b)the mean intensity errors in CTL and seven small-scale IAU experiments.

    4.3.Best Forecast Results

    The simulation results of the large- and small-scale IAU experiments indicated that the optimal relaxation times for LS_IAU and SS_IAU in Case Mangkhut (2018) were 6 h and 3 h, respectively.Therefore, the best initialization scheme of MS_IAU was the 6-h relaxation time for the large-scale increment and the 3-h relaxation time for the small-scale increment.For comparison, the all-scale IAU experiments with three different relaxation times (1 h, 3 h,and 6 h) were also considered.Eight numerical experiments were compared to test the performance of different IAU schemes on both typhoon path and intensity.These eight experiments included the CTL experiment, the BLND experiment, the best LS_IAU experiment (LS_IAU_6h), the best SS_IAU experiment (SS_IAU_3h), three AS_IAU experiments (AS_IAU_1h, AS_IAU_3h, and AS_IAU_6h), and the MS_IAU experiment (MS_IAU_6h&3h).

    Figure 11a shows the track errors versus time for the eight numerical experiments.Throughout the 36-h integration period, almost all categories of numerical experiments had an increasing trend of track errors with time, and the values of track errors ranged from 12.7 to 169.8 km.The mean values of track errors are presented in Fig.11b, which shows a mean track error of CTL of 69.5 km, representing the maximum among the eight experiments.In addition to CTL, theresults of the three experiments with IAU initialization were larger than those of the BLND experiment; among them was SS_IAU_3h.We suppose that this is mainly caused by two factors 1) compared to the large-scale environmental flow field, the small-scale increment field itself had little effect on the typhoon track performance, and 2) the vertical distribution of the small-scale increment adopted in this case was uneven.Notably, 90% of the small-scale increment consisted of high-level winds (above 399 hPa), while the small-scale, incremental information below 399 hPa was scarce and scattered.Therefore, such information was not sufficient enough to represent the characteristics of the smallscale increment field at medium and low levels.

    Fig.11.(a) Track errors of Typhoon Mangkhut (2018) versus time and (b) the mean track errors in different categories of numerical experiments.

    Although both the large-scale and small-scale increments were incorporated into the model integration in three AS_IAU experiments, the track errors of AS_IAUs with relaxation times of 1 h and 3 h were larger than that of BLND.In contrast, the track performance of AS_IAU_6h was better than that of BLND.The comparative results imply that the track performance was extremely sensitive to the relaxation time of the IAU technique.If the relaxation time was inappropriate, the IAU scheme could even negatively affect the forecast.The minimum track error was 52.5 km, obtained by the MS_IAU_6h&3h.Compared with the CTL, the track errors of MS_IAU_6h&3h decreased by 24.4%.It was also reduced by about 4% in comparison to AS_IAU_6h.

    To further analyze the relationship between the observation and simulation tracks of Typhoon Mangkhut (2018) in detail, four groups of Mangkhut (2018) tracks in the 42-h integration process are illustrated in Fig.12.The red line is obtained from the CMA best track data, while the blue,green, and pink lines represent the typhoon tracks simulated in the CTL, BLND, and MS_IAU_6h&3h experiments,respectively.Compared with the best track, all simulated typhoon tracks made a delayed landfall in China.We surmise that the physical characteristics of GRAPES_Meso may be the main reason for this phenomenon.

    Figures 13a-b illustrate the intensity errors versus time and the mean values of the intensity errors in eight numerical experiments.Their intensity errors had a range of -24.9 to 7.7 hPa.Att0, the simulated MSLPs were larger than the observation.Aftert0, the simulated MSLP graduallybecame smaller than the observation, and the difference between the simulation and observation intensities increased with time.

    Fig.12.Track of Typhoon Mangkhut (2018) from 0000 UTC on 15 September to 1800 UTC on 16 September 2018.The red line is based on the CMA best track of Typhoon Mangkhut (2018), while the blue, green, and pink lines represent the track results obtained from the CTL, BLDN, and MS_IAU_6h&3h experiments, respectively.

    It can be seen from Fig.13b that the mean intensity errors obtained from SS_IAU_3h, AS_IAU_1h, and AS_IAU_3h were larger than those of CTL, which means there may be a negative effect on intensity performance if the analysis increment is incorporated into the model through the IAU technique with an inappropriate time window.It is obvious from Fig.13b that the large-scale analysis increment could improve the performance of the typhoon intensity.Compared with CTL, the mean intensity error of LS_IAU_6h was decreased by 9.2%.Although the smallscale wind increment itself did not play an obvious role in improving the simulated typhoon intensity, the mean values of intensity errors of both AS_IAU_6h and MS_ISU_6h&3h were 0.2 hPa lower than those of LS_IAU_6h.

    In addition, to further evaluate the performance of the MS_IAU scheme, we compared the forecast results against an independent analysis of the operational ECMWF at 0.25°.For four experiments (CTL, BLND, AS_IAU_6h, and MS_IAU_6h&3h), the RMSEs of horizontal winds, U and V, were calculated between the model analyses and the ECMWF analyses, and the vertical profiles of the mean RMSEs are shown in Figs.14a-b.It can be seen from Fig.14 that the mean RMSE of CTL (pink line) and BLND(yellow line) was clearly larger than that of AS_IAU_6h(red line) and MS_IAU_6h&3h (blue line) for both U and V.The mean RMSE of MS_IAU_6h&3h was slightly smaller than that of AS_IAU_6h for analysis of U, while the RMSE of analysis V was nearly the same for MS_IAU_6h&3h and AS_IAU_6h.

    5.Discussion and Conclusion

    To prevent spurious high-frequency gravity waves caused by analysis increments, the IAU technique is applied in the numerical model.In the traditional IAU scheme, all scales of analysis increments are treated as constant additional forcing terms in the model’s prognostic equations during the same relaxation time (τ).However, because different scales of analysis information have different evolutionary speeds and life histories in the model, the traditional IAU scheme has some drawbacks.For this reason, this paper first proposes the multi-scale IAU scheme applying different scales of analysis increments, each having optimal relaxation times in the multi-scale IAU method.

    To find the optimal scheme for the multi-scale IAU technique and further compare the effects of the traditional and multi-scale IAU schemes, 20 numerical experiments with different configurations were conducted.In the experimental design, Typhoon Mangkhut (2018) was selected as the simulation focus, and the 3DVAR and 2D-DCT blending methods were used successively to obtain large- and small-scale analysis increments, respectively.After analyzing and discussing the results of 20 experiments in detail, we arrived at the following conclusions:

    1) For the IAU technique, relaxation time is a crucial parameter.To find the optimal relaxation time for the largescale increment, seven large-scale IAU (LS_IAU) experiments were conducted.One of them was without the IAU method (LS_No_IAU), and the other six were with IAUschemes, but their relaxation times were different.The surface pressure tendency of LS_No_IAU was about six times larger than that in the CTL experiment, which means that the large-scale increment introduced obvious gravity waves.The surface pressure tendency in the other six LS_IAUs was obviously smaller than that in LS_No_IAU, which indicates that the IAU technique played an important role in combating gravity waves.When a 6-h relaxation time was selected, both the track and intensity forecasts of Typhoon Mangkhut (2018) were the best.Compared to CTL, the track and intensity errors of LS_IAU_6h were reduced by about 21% and 9.8%, respectively.Therefore, we regarded 6 h as the optimal relaxation time in the large-scale IAU experiments.

    Fig.13.(a) Intensity errors of Typhoon Mangkhut (2018) versus time and (b) the mean intensity errors in different categories of numerical experiments.

    2) Similar to the LS_IAU experiments, seven smallscale IAU (SS_IAU) experiments were also presented.One of them was without IAU (SS_No_IAU), and the other six had different relaxation times in IAU schemes.The surface pressure tendency of SS_No_IAU indicated that the smallscale increment also introduced obvious gravity waves.Compared to CTL, the track errors and intensity errors of smallscale increment experiments did not show obvious improvement.The simulation results of some numerical experiments were even worse than those of CTL.We infer that this is due to the vertical distribution of the observation data.The small-scale increment adopted in this case was assimilated from the atmospheric motion vectors (AMVs), only 10% of which was distributed below 399 hPa, so AMVs could not fully reveal the characteristics of the small-scale information below 400 hPa, which is the main reason some SS_IAU experiments exerted a negative effect on the forecast performance.However, some SS_IAUs also showed positive effects on the track error of Typhoon Mangkhut (2018).Among all SS_IAU experiments, the forecast results of SS_IAU_3h were the best.Compared to CTL, the track error of SS_IAU_3h was reduced by 4.2%, and the intensity error was the same.Finally, the optimal relaxation time for the SS_IAU experiment was determined to be 3 h for TyphoonMangkhut (2018).

    Fig.14.Vertical profile of the mean analysis RMSEs of (a) U and (b) V for the CTL experiment (pink line),the BLND experiment (yellow line), AS_IAU_6h (red line), and MS_IAU_6h&3h (blue line).

    3) For the multi-scale IAU scheme, the relaxation times of the large- and small-scale increments were selected as 6 h and 3 h, respectively.Moreover, another three traditional IAU experiments with different relaxation times were conducted for comparison purposes.The simulation results show that the track error of Typhoon Mangkhut (2018) in MS_IAU_6h&3h was the smallest.Compared with CTL,the mean track error in MS_IAU_6h&3h decreased by 24.5%.Compared to AS_IAU_6h (the best simulation result in the traditional IAU schemes), the result obtained from the MS_IAU experiment was reduced by about 4%.The mean values of the intensity error in AS_IAU_6h and MS_IAU_6h&3h were equal, which was approximately 9.7%less than CTL.Moreover, the mean RMSE of the zonal wind, U, was smallest for MS_IAU_6h&3h, while that for the meridional wind, V, was nearly the same for MS_IAU_6h&3h and AS_IAU_6h.

    Overall, among all initialization schemes, the forecasting performance of MS_IAU_6h&3h was the best, which shows the superiority of the multi-scale IAU scheme.In future work, we plan to perform statistical work focusing on the optimal cutoff length and relaxation time for the multiscale IAU technique.

    Acknowledgements.This research was jointly sponsored by the Shenzhen Science and Technology Innovation Commission(Grant No.KCXFZ20201221173610028) and the key program of the National Natural Science Foundation of China (Grant No.42130605).The model data in this study are available upon request from the authors via gaoyan@gbamwf.com.

    男女国产视频网站| 日本av手机在线免费观看| 能在线免费看毛片的网站| 一级av片app| 国产日韩欧美在线精品| 老司机影院毛片| 美女大奶头视频| 少妇的逼好多水| 97热精品久久久久久| 国产精品久久久久久久久免| 国产亚洲精品av在线| 日韩一本色道免费dvd| 亚洲精品色激情综合| 毛片一级片免费看久久久久| 精品久久久久久成人av| 黄色日韩在线| 国产成人91sexporn| 国产高清有码在线观看视频| 久久久久久久国产电影| 欧美最新免费一区二区三区| 国产私拍福利视频在线观看| 2021少妇久久久久久久久久久| 国产精品久久久久久久久免| 一个人看的www免费观看视频| 少妇人妻一区二区三区视频| 久久精品久久精品一区二区三区| 国产91av在线免费观看| 波野结衣二区三区在线| 美女黄网站色视频| 国产精品熟女久久久久浪| 久久精品国产99精品国产亚洲性色| 久久久久久久久中文| 日韩欧美 国产精品| 亚洲一级一片aⅴ在线观看| 欧美成人a在线观看| 晚上一个人看的免费电影| 尾随美女入室| 久久久午夜欧美精品| 国产一区二区在线av高清观看| 毛片女人毛片| 天堂影院成人在线观看| 欧美最新免费一区二区三区| 亚洲美女视频黄频| av视频在线观看入口| 亚洲一区高清亚洲精品| 免费av观看视频| 日韩一本色道免费dvd| 国产精品日韩av在线免费观看| 久久久久网色| 一级爰片在线观看| 插逼视频在线观看| 精品一区二区免费观看| 色播亚洲综合网| 极品教师在线视频| 精品少妇黑人巨大在线播放 | 久久韩国三级中文字幕| 久久精品国产99精品国产亚洲性色| 伦精品一区二区三区| 国产极品天堂在线| 国产亚洲午夜精品一区二区久久 | 男女视频在线观看网站免费| 晚上一个人看的免费电影| 日本熟妇午夜| 亚洲精品日韩av片在线观看| 成人漫画全彩无遮挡| 一级黄片播放器| 波多野结衣巨乳人妻| 天天躁夜夜躁狠狠久久av| 国产一区二区在线av高清观看| 能在线免费观看的黄片| 成人欧美大片| ponron亚洲| 我的女老师完整版在线观看| 少妇猛男粗大的猛烈进出视频 | 亚洲人与动物交配视频| 亚洲人成网站在线观看播放| 久久热精品热| 床上黄色一级片| 简卡轻食公司| 一级爰片在线观看| 黄色配什么色好看| 国产免费福利视频在线观看| 中文字幕av成人在线电影| 欧美日韩在线观看h| 亚洲av成人精品一二三区| 免费观看a级毛片全部| 九九久久精品国产亚洲av麻豆| 国产亚洲午夜精品一区二区久久 | 欧美人与善性xxx| 精品酒店卫生间| 欧美日韩一区二区视频在线观看视频在线 | 波野结衣二区三区在线| 只有这里有精品99| 国产精品女同一区二区软件| 韩国高清视频一区二区三区| 久久久久免费精品人妻一区二区| 久久99蜜桃精品久久| 久久综合国产亚洲精品| 婷婷色av中文字幕| 日本熟妇午夜| 久久这里只有精品中国| 亚洲婷婷狠狠爱综合网| 亚洲国产精品久久男人天堂| 黄色日韩在线| 免费不卡的大黄色大毛片视频在线观看 | 国产老妇伦熟女老妇高清| 亚洲天堂国产精品一区在线| 日本黄色视频三级网站网址| 国产精品人妻久久久影院| 国产片特级美女逼逼视频| 波多野结衣巨乳人妻| 国产精品福利在线免费观看| 久久精品久久精品一区二区三区| 亚洲精品乱码久久久v下载方式| 亚洲成色77777| 久久这里只有精品中国| 美女cb高潮喷水在线观看| 欧美一级a爱片免费观看看| 国产女主播在线喷水免费视频网站 | 中国国产av一级| 欧美3d第一页| 黄色配什么色好看| 色综合亚洲欧美另类图片| 亚洲综合精品二区| 精品久久久噜噜| 男人舔奶头视频| 久久综合国产亚洲精品| 亚洲久久久久久中文字幕| 在线天堂最新版资源| 亚洲综合色惰| 老司机影院成人| 国产伦精品一区二区三区四那| 亚洲真实伦在线观看| 一夜夜www| 亚洲美女搞黄在线观看| 熟女电影av网| 亚洲国产欧美人成| 女人被狂操c到高潮| 伦理电影大哥的女人| 菩萨蛮人人尽说江南好唐韦庄 | 少妇高潮的动态图| 久久久久久伊人网av| 国产精品嫩草影院av在线观看| 国产又色又爽无遮挡免| 亚洲成人精品中文字幕电影| 免费观看性生交大片5| 亚洲图色成人| 天堂中文最新版在线下载 | 在线观看av片永久免费下载| 舔av片在线| 国产一级毛片在线| 免费av不卡在线播放| 国产av不卡久久| 韩国av在线不卡| 不卡视频在线观看欧美| 亚洲综合色惰| 乱系列少妇在线播放| 国产午夜精品论理片| 看非洲黑人一级黄片| 禁无遮挡网站| 亚洲国产精品久久男人天堂| 久久久精品大字幕| 国产美女午夜福利| 国产精品久久久久久久久免| 免费黄网站久久成人精品| 国产伦精品一区二区三区四那| 亚洲成人av在线免费| 国产真实伦视频高清在线观看| 中文精品一卡2卡3卡4更新| 女人十人毛片免费观看3o分钟| 大香蕉97超碰在线| 中文字幕亚洲精品专区| 欧美97在线视频| 国产高清不卡午夜福利| 国产一区二区三区av在线| 久久精品影院6| 午夜老司机福利剧场| 老司机福利观看| 国产免费男女视频| 中文字幕av成人在线电影| 一区二区三区四区激情视频| 午夜久久久久精精品| 日本免费一区二区三区高清不卡| 我要搜黄色片| 国产 一区 欧美 日韩| 久久精品人妻少妇| 99久久无色码亚洲精品果冻| 一级爰片在线观看| 人人妻人人澡欧美一区二区| 六月丁香七月| 青春草视频在线免费观看| 中文欧美无线码| 亚洲精品国产av成人精品| 18+在线观看网站| 亚洲在久久综合| 搡女人真爽免费视频火全软件| 一本一本综合久久| 欧美激情国产日韩精品一区| 久久久久久久久久黄片| 久久久久性生活片| 国产老妇女一区| 亚洲国产欧美在线一区| 麻豆av噜噜一区二区三区| 在线观看66精品国产| 99在线人妻在线中文字幕| 国产精品熟女久久久久浪| 免费大片18禁| 国产成人精品婷婷| 亚洲欧美日韩卡通动漫| 久久久久性生活片| 大又大粗又爽又黄少妇毛片口| 久久久欧美国产精品| 精品熟女少妇av免费看| 国产精品av视频在线免费观看| 搡老妇女老女人老熟妇| 成年免费大片在线观看| 男女那种视频在线观看| 国产精品.久久久| 自拍偷自拍亚洲精品老妇| 97在线视频观看| 国产精品久久久久久精品电影小说 | 欧美日本视频| 国内揄拍国产精品人妻在线| 国产麻豆成人av免费视频| 精品酒店卫生间| 久久精品综合一区二区三区| 国产精品久久电影中文字幕| 22中文网久久字幕| 一二三四中文在线观看免费高清| 99热精品在线国产| 色噜噜av男人的天堂激情| 亚洲成人久久爱视频| 国产精品一区二区三区四区免费观看| АⅤ资源中文在线天堂| 成人亚洲精品av一区二区| 永久免费av网站大全| 两个人视频免费观看高清| 午夜精品一区二区三区免费看| 欧美日韩在线观看h| 好男人在线观看高清免费视频| 伊人久久精品亚洲午夜| 天堂√8在线中文| 免费黄网站久久成人精品| 久久欧美精品欧美久久欧美| 最后的刺客免费高清国语| 国产精品99久久久久久久久| 五月玫瑰六月丁香| 亚洲人成网站在线播| 国内少妇人妻偷人精品xxx网站| 亚洲婷婷狠狠爱综合网| 日日摸夜夜添夜夜添av毛片| 爱豆传媒免费全集在线观看| 亚洲av电影不卡..在线观看| 久久久久久九九精品二区国产| 午夜福利在线观看吧| 久久亚洲国产成人精品v| 日韩大片免费观看网站 | 亚洲精品亚洲一区二区| 精品久久久久久久末码| 特级一级黄色大片| АⅤ资源中文在线天堂| 精品久久久久久久久久久久久| 只有这里有精品99| 精品一区二区三区视频在线| 亚洲色图av天堂| 欧美性感艳星| 天堂网av新在线| 日本wwww免费看| 精品无人区乱码1区二区| 在线天堂最新版资源| 一区二区三区四区激情视频| 欧美高清成人免费视频www| 国产精品人妻久久久影院| 亚洲不卡免费看| 久久热精品热| 男女视频在线观看网站免费| 亚洲三级黄色毛片| 国产 一区精品| 亚洲av熟女| 午夜精品一区二区三区免费看| 91久久精品电影网| 亚洲国产欧美人成| 精品无人区乱码1区二区| 内射极品少妇av片p| 村上凉子中文字幕在线| 久久久精品94久久精品| 亚洲国产色片| 日韩成人伦理影院| 久久国产乱子免费精品| 亚洲高清免费不卡视频| 国产精品电影一区二区三区| 国产亚洲5aaaaa淫片| 嫩草影院入口| av在线老鸭窝| 国产亚洲5aaaaa淫片| 一边摸一边抽搐一进一小说| 18禁在线无遮挡免费观看视频| 午夜福利视频1000在线观看| 麻豆成人av视频| 国产成人91sexporn| 亚洲av免费高清在线观看| 免费人成在线观看视频色| 国产亚洲5aaaaa淫片| 精品久久国产蜜桃| av女优亚洲男人天堂| 91狼人影院| 2021天堂中文幕一二区在线观| videossex国产| 韩国高清视频一区二区三区| 黄色日韩在线| 最近最新中文字幕免费大全7| 国产中年淑女户外野战色| 午夜免费男女啪啪视频观看| 久久99精品国语久久久| av在线播放精品| 成人亚洲精品av一区二区| 国产成人精品久久久久久| 国产视频首页在线观看| 国产v大片淫在线免费观看| 日本黄色视频三级网站网址| 日韩强制内射视频| 久久久久久久国产电影| 亚洲欧美精品自产自拍| or卡值多少钱| 国产精品永久免费网站| 午夜福利高清视频| 两个人视频免费观看高清| 亚洲av成人精品一区久久| 人人妻人人澡欧美一区二区| 午夜福利在线观看吧| 少妇的逼水好多| 日本wwww免费看| 能在线免费观看的黄片| 精品少妇黑人巨大在线播放 | 国产伦精品一区二区三区四那| 五月玫瑰六月丁香| 观看免费一级毛片| 久久久成人免费电影| 在线免费观看不下载黄p国产| 三级国产精品欧美在线观看| 亚洲最大成人手机在线| 精品国产一区二区三区久久久樱花 | 韩国av在线不卡| 又粗又爽又猛毛片免费看| 国产精品嫩草影院av在线观看| 国产白丝娇喘喷水9色精品| 国产一区二区在线av高清观看| 少妇猛男粗大的猛烈进出视频 | 99热这里只有精品一区| 精品酒店卫生间| 日韩大片免费观看网站 | 欧美日本视频| 久久久久久久久大av| 亚洲欧美一区二区三区国产| 大香蕉久久网| 日本猛色少妇xxxxx猛交久久| 亚洲av电影不卡..在线观看| 永久网站在线| 中国国产av一级| 在线播放国产精品三级| 久久欧美精品欧美久久欧美| 一级毛片我不卡| 尾随美女入室| АⅤ资源中文在线天堂| 高清毛片免费看| 亚洲成人中文字幕在线播放| 精品无人区乱码1区二区| 男女那种视频在线观看| 亚洲怡红院男人天堂| 一卡2卡三卡四卡精品乱码亚洲| 国产 一区精品| 久久精品国产99精品国产亚洲性色| 国产精品无大码| 建设人人有责人人尽责人人享有的 | 国产男人的电影天堂91| 国产色爽女视频免费观看| 国产精品国产三级国产av玫瑰| 国产综合懂色| av免费在线看不卡| 天堂影院成人在线观看| 最近最新中文字幕免费大全7| 丝袜喷水一区| 日韩大片免费观看网站 | 亚洲欧洲国产日韩| 成年av动漫网址| 国产伦精品一区二区三区视频9| 色尼玛亚洲综合影院| 国内揄拍国产精品人妻在线| 99久久人妻综合| 精品国产三级普通话版| 亚洲国产精品国产精品| 久久久成人免费电影| 天堂影院成人在线观看| 亚洲av免费在线观看| av黄色大香蕉| 亚洲自偷自拍三级| 午夜精品一区二区三区免费看| 99在线人妻在线中文字幕| 日韩一区二区视频免费看| 色哟哟·www| 秋霞伦理黄片| 国产高清国产精品国产三级 | 久久久久久久久久黄片| 成人性生交大片免费视频hd| 成人无遮挡网站| 精品久久久噜噜| 成人三级黄色视频| av国产久精品久网站免费入址| 天堂网av新在线| 国产不卡一卡二| 中文乱码字字幕精品一区二区三区 | 简卡轻食公司| 一级黄色大片毛片| 嫩草影院精品99| 国产一区有黄有色的免费视频 | 国产高潮美女av| 91久久精品国产一区二区成人| 日韩欧美在线乱码| 色吧在线观看| 嘟嘟电影网在线观看| 欧美又色又爽又黄视频| 1024手机看黄色片| 精品不卡国产一区二区三区| 少妇裸体淫交视频免费看高清| 国产精品电影一区二区三区| 国产成人免费观看mmmm| 欧美成人精品欧美一级黄| 又粗又爽又猛毛片免费看| 男人和女人高潮做爰伦理| 少妇熟女aⅴ在线视频| 综合色丁香网| 国产成人freesex在线| 最近中文字幕高清免费大全6| 久久久久久久久久久免费av| 日韩高清综合在线| 身体一侧抽搐| 国产美女午夜福利| 91aial.com中文字幕在线观看| 极品教师在线视频| 久久久久久伊人网av| 免费观看精品视频网站| 国产高清国产精品国产三级 | 插阴视频在线观看视频| 国产极品精品免费视频能看的| 毛片女人毛片| 日韩精品青青久久久久久| 国产单亲对白刺激| 热99在线观看视频| 国产 一区 欧美 日韩| 精品久久久久久久久av| 欧美日韩国产亚洲二区| 欧美zozozo另类| 国产av码专区亚洲av| 精品欧美国产一区二区三| 国内少妇人妻偷人精品xxx网站| 国产精品一二三区在线看| 久久精品国产自在天天线| 中文在线观看免费www的网站| 天堂√8在线中文| 中文资源天堂在线| 亚洲av一区综合| 色综合色国产| 亚洲四区av| 午夜福利在线在线| 国产91av在线免费观看| a级一级毛片免费在线观看| 久久久精品94久久精品| 一个人看视频在线观看www免费| 美女内射精品一级片tv| 中文天堂在线官网| 极品教师在线视频| 日韩三级伦理在线观看| 人人妻人人看人人澡| 校园人妻丝袜中文字幕| 欧美最新免费一区二区三区| 乱系列少妇在线播放| 日本色播在线视频| 又黄又爽又刺激的免费视频.| 久久久久久大精品| 亚洲精华国产精华液的使用体验| 久久久久久大精品| 在线播放无遮挡| 国产高清国产精品国产三级 | 国产精品野战在线观看| 国产私拍福利视频在线观看| 亚洲不卡免费看| 26uuu在线亚洲综合色| 国产麻豆成人av免费视频| 永久免费av网站大全| 一本一本综合久久| 国产极品精品免费视频能看的| av卡一久久| 尾随美女入室| 日韩三级伦理在线观看| 99热这里只有精品一区| 国产极品天堂在线| 非洲黑人性xxxx精品又粗又长| 亚洲av日韩在线播放| 日本一二三区视频观看| 精品国产一区二区三区久久久樱花 | 国产黄色小视频在线观看| 天堂av国产一区二区熟女人妻| 亚洲av免费在线观看| av在线老鸭窝| 亚洲自偷自拍三级| 精品久久久噜噜| 亚洲熟妇中文字幕五十中出| 大又大粗又爽又黄少妇毛片口| 亚洲自偷自拍三级| 久久久精品大字幕| 又粗又硬又长又爽又黄的视频| 精品国内亚洲2022精品成人| 久久精品影院6| 一级毛片aaaaaa免费看小| 长腿黑丝高跟| 亚洲在线自拍视频| 欧美精品国产亚洲| 国产探花在线观看一区二区| 国国产精品蜜臀av免费| 波多野结衣巨乳人妻| 久久久精品欧美日韩精品| 少妇熟女aⅴ在线视频| 丰满少妇做爰视频| 国产精品综合久久久久久久免费| 国产精品一二三区在线看| 国产一区二区在线观看日韩| 18禁动态无遮挡网站| 欧美bdsm另类| 别揉我奶头 嗯啊视频| 国产精品国产三级国产av玫瑰| 欧美一级a爱片免费观看看| 国产成人精品婷婷| 成人一区二区视频在线观看| 国产高清视频在线观看网站| 啦啦啦韩国在线观看视频| 中文精品一卡2卡3卡4更新| 啦啦啦观看免费观看视频高清| 一本—道久久a久久精品蜜桃钙片 精品乱码久久久久久99久播 | 精品国产露脸久久av麻豆 | 我要搜黄色片| 亚洲欧美日韩东京热| 91精品伊人久久大香线蕉| 精品不卡国产一区二区三区| 免费观看的影片在线观看| 亚洲电影在线观看av| 中文乱码字字幕精品一区二区三区 | 中国美白少妇内射xxxbb| 日本免费a在线| 色综合色国产| 日韩亚洲欧美综合| 国产一区二区三区av在线| 亚洲在线自拍视频| 两个人视频免费观看高清| 91久久精品电影网| 别揉我奶头 嗯啊视频| 亚洲最大成人av| 熟女电影av网| 国产精品伦人一区二区| 少妇的逼好多水| h日本视频在线播放| 99热6这里只有精品| 久久婷婷人人爽人人干人人爱| 天堂网av新在线| 最近视频中文字幕2019在线8| 亚洲性久久影院| 91久久精品国产一区二区成人| 99久久精品一区二区三区| 91午夜精品亚洲一区二区三区| 国产片特级美女逼逼视频| 亚洲欧美精品专区久久| 欧美成人精品欧美一级黄| 国产极品天堂在线| 国产一级毛片在线| 久久99热这里只有精品18| av在线老鸭窝| 中文字幕熟女人妻在线| 久久精品国产亚洲av涩爱| 国产精品久久视频播放| 日本-黄色视频高清免费观看| 国产乱来视频区| 国产精品av视频在线免费观看| 一级黄片播放器| 日韩精品有码人妻一区| 丝袜喷水一区| 欧美精品国产亚洲| 欧美日本视频| 日韩av在线大香蕉| 欧美区成人在线视频| 性色avwww在线观看| 日本黄大片高清| 非洲黑人性xxxx精品又粗又长| 成人三级黄色视频| 国产乱人偷精品视频| 国产精品人妻久久久影院| 亚洲av中文字字幕乱码综合| 在线观看美女被高潮喷水网站| 亚洲经典国产精华液单| 老司机影院成人| 日本与韩国留学比较| 国产精品久久电影中文字幕| 国产成年人精品一区二区| a级一级毛片免费在线观看| 天美传媒精品一区二区| 国产毛片a区久久久久| 国产精华一区二区三区| 亚洲精品日韩在线中文字幕| 欧美成人免费av一区二区三区| 日本wwww免费看| 可以在线观看毛片的网站| 校园人妻丝袜中文字幕| 欧美区成人在线视频| 狂野欧美白嫩少妇大欣赏| 黄片wwwwww| 天堂网av新在线| 久久综合国产亚洲精品|