• <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 ?

    Network analysis of Chinese air transport delay propagation

    2017-11-20 12:06:39MssimilinoZninSeddikBelkourZhuYnbo
    CHINESE JOURNAL OF AERONAUTICS 2017年2期

    Mssimilino Znin,Seddik Belkour,Zhu Ynbo

    aInnaxis Foundation&Research Institute,Jose′Ortega y Gasset 20,28006 Madrid,Spain

    bFaculdade de Cie?ncias e Tecnologia,Departamento de Engenharia Electrote′cnica,Universidade Nova de Lisboa,Lisboa,PortugalcAviation Data Communication Corporation,100083,Beijing,China

    Network analysis of Chinese air transport delay propagation

    Massimiliano Zanina,b,*,Seddik Belkouraa,Zhu Yanboc

    aInnaxis Foundation&Research Institute,Jose′Ortega y Gasset 20,28006 Madrid,Spain

    bFaculdade de Cie?ncias e Tecnologia,Departamento de Engenharia Electrote′cnica,Universidade Nova de Lisboa,Lisboa,PortugalcAviation Data Communication Corporation,100083,Beijing,China

    Air transport;Complex networks;Delays propagation;Functional networks;Granger causality

    The Chinese air transport system has witnessed an important evolution in the last decade,with a strong increase in the number of flights operated and a consequent reduction of their punctuality.In this contribution,we propose modelling the process of delay propagation by using complex networks,in which nodes are associated to airports,and links between pairs of them are assigned when a delay propagation is detected.Delay time series are analysed through the wellknown Granger Causality,which allows detecting if one time series is causing the dynamics observed in a second one.Results indicate that delays are mostly propagated from small and regional airports,and through flights operated by turbo-prop aircraft.These insights can be used to design strategies for delay propagation dampening,as for instance by including small airports into the system’s Collaborative Decision Making.

    1.Introduction

    The Chinese air transportation system,one of the most important and fastest-growing in the world,is receiving an increasing attention from the research community due to its distinctive characteristics and challenges.On one hand,during the last decade,China has become the second largest market,with the Beijing Capital International Airport being the second busiest airport in the world.On the other hand,China is suffering from a limited availability of airspace resources,among others due to a strong military presence.1As a result,nowadays passengers are experiencing significant delays,with an average of 40%of flights being delayed every day.According to the flight statistics of Civil Aviation Administration of China(CAAC),the appearance of delays can be attributed to one of the four main categories:airlines related(≈40%),air traffic control(≈20%),weather related(≈20%),and military uses(≈10%).These four categories are responsible for more than 90%of the flight delays in China.

    In this work,we propose the use of complex network theory to understand how delays appear and propagate through the Chinese airport network.Complex networks are powerful tools to help understanding the structures and dynamics ofcomplex systems in general,2,3and of air transportation in particular.Airport networks and air route networks have been the focus of many studies around the world-see,for instance,Refs.4,5for some examples.Specifically to the Chinese system,examples include Refs.6–10.

    The structure induced by delay propagation between airports can be seen as a complex network,which lies on top of the network created by flights and which is the result of the dynamics of the system.This is similar to what has been considered in neuroscience:flight networks are equivalent to the way different brain regions are physically connected(i.e.the connectome),while a delay network represents how information(here,information about delays)is transmitted and processed at each node.11Once such network has been reconstructed,it can be used to obtain relevant information about the propagation process:for instance,which are the most important airports,both for delay generation and propagation;or how different airlines interact to produce the observed dynamics.

    In the line of what has already been studied for the European network,12,13we here propose to reconstruct delay propagation networks for the Chinese air transportation system by applying the well-known Granger Causality metric14,15on airport delay time series.A link between two airports will thus represent a direct delay propagation,i.e.a situation in which the delays observed at one airport are partly responsible for what we observe in a second airport.By characterising the resulting network using several established metrics,we make a first step towards the understanding of the dynamics of delay propagation in the Chinese air transport network,and towards the development of adequate strategies for improving the passengers’experience.Note that the approach here proposed is fundamentally different from standard delay studies.On the one hand,we here focus on the global structure created by delay propagation,and not on the local dynamics of individual flights and airports;16,17additionally,results are obtained by observing the real dynamics of the system,and not by inferring it from large-scale synthetic models.18,19

    Beyond this introduction,this work is organised as follows.Section 2 presents the main techniques here applied,including a description of the data set(Section 2.1),the creation and preprocessing of delay time series(Section 2.2),the reconstruction of propagation networks through the Granger Causality(Section2.3),andtheanalysisoftheresultingnetworks(Section2.4).Section 3 reports the results obtained,by considering three differentnetworks:theglobaldelaypropagationone,thenetworks created by the three most important airlines,and the one created by aircraft types.Finally,Section 4 summarises our major conclusions and discusses future lines of research.

    2.Methods

    2.1.Data set description

    The data set has been provided by Aviation Data Communication Corporation(ADCC)of CAAC,and comprises high-level information for all flights crossing the Chinese airspace during the 2015 summer peak(1st to 31st August 2015).For each flight,available information includes,among others,the scheduled departure and arrival airports,scheduled and real departure and arrival time,as well as the airline code and aircraft type.Only domestic flights,i.e.departing from and arriving at a Chinese airport,have here been considered;additionally,flights involving non-ICAO airports,marked as ZZZZ,have further been discarded.Cancelled flights were also not included in the data set,which only covers executed operations.This yielded a final data set comprising 152 airports and 45.151 flights-see Fig.1 for a graphical representation.

    The delay of each flight has been calculated as the difference between the real and scheduled arrival time,and therefore represents the real delay as perceived by passengers.It has to be noted that this is different from the official announced delay,as the latter is not reported if the flight arrived within 10/15 min from the expected arrival time.

    2.2.Time series pre-processing

    Once the landing delay of each flight has been assessed,it is necessary to reconstruct the average dynamics of each airport,i.e.the average landing delay within one-hour time windows.An example of the resulting time series is represented in Fig.2(a),for the Beijing Capital International Airport.It can be appreciated that the time series is characterised by peaks appearing with a daily frequency-note the peaks at 24 h and multiples in the cross-correlation depicted in Fig.2(b).This is mainly due to the non-stationarity of delays,whose expected value is higher at the end of the afternoon,and lower during the first hours of the day-see Fig.2(c).

    In order to reduce the non-stationarity of the time series,which may result in a biased evaluation of the Granger Causality metric,we here apply a Z-Score detrend procedure.In detail,the detrended delay time series for one airport is calculated as:

    D0ed;hT being the detrended delay for daydand timeh,Ded;hT the originalaverage delay,and hD(;hTiand reD);hTT the average and standard deviation of the delays recorded at time h for all available days.A graphical representation of this transformation is depicted in Fig.3.

    The resulting time series have some interesting characteristics.Firstofall,theiraveragevalueiszerobyconstruction-note that we subtract the average delay hD(;hTi from each element.Second,positive and negative delays represent how unusual(in terms of the number of standard deviations)is the observed delay with respect to the delay expected at the same hour.Also note that,while negative delay values are usually discarded in delaypropagationstudies,heretheyencodeimportantinformation,as they represent situations in which the behaviour of the systemisbetterthanexpected;theobjectivethusincludesunderstanding if such positive situations can be propagated between different airports.Finally,and as a direct consequence of the previous point,the resulting time series are detrended:if,at a giventime,ahighdelayisobserved,butthisisalignedwithwhat expected in that time window,D0will be close to zero.

    2.3.Granger causality

    The Granger Causality(GC)test14,15has extensively been used to assess the information exchanged between different elements of a system,and more specifically,to test for causality relationships that one might intuitively assume.GC is an extre-mely powerful tool,developed by the economy Nobel Prize winner Clive Granger14(even though a similar notion might have been introduced twenty years earlier by Weiner),that have extensively been applied to economic problems20and biomedical data analysis.21–23In this work,GC will help understanding if the dynamics of two airports are related;and,in terms of delays,if the delays observed at one of them can explain the delays appearing at a second one.

    Fig.1 Map depicting the position of the 152 airports considered in this study.

    Fig.2 Average hourly landing delay at ZBAA-Beijing Capital International Airport.

    The GC is based on two very intuitive concepts:first,causes must precede the corresponding effects;and second,the effect can be better predicted using the historical information related to both the cause and the effect than when just using the effect alone.This translates to our study in the following way:an airportB‘Granger-causes” another airportAif the use of past values of the delays time series ofBhelps improving the prediction of delay time series ofA.This would indicate that there is a flow of information fromBtoA,and that thusBis(partially)responsible of the delays we observe inA.

    In mathematical terms,suppose that the delay dynamics of two airportsAandBcan be expressed as stationary time series.ThenB‘Granger-causes”Aif:

    Fig.3 Example of the detrending process of the ZBAA-Beijing Capital International Airport time series.

    where r2eAjU-T stands for the error(i.e.,the standard deviation of residuals)in forecasting the time seriesAusing the past information of the entire universeU,and r2eAjU-nB-T the error when the information about time seriesBis discarded.The forecast itself is done,in both cases,through an autoregressive-moving-average(ARMA)model.24AnF-test is then performed to assess the statistical significance of this inequality.

    The main advantage of the GC metric is that it is able to identify true causalities and discard spurious correlations.Indeed,in the case of spurious correlations,the information contained in one time series is useless to the forecast of the other,as both would be essentially the same.On the other hand,several drawbacks have to be taken into account when analysing the results yielded by GC.25First,any claim of causality must be taken with caution as a true causality can only be guaranteed considering all possible relevant source of information,26which is not usually the case in real-world experiments.Second,GC is linear and looks for linear couplings between the time series-delays,on the other hand,are known to be driven by non-linear propagation forces.Finally,GC considers the entire time series;an abnormal delay propagation,i.e.one appearing only under some special circumstances,would be smoothed and averaged,therefore loosing its statistical significance.13

    Additionally,it is necessary to stress that,while Granger Causality has successfully been applied to the study of numerous real systems,it was not designed to understand delay propagation,a process that presents several idiosyncrasies.First,delays propagate in a non-trivial way:a small delay in a flight(e.g.below 10–15 min)can usually be recovered during the enroute phase,while larger delays are linearly propagated to the destination airport.As GC is a metric that mostly detects linear causalities,it can yield unreliable results when the delay propagated is close to,or smaller than,the previously mentioned threshold.Second,a true causality can only be guaranteed considering all possible relevant sources of information,a condition that is not fulfilled when constructing a network by a pairwise node analysis.Third,the hypothesis that causes must precede the corresponding effects may not be always fulfilled in the case of delays:it is in principle possible that,due to collaborative decision making measures,future restriction capacities at one airport may generate delays in the present in another one,which,in terms of GC,would correspond to a reversal of the time arrow.Finally,GC is ill-designed to analyse time series created by discrete events;speci fically,in the case at hand,many small airports have delay values at only some speci fic times,i.e.when an aircraft arrives,such that most values of their time series are not relevant.This reduces the significance of the links detected,especially for small airports.Due to these limitations,the degree observed for small airports may be in flated;nevertheless,the global messages presented in this contribution will qualitatively be valid.

    2.4.Network reconstruction and evaluation

    Once a set of stationary time series has been obtained,one for each of thenairports considered,GC can be used to reconstruct the corresponding delay propagation network.Any network can be created by defining its adjacency matrix A,a matrix of sizen-nwithnbeing the number of nodes,whose elementai,jis equal to one when nodesiandjare connected,and zero otherwise.2–4Consequently,in order to reconstruct the delay propagation network,all pairs of airports(i,j)have to be tested by calculating the GCF-value between the corresponding time series,andai,jhas to be set to one if theF-value is statistically significant(significance level of 0.05).

    At this stage,it is important to describe what phenomena are represented by the resulting networks.First of all,a link between two airports(i,j)indicates that the delay at landing atjis driven by the delay previously observed ati.As delays are measured at landing,each link synthesises all the elements taking place between two landings:that is,not just the flight itself,but also the turn-around process at the airporti,the potential delay generated by connecting passengers,etc.Finally,a similar result may be obtained by considering the average delay at take-off-note that,in this case,the(i,j)link would include information about the turn-around process at airportjand not ati.

    Different topological metrics can then be extracted from A,each of them describing some specific structural characteristic of the network.27In this work,we consider the following ones,for being the most relevant from the point of view of delay propagation:

    (1)Link density is defined as the proportion of links that are active,with respect to the total number of potential links,i.e.

    Following the standard notation,nandlrespectively denote the number of nodes and links in the network.The higherld,the more connected is the network:delays are then easily propagated through it,with less possibilities for stopping the diffusion process.

    (2)Transitivity measures the presence of triangles in the

    network.28It is mathematically defined as the relationship between the number of triangles in the networkNΔ(three nodes with edges between each pair of them),and the number of connected triplesN3(groups of three connected nodes):

    The presence of a high number of triangles indicates that groups of three airports are strongly connected,such that a delay generated in any one of them is easily propagated to the others.

    (3)Assortativity representstheconditionalprobabilityP(k0|k),i.e.the probability of a link from a node of degreekto point to a node of degreek0.When expressed mathematically,it is defined as29:

    Mbeing the total number of links in the network.A positive assortativity indicates that nodes tend to connect with equally connected nodes.A negative value,on the other side,indicates the abundance of highly connected-isolated nodes links.From the point of view of delay propagation,a positive assortativity suggeststhat delay-propagating airports are preferentially connected between them,thus creating communities in which delays freely diffuse.

    (4)Efficiency of a network represents how easily information(here,delays)can move between two nodes,i.e.how many transitional nodes one has to pass by in order to reach the ‘destination’.The efficiency is defined by considering the inverse of the harmonic mean of the distances between pairs if nodes30:

    dijbeing the distance(or the number of jumps)between nodesiandj.

    (5)Diameter is the greatest distance between any pair of nodes.It thus indicates how many flights are needed for delays to spread across the whole system.

    (6)Information content(IC).This metric assesses the presence of regularities in the adjacency matrix,i.e.of mesoscale structures that contain small amounts of nodes.31It merges pairs of nodes that imply the lowest loss of information,until shrinking the network to a single node.This metric thus represents the amount of information encoded within the structure of the network where small values correspond to complex topology and higher values to a random-like structure.

    In some cases,it is known that the value of the metric alone is not enough to understand if the associated topological feature is relevant or not.For instance,the efficiencyEstrongly depends both on the structure of the network,and on its number of links.ComparingEfor two networks is thus meaningless,unless the value is normalised with respect to the number of links composing each network.This is accomplished by creating a large ensemble(n=1000)of random Erdo¨s-Renyi(ER)graphs,32each one of them with the same number of nodes and links as in the original network.The Z-Score of the valuemof the metricMis then calculated as:

    lMand rMrespectively being the average and the standard deviation of the ensemble values.As in the time series detrending process(see Section 2.2),large positive and negative values ofmZ–Scorerespectively indicate metrics significantly above or below the expected value.

    3.Results

    3.1.Delay propagation through airports

    As an initial analysis,we consider the reconstruction of networks representing how delays propagate through different airports.To this end,a time series has been calculated for each airport,representing the average landing delay in one-hour window;pairs of airports(i.e.nodes in the network)have then been connected by a link if a statistically significant(F-value<0.05)Granger Causality is detected between the corresponding time series.As previously discussed in Section 2.4,two airports are then connected if a delay propagation is inferred from the delay time series,allowing us to reconstruct the whole structure created by delays propagation.

    Fig.4(in all cases,the size of nodes is proportional to their number of operations,and their colour to the out-degree(from green to red))depicts the resulting network.As the network is too densely connected to allow for a visual inspection,the same figure depicts the sub-network created by the 30 busiest airports(right panel).As an initial observation,it can be appreciated that the top-30 network is relatively sparse,indicating that few delays are propagated between the most important airports.

    In order to better understand the role played by the different airports,Fig.5(a)and(b)reports the out-and in-degrees of nodes,as a function of the number of flights they operated.A clear inverse relation can be observed,indicating that the most important airports in terms of delays propagation are the smallest ones.This is aligned with what observed in other air transport networks,as for instance the European33,12and USA ones.18Fig.5(c)depicts the relation between the in-and out-degree of nodes,the former representing how strongly one airport is influenced by its neighbours,the latter how strongly it can influence other airports.A direct relation between both quantities can be observed,as illustrated by the linear fit(red dashed line,slope=0.618,r=0.643).A couple of interesting exceptions can be found,as for instance ZWHM,ZLHZ and ZUKJ,which mostly transmit delays to the network;and ZUBD,which mainly receives them.Additionally,Fig.6 depicts the in-and out-degree of nodes as a function of the average delay observed at each airport.Due to the high correlation between airport sizes and delays,results are qualitatively similar to what reported in Fig.5(a)and(b).

    In order to validate the previously presented results,the case of the Beijing Capital International Airport(ZBAA)has been further studied.Specifically,the time series of this airport and of the 21 airports Granger-causing its delays have been extracted,and a random noise added to them(Gaussian distribution with zero mean and variable variance).Fig.7 represents the evolution of the averagep-value of the causality between the 21 airports and ZBAA(Fig.7(a)),and the average fraction of statistically significant links(Fig.7(b))as a function of the variance of the noise.If the detected causality links were just the result of random fluctuations,theirp-value would be independent from the magnitude of the added noise.On the contrary,Fig.7 clearly shows that the noise confounds the detection of causalities,such that half of the links are lost for a noise level of 1.0(equivalent,by construction,to the standard deviation of the time seriesD0).

    Fig.4 Airports delay propagation networks.

    Fig.5 Analysis of nodes’role in the airport delay propagation networks.

    Fig.6 Out-and in-degree of nodes,as a function of the average delay observed in the corresponding airport.

    Finally,Table 1 reports the information corresponding to the five most important airports for delays propagation,i.e.according to the out-degree.It is important to note that some of the most important airports detected operate a very small number of flights-for instance,Kaili Huangping Airport and Hami Airport respectively operated 7 and 6 flights in the time window considered.This issue,and the problems it may cause,will be discussed in Section 4.

    Fig.7 Average p-value and average fraction of significant links between ZBAA and the 21 airports Granger-causing it,as a function of the quantity of noise introduced in the time series.

    3.2.Airlines’contribution to delay propagation

    In order to shed light on the contribution of individual airlines to the global delay propagation process,in this Section we reconstruct and analyse the delays generated by the three most important ones:Air China(ICAO code CCA),China Eastern Airlines(CES),and China Southern Airlines(CSN).It has to be noted that the relationship between individual airlines and the global delay propagation network is far from trivial.Specifically,the full delay propagation process can be seen as a multi-layer network,in which each layer represents a different airline,and the global picture is obtained as the projection of all layers into a single network.34Such projection is nevertheless not linear.A propagation between airportsAandBmay be present as a global phenomenon in the complete network,but may not be statistically significant in each layer.On the other hand,delays may be propagated betweenAandBby two or more airlines,but this may cancel out when they are considered together:a link may be present in several layers,but may disappear in the global network.

    Fig.8 represents the in-and out-degree of nodes for the three considered airlines,as a function of the number of flights operated by each one of them on those airports.The same inverse relation of Fig.5 can be observed,albeit stronger:most delays are generated and propagated by small airports,probably because of their lack of resources for delay mitigation,and because of their role in providing passengers to the more central hubs-an hypothesis that will be confirmed in Section 3.3.

    Additionally,Table 2 presents a comparison of the most important topological metrics for the three considered airlines.While they present qualitatively similar properties,the three airlines’networks are substantially different from the global delay propagation network,as described in Table 3.Specif ically,they present a higher transitivity and a lower assortativity and Information Content.The propagation of delays within each airline is thus characterised by a less random structure,composed of many triangles centred on some focal small airports.This suggests a process in which delays are generated at small airports,for then being propagated to one of the hubs of the system and then back again to small airports.

    It is also worth noting that the degree density of the individual airline networks is higher than what obtained for the global propagation network,approx.0.55 vs.0.296.Two main causes are behind such behaviour.First of all,CC is not a linear operation,such that the global network is not simply the sum of the links observed for different airlines;as an example,if one finds a causalitya-bfor one airline,and a second causalityb-afor another airline,the net result may be the absence of any causality relation betweenaandb.The interested reader may refer to Ref.34 for a throughout discussion of this problem.Second,the three considered airline networks are smaller than the global network,as no airline operated in all the Chinese airports;approx.100 vs.152 nodes.Even if all links observed for individual airlines were present in the global network,the larger size of the latter would imply a smaller link density.

    Table 1 Most central airports,according to the out-degree.

    Fig.8 Degree of nodes vs number of monthly flights they operated,for the three most important Chinese airlines.

    Table 2 Synthesis of airlines’topological metrics.

    Table 3 Synthesis of topological metrics.

    3.3.Delay propagation through aircraft

    As a last topic,we study how delays propagate through different aircraft types.As aircraft size and engine type are related to the kind of role they execute within the network,such analyses will yield us information about which roles are most important for delays propagation.

    All flights have been categorised in four groups,as described by Table 4.The fourth group,comprising mostly executive jets and very small aircraft,has been discarded for not being relevant from a commercial point of view.Afterwards,three time series have been created,corresponding to the hourly average landing delay of all flights executed by these three groups of aircraft,and the GC calculated.

    The network only comprises one statistically significant link,C-B(turbo prop-narrow body,p-value of 0.0225).This suggests that there is a strong delay propagation from turbo prop aircraft,which collect passengers from small and regional airports,to medium range national flights.This is clearly aligned with the results presented in Sections 3.1 and 3.2:any delay generated in regional(thus small)airports is propagated to the whole system by the hub-and-spoke structure.Small airports are thus key to understand the driving forces behind delays propagation.

    Table 4 Classification of aircraft.

    4.Conclusions and discussion

    In this contribution,we proposed the use of the well-known GC metric to describe the structure underlying the phenomenon of delay propagation in the Chinese air transport system.By obtaining detrended time series representing average landing delays at different airports,and by calculating the presence of a causality between pairs of them,we reconstructed the complex network representing delay propagations.Such network has then been studied using standard network metrics,to detect its specific characteristics,and the presence of notable nodes.While this approach is common in other fields of research,the most important of which being neuroscience,11,35it has seldom been used in air transport.

    The proposed approach allowed us to highlight the importance of small airports,which mostly contribute to the propagation of delays.The most common scenario suggested by the presented results involves a turbo-prop aircraft arriving late at a small airport.As not enough buffers are present,the next flight will propagate the delay back to the hub-but such delay would be diluted over all the flights operating there.Finally,the same aircraft undertakes another trip,thus propagating the delay to a second small airport,and creating a causality links between the first and the last one.This has been con-firmed by the analysis of the delay propagation when considering aircraft types:a statistically significant causality relation has been detected from turbo-props to narrow body aircraft,suggesting that a delay appearing at small airports is propagated,through a flight operated by a turbo-prop aircraft,to the big hubs of the system.This behaviour has been observed in the three most important airlines,suggesting that it is not due to specific airline procedures,but is instead due to a systemic cause.

    These insights can be used to design strategies to improve the system,and dampen the delay propagation.Specifically,our analysis suggests that big airports are operating correctly,and that,instead,the connections incoming from small airports should be improved.This may be achieved,on one hand,by extending the CDM process to small airports,such that delays incoming from them can be taken into account while planning flights at hubs.On the other hand,the technology available at small airports may need to be improved,in order to allow them to cope with,for instance,bad weather.

    1.Dong N.Current situation,cause of flight delays and governance path of china’s civil aviation.J Beijing Univ Aeronautics Astronautics(Soc Sci Ed)2013;26:25–32.

    2.Newman ME.The structure and function of complex networks.SIAM Rev2003;45(2):167–256.

    3.Boccaletti S,Latora V,Moreno Y,Chavez M,Hwang D-U.Complex networks:structure and dynamics.Phys Rep2006;424(4):175–308.

    4.Zanin M,Lillo F.Modelling the air transport with complex networks:a short review.Eur Phys J-Spec Top2013;215(1):5–21.

    5.Lordan O,Sallan JM,Simo P.Study of the topology and robustness of airline route networks from the complex network approach:asurveyand research agenda.JTranspGeog2014;37:112–20.

    6.Lin CC,Chen YC.The integration of Taiwanese and Chinese air networks for direct air cargo services.Transport Res A-Pol2003;37(7):629–47.

    7.Wang FH,Jin FJ,Zeng G.Geographic patterns of air passenger transport in China.Scientia Geographica Sinica2003;23(5):519–25.

    8.Li W,Cai X.Statistical analysis of airport network of China.Phys Rev E2004;69(4):046106.

    9.Wang J,Mo H,Wang F,Jin F.Exploring the network structure and nodal centrality of China’s air transport network:a complex network approach.J Transp Geog2011;19(4):712–21.

    10.Cai KQ,Jun Z,Du WB,Cao XB.Analysis of the Chinese air route network as a complex network.Chin Phys B2012;21(2):028903.

    11.Bullmore E,Sporns O.Complex brain networks:graph theoretical analysis of structural and functional systems.Nat Rev Neurosci2009;10(3):186–98.

    12.Cook A,Tanner G,Zanin M.Towards superior air transport performance metrics–imperatives and methods.J Aerospace Oper2013;2(1–2):3–19.

    13.Belkoura S,Zanin M.Phase changes in delay propagation networks.Seventh international conference on research in air transport;2016.

    14.Granger CW.Some recent development in a concept of causality.J Econometrics1988;39(1):199–211.

    15.Granger CW.Causality,cointegration,and control.J Econ Dyn Con1988;12(2):551–9.

    16.Beatty R,Hsu R,Berry L,Rome J.Preliminary evaluation of flight delay propagation through an airline schedule.Air Traf fic Control Quarterly1999;7(4):259–70.

    17.Fan L,Wu F,Zhou P.Ef ficiency measurement of Chinese airports with flight delays by directional distance function.J Air Transport Manage2014;34:140–5.

    18.AhmadBeygi S,Cohn A,Guan Y,Belobaba P.Analysis of the potential for delay propagation in passenger airline networks.J Air Transport Manage2008;14:221–36.

    19.Fleurquin P,Ramasco JJ,Eguiluz VM.Systemic delay propagation in the US airport network.Sci Rep2013;3(3):1159.

    20.Hoover KD.Causality in macroeconomics.Cambridge:Cambridge University Press;2001.

    21.Brovelli A,Ding M,Ledberg A,Chen Y,Nakamura R,Bressler SL.Beta oscillations in a large-scale sensorimotor cortical network:Directional influences revealed by granger causality.P Natl Acad Sci USA2004;101(26):9849–54.

    22.Kamin′ski M,Ding M,Truccolo WA,Bressler SL.Evaluating causal relations in neural systems:Granger causality,directed transfer function and statistical assessment of significance.Biol Cybern2001;85(2):145–57.

    23.Roebroeck A,Formisano E,Goebel R.Mapping directed influence over the brain using granger causality and fMRI.Neuroimage2005;25(1):230–42.

    24.Brockwell PJ,Davis RA.Time series:theory and methods.New York:Springer Science&Business Media;2013.

    25.Bressler SL,Seth AK.Wiener–granger causality:a well established methodology.Neuroimage2011;58(2):323–9.

    26.Granger CW.Testing for causality:A personal viewpoint.J Econ Dyn Con1980;2:329–52.

    27.Costa LdF,Rodrigues FA,Travieso G,Villas BPR.Characterization of complex networks:a survey of measurements.Adv Phys2007;56(1):167–242.

    28.Newman ME.Scientific collaboration networks.I.Network construction and fundamental results.Phys Rev E2001;64(1):016131.

    29.Newman ME.Assortative mixing in networks.Phys Rev Lett2002;89(20):208701.

    30.Latora V,Marchiori M.Efficient behavior of small-world networks.Phys Rev Lett2001;87(19):198701.

    31.Zanin M,Sousa PA,Menasalvas E.Information content:assessing meso-scale structures in complex networks.EPL(Europhys Lett)2014;106(3):30001.

    32.Bolloba′s B.Random graphs.Cambridge:Cambridge University Press;2001.

    33.Jetzki M.The propagation of air transport delays in Europe[dissertation].Aachen:RWTH Aachen University;2009.

    34.Zanin M.Can we neglect the multi-layer structure of functional networks?Phys A:Stat Mech Appl2015;430:184–92.

    35.Zanin M,Papo D,Sousa PA,Menasalvas E,Nicchi A,Kubik E,et al.Combining complex networks and data mining:why and how.Phys Rep2016;635:1–44.

    23 May 2016;revised 22 September 2016;accepted 14 December 2016

    Available online 14 February 2017

    *Corresponding author at:Innaxis Foundation&Research Institute,Jose′Ortega y Gasset 20,28006 Madrid,Spain.

    E-mail address:mzanin@innaxis.org(M.Zanin).

    Peer review under responsibility of Editorial Committee of CJA.

    亚洲精品成人av观看孕妇| 高潮久久久久久久久久久不卡| 大陆偷拍与自拍| 午夜视频精品福利| 在线观看免费视频日本深夜| 在线 av 中文字幕| 9191精品国产免费久久| 亚洲国产av影院在线观看| 91国产中文字幕| 99国产精品一区二区蜜桃av | 日本wwww免费看| 老汉色av国产亚洲站长工具| 黄色a级毛片大全视频| 日韩有码中文字幕| 国产精品自产拍在线观看55亚洲 | 12—13女人毛片做爰片一| 不卡av一区二区三区| 怎么达到女性高潮| 菩萨蛮人人尽说江南好唐韦庄| 1024视频免费在线观看| 天天操日日干夜夜撸| 人人妻人人澡人人爽人人夜夜| av在线播放免费不卡| 精品一品国产午夜福利视频| 成人国产一区最新在线观看| 香蕉久久夜色| 丰满人妻熟妇乱又伦精品不卡| 色婷婷av一区二区三区视频| 视频在线观看一区二区三区| 久久人妻熟女aⅴ| 国产一区二区三区在线臀色熟女 | 黄片大片在线免费观看| 国产精品二区激情视频| tube8黄色片| 岛国在线观看网站| 大片免费播放器 马上看| 久久午夜综合久久蜜桃| 99久久精品国产亚洲精品| 人人妻人人澡人人看| 日本一区二区免费在线视频| 如日韩欧美国产精品一区二区三区| 亚洲av欧美aⅴ国产| 亚洲av第一区精品v没综合| 欧美黄色片欧美黄色片| 亚洲成人国产一区在线观看| 欧美变态另类bdsm刘玥| 国产免费视频播放在线视频| 免费不卡黄色视频| 日韩人妻精品一区2区三区| 搡老岳熟女国产| 曰老女人黄片| 日韩免费高清中文字幕av| 久久精品国产99精品国产亚洲性色 | 日本a在线网址| 极品教师在线免费播放| 黑丝袜美女国产一区| 亚洲男人天堂网一区| 激情视频va一区二区三区| 999久久久国产精品视频| 一二三四在线观看免费中文在| 国产无遮挡羞羞视频在线观看| 曰老女人黄片| 久久99热这里只频精品6学生| 精品久久蜜臀av无| 满18在线观看网站| 女性生殖器流出的白浆| 精品人妻在线不人妻| 精品亚洲成国产av| 天天躁夜夜躁狠狠躁躁| 麻豆乱淫一区二区| 亚洲一区中文字幕在线| av不卡在线播放| 在线观看舔阴道视频| 精品一区二区三区av网在线观看 | 大码成人一级视频| 日本wwww免费看| 99久久国产精品久久久| 日本一区二区免费在线视频| 久久狼人影院| av视频免费观看在线观看| 精品亚洲成国产av| 纯流量卡能插随身wifi吗| 又紧又爽又黄一区二区| 免费在线观看视频国产中文字幕亚洲| 王馨瑶露胸无遮挡在线观看| 亚洲av片天天在线观看| 色视频在线一区二区三区| 久久久精品94久久精品| 最新美女视频免费是黄的| 国产高清videossex| 精品第一国产精品| 精品一区二区三区av网在线观看 | 热99国产精品久久久久久7| 高清欧美精品videossex| 下体分泌物呈黄色| 在线av久久热| 精品熟女少妇八av免费久了| 国产精品.久久久| 精品亚洲成a人片在线观看| 男女午夜视频在线观看| 热99国产精品久久久久久7| svipshipincom国产片| 国产成人免费观看mmmm| 久久免费观看电影| 一区二区三区精品91| 日韩大码丰满熟妇| 成年人黄色毛片网站| 精品国产一区二区久久| 成人国产av品久久久| 亚洲精品国产色婷婷电影| 搡老乐熟女国产| 久久国产精品人妻蜜桃| 丰满人妻熟妇乱又伦精品不卡| 免费久久久久久久精品成人欧美视频| 可以免费在线观看a视频的电影网站| 中亚洲国语对白在线视频| 日韩视频一区二区在线观看| av视频免费观看在线观看| 日韩成人在线观看一区二区三区| 99久久人妻综合| 亚洲国产欧美日韩在线播放| 婷婷成人精品国产| 在线观看www视频免费| 久久久精品区二区三区| 欧美亚洲日本最大视频资源| 精品国产亚洲在线| 成人特级黄色片久久久久久久 | 亚洲精品中文字幕在线视频| 亚洲欧美精品综合一区二区三区| 五月天丁香电影| 一边摸一边抽搐一进一出视频| 一二三四在线观看免费中文在| 欧美亚洲日本最大视频资源| 97在线人人人人妻| 在线观看免费视频网站a站| 欧美黑人欧美精品刺激| 大陆偷拍与自拍| 高清视频免费观看一区二区| 国产精品香港三级国产av潘金莲| 亚洲国产毛片av蜜桃av| 一区二区日韩欧美中文字幕| 国产一区二区三区综合在线观看| 精品国产一区二区三区久久久樱花| 这个男人来自地球电影免费观看| 国产精品偷伦视频观看了| 久久天躁狠狠躁夜夜2o2o| 桃红色精品国产亚洲av| 国产极品粉嫩免费观看在线| 99久久精品国产亚洲精品| 色综合欧美亚洲国产小说| 国产在线精品亚洲第一网站| 精品少妇久久久久久888优播| av片东京热男人的天堂| 国产极品粉嫩免费观看在线| 黄色丝袜av网址大全| 女警被强在线播放| 国产成人欧美在线观看 | 国产午夜精品久久久久久| 精品久久久久久久毛片微露脸| 人妻 亚洲 视频| 丝袜喷水一区| 在线观看免费视频网站a站| 天天躁夜夜躁狠狠躁躁| 午夜免费成人在线视频| 嫩草影视91久久| 免费黄频网站在线观看国产| 欧美老熟妇乱子伦牲交| 99精国产麻豆久久婷婷| 日韩人妻精品一区2区三区| 别揉我奶头~嗯~啊~动态视频| 国产97色在线日韩免费| 丰满人妻熟妇乱又伦精品不卡| xxxhd国产人妻xxx| 国产亚洲精品一区二区www | 久久精品国产99精品国产亚洲性色 | 午夜福利视频精品| 91精品国产国语对白视频| 精品高清国产在线一区| 精品少妇内射三级| 午夜福利视频精品| 91九色精品人成在线观看| 在线观看舔阴道视频| 亚洲色图av天堂| 在线观看www视频免费| 亚洲精品在线观看二区| 精品一品国产午夜福利视频| 亚洲男人天堂网一区| 国产精品av久久久久免费| 狠狠狠狠99中文字幕| 交换朋友夫妻互换小说| 制服诱惑二区| 国产成人欧美在线观看 | bbb黄色大片| 国产成人欧美在线观看 | 三级毛片av免费| 搡老乐熟女国产| 91九色精品人成在线观看| 黄网站色视频无遮挡免费观看| 亚洲成人免费电影在线观看| 国产亚洲一区二区精品| 麻豆乱淫一区二区| 天天躁狠狠躁夜夜躁狠狠躁| 国产91精品成人一区二区三区 | 国产精品一区二区免费欧美| 淫妇啪啪啪对白视频| 午夜日韩欧美国产| 午夜两性在线视频| 国产日韩欧美亚洲二区| 高清视频免费观看一区二区| 狠狠狠狠99中文字幕| 日韩有码中文字幕| 亚洲欧洲日产国产| 91精品国产国语对白视频| 九色亚洲精品在线播放| 精品亚洲乱码少妇综合久久| 天堂动漫精品| 日韩欧美一区视频在线观看| 两个人免费观看高清视频| 国产日韩欧美在线精品| 97在线人人人人妻| 国产xxxxx性猛交| 色在线成人网| 人人妻,人人澡人人爽秒播| 涩涩av久久男人的天堂| 免费在线观看影片大全网站| 999久久久国产精品视频| 欧美成人午夜精品| 成人特级黄色片久久久久久久 | 成年动漫av网址| 亚洲人成77777在线视频| 欧美成狂野欧美在线观看| 欧美另类亚洲清纯唯美| 男男h啪啪无遮挡| 亚洲欧美激情在线| 免费看a级黄色片| 精品乱码久久久久久99久播| 啦啦啦 在线观看视频| 在线看a的网站| 麻豆国产av国片精品| 黑人欧美特级aaaaaa片| 一本大道久久a久久精品| 日韩制服丝袜自拍偷拍| 国产精品香港三级国产av潘金莲| 欧美老熟妇乱子伦牲交| 99精国产麻豆久久婷婷| 免费女性裸体啪啪无遮挡网站| 视频在线观看一区二区三区| 一区二区av电影网| 啦啦啦免费观看视频1| videos熟女内射| 宅男免费午夜| 国内毛片毛片毛片毛片毛片| 成人手机av| 熟女少妇亚洲综合色aaa.| 热re99久久国产66热| 日韩中文字幕欧美一区二区| 国产精品麻豆人妻色哟哟久久| 人妻一区二区av| 又大又爽又粗| 欧美黄色淫秽网站| 天堂8中文在线网| 一级黄色大片毛片| 巨乳人妻的诱惑在线观看| 日韩三级视频一区二区三区| 欧美精品一区二区大全| 午夜精品久久久久久毛片777| 日本av手机在线免费观看| 老司机在亚洲福利影院| 午夜福利欧美成人| 视频区欧美日本亚洲| 香蕉国产在线看| 高清在线国产一区| 黄色视频在线播放观看不卡| 久久国产精品男人的天堂亚洲| 视频区欧美日本亚洲| 丝袜人妻中文字幕| 老司机午夜福利在线观看视频 | 一本综合久久免费| 亚洲伊人色综图| 国产在线免费精品| 18在线观看网站| 免费不卡黄色视频| 一区二区三区国产精品乱码| 欧美亚洲日本最大视频资源| a级片在线免费高清观看视频| 国产精品久久久久久人妻精品电影 | 精品午夜福利视频在线观看一区 | 男人舔女人的私密视频| 国产欧美亚洲国产| 亚洲五月婷婷丁香| 成人影院久久| 亚洲欧美日韩高清在线视频 | 91麻豆精品激情在线观看国产 | 国产成人av激情在线播放| 日日摸夜夜添夜夜添小说| 变态另类成人亚洲欧美熟女 | 侵犯人妻中文字幕一二三四区| 少妇精品久久久久久久| 岛国毛片在线播放| 久久午夜亚洲精品久久| 最近最新中文字幕大全电影3 | 十八禁网站免费在线| 青草久久国产| 人人澡人人妻人| 一边摸一边做爽爽视频免费| 欧美午夜高清在线| 国产av精品麻豆| 国产精品一区二区在线不卡| 黑人欧美特级aaaaaa片| 亚洲精品成人av观看孕妇| 精品国产一区二区三区四区第35| 国产一区二区三区在线臀色熟女 | 又紧又爽又黄一区二区| 国产淫语在线视频| 亚洲七黄色美女视频| 桃花免费在线播放| 欧美日韩国产mv在线观看视频| 老汉色∧v一级毛片| 国产三级黄色录像| 国产黄色免费在线视频| 精品午夜福利视频在线观看一区 | 在线观看66精品国产| av在线播放免费不卡| 丝袜美腿诱惑在线| 一本—道久久a久久精品蜜桃钙片| 成人永久免费在线观看视频 | 久久久精品国产亚洲av高清涩受| 精品高清国产在线一区| 老熟女久久久| 在线看a的网站| 日本vs欧美在线观看视频| 久久国产亚洲av麻豆专区| 少妇粗大呻吟视频| 18禁国产床啪视频网站| 免费人妻精品一区二区三区视频| 亚洲精品国产色婷婷电影| 美女国产高潮福利片在线看| 日韩成人在线观看一区二区三区| 性高湖久久久久久久久免费观看| 一进一出抽搐动态| 美国免费a级毛片| 日本欧美视频一区| 母亲3免费完整高清在线观看| 精品一区二区三区av网在线观看 | 国产又爽黄色视频| 亚洲av日韩在线播放| 国产激情久久老熟女| 久热爱精品视频在线9| 黄片小视频在线播放| 后天国语完整版免费观看| 亚洲精品久久午夜乱码| 蜜桃国产av成人99| 色94色欧美一区二区| 国产精品欧美亚洲77777| 久久人妻av系列| 久久精品亚洲精品国产色婷小说| 国产精品影院久久| 欧美午夜高清在线| 男女边摸边吃奶| aaaaa片日本免费| 精品卡一卡二卡四卡免费| 亚洲成人免费电影在线观看| 在线观看免费视频网站a站| 人人妻人人澡人人爽人人夜夜| 国产欧美日韩综合在线一区二区| 一本久久精品| tube8黄色片| 国内毛片毛片毛片毛片毛片| 一本—道久久a久久精品蜜桃钙片| 欧美国产精品va在线观看不卡| 国产不卡一卡二| 亚洲国产欧美一区二区综合| 久久国产精品人妻蜜桃| 欧美日韩亚洲高清精品| 成人免费观看视频高清| 自拍欧美九色日韩亚洲蝌蚪91| 九色亚洲精品在线播放| 91精品三级在线观看| 热99re8久久精品国产| 大片免费播放器 马上看| 黄网站色视频无遮挡免费观看| 国产欧美日韩一区二区三| 亚洲精华国产精华精| 多毛熟女@视频| 欧美日韩精品网址| 女人被躁到高潮嗷嗷叫费观| 极品教师在线免费播放| 久久久久视频综合| 成年版毛片免费区| 热99久久久久精品小说推荐| 久久精品亚洲av国产电影网| 亚洲精品久久午夜乱码| av线在线观看网站| 久久精品国产亚洲av高清一级| 在线播放国产精品三级| 不卡一级毛片| 久久久国产精品麻豆| 十分钟在线观看高清视频www| 成年人午夜在线观看视频| a级片在线免费高清观看视频| 欧美日韩视频精品一区| 日日摸夜夜添夜夜添小说| 极品少妇高潮喷水抽搐| 18禁美女被吸乳视频| 黄色丝袜av网址大全| 丝袜在线中文字幕| 午夜福利免费观看在线| 亚洲国产av新网站| a级毛片黄视频| 国产伦理片在线播放av一区| 精品一区二区三区四区五区乱码| 一边摸一边抽搐一进一出视频| 男女床上黄色一级片免费看| 午夜老司机福利片| 老司机亚洲免费影院| 亚洲视频免费观看视频| 国产成人av教育| 九色亚洲精品在线播放| 人人妻人人澡人人爽人人夜夜| 国产成人av教育| 9热在线视频观看99| 国产精品九九99| 成人三级做爰电影| 热re99久久精品国产66热6| 精品国产一区二区三区久久久樱花| 大码成人一级视频| 日本黄色视频三级网站网址 | 国产又色又爽无遮挡免费看| 如日韩欧美国产精品一区二区三区| 真人做人爱边吃奶动态| 在线观看免费视频网站a站| 亚洲一卡2卡3卡4卡5卡精品中文| 久久精品亚洲精品国产色婷小说| 亚洲视频免费观看视频| 色婷婷av一区二区三区视频| 国产精品98久久久久久宅男小说| 国产欧美日韩精品亚洲av| 成人国产av品久久久| 日韩免费av在线播放| 国产欧美日韩精品亚洲av| 国产97色在线日韩免费| 一区二区三区乱码不卡18| 每晚都被弄得嗷嗷叫到高潮| 国产在线观看jvid| 久久香蕉激情| 国产老妇伦熟女老妇高清| 岛国毛片在线播放| 国产精品免费大片| 欧美日韩亚洲高清精品| av网站免费在线观看视频| 久久午夜亚洲精品久久| 制服诱惑二区| 我要看黄色一级片免费的| 精品乱码久久久久久99久播| 亚洲人成伊人成综合网2020| 成年版毛片免费区| 亚洲欧美一区二区三区黑人| 精品卡一卡二卡四卡免费| 国产av一区二区精品久久| 亚洲国产av影院在线观看| 777久久人妻少妇嫩草av网站| 亚洲中文av在线| 亚洲欧洲日产国产| 国产深夜福利视频在线观看| 亚洲精华国产精华精| 色播在线永久视频| 久久av网站| 首页视频小说图片口味搜索| 妹子高潮喷水视频| 欧美亚洲 丝袜 人妻 在线| 啦啦啦视频在线资源免费观看| 80岁老熟妇乱子伦牲交| 两人在一起打扑克的视频| 亚洲欧美激情在线| 亚洲精品中文字幕一二三四区 | 亚洲全国av大片| 亚洲精品中文字幕一二三四区 | 日韩欧美一区视频在线观看| 久久亚洲真实| 午夜福利影视在线免费观看| 亚洲熟妇熟女久久| 正在播放国产对白刺激| 国产深夜福利视频在线观看| 一边摸一边抽搐一进一小说 | 中文字幕高清在线视频| 久久天堂一区二区三区四区| 欧美一级毛片孕妇| 亚洲精品在线观看二区| 欧美成狂野欧美在线观看| 一二三四在线观看免费中文在| 精品久久久久久久毛片微露脸| 男女高潮啪啪啪动态图| 亚洲美女黄片视频| 午夜精品久久久久久毛片777| 国产精品1区2区在线观看. | 999久久久精品免费观看国产| 男女免费视频国产| 午夜福利视频精品| 国产伦理片在线播放av一区| 亚洲综合色网址| 亚洲全国av大片| 在线观看免费日韩欧美大片| 亚洲午夜精品一区,二区,三区| 午夜福利在线免费观看网站| 国产在线精品亚洲第一网站| avwww免费| 成年女人毛片免费观看观看9 | 最近最新中文字幕大全电影3 | 精品久久久久久久毛片微露脸| 欧美成人免费av一区二区三区 | 国产激情久久老熟女| 亚洲av成人不卡在线观看播放网| 大型av网站在线播放| 亚洲欧美日韩高清在线视频 | 欧美人与性动交α欧美精品济南到| 免费日韩欧美在线观看| 午夜福利一区二区在线看| 国产极品粉嫩免费观看在线| 777久久人妻少妇嫩草av网站| 欧美激情高清一区二区三区| 国产免费视频播放在线视频| 国产一卡二卡三卡精品| 高清av免费在线| 亚洲一区二区三区欧美精品| 天堂8中文在线网| 欧美老熟妇乱子伦牲交| 欧美成人免费av一区二区三区 | 一级毛片电影观看| 老熟女久久久| 免费观看a级毛片全部| 麻豆乱淫一区二区| 亚洲黑人精品在线| 亚洲专区字幕在线| 人人妻人人澡人人爽人人夜夜| 日本av手机在线免费观看| 久久毛片免费看一区二区三区| 午夜激情久久久久久久| 午夜福利视频在线观看免费| 国产欧美日韩一区二区精品| 麻豆成人av在线观看| 国产成人精品久久二区二区91| 国产亚洲精品一区二区www | 精品少妇黑人巨大在线播放| 久久影院123| 精品人妻在线不人妻| 自拍欧美九色日韩亚洲蝌蚪91| 十八禁高潮呻吟视频| 中文字幕人妻熟女乱码| 日本黄色视频三级网站网址 | 亚洲一码二码三码区别大吗| 在线观看舔阴道视频| 国产欧美亚洲国产| 正在播放国产对白刺激| 国产精品久久久av美女十八| 国产精品亚洲av一区麻豆| 精品人妻1区二区| 99精品久久久久人妻精品| 亚洲成人免费电影在线观看| 国产福利在线免费观看视频| av不卡在线播放| 香蕉丝袜av| 超碰成人久久| 免费观看人在逋| 又黄又粗又硬又大视频| 91av网站免费观看| 国产在视频线精品| 久久人妻福利社区极品人妻图片| 精品国产国语对白av| 女性生殖器流出的白浆| 两个人看的免费小视频| 国产欧美日韩一区二区三| 多毛熟女@视频| 久久狼人影院| 国产成人欧美在线观看 | 亚洲色图av天堂| 精品国产乱码久久久久久男人| 午夜91福利影院| 精品久久久久久久毛片微露脸| 国产成人免费无遮挡视频| 国产在线视频一区二区| 淫妇啪啪啪对白视频| 99国产精品免费福利视频| 91老司机精品| 精品人妻熟女毛片av久久网站| 日本av手机在线免费观看| 99re6热这里在线精品视频| 国产在线精品亚洲第一网站| 男女床上黄色一级片免费看| 我的亚洲天堂| 久久久水蜜桃国产精品网| 50天的宝宝边吃奶边哭怎么回事| 国产欧美日韩精品亚洲av| 亚洲情色 制服丝袜| 一区二区日韩欧美中文字幕| 亚洲国产欧美在线一区| 成年动漫av网址| 国产精品 欧美亚洲| 欧美黑人精品巨大| 中文字幕最新亚洲高清| 精品人妻在线不人妻| 丰满迷人的少妇在线观看| 99精品欧美一区二区三区四区| 国产成人免费无遮挡视频| 国产亚洲欧美在线一区二区| 91精品国产国语对白视频| 国产精品.久久久| 国产1区2区3区精品| 岛国毛片在线播放| 黄色毛片三级朝国网站| 女人久久www免费人成看片| 一级毛片女人18水好多| 亚洲av美国av| 脱女人内裤的视频| 99re在线观看精品视频| 久久久国产一区二区|