• <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 ?

    Performance of a data-driven technique applied to changes in wave height and its effect on beach response

    2016-09-07 07:31:36JosHorrilloCrlloHrshinieKrunrthnShunqiPnDominicReeve
    Water Science and Engineering 2016年1期

    Jos′e M.Horrillo-Crllo*,Hrshinie KrunrthnShun-qi Pn,Dominic Reeve

    aEnergy&Environment Research Group,Zienkiewicz Centre for Computational Engineering,Swansea University,Bay Campus,Swansea SA1 8EN,UKbHydro-environmental Research Centre,School of Engineering,Cardiff University,Cardiff CF24 3AA,UK

    ?

    Performance of a data-driven technique applied to changes in wave height and its effect on beach response

    Jos′e M.Horrillo-Caraballoa,*,Harshinie Karunarathnaa,Shun-qi Panb,Dominic Reevea

    aEnergy&Environment Research Group,Zienkiewicz Centre for Computational Engineering,Swansea University,Bay Campus,Swansea SA1 8EN,UK
    bHydro-environmental Research Centre,School of Engineering,Cardiff University,Cardiff CF24 3AA,UK

    Available online 2 March 2016

    Abstract

    In this study the medium-term response of beach profiles was investigated at two sites:a gently sloping sandy beach and a steeper mixed sand and gravel beach.The former is the Duck site in North Carolina,on the east coast of the USA,which is exposed to Atlantic Ocean swells and storm waves,and the latter is the Milford-on-Sea site at Christchurch Bay,on the south coast of England,which is partially sheltered from Atlantic swells but has a directionally bimodal wave exposure.The data sets comprise detailed bathymetric surveys of beach profiles covering a period of more than 25 years for the Duck site and over 18 years for the Milford-on-Sea site.The structure of the data sets and the data-driven methods are described.Canonical correlation analysis(CCA)was used to find linkages between the wave characteristics and beach profiles.The sensitivity of the linkages was investigated by deploying a wave height threshold to filter out the smaller waves incrementally.The results of the analysis indicate that,for the gently sloping sandy beach,waves of all heights are important to the morphological response.For the mixed sand and gravel beach,filtering the smaller waves improves the statistical fit and it suggests that low-height waves do not play a primary role in the medium-term morphological response,which is primarily driven by the intermittent larger storm waves.

    ?2016 Hohai University.Production and hosting by Elsevier B.V.This is an open access article under the CC BY-NC-ND license(http:// creativecommons.org/licenses/by-nc-nd/4.0/).

    Beach profile;Canonical correlation analysis;Data-driven technique;Empirical orthogonal function;Forecast;Statistical model;Wave height threshold

    1.Introduction

    Shorelines are changing due to construction and development,as part of strategic approaches to flood and coastal management and shoreline management plans.As a result,in recent years there has been a greater need for more robust methodologies for incorporating risk assessments within coastal engineering design.The anticipated changes in climate arising from a rise in global temperatures will change wave and water level conditions,affecting the susceptibility of coastlines and beaches(Sutherland and Gouldby,2003).The stability of many coastlines and beaches depends on the characteristics of the area in which they are located.Taking this into account,there is a need to understand how coastlines, and,in particular,beaches,react during storms,and how beach profiles respond to sequences of waves and storms.Without this knowledge,it is extremely difficult to provide accurate assessments of how coastlines and beach profiles interact and how these interactions are likely to change over time. Furthermore,it is important that shoreline management plans include this information so that coastal managers are able to have confidence in predictions of beach behaviour.Such plans typically require possible morphological changes to be assessed over a period of up to 100 years into the future.This requirement is difficult to meet with current forecastingmethods,and estimates are often determined in an ad hoc manner on a case by case basis.Process-based morphological models have been developed to estimate meso-scale coastal morphological changes,but are yet to be used as a routine part of coastal management.The process-based models have performed well for short-term predictions,but they encounter difficulties when applied to meso-scale cases.Not only are these models difficult to operate,they can also suffer from instability and require significant computing and data resources for medium-to long-term prediction.As a result,some of the broad morphological tendencies observed in practice are difficult to reproduce reliably(de Vriend et al.,1993;Pan et al.,2010).

    This has encouraged the development of data-driven techniques that are based entirely on analysis and extrapolation of observations(R′o.zy′nski,2003;Haxel and Holman, 2004;Reeve et al.,2016).With the growing amount of observations available from coastal monitoring programs in the UK and elsewhere and with the development of more sophisticated statistical analysis techniques,data-driven methods offer an additional alternative to the traditional methods available for meso-scale prediction.The underlying argument for employing data-driven methods is as follows: Predicting beach morphology is difficult.Predicting the wave characteristics is also difficult but is being done on an almost routine basis for navigation,logistics,and coastal management applications.If we can establish a strong relationship between the wave characteristics and beach response from historical records,then we can use this relationship,together with forecasts of the wave characteristics,to create forecasts of beach morphology.The method should work if(a)a strong relationship can be established between some measurements of wave activity and beach response,and(b)the conditions experienced in the forecast period are statistically similar to those upon which the relationship has been determined. This leaves open the question of which measurements of wave activity and beach morphology will provide the best linkage.

    One method that has proven to be very useful in this context is canonical correlation analysis(CCA).CCA belongs to the family of methods based on correlation techniques and measures the relationship between the observed values of two sets of variables.It has been used with measurements from the Field Research Facility(FRF)at the Duck site in North Carolina,USA,by Larson et al.(2000)to detect coherent patterns in the wave and beach profile data and then to use these to predict the beach profiles on the basis of the waves alone.The researchers used a parametric description of waves based on the wave height and wave energy,and found the best results with wave height.Horrillo-Caraballo and Reeve(2008) extended this study to investigate how the choice of distribution function used to describe the wave height can influence the quality of predictions.R′o.zy′nski(2003)used CCA to evaluate the evolution patterns of multiple longshore bars and the interactions between them in Lubiatowo,Poland.The sensitivity of the method to the data sampling rate and the duration of the records were investigated by Horrillo-Caraballo and Reeve(2010).They analysed the quality of forecasts made using CCA for beaches at Duck,USA and Milford-on-Sea,UK.They concluded that there is no strong reduction in prediction accuracy over the forecast period and that there is an increase in the forecast error when the duration and density of the records used to determine the regression matrix are degraded.Reeve and Horrillo-Caraballo(2014) used data-driven methods to forecast the behaviour of beaches with different exposures caused by nearby structures.They found that CCA could obtain strong correlation between the local beach behaviour and the offshore wave conditions, thereby encapsulating the effects of diffraction on beach evolution.This study extended that investigation to examine the sensitivity of the relationship between the wave height and beach response by invoking a graduated threshold for the wave height.

    Fig.1.Location of study sites.

    We studied two different sites:the Duck site,located in North Carolina,on the east coast of the USA,which has sandy beaches,and the Milford-on-Sea site,located at Christchurch Bay,on the south coast of England,where beaches are classified as mixed sand and gravel beaches(Fig.1).

    At the Duck site,the US Army Corps of Engineers has maintained a long-term campaign of measurement of waves, tides,currents,local meteorology,and beach response since its creation(FRF,2007).For this study we used historical wave records and monthly beach profiles covering a period of 25 years.For the Milford-on-Sea site,the approximately biannual beach profile surveys and continuous wave recordings cover a period of over 18 years.This measurement program is part of the local shoreline management plan and the response of the mixed sand and gravel beaches in this area have been monitored since 1987(Bradbury et al.,2003).

    The two data sets have been described in detail by Horrillo-Caraballo and Reeve(2010)and the interested reader may find more information in that paper.Just as different sediment transport equations are required to describe the movement of sand and gravel in process-based models,so it might be thought that different statistical methods might be required to describe the two different beach types.Horrillo-Caraballo and Reeve(2010)demonstrated that this was not the case and that CCA could perform well with both beach types.Furthermore, degrading the sampling rate of the Duck data to make it similar to the Milford-on-Sea measurements resulted in only a small increase in forecast errors.Earlier studies have focused on establishing the relationship between the wave height or wave energy and beach response,on the basis that the active beach profile is shaped predominantly by wave action.Here, we extended this concept based on the hypothesis that the significant changes in beach morphology are caused by the events that cause the greatest sediment transport,that is,larger waves.We revisited two sites that were the subject of earlier studies to investigate whether imposing a threshold on the wave height can improve the correlation between waves and beach response,thereby leading to an improvement in forecasting ability.

    This paper is structured as follows:In section 2 a brief outline of CCA is provided.The measurements at the two study sites are described in section 3.The analysis methodology is given in section 4,followed by a presentation of the main results in section 5.The paper closes in section 6 with conclusions.

    2.Canonical correlation analysis

    CCA is one of a family of correlation techniques but is distinct from factor/principal component analysis despite certain conceptual and terminological similarities.CCA is used to investigate the intercorrelation between two sets of variables,whereas factor/principal component analysis or the empirical orthogonal functions(EOFs)method identifies the pattern of relationships within one data set(Clark,1975). Before CCA is performed,it is customary to smooth the observations to reduce noise.Here,we expanded each data set into EOFs,truncated the EOF expansion,and then recombined the EOFs to create filtered data sets.If the two original data sets are denoted as Y(for example,if the wave height probability density distributions constitute a data matrix with size nt×ny),and Z(for example,if the time sequence of beach profiles constitute a data matrix with size nt×nz),a regression matrix can be derived between the two matrices,representing the established correlation between the dominant patterns in the two variables.This means that if the variable Y is known for some future time,the other variable Z can be predicted using the regression matrix(Larson et al.,2000).The predictions Zp,based on a predicted wave matrix Yp,are given by the following: where A is a regression matrix that defines the relationship between the two variables based on historical measurements. A detailed explanation of CCA and the solutions of Eq.(1)can be found in R′o.zy′nski(2003).The variable Zpcontains the beach profile measurements and the variable Ypcontains the corresponding values of wave conditions represented as a probability density function(pdf).As wave measurements are generally easier to obtain than beach profile measurements, Eq.(1)can be used to forecast beach profiles from wave conditions(Larson et al.,2000).The quality of the data used to define the regression matrix will of course constrain the quality of the prediction obtained.

    3.Field data

    3.1.Duck site

    3.1.1.Beach profiles

    The US Army Corps of Engineers coastal monitoring facility,i.e.,FRF,at Duck,North Carolina,USA,has been monitoring the nearshore area for more than 33 years, surveying some of the area at least twice a month,resulting in hundreds of surveys.

    The area is characterized by regular shore-parallel contours, a moderate slope,and bars in the surf zone.An outer storm bar is present at about 4.5 m of depth,relative to the mean water level,and an inner bar is present between 1.0 and 2.0 m of depth relative to the mean water level.The sediment includes a medium-to-fine sand mixture with a grain size decreasing from 1 mm on the foreshore to 0.1-0.2 mm offshore.According to Birkemeier et al.(1981),the tide in the area is considered micro-tidal,with a tidal range between 0.7 m for neap tides and 1.5 m for spring tides.

    The beach profile data used for this study were obtained from the FRF webpage labelled as Profile 62(FRF,2007), covering the period from July 1981 to January 2006(226 profiles).The elevations of the profile data were measured relative to the US 1929 National Geodetic Vertical Datum (NGVD29).Profiles were interpolated to regular spacing at a 2-m resolution using the spline method described in Li et al. (2005).The interpolation was carried out over a domaindefined as 70 m from the main baseline(dune level position), up to 910 m offshore(around 8 m of water depth),in order to maximize the number of usable profiles and to extend the profiles out to the depth of closure.The depth of closure at this site has been estimated to be in water depths of between 4 and 6 m by Larson and Kraus(1994).The resulting data set is shown in Fig.2 as a colour-coded plot of elevations.

    3.1.2.Wave data

    Wave measurements may also be accessed from the FRF webpage.Significant wave height(Hs),and peak spectral wave period(Tp),were obtained from a directional waverider buoy located approximately 3 km offshore at a water depth of 17.4 m(FRF,2007).The wave data obtained from the buoy are regularly recorded every 6 h and,during some particular periods,hourly values have been recorded(Fig.3).

    3.2.Milford-on-Sea site

    3.2.1.Beach profiles

    Christchurch Bay encloses a shallow embayment with an average depth of approximately 7 m relative to the chart datum (CD),demarcated by Hengistbury Head to the west and Hurst Spit to the east(Fig.1(b)).Waves undergo shoaling and refraction in this area due to the shallow and wide shoreface of Christchurch Bay.The bay is characterized by dominant waves from the southwest(http://www.scopac.org.uk/scopac_ sedimentdb/chrst/chrstmap.pdf),and Hengistbury Head has played a mooring role for Christchurch Bay,resulting in a classical spiral beach planform(Halcrow Group,1999). Christchurch Bay is more exposed to swell waves originating in the Atlantic Ocean than the wind waves coming from the east and southeast;this,combined with the shallow bathymetry,means that storm waves are more depth-limited in this area(http://www.scopac.org.uk/scopac_sedimentdb/chrst/ chrstmap.pdf).

    The beach profile data used for this study were accessed from the Channel Coastal Observatory(CCO)webpage(CCO, 2006)and are labelled as Profile 5f00107.Surveys for the period from November 1988 to January 2006(50 profiles) were obtained.All profile elevation data are referenced to the Ordnance Survey Datum Newlyn(ODN).Only surveys extending from the dune region out to a water depth of the mean low water level(MLWL)were incorporated into the analysis.Each profile was interpolated to regular spacing with a 0.5 m resolution according to the method of Li et al.(2005). Fig.4 displays the subset of measured profiles along Profile 5f00107 that were used in CCA.

    Fig.2.Historic bathymetric surveys of Duck profile from July 2,1981 to January 13,1998.

    Fig.3.Hsmeasurements at Duck site from July 2,1981 to January 13, 1998.

    3.2.2.Wave data

    Measurements from a waverider buoy deployed at approximately 10-12 m of water depth are available.Hsand Tphave been obtained from this buoy.The predominant wave direction in this area is from the southwest.Data from the wave buoy site have been compared with the synthetic offshore wave data from the UK Met Office that have subsequently been transformed to the wave buoy site,through numerical modelling.Bradbury et al.(2004)compared a one-month sample of measured and synthetic data.They noted that there was a clear relationship between measured and modelled conditions,concluding that their confidence in the offshore synthetic wave data,the numerical wave transformation process,and the wave buoy measurements was high (Fig.5).

    Fig.4.Historic bathymetric surveys of Milford-on-Sea profile from November 1,1988 to January 13,1998.

    The beach profiles at the Duck site and the Milford-on-Sea site present some clear differences.The beach material at the Duck site is sand,whereas,at the Milford-on-Sea site,it is a mixture of sand and gravel.Due to these differences,thecharacteristics of the beaches of the two sites also differ (Horrillo-Caraballo and Reeve,2010).The extent of the survey data is one of the significant differences between the sites.The Duck beach profile measurements extend considerably further offshore,to what is considered the depth of closure,while at the Milford-on-Sea site,the beach profile measurements extend to the low water level(LWL).While this makes comparison of beach volumes over time problematic,it is practical to assume that the LWL and the mean high water spring(MHWS)are quite sensitive to nearshore processes on an annual and biannual time scale and can be used as indicators of morphological changes in the beach profile.The active part of the upper beach was the focus of this study.The surveys of both sites covered this section of the beach.

    Fig.5.Hsmeasurements at Milford-on-Sea site from November 1, 1988 to January 13,1998.

    4.Methodology

    The measurements at the two sites have been processed to generate series with the same sampling rate for both profiles and wave conditions,a requirement for the application of CCA.At the Duck site,the first 155 profiles were used in CCA and the remaining 71 profiles were used for comparison against the predictions obtained based on the regression matrix.For the Milford-on-Sea site,the first 29 beach profiles were analysed using CCA and the remaining 21 profiles were used for comparison with the predictions obtained from CCA on the basis of the regression matrix.Table 1 shows a summary of the data used for the analysis.

    We followed the procedure detailed in Horrillo-Caraballo and Reeve(2010)and described the waves with a nonparametric empirical distribution function.The wave conditions for each interval were converted into a single function by creating a pdf of significant wave height.CCA was then applied to the sequence of beach profiles and Hspdfs.Figs.6 and 7 show the empirical wave height distributions for the two sites.For the Duck site there is more information,as the FRF was established at the beginning of the 1980s and because of the relatively high frequency of sampling.The data at the Milford-on-Sea site are not as frequent as at the Duck site.It has been monitored from the late-1980s and profiles are measured generally twice a year.The empirical wave height distributions were created by combining the separate distributions calculated from each interval.In order to investigate the importance of large waves to the response of the beach morphology,the empirical wave height distributions were calculated for different wave height thresholds(TH),which were set at intervals of 1 m,yielding six different pdfs for the Duck site and four for the Milford-on-Sea site.The height of individual peaks in the distribution could change as TH varied because TH was applied to the pdfs for each interval before the combining.For the Duck site,TH was increased to a maximum of 5 m,while for the Milford-on-Sea site the maximum TH was 3 m.

    5.Results

    The first step in CCA was to correlate the profile response to the offshore wave height.The purpose was to investigate whether there was any relationship between the wave statistics and the observed pattern of variation in the profile response. CCA was carried out using three canonical modes,following the procedure of Horrillo-Caraballo and Reeve(2010),who found that this led to the best performance of the prediction analysis.

    Comparisons are shown for the Duck and Milford-on-Sea data sets using the data defined in Table 1 and the predictions of eight years into the future(up to 2006),where the prediction for each particular date was made on the basis of the regression matrix and the pdf of significant wave heights in the period between that date and the date of the previous survey.For the Duck site,the CCA regression matrix was based on the data from July 2,1981 to January 13,1998.For the Milford-on-Sea site,the CCA regression matrix was determined using beach profiles and significant wave heights for the period from November 1,1988 to January 13,1998, and the profile predictions for the two sites were calculated using wave data over the period from January 13,1998 to January 30,2006(see Table 1).

    Prior to the analyses,the temporal mean was subtracted from all data sets.According to Fig.8(a),the first three spatial EOFs obtained from the profile data sets at the Duck site explained 69%of the variation in the data.The first spatial eigenfunction(E1)explained 35%of the variation around themean,the second(E2)explained 21%,and the third(E3) explained 13%.At the Milford-on-Sea site(Fig.8(b)),79%of the variation is explained by the first three spatial EOFs.E1 explained 43%of the variation,E2 explained 25%,and E3 explained 11%.The difference in percentages between the two sites can be interpreted as follows:The data from the Milfordon-Sea site need fewer modes to describe the entire data set and so contain less complexity than the data from the Duck site.This is partly explained by the length of monitoring at the two sites and the frequency of observations.The Duck site captures shorter-term fluctuations and is also more likely to contain extreme events due to its greater duration.For both sites,the EOFs describing the profile shapes show a rich structure due to the presence of bars,but the shapes of the EOF modes can be used to determine mean properties of the impact of the bar over the profile(Larson et al.,2000).Additionally, the temporal EOFs(not shown in this paper)can be used to determine trends of profile changes,oscillatory cycles,and relate them to coastal processes(Horrillo-Caraballo et al., 2015).

    Table 1 Summary of data used in this study.

    Fig.6.Empirical wave height distribution with different TH values at Duck site.

    The root mean square error(RMSE)was used to determine the quality of fit between measured and predicted profiles. Two cases are shown:the RMSE between the prediction and measurement as a time averaged function over the profile,and the RMSE as a spatially averaged function over the forecast period.

    Fig.9(a)shows results for the Duck site,for the cases with TH=0 m,TH=1 m,TH=2 m,TH=3 m,TH=4 m,and TH=5 m.The general behaviour of the results from the use of different wave height thresholds is similar:The time-averaged RMSE increases where major changes in the elevations are present.This is due largely to the movement and change of position of the bars in the beach profile.If no wave height threshold is considered(TH=0 m),the time-averaged RMSE decreases between 70 m and 180 m of the cross-shore distance with respect to the other thresholds,but the time-averaged RMSE increases when predicting the first bar and the offshore area(between 470 m and 850 m of the cross-shore distance).The area between 270 m and 420 m is more affected by larger waves and the change of position of the bars.

    Fig.10(a)shows the results for the Milford-on-Sea site,for the cases with TH=0 m,TH=1 m,TH=2 m,and TH=3 m.At the upper beach,where there is very little movement,the time-averaged RMSE is,not surprisingly,very small.Further seaward the RMSE rises,with an average valueof about 0.7 m,ranging between 0 and 1.3 m.The time averaged RMSE is largest at the upper beach crest and at around 25-27 m of the cross-shore distance.As at the Duck site,the general trends with use of different wave height thresholds are similar:The time-averaged RMSE increases where major changes in the elevations are present.If CCA is carried out with no threshold,the time-averaged RMSE increases between 17 m and 47 m of the cross-shore distance with respect to the other thresholds.The time-averaged RMSEs at the Duck site are generally smaller across the profile when no threshold is employed.However,at the Milford-on-Sea site,setting a wave height threshold leads to an improvement in forecast error over 20-50 m of the crossshore distance.

    Fig.7.Empirical wave height distribution with different TH values at Milford-on-Sea site.

    An investigation of the dependence of forecast error upon the forecast window was also made.Figs.11 and 12 show the spatially averaged RMSE for the Duck and Milford-on-Sea sites over the eight-year forecast period.They demonstrate one of the advantages of data-driven methods over process based methods.As the forecast window lengthens,there is no evidence of a commensurate increase in error,as one often finds with a time-stepping solution of differential equations. Instead,the general trend of errors is remarkably consistent.At the Duck site,there are two periods where there is an increase in the spatially averaged RMSE that can be traced to unusual sequences of storms(Capobianco et al.,1997;FRF,2015). This highlights one of the disadvantages of data-driven methods when an unusual wave event occurs that is notrepresented in the data used to compute the regression matrix: Large discrepancies can be expected between forecasts and actual beach profiles.The spatially averaged RMSE at the Milford-on-Sea site(Fig.12)fluctuates between 0.2 m and 1.0 m,with a peak after the beginning of 2004.The explanation for this peak can be related to a gap in the wave records at the end of 2003.In this case,the calculation of the empirical wave height distribution for this time of the year is uncertain. Fig.12 also clearly shows that the condition of TH=1 m provides consistently better forecast accuracy than the condition of TH=0 m.

    Fig.8.First three spatial EOFs determined from measured beach profiles at two sites.

    Fig.9.Time-averaged RMSE for predictions made with different TH values and mean,maximum,and minimum elevations of measured beach profile at Duck site.

    Almeida et al.(2011)concluded in their study in Praia de Faro,in southern Portugal,that waves higher than 2.3 m were responsible for the main morphological changes in the berm and beach face and waves higher than 3.2 m were responsible for the changes in the sub-tidal area and long-shore bars.We have found a rather more nuanced picture from our analysis.

    Based on the wave height threshold analysis,there is a difference in the relationship between waves and morphological response at the two sites.Removing the smaller waves leads to an increase in the forecast error for the upper beach and a decrease in the forecast error for the lower beach at the Duck site,while it leads to a reduction of the forecast error across the whole profile at the Milford-on-Sea site.An initial interpretation of these results is that,for the more gently sloping sandy beach at the Duck site,the contribution of smaller waves to the morphological evolution of the upper beach should not be ignored,whereas for the mixed beaches at the Milford site the contribution of smaller waves is less important.

    Fig.10.Time-averaged RMSE for predictions made with different TH values and mean,maximum,and minimum elevations of measured beach profile at Milford-on-Sea site.

    6.Conclusions

    In this study we have investigated the medium-term prediction of beach profiles using a data-driven technique based on CCA.Measurements from two sites have been used:the Duck site,in the USA,and the Milford-on-Sea site,in the UK. The Duck site is dominated by mildly sloping sandy beaches while the Milford-on-Sea site is a mixed sand and gravel beach with a steep gravel upper beach.

    Data-driven techniques rely on finding a strong correlation between measurements of the processes causing beach changes and the beach response to these processes.Once such a correlation is established,then it can be used,in conjunction with knowledge of the forcing processes,to formulate predictions of the corresponding beach profile.Here,we have used CCA to establish correlation properties between sequences of wave conditions and corresponding measurements of beach profiles.The correlation was determined from the initial portion of the measurements.Forecasts,or,more accurate,hindcasts of beach profiles were made for the remaining portion of the measurements using the correlation and wave information for the period up to the date at which the beach profile was required.A validation of the forecast was made against the corresponding measured beach profiles. Forecasts were performed for periods of up to eight years in advance.

    The choice of metric for processes and beach response has been investigated elsewhere(e.g.,Larson et al.,2000; Horrillo-Caraballo and Reeve,2008),and it has been found that the empirical wave height distribution and beach profile are good metrics.

    Fig.11.Spatially averaged RMSE for predictions made with different TH values and measured Hsat Duck site.

    Fig.12.Spatially averaged RMSE for predictions made with different TH values indicated by label key and measured Hsat Milford-on-Sea site.

    Here,our primary interest has been in investigating the hypothesis that changes in beach profile morphology are driven primarily by the largest,most energetic waves.To this end we have performed repeated analyses with wave heights exceeding a series of thresholds.A secondary aim has been to test the dependence of forecast error on the length of the forecast window.Our specific question was,is there a degradation in forecast quality within the forecast period?

    We have found that the RMSE of the forecasts is dependent on the wave height threshold(TH).At the Milford-on-Sea site the forecasts were improved by setting a TH of 1 m.Further improvement in the error does not occur uniformly with increasing TH,which might be expected if beach response were being driven purely by the largest wind waves.At the Duck site,forecasts for the upper beach are best when no threshold is employed,whereas for the lower beach a small improvement in forecast error is found when a threshold is imposed.It is concluded that small-amplitude waves play an important part in shaping the upper beach profile at the Duck site,and have a rather less significant role in influencing the evolution of the lower beach,suggesting that a small amplitude swell is not a significant driver of the mediumterm evolution of the beach morphology.

    The forecast errors at both sites do not exhibit growth with the forecast period,as is often found with time-stepping process-based models due to the accumulation of numericalinaccuracies.However,the forecast error can vary over time, particularly when an unusual wave condition occurs that is not represented in the data used to determine the correlation.

    As coastal monitoring programs continue,and new programs begin,more data will become available,and means of extracting information useful for coastal management can be anticipated.Further research is required to determine the best metrics for correlation studies,limitations on forecast periods, the spatial and temporal distribution of forecast errors,and the dependence of forecast errors on beach type and exposure.The data-driven method based on CCA described here is one example of the sophisticated statistical techniques that can be brought to bear on this challenge,and it has been demonstrated how medium-term forecasts can be formulated.

    Acknowledgements

    The authors would like to thank the staff at the FRF,Field Data Collections and Analysis Branch,of the US Army Corps of Engineers in Duck,North Carolina,USA and at the Channel Coastal Observatory in Southampton,UK for making the field data available for both sites.

    References

    Almeida,L.P.,Ferreira,O.,Pacheco,A.,2011.Thresholds for morphological changes on an exposed sandy beach as a function of wave height.Earth Surf.Process.Landf.36(4),523-532.http://dx.doi.org/10.1002/esp.2072. Birkemeier,W.A.,DeWall,A.E.,Gorbics,C.S.,Miller,H.C.,1981.A User's Guide to CERC's Field Research Facility.Miscellaneous Report No.81-7. U.S.Army,Corps of Engineers,Coastal Engineering Research Center (U.S.).National Technical Information Service,Springfield.

    Bradbury,A.P.,Colenutt,A.J.,Cross,J.,Eastick,C.,Hume,D.,2003.Evaluation of coastal process impacts arising from nearshore aggregate dredging for beach recharge:Shingles Banks,Christchurch Bay.In:The International Conference on Coastal Management.Institution of Civil Engineers,Brighton,pp.98-112.

    Bradbury,A.P.,Mason,T.E.,Holt,M.W.,2004.Comparison of the performance of the Met Office UK:Waters wave model with a network of shallow water moored buoy data.In:Proceedings of the 8th International Workshop on Wave Hindcasting and Forecasting.North Shore Oahu,Hawaii.

    Capobianco,M.,Larson,M.,Nicholls,R.J.,Kraus,N.C.,1997.Depth of closure:A contribution to the reconciliation of theory,practice,and evidence.In:Proceedings of the 3rd Coastal Dynamics.American Society of Civil Engineers,Reston,pp.506-515.

    Channel Coastal Observatory(CCO),2006.Channel Coastal Observatory, England.http://www.channelcoast.org[Retrieved June 18,2006].

    Clark,D.,1975.Understanding Canonical Correlation Analysis.Concepts and Techniques in Modern Geography,No.3.Geo Abstracts Ltd.,Norwich.

    de Vriend,H.J.,Capobianco,M.,Chesher,T.,de Swart,H.E.,Latteux,B., Stive,M.,1993.Approaches to long-term modelling of coastal morphology:A review.Coast.Eng.21(1-3),225-269.http://dx.doi.org/ 10.1016/0378-3839(93)90051-9.

    Field Research Facility(FRF),2007.Field Data Collections and Analysis Branch.US Army Corps of Engineers,Duck.http://www.frf.usace.army. mil/frf_home.shtml[Retrieved Aug.17,2007].

    Field Research Facility(FRF),2015.Field Data Collections and Analysis Branch.US Army Corps of Engineers,Duck.http://www.frf.usace.army.mil/AnuRpt/programs/storms/storm_table.txt[Retrieved Jun.23,2015].

    Halcrow Group,1999.Poole and Christchurch Bay Shoreline Management Plan.Volume 2:Physical Environment.Poole and Christchurch Bays Coastal Group,Swindon.

    Haxel,J.H.,Holman,R.A.,2004.The sediment response of a dissipative beach to variations in wave climate.Mar.Geol.206(1-4),73-99.http://dx.doi. org/10.1016/j.margeo.2004.02.005.

    Horrillo-Caraballo,J.M.,Reeve,D.E.,2008.An investigation of the link between beach morphology and wave climate at Duck,NC,USA.J.Flood Risk Manag. 1(2),110-122.http://dx.doi.org/10.1111/j.1753-318X.2008.00013.x.

    Horrillo-Caraballo,J.M.,Reeve,D.E.,2010.An investigation of the performance of a data-driven model on sand and shingle beaches.Mar.Geol. 274(1),120-134.http://dx.doi.org/10.1016/j.margeo.2010.03.010.

    Horrillo-Caraballo,J.M.,Reeve,D.E.,Karunarathna,H.,Pan,S.,Burningham,H., 2015.Multi-scale statistical analysis of beach profiles on the Suffolk Coast, UK.In:Proceedings of the Coastal Sediments 2015.World Scientific Publishing,Singapore.

    Larson,M.,Kraus,N.C.,1994.Temporal and spatial scales of beach profile changes,Duck,North Carolina.Mar.Geol.117(1-4),75-94.http://dx. doi.org/10.1016/0025-3227(94)90007-8.

    Larson,M.,Capobianco,M.,Hanson,H.,2000.Relationship between beach profiles and waves at Duck,NC,determined by canonical correlation analysis.Mar.Geol.163(1-4),275-288.http://dx.doi.org/10.1016/ S0025-3227(99)00119-X.

    Li,Y.,Lark,M.,Reeve,D.E.,2005.The multi-scale variability of beach profiles at Duck,N.C.,USA,1981-2003.Coast.Eng.52(12),1133-1153. http://dx.doi.org/10.1016/j.coastaleng.2005.07.002.

    Pan,S.,Reeve,D.E.,Davidson,M.,O'Connor,B.,Vincent,C.,Dolphin,T., Wolf,J.,Thorne,P.,Bell,P.,Souza,A.,et al.,2010.Larger-scale morphodynamic impacts of segmented shore-parallel breakwaters on coasts and beaches:An overview of the LEACOAST2 project.Shore Beach 78(4),35-43.

    Reeve,D.E.,Horrillo-Caraballo,J.M.,2014.Analysis and prediction of the differential beach response within a coastal defence scheme:Herne Bay. UK.Proc.ICE Marit.Eng.167(MA1),29-41.

    Reeve,D.E.,Karunarathna,H.,Pan,S.Q.,Horrillo-Caraballo,J.M.,R′o.zy′nski,G., Ranasinghe,R.,2016.Data-driven and hybrid coastal morphological prediction methods for meso-scale forecasting.Geomorphology 256,49-67. http://dx.doi.org/10.1016/j.geomorph.2015.10.016.

    R′o.zy′nski,G.,2003.Data-driven modelling of multiple longshore bars and their interaction.Coast.Eng.48(3),151-170.http://dx.doi.org/10.1016/ S0378-3839(03)00024-3.

    Sutherland,J.,Gouldby,B.,2003.Vulnerability of coastal defences to climate changes.Proc.Inst Civ.Eng.Water Marit.Eng.156(2),137-145.http:// dx.doi.org/10.1680/wame.2003.156.2.137.

    29 June 2015;accepted 6 January 2016

    This work was supported by the UK Natural Environment Research Council(Grant No.NE/J005606/1),the UK Engineering and Physical Sciences Research Council(Grant No.EP/C005392/1),and the Ensemble Estimation of Flood Risk in a Changing Climate(EFRaCC)project funded by the British Council under its Global Innovation Initiative.

    *Corresponding author.

    E-mail address:j.m.horrillo-caraballo@swansea.ac.uk(Jos′e M.Horrillo-Caraballo).

    Peer review under responsibility of Hohai University.

    http://dx.doi.org/10.1016/j.wse.2016.02.006

    1674-2370/?2016 Hohai University.Production and hosting by Elsevier B.V.This is an open access article under the CC BY-NC-ND license(http://creativecommons.org/licenses/by-nc-nd/4.0/).

    精品国产亚洲在线| 中国美女看黄片| 日韩大码丰满熟妇| 国产免费现黄频在线看| 免费高清在线观看日韩| av有码第一页| 亚洲精品自拍成人| 老司机午夜十八禁免费视频| 亚洲综合色网址| 一级毛片电影观看| 久久ye,这里只有精品| 欧美大码av| 人人妻人人添人人爽欧美一区卜| 中文字幕av电影在线播放| 一区二区三区乱码不卡18| 精品第一国产精品| 啦啦啦视频在线资源免费观看| 欧美性长视频在线观看| 亚洲欧美精品综合一区二区三区| 亚洲精品国产区一区二| av天堂在线播放| 97在线人人人人妻| 国产淫语在线视频| 亚洲,欧美精品.| 美女高潮喷水抽搐中文字幕| 亚洲av电影在线进入| 狠狠狠狠99中文字幕| 中文欧美无线码| 亚洲五月婷婷丁香| 日本av免费视频播放| 国产麻豆69| 国产精品久久电影中文字幕 | 国产一区二区激情短视频| 亚洲av电影在线进入| 老鸭窝网址在线观看| 亚洲人成电影免费在线| 精品熟女少妇八av免费久了| 亚洲国产欧美在线一区| 精品一区二区三区av网在线观看 | 午夜两性在线视频| 三上悠亚av全集在线观看| 精品亚洲乱码少妇综合久久| 精品视频人人做人人爽| 精品人妻熟女毛片av久久网站| 久久人妻福利社区极品人妻图片| 嫁个100分男人电影在线观看| 国产区一区二久久| 精品一区二区三区av网在线观看 | 一个人免费看片子| 一区福利在线观看| 国产一区二区三区综合在线观看| 日本a在线网址| 久久精品成人免费网站| 午夜老司机福利片| 高清av免费在线| 国产av国产精品国产| 超碰成人久久| 色老头精品视频在线观看| 黄网站色视频无遮挡免费观看| 丝袜在线中文字幕| tube8黄色片| 少妇被粗大的猛进出69影院| 久久久水蜜桃国产精品网| 亚洲av第一区精品v没综合| 久久亚洲真实| 久9热在线精品视频| 欧美黄色淫秽网站| 欧美国产精品一级二级三级| 亚洲中文日韩欧美视频| 曰老女人黄片| 黑人巨大精品欧美一区二区蜜桃| 男女无遮挡免费网站观看| 欧美激情极品国产一区二区三区| 国产成人精品久久二区二区91| 免费观看a级毛片全部| 久久久久久亚洲精品国产蜜桃av| 免费av中文字幕在线| 丁香六月欧美| 又黄又粗又硬又大视频| 黄色成人免费大全| 男人舔女人的私密视频| 亚洲欧美色中文字幕在线| 在线永久观看黄色视频| 一本一本久久a久久精品综合妖精| 国产一区二区三区在线臀色熟女 | 水蜜桃什么品种好| 国产亚洲一区二区精品| 日韩制服丝袜自拍偷拍| 日韩制服丝袜自拍偷拍| 亚洲少妇的诱惑av| 国产精品麻豆人妻色哟哟久久| 黄片小视频在线播放| 欧美日韩中文字幕国产精品一区二区三区 | a级毛片在线看网站| 国产欧美日韩一区二区精品| 最新的欧美精品一区二区| 久久久精品国产亚洲av高清涩受| 欧美黄色淫秽网站| 国产成人啪精品午夜网站| 久久热在线av| 色播在线永久视频| 曰老女人黄片| 999久久久精品免费观看国产| 男人舔女人的私密视频| 天天影视国产精品| 亚洲男人天堂网一区| 黑丝袜美女国产一区| 纵有疾风起免费观看全集完整版| 美女主播在线视频| 丰满人妻熟妇乱又伦精品不卡| 国产免费视频播放在线视频| 美女国产高潮福利片在线看| 女性被躁到高潮视频| 国产无遮挡羞羞视频在线观看| 精品人妻1区二区| 欧美午夜高清在线| 国产一区二区三区在线臀色熟女 | www日本在线高清视频| 国产老妇伦熟女老妇高清| 国产亚洲av高清不卡| 欧美+亚洲+日韩+国产| 久久亚洲精品不卡| 黄色视频,在线免费观看| 制服诱惑二区| 欧美另类亚洲清纯唯美| 真人做人爱边吃奶动态| 美女视频免费永久观看网站| 国内毛片毛片毛片毛片毛片| 18禁黄网站禁片午夜丰满| 国产日韩一区二区三区精品不卡| 欧美激情久久久久久爽电影 | 男人操女人黄网站| 精品欧美一区二区三区在线| 国产高清videossex| 日韩视频一区二区在线观看| 亚洲黑人精品在线| 操美女的视频在线观看| 中文字幕最新亚洲高清| 老司机深夜福利视频在线观看| 热re99久久精品国产66热6| 免费黄频网站在线观看国产| 亚洲精品美女久久av网站| cao死你这个sao货| 日韩有码中文字幕| 午夜久久久在线观看| 啦啦啦在线免费观看视频4| 九色亚洲精品在线播放| 在线观看人妻少妇| 亚洲精品粉嫩美女一区| 啪啪无遮挡十八禁网站| 美女主播在线视频| 亚洲欧洲精品一区二区精品久久久| netflix在线观看网站| 免费在线观看影片大全网站| 波多野结衣av一区二区av| 国产精品九九99| 久久亚洲真实| 18禁裸乳无遮挡动漫免费视频| 成人影院久久| 两性午夜刺激爽爽歪歪视频在线观看 | 亚洲精品国产色婷婷电影| 丰满少妇做爰视频| 成人特级黄色片久久久久久久 | 国产免费现黄频在线看| 18在线观看网站| 热99re8久久精品国产| 夫妻午夜视频| 亚洲午夜理论影院| 久久国产精品影院| 日本五十路高清| 久久久久精品人妻al黑| 黄频高清免费视频| 99精品久久久久人妻精品| 久久ye,这里只有精品| 一本综合久久免费| 女人被躁到高潮嗷嗷叫费观| 国产免费av片在线观看野外av| 性色av乱码一区二区三区2| 两性夫妻黄色片| 国产亚洲欧美精品永久| 亚洲欧美一区二区三区久久| 少妇粗大呻吟视频| 他把我摸到了高潮在线观看 | 丝袜美足系列| 另类精品久久| 精品一区二区三卡| 免费女性裸体啪啪无遮挡网站| 欧美日韩视频精品一区| 2018国产大陆天天弄谢| 精品少妇内射三级| 国产精品99久久99久久久不卡| 男人舔女人的私密视频| 国产精品二区激情视频| 久久免费观看电影| aaaaa片日本免费| 在线观看免费午夜福利视频| 免费久久久久久久精品成人欧美视频| 少妇的丰满在线观看| 脱女人内裤的视频| 亚洲精品成人av观看孕妇| 日本欧美视频一区| 国产欧美日韩精品亚洲av| 一边摸一边抽搐一进一小说 | 高清av免费在线| 久久精品亚洲精品国产色婷小说| 亚洲国产成人一精品久久久| 久久中文看片网| 色精品久久人妻99蜜桃| 国产97色在线日韩免费| 国产成人免费无遮挡视频| 亚洲精品在线美女| 国产91精品成人一区二区三区 | 老司机亚洲免费影院| 亚洲av成人不卡在线观看播放网| 十八禁网站免费在线| 视频在线观看一区二区三区| av又黄又爽大尺度在线免费看| 少妇猛男粗大的猛烈进出视频| 妹子高潮喷水视频| 精品国产一区二区三区四区第35| 亚洲性夜色夜夜综合| 一区二区av电影网| 免费在线观看日本一区| 国产精品久久久久久精品古装| 成年动漫av网址| 亚洲av第一区精品v没综合| 最新在线观看一区二区三区| 真人做人爱边吃奶动态| 狠狠精品人妻久久久久久综合| 99re6热这里在线精品视频| 18在线观看网站| 一区二区三区精品91| 亚洲中文日韩欧美视频| 在线播放国产精品三级| 多毛熟女@视频| 国产亚洲午夜精品一区二区久久| 国产精品秋霞免费鲁丝片| 欧美日韩视频精品一区| 桃红色精品国产亚洲av| 欧美亚洲 丝袜 人妻 在线| 国产成人免费观看mmmm| 亚洲欧洲精品一区二区精品久久久| 777久久人妻少妇嫩草av网站| 男女无遮挡免费网站观看| 国产成人精品久久二区二区91| 人妻久久中文字幕网| 久久久久网色| 国产淫语在线视频| 国产成人精品无人区| 国产成人免费观看mmmm| 亚洲专区字幕在线| tube8黄色片| 天天操日日干夜夜撸| 久久久久视频综合| 成人永久免费在线观看视频 | 精品国产一区二区久久| 一区二区三区国产精品乱码| 亚洲伊人色综图| 丝袜美腿诱惑在线| 久久精品国产99精品国产亚洲性色 | 国产精品成人在线| www.精华液| 亚洲午夜理论影院| videos熟女内射| 免费观看人在逋| 日韩大片免费观看网站| 国产精品电影一区二区三区 | 97在线人人人人妻| 国产有黄有色有爽视频| 狂野欧美激情性xxxx| 欧美性长视频在线观看| 巨乳人妻的诱惑在线观看| 亚洲av第一区精品v没综合| 大码成人一级视频| 精品国产国语对白av| videos熟女内射| 男女免费视频国产| 成人三级做爰电影| 在线 av 中文字幕| 欧美黑人精品巨大| 欧美精品人与动牲交sv欧美| 亚洲精品乱久久久久久| 国产精品久久久久成人av| 国产精品久久久av美女十八| 亚洲欧美一区二区三区黑人| 欧美中文综合在线视频| 老熟妇乱子伦视频在线观看| 精品人妻在线不人妻| 久久精品成人免费网站| 国产日韩一区二区三区精品不卡| 亚洲精品国产区一区二| 国产午夜精品久久久久久| 人人妻人人澡人人看| netflix在线观看网站| 午夜福利视频在线观看免费| 天堂中文最新版在线下载| 精品午夜福利视频在线观看一区 | 国产又爽黄色视频| 母亲3免费完整高清在线观看| 亚洲国产中文字幕在线视频| 一边摸一边抽搐一进一出视频| 亚洲精品久久成人aⅴ小说| 视频在线观看一区二区三区| 看免费av毛片| 中文字幕制服av| 亚洲精品一二三| 精品熟女少妇八av免费久了| 99国产精品99久久久久| 久久久国产欧美日韩av| 亚洲一码二码三码区别大吗| 欧美国产精品一级二级三级| 亚洲成人免费电影在线观看| 人妻一区二区av| 在线观看免费午夜福利视频| 日韩免费高清中文字幕av| h视频一区二区三区| 大香蕉久久成人网| 精品少妇内射三级| 精品视频人人做人人爽| 亚洲一区二区三区欧美精品| 亚洲 欧美一区二区三区| 国产日韩欧美亚洲二区| 久久99一区二区三区| 久久人妻av系列| 亚洲专区中文字幕在线| 俄罗斯特黄特色一大片| 美女扒开内裤让男人捅视频| 另类精品久久| 国产有黄有色有爽视频| 久久久久精品国产欧美久久久| 少妇粗大呻吟视频| 成在线人永久免费视频| 精品午夜福利视频在线观看一区 | 蜜桃在线观看..| 亚洲熟女精品中文字幕| 日韩免费高清中文字幕av| 两个人免费观看高清视频| 欧美日韩一级在线毛片| 无限看片的www在线观看| 午夜福利一区二区在线看| 在线观看免费午夜福利视频| 国产老妇伦熟女老妇高清| 亚洲av美国av| 无人区码免费观看不卡 | 99国产综合亚洲精品| 高清在线国产一区| 久久久久久久大尺度免费视频| 国内毛片毛片毛片毛片毛片| 久久久精品94久久精品| 中文字幕制服av| 国产成人精品久久二区二区91| 我的亚洲天堂| 国产野战对白在线观看| 中亚洲国语对白在线视频| 看免费av毛片| 欧美av亚洲av综合av国产av| 十八禁人妻一区二区| 99久久99久久久精品蜜桃| 巨乳人妻的诱惑在线观看| 亚洲熟女精品中文字幕| 两性午夜刺激爽爽歪歪视频在线观看 | 精品福利观看| 99国产精品免费福利视频| 老司机午夜十八禁免费视频| 精品福利观看| 国产三级黄色录像| 自拍欧美九色日韩亚洲蝌蚪91| 天堂俺去俺来也www色官网| 久久久欧美国产精品| 免费观看a级毛片全部| 欧美日韩国产mv在线观看视频| av在线播放免费不卡| 欧美乱妇无乱码| 搡老熟女国产l中国老女人| 两个人免费观看高清视频| 亚洲熟女精品中文字幕| 午夜福利影视在线免费观看| 另类亚洲欧美激情| 亚洲国产av影院在线观看| 国产精品1区2区在线观看. | 欧美+亚洲+日韩+国产| 亚洲国产精品一区二区三区在线| 我的亚洲天堂| 国产真人三级小视频在线观看| 欧美日韩一级在线毛片| 欧美激情久久久久久爽电影 | 老司机影院毛片| 狂野欧美激情性xxxx| 建设人人有责人人尽责人人享有的| 日韩熟女老妇一区二区性免费视频| 日韩中文字幕视频在线看片| 天堂俺去俺来也www色官网| 日韩欧美免费精品| 欧美久久黑人一区二区| av天堂在线播放| 欧美亚洲 丝袜 人妻 在线| 99久久99久久久精品蜜桃| 亚洲视频免费观看视频| 色婷婷av一区二区三区视频| 欧美成人午夜精品| 中文字幕另类日韩欧美亚洲嫩草| 亚洲精品国产色婷婷电影| 久久精品国产a三级三级三级| 欧美精品一区二区大全| 精品高清国产在线一区| 两个人免费观看高清视频| 黄色片一级片一级黄色片| 久久 成人 亚洲| 国产精品久久久人人做人人爽| 99香蕉大伊视频| 欧美国产精品va在线观看不卡| 在线天堂中文资源库| √禁漫天堂资源中文www| 亚洲全国av大片| 色精品久久人妻99蜜桃| 99久久99久久久精品蜜桃| 在线永久观看黄色视频| 国产区一区二久久| 成人特级黄色片久久久久久久 | 在线观看免费午夜福利视频| 色老头精品视频在线观看| 亚洲精品国产精品久久久不卡| 亚洲国产中文字幕在线视频| www.自偷自拍.com| 色精品久久人妻99蜜桃| 午夜日韩欧美国产| 日本精品一区二区三区蜜桃| 日韩中文字幕视频在线看片| 亚洲精品中文字幕在线视频| cao死你这个sao货| av国产精品久久久久影院| 久久狼人影院| 这个男人来自地球电影免费观看| 日韩免费高清中文字幕av| 一二三四社区在线视频社区8| 国产精品亚洲av一区麻豆| 在线观看免费视频日本深夜| 欧美激情高清一区二区三区| 怎么达到女性高潮| 国产主播在线观看一区二区| 亚洲国产av新网站| 侵犯人妻中文字幕一二三四区| 精品国产一区二区三区久久久樱花| 国产高清国产精品国产三级| 国产男靠女视频免费网站| 制服诱惑二区| 在线十欧美十亚洲十日本专区| 久久精品人人爽人人爽视色| 国产精品免费一区二区三区在线 | 在线观看舔阴道视频| 深夜精品福利| 亚洲精品国产色婷婷电影| 国产真人三级小视频在线观看| 国产精品秋霞免费鲁丝片| 丰满饥渴人妻一区二区三| 亚洲一码二码三码区别大吗| 亚洲成人手机| 啦啦啦 在线观看视频| 男女高潮啪啪啪动态图| 午夜免费成人在线视频| 美女主播在线视频| 亚洲欧美激情在线| 久久亚洲真实| 一本综合久久免费| 国产亚洲欧美在线一区二区| 制服人妻中文乱码| 丁香欧美五月| 精品熟女少妇八av免费久了| 亚洲成人免费电影在线观看| av网站免费在线观看视频| 亚洲va日本ⅴa欧美va伊人久久| 亚洲av日韩精品久久久久久密| 天堂8中文在线网| 欧美黑人精品巨大| 777米奇影视久久| 亚洲伊人色综图| 99精品久久久久人妻精品| 热re99久久精品国产66热6| 老鸭窝网址在线观看| 国产精品久久久久成人av| 亚洲九九香蕉| 在线观看免费高清a一片| 中文字幕人妻丝袜一区二区| 国产精品欧美亚洲77777| 精品国产一区二区三区四区第35| 国产成人免费观看mmmm| 狠狠精品人妻久久久久久综合| 免费人妻精品一区二区三区视频| 欧美日韩av久久| 69av精品久久久久久 | 国产精品免费一区二区三区在线 | 亚洲成人手机| 久久国产精品男人的天堂亚洲| 国产成人免费观看mmmm| 天天躁夜夜躁狠狠躁躁| 午夜精品久久久久久毛片777| 激情视频va一区二区三区| 久久久久久久久久久久大奶| 欧美性长视频在线观看| 成在线人永久免费视频| 一区二区三区国产精品乱码| 啦啦啦 在线观看视频| 国产aⅴ精品一区二区三区波| 久久亚洲精品不卡| 日本撒尿小便嘘嘘汇集6| www.自偷自拍.com| 一进一出抽搐动态| 亚洲国产成人一精品久久久| 中文欧美无线码| 变态另类成人亚洲欧美熟女 | 亚洲国产av影院在线观看| 亚洲国产成人一精品久久久| 日本av免费视频播放| 久久久久久免费高清国产稀缺| 亚洲人成电影观看| 欧美日韩精品网址| 叶爱在线成人免费视频播放| 大片免费播放器 马上看| 国产在线免费精品| 黄色毛片三级朝国网站| 深夜精品福利| 久久人人爽av亚洲精品天堂| 黄片小视频在线播放| av一本久久久久| 国产不卡av网站在线观看| 亚洲国产中文字幕在线视频| 国产精品久久久av美女十八| 操美女的视频在线观看| 乱人伦中国视频| 国产精品秋霞免费鲁丝片| 成人国产一区最新在线观看| av福利片在线| 18禁观看日本| 国产精品美女特级片免费视频播放器 | 91国产中文字幕| 女同久久另类99精品国产91| 夜夜爽天天搞| 亚洲欧美精品综合一区二区三区| 男女下面插进去视频免费观看| 99国产精品一区二区三区| 俄罗斯特黄特色一大片| 免费一级毛片在线播放高清视频 | 自线自在国产av| 亚洲天堂av无毛| 亚洲欧美激情在线| 日韩大码丰满熟妇| 精品少妇一区二区三区视频日本电影| 肉色欧美久久久久久久蜜桃| 香蕉久久夜色| 99国产综合亚洲精品| 一区二区三区乱码不卡18| videosex国产| 精品少妇内射三级| 国产午夜精品久久久久久| 一级毛片女人18水好多| 蜜桃在线观看..| 色综合欧美亚洲国产小说| 国产不卡一卡二| 成年女人毛片免费观看观看9 | 我的亚洲天堂| 久久中文看片网| 男女高潮啪啪啪动态图| 他把我摸到了高潮在线观看 | 亚洲少妇的诱惑av| 亚洲人成伊人成综合网2020| 午夜精品国产一区二区电影| 久久香蕉激情| 丝袜在线中文字幕| 一区在线观看完整版| 狠狠精品人妻久久久久久综合| 女警被强在线播放| 国产在线一区二区三区精| 极品人妻少妇av视频| 午夜免费成人在线视频| 国产一区二区在线观看av| 女人高潮潮喷娇喘18禁视频| 菩萨蛮人人尽说江南好唐韦庄| 久久精品国产a三级三级三级| 99精品久久久久人妻精品| 纯流量卡能插随身wifi吗| 操出白浆在线播放| 国产精品影院久久| 首页视频小说图片口味搜索| 国产精品麻豆人妻色哟哟久久| 自拍欧美九色日韩亚洲蝌蚪91| 国产日韩欧美在线精品| 中文字幕色久视频| 久久狼人影院| 黑人猛操日本美女一级片| 深夜精品福利| 在线播放国产精品三级| 久久精品aⅴ一区二区三区四区| 色尼玛亚洲综合影院| 国产高清激情床上av| 黄色丝袜av网址大全| 国产伦理片在线播放av一区| 久久99热这里只频精品6学生| 久久久水蜜桃国产精品网| 精品少妇久久久久久888优播| 岛国毛片在线播放| 日日摸夜夜添夜夜添小说| 精品久久久久久久毛片微露脸| 欧美日韩国产mv在线观看视频| 香蕉丝袜av| 久久久久网色| 中国美女看黄片| 一区二区三区激情视频| 两性夫妻黄色片| 国产高清视频在线播放一区| 桃花免费在线播放| 亚洲国产毛片av蜜桃av| 国产精品免费大片| 午夜福利乱码中文字幕| 黄色丝袜av网址大全| 国产精品免费一区二区三区在线 |