Title - 8th International Coastal Symposium (Instructions for by s42gs6

VIEWS: 22 PAGES: 5

									Journal of Coastal Research       SI 56            pg - pg            ICS2009 (Proceedings)            Portugal              ISSN




Application of remote sensing video systems to coastal defence
monitoring
D. Rihouey, J. Dugor, D. Dailloux, D. Morichon
LaSAGeC2
University of Pau, Anglet
64600, France
didier.rihouey@univ-pau.fr



                       ABSTRACT

                       RIHOUEY, D., DUGOR, J. DAILLOUX, D., and MORICHON, D., 2009. Application of remote sensing video systems
                       to engineering works monitoring. Journal of Coastal Research, SI 56 (Proceedings of the 10th International
                       Coastal Symposium), pg – pg. Lisbon, Portugal, ISBN

                       Shoreline stability is an important issue along much of the Mediterranean’s Coasts. European project
                       MEDDOOC INTERREG BEACHMED-e focused on the strategic management of beach protection for the
                       sustainable development of the Mediterranean coastal zone. In the framework of this project, a video system has
                       been installed in Valras (Gulf of Lions / France) to monitor coastal evolutions and recent protection works. Over
                       the past 20 years, coastal video remote sensing techniques represent an efficient alternative tool to classical in
                       situ surveying techniques. Coastal video monitoring is based on Time exposure images (Timex) acquisition and
                       photogrammetry technique which allows transforming 2D image coordinates into the corresponding 2D real
                       world coordinates. This paper presents the use of video monitoring technique to estimate the impact of
                       engineering works. Important beach retreat has been observed for decades along the 3 km of Valras beaches and
                       12 breakwaters have been built until 2007. From January to May 2008, a new similar protection, a submerged
                       breakwater and 95,000 m3 sand nourishment have been added. As Empirical Orthogonal Function (EOF)
                       analysis has become an established method for investigating temporal beach fluctuation, EOF is performed using
                       weekly video monitored shorelines over a 6 months period. The results show: (1) the natural erosion / accretion
                       phases of the already protected shoreline, (2) the impact of beach nourishment and the trend toward the
                       equilibrium position of the restored shoreline, (3) the efficiency of video monitoring for shoreline management.

                       ADITIONAL INDEX WORDS: Video monitoring, Empirical Orthogonal Function, Engineering work



                     INTRODUCTION                                         The use of shore based video systems has been introduced by
   Coastal zones of the Mediterranean Sea are territory of             the Coastal Imaging Lab, University of Oregon, in the beginning
particular interest for sustainable strategic development.             of the 1990s. Coastal video monitoring is based on Time exposure
Quantitative analysis of sandy beaches evolution play an essential     images (Timex) acquisition and photogrammetry technique which
part in the integrated management of coastal zones. They are           allows transforming image coordinates into the corresponding real
especially critical when planning coastal defence and assessing        world coordinates (HOLLAND et al., 1997). The so-called ARGUS
their efficiency. The European project MEDDOOC INTERREG                system have been recently used in the Coast View project
BEACHMED-e and more precisely the OpTIMAL (Optimisation                (DAVIDSON and MEDINA, 2007) for developing video-derived
of Integrated Monitoring Techniques Applied to Coastlines)             Coastal State Indicators (ICSs) in support of coastal management.
subproject focused on the development and the implementation of        With the increasing offer of video cameras and video
measurements techniques to characterize erosion for the                technologies, several systems have been developed in the past
sustainable use of the resources.                                      years and tested in the framework of the OpTIMAL project.
   Monitoring of near-shore systems has traditionally relied on in        In December 2007, a Kosta System video station has been
situ measurements of waves, currents, sediment transport and           installed in Valras (Gulf of Lions / France) to monitor coastal
morphological changes. These technologies provide data of high         evolutions and recent protection works. The aim of this paper is to
quality, but have limited resolution in time and space because of      present the use of video monitoring technique to estimate the
the expense and logistical difficulties associated with deployment.    impact of engineering works by combining weekly shoreline
Satellite and airborne remote sensing techniques have improved         detection and Empirical Orthogonal Function (EOF) analysis.
the spatial coverage of measurements with reasonable resolution.           The paper is set out as following: Section 2 gives details on the
However, the use of these techniques is not cost-efficient for the     study site and the description of the Valras video station. Section 3
purpose of long term, high resolution monitoring of the near-shore     describes the methodology used to automatically detect the
processes imply in coastal management. Over the past 20 years,         shoreline position from the video system, as well as the EOF
shore-based video remote sensing systems have provided an              technique used for this study. The results and the conclusions are
alternative low cost tool to remotely survey coastal areas.            presented respectively in Section 4 and 5.




                                           Journal of Coastal Research, Special Issue 56, 2009
                   Short Running Head (e.g. Beach Management and Safety. Use Times New Roman 9 font normal)



        STUDY SITE AND VIDEO SYSTEM                                                               METHODS
   Valras is a beach resort located on the Gulf of Lion,
Mediterranean Sea, France. The Gulf of Lion is a microtidal area        Shoreline detection
(spring tidal range of about 0.4 m) and the wave climate is                Images acquired by the video station usually show a strong
characterized by a significant annual wave height of 0.83 m, a          contrast between blue sea water and yellow sandy beaches.
peak period of about 5 seconds, and a south-east direction. The         Different algorithms have already been developed and proved
sandy beaches of Valras stretch between the Orb and the Aude            their efficiency to automatically detect the position of the
river-mouth (Figure 1). The 3 km of Valras coastline are highly         shoreline (PLANT and HOLMAN, 1997; OSORIO, 2005;
eroded since the building of the harbour jetty entrance in the          AARNINKHOF, 2003). They are based on color information or on
1960s. The longshore drift, induced by East wind storm, is              the brightness level recorded by the CCD sensor of the video.
stopped by this jetty located on the west side of the Orb river-        However, these techniques can be limited when the contrast
mouth. Hence, important beach retreat has been observed and 12          between water and sand is blurred (fog, rain, high turbidity level,
breakwaters have been built until 2007. From January to May             algal bloom, sun reflections). In this study, the technique used is
2008, a similar protection, a submerged breakwaters and                 based on a mathematical approach. The RGB and HSV models are
95,000 m3 sand nourishment have been added.                             combined as six independent parameters treated by a clustering
                                                                        algorithm based on a k-mean function (MORICHON et al., 2008;
                                                                        DAILLOUX, 2008). This function aims at discriminating two
                                                                        clusters respectively referred to as the wet area (sea water) and the
                                                                        dry area (sand). Then, a selection criteria allows to automatically
                                                                        detect the two parameters which centroïds are the most distant,
                                                                        with the less diffusive distribution, and the best ratio between the
                                                                        number of points in each cluster. Finally, the clustering algorithm
                                                                        is used on the selected couple to detect the position of the edge
                                                                        between water and sand (Figure 3).



 Figure 1. Field of study

  Since December 2007, a Kosta System video station, composed
of 6 cameras (0.8 Mega pixels), was installed on a top of a 47m
high building. The accuracy of the video system is governed by
the nearly rectangular dimension of a pixel foot print (LIPPMAN          Figure 3. (a) Initial selected region, (b) k-mean function
and HOLMAN, 1989). The Valras station configuration aims to              analysis, and (c) plot of the detected shoreline.
remotely monitor the shoreline over a 3 km radius. In this study
the system is used to remotely survey the constructed area                 Based on this technique, a serie of shorelines were detected on
covering a coastline of 1 km on each side of the station. Inside this   rectified plan-view images. The first step consisted in selecting the
area, the cross-shore resolution is below 1 m per pixel, and the        most appropriate images for the analysis. The images were
longshore resolution range from the centimeter (near the station)       selected on a weakly basis, during calm days (and after periods of
to 10 m (at the end of the domain). Figure 2 shows the final            calm conditions), for a weak sea water level variation according to
projected image merged from the 6 cameras images, and the               the tidal gauge located at the Sète Harbor. Between December
remotely surveyed areas for this study.                                 2007 and June 2008, 19 shorelines detection were performed. The
                                                                        maximum sea level range recorded by the gauge for the 19
                                                                        shorelines was about 20 cm. The error induced during the video
                                                                        detection of the shoreline by the sea level variation is proportional
                                                                        to the slope of the beach.

                                                                        EOF Analysis
                                                                           The empirical orthogonal function (EOF) technique is used to
                                                                        define the patterns of spatial and temporal behaviour of the
                                                                        shoreline position. The EOF method determines the shape of the
                                                                        expansion functions directly from the data, rather than from an a
                                                                        priori selection of shape functions. Developed from the early to
                                                                        mid 1900s (PEARSON, 1901; HOTELLING, 1933), the EOF
                                                                        technique has become widely known and used across a broad
                                                                        range of scientific disciplines with the advent of electronic
                                                                        computing. In the coastal sciences the first application of the EOF
                                                                        technique was realised by WINANT et al. (1975) to analyse beach
                                                                        profile measurements. After this pioneering study, EOF has
                                                                        become a commonly applied technique in morphological research
Figure 2. Rectified plan-view image of the Valras video station,        to investigate beach response over time scales of month to decades
and the studied areas (black squares).                                  (LARSON et al., 2003 ; RIHOUEY, 2004 ).




                                            Journal of Coastal Research, Special Issue 56, 2009
                  Author’s last name (e.g. Smith or Smith and Jones or Smith et al. Use Times New Roman 9 font normal)



   The EOF technique may be described briefly as follows. We                                       eigenfunction (Figure 6.a to 9.a) corresponds to the shoreline
denote the discrete shoreline position by y(xl,tk), where xl is long-                              mean position over the period. The related temporal function
shore position and tk the time of the data points, with 1 ≤ l ≤ L and                              describes fluctuations in cross-shore position of the mean
1 ≤ k ≤ K . Thus, the idea of EOF analysis is to expand y as a                                     shoreline shape. Thus, the East part of Valras shoreline oscillates
                                                                                                   around an equilibrium position (Figure 6.b and 7.b) whereas the
linear combination of functions of space and time:
                                                                                                   West part migrates offshore (Figure 8.b and 9.b) during the beach
                                   L

                               ∑C
                                                                                                   nourishment and then tends toward an equilibrium position.
           y(xl , t k ) =                      p   (t k ) e p (xl )                    (1)         Indeed, low value of c1(t) (dot on Figure 8.b) corresponds to weak
                               p =1                                                                beach width (Figure 8.g) whereas high value of c1(t) (dot on
where ep, often referred to as the spatial eigenfunctions, are                                     Figure 9.b) corresponds to largest beach width (Figure 9.g).
determined directly as the eigenfunctions of the correlation matrix                                   Figures 4.b and 5.b represent the correlation between beach
of the data A, together with their corresponding eigenvalues λp:                                   surface evolution over the period (Figure 4.a and 5.a) and the 1st
                                                                                                   temporal eigenfunction. Thus, the 1st temporal eigenfunction is
          A ep = λp ep                                                                 (2)         highly correlated with the beach surface evolution of the East part
                                                                                                   (r2 =0.92) and the West part (r2 =0.99) of Valras. The error
Hence, if A is real and symmetric it has L real eigenvalues, and its                               induced by sea level variation on beach surface calculation has
eigenvectors may be chosen as mutually orthonormal, that is,                                       been evaluated (GAUFRES et al., 2008) and are negligible (<8%) in

          ∑ e (x ) e (x ) = δ
            L                                                                                      comparison with surface evolution (Figure 4.a and 5.a).
                  p   l        q       l               pq                              (3)
           l =1
Where δpq is the Kronecker delta. The correlation matrix A is
calculated from the data. It has L×L elements aij of the form:
                           K

                          ∑ y(x , t ) y(x , t )
                      1
          ai j =                           i       k        j       k                  (4)
                      K   k =1
 Finally, the temporal coefficients cp (t) called also weightings,
may be calculated from equation (1) and (3):

                ( ) ∑ y(x , t ).e (x )
                                                                                                    Figure 4. (a) East beach surface evolution (m²) and
          c p tk =                         l       k    p       l                      (5)          corresponding error bar. (b) Correlation between beach surface
                          p                                                                         evolution and the 1st temporal mode.
   Several properties of real symmetric matrices may be used to
aid the interpretation of some of the calculated quantities. For
instance, the trace of A is equal to the mean square value (or
‘energy’) of the data. It is also true that the sum of the eigenvalues
is equal to the mean square of the data, and so each individual
eigenvalue, λp, represents the relative contribution of mode p to
the total variability. Also, when a real symmetric matrix is
diagonalised, the diagonal elements are the eigenvalues, and the
matrix rows and columns can be rearranged via elementary matrix
operations so that the eigenvalues are in decreasing order,
λ1 > λ2 >…> λL. Then, the eigenvector e1 is the vector that                                         Figure 5. (a) West beach Surface evolution (m²) and
accounts for most of the mean square value and e2 represents the                                    corresponding error bar. (b) Correlation between beach Surface
maximum energy from what is left and the same with the other                                        evolution and the 1st temporal mode.
eigenvectors. The shape functions defined by the EOF analysis
can be interpreted as various ‘modes’ of variation in analogy with                                    The subsequent eigenfunctions represent the variation about the
Fourier analysis. As an example, if the data is not de-meaned, the                                 mean. For both EOF Analysis, the 2nd and the 3rd eigenfunctions
first eigenfunction is equivalent to the time mean computed                                        together account for over 85% of the variance.
directly from the data (ARANUVACHAPUN and JOHNSON, 1979).                                             East to the video station, the 2nd (Figure 6.c, 7c) and the 3rd
The second eigenfunction represents the first ‘mode of variation’                                  (Figure 6.e, 7.e) spatial modes relate to salient morphologies. The
about the time mean, and so on.                                                                    corresponding temporal functions describe progradation /
                                                                                                   destruction cycles of this sedimentary structures. Indeed, low
                                       RESULTS                                                     value of c2(t) (dot on Figure 6.d) corresponds to “flat shaped”
  First, the 19 detected shorelines are projected in a common                                      shoreline observed post storm (Figure 6.g) whereas high value of
coordinate system in which the x axis is parallel to the longshore                                 c2(t) (dot on Figure 7.d) corresponds to well developed salient
direction, and the y axis is parallel to the cross-shore direction. In                             (Figure 7.g).
order to separate the already protected part of the beach (East to                                    West to the video station, the 2nd (Figure 8.c, 9c) and the 3rd
the station) and the nourished part (West of the station), EOF                                     (Figure 8.e, 9.e) spatial modes relate to salient morphologies but
analysis has been performed respectively for the East part and the                                 also to beach nourishment. The 3rd temporal function (Figure 8.f,
West part of Valras. For both EOF Analysis, over 98 % of the                                       9.f) represents the beginning of the nourishment carried out from
mean square of the data is contained in the first function and                                     January to March and located between 300 m and 600 m (Figure
99.5% of the mean square of the data is captured by the first three                                10.a, 10.b). The 2nd temporal function (Figure 8.d, 9.d) represents
modes. As the data are not de-meaned, the first spatial                                            the end of the nourishment carried out from March to May and
                                                                                                   located between 600 m and 900 m (Figure 10.c, 10.d).



                                                                        Journal of Coastal Research, Special Issue 56, 2009
                 Short Running Head (e.g. Beach Management and Safety. Use Times New Roman 9 font normal)




Figure 6. Spatial EOFs for shoreline of the East part of Valras:     Figure 7. Spatial EOFs for shoreline of the East part of Valras:
(a) 1st mode, (c) 2nd mode, (e) 3rd mode. Related temporal           (a) 1st mode, (c) 2nd mode, (e) 3rd mode. Related temporal
function: (b) 1st mode, (d) 2nd mode, (f) 3rd mode. Shoreline        function: (b) 1st mode, (d) 2nd mode, (f) 3rd mode. Shoreline
reconstructed with the first 3 eigenfunctions (g) for the 6th of     reconstructed with the first 3 eigenfunctions (g) for the 18th oh
January (post storm).                                                March (well developed salient).




Figure 8. Spatial EOFs for shoreline of the West part of Valras:     Figure 9. Spatial EOFs for shoreline of the West part of Valras:
(a) 1st mode, (c) 2nd mode, (e) 3rd mode. Related temporal           (a) 1st mode, (c) 2nd mode, (e) 3rd mode. Related temporal
function: (b) 1st mode, (d) 2nd mode, (f) 3rd mode. Shoreline        function: (b) 1st mode, (d) 2nd mode, (f) 3rd mode. Shoreline
reconstructed with the first 3 eigenfunctions (g) for the 6th of     reconstructed with the first 3 eigenfunctions (g) for the 5th of
January (post storm).                                                April (during the beach nourishment).




                                         Journal of Coastal Research, Special Issue 56, 2009
               Author’s last name (e.g. Smith or Smith and Jones or Smith et al. Use Times New Roman 9 font normal)




Figure 10. Timex images from the Valras video station : (a) before the engineering work in December 2007, and (c) during the en
engineering work in February (b) and in April (c), after the engineering work in June (d)

                                                                      HOTELLING, H., 1933. Analysis of a complex of statistical
       CONCLUSION AND PERSPECTIVES                                       variables into principle components. Journal of Educational
                                                                         Psychology 24, 417–441.
   An application of remote sensing video systems to coastal          HSU, T.W., OU, S.H. and WANG, S.K., « On the prediction of
defence monitoring has been presented. EOF technique was                 beach changes by a new 2-D empirical eigenfunction model »
applied to shoreline contours extracted from 10 min times-average        Coastal Engineering, Vol. 23, p. 255-270, 1994.
images at constant sea level (CSI). The results shows that this       GAUFRES, P., ANDRES, B., TROUVILLIEZ, A., RIHOUEY, D. DUGOR,
approach allows to quantify: (1) the beach surface, (2) the natural      .J, 2008. Application de la video pour le suivi de l’impact des
erosion / accretion phases of the already protected shoreline, (3)       travaux de defence du littoralde Valras-Plage (Hérault).
the impact of beach nourishment and the trend toward the                 Proceedings of the 10th Journée Nationnales du Génie Côtier-
equilibrium position of the restored shoreline. The technique            Génie-Civil (Sofia Antipolis, Franc), pp. 543-552.
presented in this study seems to be a suitable tool to coastal        KROON, A., DAVIDSON, M.A., AARNINKHOF, S.G.J., ARCHETTI, R.,
engineering purpose, as previously shown by Kroon et al. (2007).         ARMAROLI, C., GONZALEZ, M., MEDRI, S., OSORIO, A.,
Nevertheless, the short duration of the video monitoring period          AAGAARD, T., HOLMAN, R.A., SPANHOFF, R., 2007.
still not allows an accurate appreciation of the coastal defence         Application of remote sensing video systems to coastline
efficiency. Moreover, the three first modes together account for         management problems. Coastal Engineering 54, 493–505
over 99.5% of the mean square of the data. This is a large            LARSON, M., CAPOBIANCO, M., JANSEN, H., ROSZYNSKI, G.,
proportion of the variance for analyses of this kind (Reeve et al.,      SOUTHGATE, H., STIVE, M., WIJNBERG, K., 2003. Analysis and
2008) and, as expected, the EOFs describe the variability in the         modelling of field data on coastal morphological evolution
shoreline behaviour efficiently. In the near future, a considerable      over yearly and decadal time scales: part 1. Background and
interest will be brought in continuing the Valras video monitoring,      linear techniques. Journal of Coastal Research 19, 760–775.
as EOF analysis is a suitable tool to initiate the development of a   MORICHON, D., DAILLOUX, D., AARNINKHOF, S., ABADIE, S.,
data-driven model for predicting nearshore morphological                 2007. Using a shore based video system to hourly monitor
evolution (AUBREY, 1980; HSU et al., 1994; RIHOUEY, 2007;                storm water plumes (Adour River, Bay of Biscay). Journal of
REEVE et al., 2008).                                                     Coastal Researc, 24, 133-140.
                                                                      OSORIO, A., 2005. Video-derive techniques and methodologies for
                                                                         coastal management: Universidad de Cantabria Phd Thesis. .
                        REFERENCE                                     PEARSON, K., 1901. On lines and planes of closest fit to systems of
AARNINKHOF S.G.J., TURNER I.L., DRONKERS T.D.T.,                         points in space. Philosophical Magazine 6 (2), 559–572.
   CALJOUW M., NIPIUS L., 2003. A video-based technique               PLANT, N.G, HOLMAN R.A, (1997). Intertidal beach profile
   for mapping intertidal beach bathymetry. Coastal                      estimation using video image. Marine Geology 140, 1-24.
   Engineering, 49 (4), 275-289.                                      REEVE, D.E., HORRILLO-CARABALLO J.M., MAGAR, V, 2008.
AUBREY, D.G., 1980. The statistical Prediction of Beach Changes          Statistical analysis and forecasts of long-term sandbank
   in southern California. Journal of Geophysical Research, 85           evolution at Great Yarmouth, UK. Estuarine, Coastal and
   (6), 1980, p. 3264-3276.                                              Shelf Science 79 387–399
ARANUVACHAPUN, S., JOHNSON, J.A., 1979. Beach profile at              RIHOUEY, D., 2004. Analyse statistique de l’évolution
   Gorleston and Great Yarmouth. Coastal Engineering 2, 201–             morphodynamique des plages sableuses. Application aux sites
   213.                                                                  d’étude du Programme National d’Environnement Côtier et
DAILLOUX, D., 2008. Video measurements of the Adour plume                aux plages d’Anglet : Université de Pau et des Pays de
   dynamic and its surface water optical characteristics:                l’Adour, Ph.D. thesis, 196p.
   Université de Pau et des Pays de l’Adour, Ph.D. thesis, 169p.      WINANT, C.D., INMAN, D.L., NORDSTROM, C.E., 1975.
DAVIDSON, M.A., VAN KONINGSVELD, M., DE KRUIF, A., RAWSON,               Description of seasonal beach changes using empirical
   J., HOLMAN, R.A., LAMBERTI, A., MEDINA, R., KROON, A.,                eigenfunctions. Journal of Geophysical Research 80 (15),
   AARNINKHOF, S.G.J., 2007. The CoastView project:                      1979–1986.
   developing video-derived coastal state indicators in support of
   coastal zone management. Coastal Engineering 54, 463–475.
HOLLAND, K.T., HOLMAN, R.A., LIPPMANN, T.C., STANLEY, J.,                               ACKNOLEDGEMENT
   PLANT, N., 1997. Practical use of video imagery in nearshore         The authors acknowledge The Beachmed-e Project and the
   oceanographic field studies. IEEE Journal of Oceanic               Hérault council (CG34) for the founding. We also want to thank
   Engineering 22 (1), 81–92.                                         the DRE of Languedoc Roussillon for tidal data and the CETMEF
                                                                      for technical support.



                                           Journal of Coastal Research, Special Issue 56, 2009

								
To top