This file is part of the following reference: Access to this file is available from:

This file is part of the following reference: Page, Geoff (27) Mass Transport Evaluation using consolidated VHF Radar and Acoustic Doppler Current Profiler data. Masters (Research) thesis, Jaes Cook University. Access to this file is available fro: http://eprints.jcu.edu.au/285

5 Data Analysis and Results This chapter details how the radar tie-series acquisitions obtained over the six week deployent were analysed and copared with ADCP easureents. The first section in this chapter exaines the quality of the data obtained fro both the radar and the additional instruentation. Fro the very start of the data analysis process it was evident that the radar was not perforing to the expected range specifications. Although unknown throughout ost of the research effort, the PortMap radar systes used in Venice had a prograing fault within the real-tie coputer software. This was reported by the anufacturer in Noveber, 26 after ost of the data analysis had already taken place. The technical details of this fault will be covered in the final section of this chapter, Section 5.8. The fault effectively caused the range resolution of the syste to be 2 instead of. This reduced the already poor range of the radar by a factor of 5. As such, the range of the syste was insufficient for the purpose of evaluating ass transport through the Lido inlet. Sections 5. through to 5.7 of this chapter describe how current easureents were obtained fro the radar spectra and how this could be used to produce surface current easureents along a transect. This was perfored on the understanding that the PortMap systes had a range resolution of. Although now known to be incorrect because of the range resolution error, these sections detail how the analyses and algoriths were developed to provide surface current easureents across the Lido channel transect. As previously explained, the final section describes the technical details of the PortMap systeʼs fault. It also shows how the direction finding algorith, DFind was odified to allow for this fault to enable it to create short-range radial current easureents using a range resolution of 2, assuing that the doinant energy is fro signals of this range resolution. As will be seen in the Discussion chapter, sufficient surface current easureents were obtained 56

within a short range to observe an interesting surface current dynaic within the Lido channel. 5. Quality of Data The quality of the data obtained fro the PortMap radar, ADCP, Tide Gauge and Meteorological station (Weather station) are exained in this section. Events causing a coplete loss of data where systes were unavailable due to aintenance disruptions or syste failure are also reported in this section. 5.. ADCP, Tide Gauge and Meteorological Data Quality This additional data was provided by the partner organisations, the National Institute of Oceanography and Applied Geophysics (OGS - Trieste) and The Institute of Marine Science (CNR-ISMAR - Venice). It was obtained using coercial instruents and sensors. This provided high quality, reliable data fro instruents that have been deployed and used successfully over a long period of tie and for previous studies. For the period of investigation, st October - th Noveber, 25 there were no lapses in data (% data loss) fro either the Tide Gauge or Meteorological data sets. These were recorded at 5 inute intervals. The tie-series plot shown on the following page (Figure 5.) represents the wind speed, direction and wind vector data obtained fro the weather station for the st October. The self-contained ADCP that was deployed on the sea floor within the channel was retrieved on the 2th October to download the logged data, and for aintenance and cleaning. This resulted in a lapse of ADCP record data on this date for a period of 3 hrs, 4 inutes fro 2:2:2 UTC to 6::2 UTC. This was considered to be negligible as this represents a ere.36% data loss over the period of interest. A tie-series plot of the tide gauge data for the Lido Inlet is displayed in Figure 5.2. This is plotted together with the velocity of the ADCP bin closest to the 57

surface, resolved into the axis of the Lido channel. This gives an overview of the tidal level and corresponding tidal strea velocities within the channel throughout the observation period. Periods of Neap and Spring tides can be clearly observed in this plot. 7 Wind Speed! Oceanographic Platfor /s 6 5 4 3 2 274 274. 274.2 274.3 274.4 274.5 274.6 274.7 274.8 274.9 275 Day Nuber Wind Direction! Oceanographic Platfor 27 Degrees 8 9 274 274. 274.2 274.3 274.4 274.5 274.6 274.7 274.8 274.9 275 Day Nuber Wind Vector! Oceanographic Platfor 5.s - 274 274. 274.2 274.3 274.4 274.5 274.6 274.7 274.8 274.9 275 Day Nuber Figure 5. Tie-series of Wind Speed, Wind Direction and Wind Vector for st October 58

Tide Gauge! Lido Tidal Height,.5!.5 275 28 285 29 295 3 35 3 35 Day Nuber 2 Lido Channel ADCP! Bin 9.5 Current Speed, /s.5!.5!!.5 275 28 285 29 295 3 35 3 35 Day Nuber Figure 5.2 Tidal Level and ADCP Bin 9 Velocities for the Lido Channel In Figure 5.2, the discontinuity on DAY 285 in the Lido Channel ADCP plot represents the period that the ADCP was being serviced. With such a high quality easureent signal, the velocity data obtained fro the ADCP provides a sound reference with which to verify the accuracy of the easureents obtained using the PortMap Ocean Surface Current Radar for the ADCP location. 59

5..2 Radar Data Quality The availability and quality of the easureent of surface currents fro the PortMap syste depend on obtaining received signals with a good signal-tonoise ratio (SNR) throughout the ranges of interest. During and iediately following the deployent period, analysis of the radar data took place to observe the signal-to-noise ratio of the recorded signals. This was priarily used to gauge the range at which reliable current data could still be obtained fro the radar signals. A post-processing utility, PMAP2DAT.EXE, developed at Jaes Cook University, was used to generate power spectru files fro the binary tie-series files written by the PortMap radar. PMAP2DAT was used to output ASCII.dat files containing a 248 point power spectru for each antenna, for all 5 recorded range cells. These processed data files were then used by a MATLAB graphical user interface tool developed as part of this Thesis work to present the power spectra graphically as a waterfall plot (Figure 5.3). Power Spectru! Antenna Power Spectru! Antenna 2!2!2!4!4!6!6!8!8!! Power!2 Power!2!4!4!6!6!8!8!2!2!22!22!4!3!2! 2 3 4 Frequency Power Spectru! Antenna 3!2!4!6!8!4!3!2! 2 3 4 Frequency Power Spectru! Antenna 4!2!4!6!8 Power!!2!4!6!8!2!22 Power!!2!4!6!8!2!22!4!3!2! 2 3 4 Frequency!4!3!2! 2 3 4 Frequency Figure 5.3 Noralised power spectru plots (ranges to 5) - Good SNR, Index 6 6

For a single 4 inute, 26 second acquisition, the power spectru is soothed using a digital low-pass filter and displayed for a specified nuber of range cells (typically range cells to 5). The data are also noralised and vertically shifted downward with increasing range to allow for easier inspection. As such, the closest range cell is at the top of the graph. This graphical representation for each acquisition was visually inspected to estiate the effective range of the radar. The priary indication of good signalto-noise ratio in the recorded data was the obvious presence of Bragg scatter peaks either side of the Hz centre over a nuber of ranges. As shown in Figure 5.3, a record with good signal-to-noise ratio shows obvious Bragg peaks extending out to 5 range cells. A record with very poor signal-to-noise ratio such as that shown in Figure 5.4 shows no discernible Bragg peak beyond the first one or two range cells. Power Spectru! Antenna Power Spectru! Antenna 2!2!2!4!4!6!6 Power!8 Power!8!!!2!2!4!4!6!6!4!3!2! 2 3 4 Frequency Power Spectru! Antenna 3!4!3!2! 2 3 4 Frequency Power Spectru! Antenna 4!2!2!4!4!6!6 Power!8 Power!8!!!2!2!4!4!6!6!4!3!2! 2 3 4 Frequency!4!3!2! 2 3 4 Frequency Figure 5.4 Noralised power spectru plots (ranges to 5) - Poor SNR, Index 6

An estiate of range was evaluated for every record obtained fro the radar throughout the six week deployent period. This estiated range was a value fro to 6, with a value of 6 indicating that the radar could see in excess of 6 range cells. A rating zero indicated the worst case where the radar had no perceived usable data in this sapling period. As previously indicated, the range resolution during this process was understood to be at the tie of this analysis. The range estiates for each station were tabled and copared. The histogra in Figure 5.5 shows the distribution of range quality for each station, and for each look direction (left ost colour-bar represents the look direction within the channel). Ties that the stations collected no data due to failure are ignored. For each station, the radar is ostly perceived to be able to gather data within the first 5 range cells only. Of particular note is the significant nuber of ties that the Lido station is unable to observe any returned Bragg peaks. 55 Range Distribution! Sabbioni Boresight 28 T 55 Range Distribution! Lido Boresight 336 T 5 Boresight 22 T 5 Boresight 36 T 45 Boresight 6 T Boresight T 45 Boresight 96 T Boresight 56 T 4 4 35 35 Counts 3 25 Counts 3 25 2 2 5 5 5 5 2 3 4 5 6 Ranges 2 3 4 5 6 Ranges Figure 5.5 Distribution of range quality, Sabbioni and Lido Stations 62

5..3 Radar Failure Events Although efforts were in place to onitor the stations daily to inspect and correct faults, soe data went unrecorded due to syste faults. Aside fro the Lido station fault that resulted in poor data records, there were only two coplete failure events that occurred during the deployent. The workstation coputer at the Sabbioni station copletely failed at 2 local tie on the 24th October due to a rodent interfering with the coputerʼs ainboard. This was not locally repairable, and a replaceent workstation coputer had to be shipped fro Gerany. This station was brought back online on the 28th October, at 55 local tie. This resulted in a coplete loss of data for a four day period. At the Lido site, inor faults with the antenna controller caused two inor disruptions to the noral sequence of data recording. In the first instance, water ingress caused the power to the antenna to be disrupted on the 3th October. Although data were still recorded, they were only obtained fro a single look direction until the fault was fixed the following day. Siilarly, a echanical fault with the otor controller caused the antenna to becoe stuck in a single look direction the following day. No coplete syste failures occurred at the Lido station. 5..4 Obtaining Surface Current Measureents Fro the visual inspection of the recorded radar spectra, we see that ost of the tie the radar has a range of 5 or less ( range resolution) and is not perforing to specification. Since the Lido channel is approxiately 9 wide, it is obvious and readily seen in Figure 5.6 that the radar syste is not providing adequate, overlapping coverage of the channel fro both stations. Overlapping coverage of the easureent area is necessary to resolve the two diensional velocity of the surface current. Without this, it is only possible to obtain radial coponent current easureents fro each radar station. 63

#, 28 T Sabbioni #7, T #2, 336 T #4, 36 T #3, 22 T #5, 6 T Lido #6, 96 T #8, 56 T Figure 5.6 Estiated Maxiu Radar Range (Iage ap courtesy of Cnes/Spot, DigitalGlobe, TerraMetrics and GoogleEarth) In order to obtain surface current easureents within the channel, we can ake the assuption that the current flow within the channel is constrained in the direction of the axis of the channel (toward the lagoon, 36 ). Fro this, if a region of interest is sufficiently within the channel, the radial velocity obtained fro the radar easureent can be resolved in the direction of the channel. This liits our use of the radar data to that obtained fro the first two look directions, up the channel (28 and 336 ) and across the channel (22 and 36 ). The next section explains how surface current easureents are obtained for the channel using a cobination of PMAP2DAT and additional MATLAB software. 64

5.2 Radar Analysis - PMAP2DAT Direction Finding An Ocean Surface Current Radar can resolve aziuth direction by either of two ethods, bea-foring or direction finding. The use of direction finding to resolve the aziuth of received backscatter signals fro the ocean is described by Barrick et al. (977). Direction finding techniques are also described in ore detail in Section 5.6 which presents the developent of the DFind direction finding algorith. In this section, we describe the use of the PMAP2DAT postprocessing utility which uses direction finding techniques to derive surface current easureents fro the binary (.SORT) tie-series data. 5.2. PMAP2DAT Direction Finding According to processing paraeters specified by the user, PMAP2DAT obtains current easureent data fro the binary (.SORT) tie-series data recorded by the PortMap radar. This process is perfored sequentially in three stages by batch processing files, as specified by the user. These stages include the generation of direction finding radial currents, the re-gridding of these radials onto user specified gridpoints, and the conversion of these results into ASCII.dat files to be iported into other software packages. For each acquisition file, the direction finding algorith uses Fast Fourier Transfors to create a 248 point frequency spectru for each range of the 248 tie-series saples recorded fro each antenna receiver. A weighted su of the power spectru for all the antennas is created and then soothed. A representation of this spectru is shown in Figure 5.7. This sued power spectru is used to deterine the location of the Bragg peaks within the spectru. The Bragg peaks are identified and classified by the aount of energy contained in the peaks found, and whether the peaks are above a pre-deterined noise threshold. Various other ethods are also eployed in PMAP2DAT to copare peaks, including coparisons between peaks either side of Hz. 65

8 x 4 4 Points 2925 7 6 Relative Power Density 5 4 3 2!4!3!2! 2 3 4 Doppler Frequency (Hz) Figure 5.7 Exaple power spectru with Bragg peaks of interest Once the Bragg peak locations are identified in the spectru, the algorith uses these frequency bounds to process the original spectru fro each antennae. Within a Bragg peak region of interest, direction finding techniques are used to identify the aziuth of origin for each frequency spectru point. This uses the properties of the linear antenna array, together with the I and Q values obtained for each antenna spectru to ʻsteerʼ the array. The aziuth angle is the angle at which the signal agnitude is a axiu. Each frequency point that is processed using direction finding has an aziuth angle and a corresponding radial velocity as per the Bragg wave equation (2.7). The easureents are all collated into range radials, i.e. radial easureents for each range cell. These range radials are written to an interi binary file to be processed by the re-gridding operation. 5.2.2 Re-gridding Radial Currents The second processing stage re-grids the range radials obtained for each range cell onto a grid of latitude and longitude coordinates, as specified by a user gridpoint file. For the Venice deployent, a cartesian coordinate syste with the Sabbioni station at the origin was chosen for the re-gridding of the radials. A square grid for each radar station is used to re-grid the radial velocities. As discussed in Section 5., this is restricted to points within the 66

channel where it is assued that the current flow is constrained to the direction of the axis of the channel. Grid points that are at angles ill-conditioned for resolving the radial current in the direction of the channel are reoved. Points excluded were within +/- 5 of a line perpendicular to the channel direction passing through each radar station. Points where the current was not considered to be constrained within the channel were also excluded. The resulting grid for each station is shown in Figure 5.8. 8 6 4 Distance fro Sabbioni () 2 A 5!2!4 5!6 Sabbioni - blue asterisks Lido - red circles A - Moored ADCP Location 2 c/s c/s = 5 on axis scales!4!2!!8!6!4!2 Distance fro Sabbioni () Figure 5.8 Channel Grid points for Sabbioni and Lido 67

This linear grid was translated into the required coordinates of latitude and longitude using spherical trigonoetry (great circles). A diagra representing the geoetry used for this is shown in Figure 5.9. North Pole C b a Sabbioni A α c B Grid Point Figure 5.9 Spherical Trigonoetry for Latitude/Longitude Conversion Assuing that the distance between Sabbioni and a selected gridpoint is considered sall in relation to the radius of the earth, we can use a great circle passing through the Sabbioni station and the gridpoint to calculate the latitude and longitude of the gridpoint. Using the spherical trigonoetry forulae cos(a) = cos(b) cos(c) + sin(b) sin(c) cos(a) (5.) sin(a) sin(a) = sin(b) sin(b) = sin(c) sin(c) (5.2) let b = Latitude of Sabbioni, the radius of the earth, Re = 637 k and channelx and channely represent the X and Y distance fro Sabbioni. 68

Now, A = π/2 α (5.3) where If channely α = tan channelx (5.4) channelx is less than zero then α = α + π The angle c (radians) is given by c = channelx 2 + channely 2 /Re (5.5) and equation (5.) can now be used directly to solve for cos(a). Using this result, a = cos [cos(a)] and sin(a) = cos 2 (a) (5.6, 5.7) Now, using Equation (5.2), we ay solve for sin(c) :- sin(c) = sin(c) sin(a) sin(a) (5.8) The latitude and longitude of the selected gridpoint are thus given by :- GridP ointlat = a (5.9) GridP ointlong = LongSab + C (5.) A MATLAB script was used to generate a separate gridpoint file (.grid) containing the Latitudes and Longitudes of the gridpoints for each radar station. PMAP2DAT now re-processes the radial currents obtained for each range cell and perfors an inverse distance weighted su of the radial velocities within a region surrounding each gridpoint contained in each stationʼs respective.grid file. Once processed, these gridpoint radials are output to a second interi binary file. 5.2.3 Archiving Data The final processing step takes the radial currents for each of the specified gridpoints, and outputs these to an ASCII.dat file. These files consist of 69

gridpoint coordinates relative to the origin (Sabbioni Station), surface current speed (/s) and bearing (degrees East of True North). This file forat is suitable for iporting into other software tools such as MATLAB. MATLAB is used for all further analyses. Figure 5. shows an exaple of a radial current ap generated by MATLAB using an archived radials file produced by PMAP2DAT. 8 Radials Plot: Sabbioni & Lido Stations, Separate!34! 6 4 Distance fro Sabbioni () 2 A!2!4!6 Sabbioni data (blue) Lido data (red) A - Moored ADCP Location c/s = 5 on axis scales 2 c/s!4!2!!8!6!4!2 Distance fro Sabbioni () Figure 5. Radial Current Map 5.2.4 Channel Currents The radial current easureents obtained using PMAP2DAT are now further processed to obtain the surface current within the channel. As shown in the section on radar quality in Section 5., there is not adequate overlapping coverage of the channel fro both stations. We ake the assuption that if the region of interest is constrained within the channel, then the surface current is in 7

the direction of the axis of the channel. We can therefore use the radial coponent current easureents to deterine the channel current. The channel current is deterined as follows. A radial current easureent for a gridpoint can be used to deterine the channel current at that point. Using the diagra shown in Figure 5., the channel current resolved in the direction of the radial can be written as V radial = V channel cos(θ) (5.) Vchannel Vradial θ Figure 5. Channel Current with Radial Coponent Now, taking the positive channel current to be in the direction of a flood tide (i.e. 36 ) we can evaluate θ as the difference between the radial current direction and the channel axis direction as θ = θ radial θ channel (5.2) Therefore, the agnitude of the channel current is given by V radial V channel = cos (θ radial θ channel ) (5.3) where a positive value for Vchannel is a current towards the lagoon (flood current). Solving for the channel current, Vchannel in equation 5.3 can becoe illconditioned as the difference between the radial angle and the channel direction approaches 9. To prevent this fro occurring, channel gridpoints that would 7

create a difference angle greater than 75 were reoved fro the grid as shown previously in Figure 5.8. This process is perfored for each radial current velocity produced by PMAP2DAT to create a vector ap of currents within the channel. Figure 5.2 shows the channel currents for the corresponding radials that were processed by PMAP2DAT in Figure 5.. 8 Current Plot: Sabbioni & Lido Stations, Separate!34! 6 4 Distance fro Sabbioni () 2 A!2!4!6 Sabbioni data (blue) Lido data (red) A - Moored ADCP Location c/s = 5 on axis scales 2 c/s!4!2!!8!6!4!2 Distance fro Sabbioni () Figure 5.2 Channel Currents derived fro Radial Measureents 5.2.5 Developent of User Interface A large quantity of data was produced fro the six week deployent. To efficiently process and analyse this data, batch processing for PMAP2DAT was used together with a graphical user interface (GUI) developed in MATLAB to perfor the above operations. 72

Once the entire data set had been batch processed using PMAP2DAT to produce radial current data files, the CurrentsGUI user interface was used to visualise and scroll through current vector aps fro any date or tie within the data series. A screenshot of CurrentsGUI shown below in Figure 5.3 shows the interface, and the controls used to view the current data. As will be described in Sections 5.4 and 5.5, this tool was further developed to process sequences of vector aps for producing tie-series of average current easureents within a 2 radius of the ADCP location and for creating a channel transect of surface currents. Figure 5.3 CurrentsGUI Graphical User Interface 73

5.3 Spatial Separation Correlation - Radar Spectra In order to verify the range resolution of the radar syste, an exercise was undertaken to observe the correlation between bea-fored radar spectra at a nuber of points within a region that spanned ultiple range cells. Fro Lido tie-series data, radar spectra were generated using bea-foring to each location of a 25 point grid covering 25 x 25. The spacing of the points on this grid is 5 in both the x and y direction, as shown in Figure 5.4. 5 Figure 5.4 25-Point Correlation Grid Locations (Iage Map courtesy of GoogleEarth and DigitalGlobe) A radar spectru was produced for each point on this grid by using beaforing techniques. The spectra were produced using a DC filter on the antenna tie-series signals to eliinate any strong Hz spectral peaks. 74

Using the spectra for each point in the grid, the auto-correlation between all points was calculated as k A ijk A ijk n (5.4) where k is the aplitude spectru index, i,j are the spatial location coordinates, and n is the nuber of spectru points in the coplete su (i.e. n = 248 * 25). Next, a correlation index was calculated using the spectru and coplex conjugate of the coparison spectru for each value of increasing spatial separation. In this 5 x 5 grid, the spatial separation increases fro 2 through to 4. For exaple, the calculation for a spatial separation of is the su k k A ijk A i+,j,k n A ijk A i,j+,k n, spatial step = (5.5) and for increasing spatial separation, for exaple spatial step of 2 k k A ijk A i+2,j,k n A ijk A i,j+2,k n, spatial step = 2 (5.6) The values obtained using these correlation functions for spatial step fro to 4 are plotted against spatial step nuber. The resulting plot of spatial separation correlation obtained for the 25 point grid spectru data for a chosen day and tie (DAY29, 25) is shown in Figure 5.5 below. Although very coarse, this plot shows that the spectral correlation for adjacent points diinishes with increasing distance of separation. It ust be noted too that no distinction is ade between the correlation in the direction of range or aziuth. 75

3.5 x 4 25 Point Spatial Separation DAY29,25 3 2.5 2 Correlation.5.5.5.5 2 2.5 3 3.5 4 Spatial Separation Figure 5.5 25 Point Spatial Separation Correlation A odified approach was undertaken to overcoe the liitations of using only 25 points in the coparison. It was also considered that high current shear ay exist between the gridpoints on the coarse 25 point grid (5 spacing) so a new grid was used with additional gridpoints placed on a spacing within the existing grid. Over the sae region of interest, bea-fored radar spectra were obtained for each of the 44 gridpoints on this fine scale grid. The spatial step correlations were also calculated and plotted separately for increasing separation in the direction of range, and for across the aziuth. Using the sae data as before, the correlation values for the 44 point grid are shown in Figure 5.6 below. 76

6 Spatial Separation Correlation! Range 9 Spatial Separation Correlation! Aziuth 55 85 5 8 45 Correlation 4 Correlation 75 35 7 3 65 25 2 2 4 6 8 2 4 6 8 2 Spatial Step 6 2 4 6 8 2 4 6 8 2 Spatial Step Figure 5.6 44 Point Spatial Separation Correlation We can see fro the first graph in Figure 5.6 that the spectral correlation in the direction of range for spatial steps through declines slightly. The correlation then decreases rapidly for spatial separation steps above. As the gridpoint spacing chosen was a tenth of the range resolution of the radar, this shows that the range resolution of the syste is indeed dictated by the bandwidth and is. Fro the second plot, we see that the correlation in the direction across the aziuth decreases with increasing spatial separation. 77

5.4 Averaging GridPoint Radar Measureents The surface current aps within the channel, as obtained in Section 5.2 can be further processed to provide secondary current data for regions within the channel. In this section, we create an average surface current easureent for a circular zone of the channel surrounding the deployed ADCP. This will allow the coparison of radar current easureents within this zone with those obtained by the ADCP. 5.4. ADCP Region Averaging Surface current data fro radar gridpoints within a 2 radius of the ADCP location are used to create an average surface current easureent for the ADCP region. The region of interest is shown in Figure 5.7 by the circular dotted line surrounding the ADCP location arked as point A. 8 Current Plot: Sabbioni & Lido Stations, Separate!34! 6 4 Distance fro Sabbioni () 2 A!2!4!6 Sabbioni data (blue) Lido data (red) A - Moored ADCP Location c/s = 5 on axis scales 2 c/s!4!2!!8!6!4!2 Distance fro Sabbioni () Figure 5.7 ADCP Region of Averaging 78

The ean and standard deviation are calculated for valid radar surface current easureents within the circular zone. Any data points having a surface current easureent differing in value ore than twice the standard deviation fro the ean are discarded. The reaining data points that are within twice the standard deviation of the ean are inverse-distance-weighted sued using the forula Z = n Z i i= d p i n i= d p i (5.7) where Zi is the radar surface current at each valid data point, di is the distance of each data point fro the origin (ADCP location), and p is the chosen weighting factor of 2. The variance for the points used for this inverse-distance-weighted ean is calculated :- [ ] n (Zi Z) 2 i= S 2 z = d p i n i= d p i (5.8) where Z is the inverse distance weighted ean, as calculated using Equation 5.7. The standard deviation is obtained directly fro the result obtained for the variance. As an exaple, the ean current within the ADCP zone for the current vector ap shown in Figure 5.7 above is calculated and then plotted in Figure 5.8. 79

Current within ADCP Zone! Sabbioni Data.5.5!.5 Current (/s), Positive = Flood Tide!!.5 Current within ADCP Zone! Lido Data.5.5!.5!!.5 Current (/s), Positive = Flood Tide Figure 5.8 Average Channel Current within the ADCP Zone The graphical user interface, CurrentsGUI uses the process shown above to calculate the inverse-distance-weighted ean for either a single current ap, or sequentially for a range of current aps as specified by start day/tie and end day/tie. The software processes all the current aps within the chosen date/ tie range and produces an output file consisting of a tie-series of ADCP region averages and corresponding standard deviations. A quality index which is the nuber of data points used to calculate the inverse-distance-weighted ean and standard deviation is also recorded. The six week tie-series obtained for the average channel current within the ADCP zone is shown in Figure 5.9. This plot shows the tidal level, ADCP zone current easured using Sabbioni Radar data and the Lido Radar data respectively. The ADCP zone average current signals are considerably noisy. The Sabbioni average current data is also noisier and has ore discontinuities than that obtained using the Lido data. This is because the Sabbioni station has insufficient range to cover the ADCP region properly, as shown in Section 5.. The ADCP zone currents obtained fro the radar current aps can now be copared with the ADCP easured currents for the bin closest to the surface (Bin 9) which is at a depth of. The scatter plot shown in Figure 5.2 shows 8

that the results obtained using the radar do not correlate well with the ADCP results, and that the radar is consistently easuring lower current speeds than the ADCP. 5.4.2 Noise Reduction An iterative process of binoial weighting and interpolation can be used on the ADCP zone tie-series to reduce the noise in the signal. For each value in the ADCP zone tie-series, a binoially weighted value can be calculated using the forula U i = u i 2 + 4u i + 6u i + 4u i+ + u i+2 6 where i is the index of the current tie-series value. Now, if a binoially weighted value differs by ore than.s - fro the original value, the original value is discarded and replaced by an interpolated value being the ean of the adjacent points in the tie-series. 8

.5 Tide Station! Lido 275 28 285 29 295 3 35 3 35 Day Nuber!.5 3 Radar (ADCP Region)! Sabbioni 275 28 285 29 295 3 35 3 35 Day Nuber Radar (ADCP Region)! Lido 275 28 285 29 295 3 35 3 35 Day Nuber Figure 5.9 ADCP Zone Average Current Tie-series Current speed in the direction of the channel (.s - ) 2! Tide Height () Velocity (.s - )!2 2!!2 Velocity (.s - ) 82

.5 RAW Radar Data vs ADCP Current 45.5 Radar Current (.s - )!.5! Gradient =.3737 Y!Intercept =!.687!.5!.5!!.5.5.5 ADCP Current (.s - ) Figure 5.2 Radar Current (ADCP Zone) vs ADCP Current (Bin 9) This binoial weighting and interpolation process is then repeated. This second tie, a point in the first pass interpolated series is discarded if it differs by ore than.3 s - fro the second binoially weighted value. Once points have been reoved and interpolated a second tie, the final output is a binoially weighted sooth curve. This process is suarised in the flow chart shown in Figure 5.2, with the resulting processed tie-series for the ADCP zone radar current data shown in Figure 5.22 on the following page. Create Binoial Series - # Create Binoial Series - #2 Original Series If > Interpolate If >.3 Interpolate Final Soothing Binoial Series Final Series Figure 5.2 Binoial Weighting and Interpolation Algorith 83

2.5.5!.5!!.5 ADCP Zone Tie!series, Raw and 2!Pass Interpolated and Soothed RAW current tie-series Processed current tie-series 275 28 285 29 295 3 35 3 35 Day Figure 5.22 ADCP Zone Average Current Tie-series - RAW and Processed Current /s 84

Coparing the interpolated and soothed ADCP zone radar currents with those easured by the ADCP reveals that although we have reduced the noise in the radar current signal, the radar is still consistently reporting lower current speeds than the ADCP. The variation between the two instruents also increases for ADCP current speeds greater than.5.s -..5 2 Pass Interp. Binoial Radar vs ADCP Current 45.5 Radar Current (.s - )!.5! Gradient =.3485 Y!Intercept =!.7835!.5!.5!!.5.5.5 ADCP Current (.s - ) Figure 5.23 Interpolated/Soothed Radar Current (ADCP Zone) vs ADCP Current (Bin 9) 85

5.5 Generating Three Diensional Transects In this section, we describe the processing of surface current easureents obtained fro both the Sabbioni and Lido stations to create a surface current velocity profile across a transect of the Lido inlet. This transect profile can be cobined with the ADCP current data to provide three diensional profiling of currents within the Lido channel. 5.5. Transect Averaging The ethod used for transect averaging is siilar to the ADCP region averaging presented in Section 5.4. Average surface current easureents along the transect are calculated using data fro points within a rectangular zone surrounding each transect point. The transect used aligns with the ADCP location in the iddle of the channel, as shown by the thick blue line crossing the channel in Figure 5.24. %## 9:22/-+;<3+=5,66)3-)>?)@35+,+)3-*A5/B,2,+/!C#"####!! &## "## ()*+,-./2345,66)3-)748 $## # '!$##!"##!&##!#.4D*EF#43-,G)**.,</*!!"##!!$##!!###!%##!&##!"##!$## # ()*+,-./2345,66)3-)748 Figure 5.24 Channel Transect - Averaging Region 86

Points spaced apart along the transect are used as the centre points for the average current calculation. For each of these points, a rectangle is fored which is bounded by parallel lines 2 either side of the transect line, together with lines in the direction of the channel, 7.7 either side of the transect point. This geoetry is shown below for the second point along the transect (Figure 5.25). -4, 4 4-37.5, -227.5 4.4 Figure 5.25 Transect Bounding Box Geoetry The algorith for generating the transect profile takes the vector ap points as shown in Figure 5.24. A subset of data points is created that are within 2 perpendicular distance to the transect line. This subset is then sequentially processed for each point along the transect to identify data points that are within 7.7 perpendicular distance to a vector in the direction of the channel that passes through the selected transect point. The deterination of perpendicular distance in each case is calculated using the vector ethod below. 87

For three points, P (x, y), P2 (x2, y2), P3 (x, y) as shown in Figure 5.26, the perpendicular distance, d can be calculated using a vector ethod. v P 2 (x 2, y 2 ) P (x, y ) r d P 3 (x, y ) Figure 5.26 Vector Method for deterining perpendicular distance If the line is specified by two points, P and P2, then a vector v perpendicular to the line is given by [ v = y 2 y (x 2 x ) ] (5.9) and the vector r is given by r = [ ] x x y y (5.2) Then the distance fro P3 to the line is given by projecting r onto v, giving d = ˆv r = (x 2 x )(y y ) (x x )(y 2 y ) (x2 x ) 2 + (y 2 y ) 2 (5.2) or d = det(p 2 P P P ) P 2 P (5.22) Once data points are identified within each rectangular section of the transect, the inverse-distance-weighted ean and standard deviation are then calculated using the inverse-distance-weighting forulae 5.7 and 5.8 fro Section 5.4. 88

Figure 5.27 shows an exaple transect plot for the vector ap shown in Figure 5.24. The first two plots show the transect processed using individual station data whereas the third plot shows the transect profile if current data fro both stations is cobined. Transect Currents! Sabbioni Data.5 Current /s!.5! 2 3 4 5 6 7 8 9 Transect data point # (spacing ) Transect Currents! Lido Data.5 Current /s!.5! 2 3 4 5 6 7 8 9 Transect data point # (spacing ) Transect Currents! Cobined Data.6.4.2 Current /s!.2!.4!.6 2 3 4 5 6 7 8 9 Transect data point # (spacing ) Figure 5.27 Transect Profile for Day 34_ (Transect data point # is 5 fro Lido Breakwater) 89

Siilar to the ADCP region averaging, the graphical user interface, CurrentsGUI is also used to ipleent the transect processing. It is used to produce a tie-series of transect easureents between chosen dates and ties. These tie-series transect values ay also be used together with the ADCP easureents and water depth profile to for a 3-Diensional transect view such as that shown in Figures 5.28 and 5.29. Day 292.467, hrs Day 292.5, 2 hrs Day 292.4444, 4 hrs Day 292.5278, 24 hrs Day 292.4722, 2 hrs Day 292.5556, 32 hrs.s - = 2 on x-axis scale Figure 5.28 3-Diensional Transect view, Day 292.467-292.5556 ADCP current vectors indicate the vertical profile at 4 across the channel. (Iage Texture Map Courtesy of Google Earth) 9

Day 292.5833, 4 hrs Day 292.6667, 6 hrs Day 292.6, 42 hrs Day 292.6944, 64 hrs Day 292.6389, 52 hrs Day 292.7222, 72 hrs.s - = 2 on x-axis scale Figure 5.29 3-Diensional Transect view, Day 292.5833-292.7222 ADCP current vectors indicate the vertical profile at 4 across the channel. (Iage Texture Map Courtesy of Google Earth) 9

The 3-Diensional transect view clearly shows the transect surface current profile, the ADCP current profile and the bathyetry of the transect. It ust be noted that the scale used for the vertical diension is different fro the horizontal diensions. The iage sequence shown in Figures 5.28 and 5.29 is during a strong ebb tide during spring tides. In this sequence, when strong currents are easured by the ADCP, the radar is clearly seen to under-report the current velocity. This further shows the effect seen in the scatter plots presented in Section 5.4 where the radar easureents fail to correlate well with the ADCP for higher current speeds. Using the sae technique as for the ADCP zone averaging in Section 5.4, each tie-series for individual transect points can be 2-pass interpolated and binoially soothed. In Figure 5.3, the interpolated and soothed average current for transect point 4 (3 fro Lido side) is copared with the ADCP current (Bin 9). This shows a siilar result to that obtained for the ADCP region averaging. The spatial region used for each of these radar averages is centred approxiately on the sae location though a different shape..5 Interpolated & Soothed Transect Point 4 vs ADCP Current (Bin 9) 45 (.s - ).5 Radar Current!.5! Gradient =.3564 Y!Intercept =!.82768!.5!.5!!.5.5.5 ADCP Current (.s - ) Figure 5.3 Transect Point 4 vs ADCP Current (Bin 9) 92

2 Interpolated & Binoially Soothed Transect Data.5.5!.5!!.5!2 3 32 33 34 35 36 37 38 39 Day ADCP Nuber Current Figure 5.3 Surface Current Transect Tie-series (all points on transect) Radar Current (.s - ) 93

There is still a significant level of noise in the transect current velocity signal (Figure 5.3), even after each point on the transect is interpolated and soothed. This, together with the constant under-reporting of current speeds fro the radar easureents led to the investigation of sources of noise in the radar signal. The next section describes soe of these noise artefacts and a different approach to obtaining surface current easureents fro the radar spectra. 94

5.6 Radar Analysis - Direction Finding with Phase Correlation Discriination Close inspection of the radar spectra reveals that spurious noise peaks appear randoly within the range spectra. At ties these peaks appear outside of the area we would expect to see Bragg peaks, but at other ties these peaks are in superposition with Bragg peaks. This noise can affect the derivation of current easureent fro the spectra, so a ethod of characterising the noise was developed and used to iprove the accuracy of current easureents. The range spectru fro each of the antennae is used to calculate the su of the power spectra for all the antennae. For the exaple given in Figure 5.32, we can identify two ain energy areas, as highlighted. One is approxiately between -2 to - Hz, the other around +3 Hz. Now, we would expect Bragg wave signals to be shifted by the underlying current velocity fro their +/-.2587 Hz positions and so, could reasonably expect that the -2 to - Hz energy is the Bragg wave signal. The signal situated at 3 Hz is therefore considered to be noise. It does not represent Bragg energy and its origin ay be due to outside interference or internal electrical noise within the radar syste. 8 Power Spectru! Su of 4 Antennas 6 Relative Power Density 4 2 8 6 4 2!4!3!2! 2 3 4 Doppler Frequency (Hz) Figure 5.32 Power Spectru, Su of 4 Antennae - Day 274, 925 Looking in ore detail at the -2 to - Hz energy band, we copare the phase of each antennaʼs signal with another. First, the phase values for antenna # are plotted against the phase values for antenna #2 for each point on the spectru in this frequency band, as shown in the first plot of Figure 5.33. 95

4 Antenna Phase VS Antenna 2 Phase!2.Hz to!.hz 4 Antenna 2 Phase VS Antenna 3 Phase!2.Hz to!.hz 3 3 Phase (rad) 2! Phase (rad) 2!!2!2!3!3!4!4!3!2! 2 3 4 Phase (rad)!4!4!3!2! 2 3 4 Phase (rad) 4 Antenna 3 Phase VS Antenna 4 Phase!2.Hz to!.hz 4 Antenna 4 Phase VS Antenna Phase!2.Hz to!.hz 3 3 Phase (rad) 2! Phase (rad) 2!!2!2!3!3!4!4!3!2! 2 3 4 Phase (rad)!4!4!3!2! 2 3 4 Phase (rad) Figure 5.33-2 to - Hz Region Phase Correlation Siilarly, the phase relationships between the other antennae are also plotted. The phase relationship plot for the 3Hz energy band is shown below in Figure 5.34. 3 Antenna Phase VS Antenna 2 Phase +2.7Hz to +3.2Hz 4 Antenna 2 Phase VS Antenna 3 Phase +2.7Hz to +3.2Hz 2 3 Phase (rad)! Phase (rad) 2!!2!2!3!3!4!4!3!2! 2 3 4 Phase (rad)!4!4!3!2! 2 3 Phase (rad) 4 Antenna 3 Phase VS Antenna 4 Phase +2.7Hz to +3.2Hz 4 Antenna 4 Phase VS Antenna Phase +2.7Hz to +3.2Hz 3 3 Phase (rad) 2!!2 Phase (rad) 2!!2!3!3!4!4!3!2! 2 3 4 Phase (rad)!4!4!3!2! 2 3 4 Phase (rad) Figure 5.34 3Hz Region Phase Correlation 96

For the 3 Hz noise peak, the phase of the signals fro each antenna is strongly, linearly correlated. This shows that the origin of these spurious signals is not fro external interference, as if this were the case they would originate fro a specific direction or directions. When direction finding techniques are used on these spurious signals, they appear to originate fro either +/- 45 degrees to the aziuth of the antenna array regardless of the orientation, because the phases are always linearly correlated. Since PMAP2DAT does not have the function to discern this phase correlation when selecting energy peaks, a separate analysis algorith was written in MATLAB to produce radial current easureents fro range spectra. The MATLAB algorith uses the phase correlation property of peaks to discard peaks that are due to spurious internal noise. PMAP2DAT is still used to batchprocess the range tie-series into range spectru.dat files. 5.6. MATLAB Direction Finding Algorith The first step undertaken by the MATLAB Direction Finding algorith, DFind is to iport a range spectru file containing the range spectru for each of the ranges to be analysed. As the radar isnʼt satisfactorily obtaining signals fro beyond range cells, the DFind algorith is liited to range cells to increase the processing speed. Processing is undertaken one range cell at a tie. The exaple shown will use the coplex spectra fro the 5th range cell obtained by the Lido station. Initially, the power spectru for each antenna is generated and sued together (Figure 5.35). 97

x 4 Power Spectru! Su of 4 Antennas 8 6 4 Relative Power Density 2 8 6 4 2!4!3!2! 2 3 4 Doppler Frequency (Hz) Figure 5.35 Power Spectru (Day 29, 445 Range 5) - Su of 4 Antennas The power spectru su for all antennae is then soothed, and the largest peak on the negative half of the spectru is identified. The width of the selected peak is deterined by the points at which the signal falls 6 db below the ain peak. x 4 Soothed Power Spectru! Su of 4 Antennas 6 5 Relative Power Density 4 3 2!4!3!2! 2 3 4 Doppler Frequency (Hz) Figure 5.36 Soothed Power Spectru (Range 5) - Selection of Positive Peak The selected peak is then copared with the noise level of the spectru to deterine if it satisfies a iniu signal-to-noise criterion. The Rayleigh noise figure for the spectru is deterined using a ethod developed by Heron and Heron (2). Using this ethod, the sued power spectru is rank ordered as per the first plot shown in Figure 5.37. 98

2 Spectral Aplitude 5 5..2.3.4.5.6.7.8.9 q 2 Spectral Aplitude 5 5.5.5 2 2.5 3 3.5 4 sqrt(!2ln(q)) Figure 5.37 Rank Ordered Spectru, and [p, 2 ln(q)] Using the relation for Rayleigh distributed rando noise, σ 2 = p 2 2 ln(q) (5.23) we ay use the relationship between p and q to deterine the Rayleigh noise figure as a least-squares fit through data points [p, 2 ln(q)]. As such, the second plot in Figure 5.37 shows a least-squares fit to deterine the slope, σ of the linear portion. For a chosen spectral peak to be considered having sufficient signal level, it ust be six ties the slope value of σ. Once the peak is selected and satisfies the signal-to-noise criterion, it is checked for the phase correlation effect shown above. The phase of the coplex spectru for antennas and 2 is sorted and plotted in ascending order. A least-squares fit is plotted for this coparison and if the standard deviation of the residuals is greater than the peak is rejected. This is the case for the data shown in Figure 5.38 so the initially selected peak is rejected. 99

(rad) Phase # 2 9 8 7 6 5 4 3 Residuals 2.5.5!.5!!.5!2 2 2 3 4 5 6 7!2.5 2 3 4 5 6 7 Phase #2 (rad) Phase #2 (rad) Figure 5.38 Phase Coparison Least-Squares Fit (Std. Deviation of Residuals > ) Once a peak is rejected, the next ost significant peak on that half of the spectru is selected, as shown in the case of Figure 5.39. x 4 4 Soothed Power Spectru! Su of 4 Antennas 3.5 Relative Power Density 3 2.5 2.5.5!4!3!2! 2 3 4 Doppler Frequency (Hz) Figure 5.39 Soothed Power Spectru (Range 5) - Selection of Next Peak This next peak is again copared with the noise figure, and exained for phase correlation. As shown in Figure 5.4, the phases are not well correlated between antennas and 2 for this second selected peak, so it is accepted as being a valid peak for the negative half of the spectru.

25 2 5 (rad) 5 Phase # Residuals!5 5!!5 2 3 4 5 6 7 (rad)!5 2 3 4 5 6 7 (rad) Phase #2 Phase #2 Figure 5.4 Phase Coparison Least-Squares Fit - Peak B (Standard Deviation of Residuals < ) This process is repeated for the positive half of the spectru. If valid peaks exist in both halves of the spectru, the strongest peak is selected to be used in the direction finding process to deterine radial current velocities. If there are no valid peaks in either half of the spectru, no current easureents are obtained for that range spectru. The peak that is selected for direction finding is defined by a start and stop frequency point. The coplex I and Q values corresponding to each frequency point of the antenna spectru are used to deterine the aziuth of signal origin for each frequency point. Using antenna array pattern ultiplication, the values for I and Q can be used to plot the agnitude of the signal for various steering angles, thus deterining the direction at which this is a axiu. x θ λ 2 Figure 5.4 Antenna Array Steering Geoetry

For the linear antenna array shown in Figure 5.4 with half-wavelength spacing of the eleents, the agnitude for phase angle ψ can be calculated as follows :- g = a e jφ + a 2 e jφ 2+ψ + a 3 e jφ 3+2ψ + a 4 e jφ 4+3ψ (5.24) where a, a 2, a 3, a 4 and φ, φ 2, φ 3, φ 4 are the agnitudes and phases coputed fro the I and Q values for each receiver at a particular spectru frequency point. As the phase angle, ψ is varied fro -9 to +9, this corresponds to a physical steering angle of cos θ = x λ 2 = ψ 2π λ 2 λ = ψ π (5.25) where the physical steering angle, θ is calculated as θ = cos ψ π (5.26) As the phase angle, ψ, is varied to ʻsteerʼ the antenna through all forward angles of aziuth, the agnitude, g can be plotted against steering angle, θ as shown in Figure 5.42. 8 Direction Finding Magnitude 7 6 Magnitude, g 5 4 3 2!!8!6!4!2 2 4 6 8 Steering Angle (deg) Figure 5.42 Direction Finding Magnitude vs Steering Angle In figure 5.42, the peak agnitude is deterined. If this peak doesnʼt fall within +/- 3 of the boresight direction, it is discarded. If the peak falls within +/- 3 2

of the boresight as shown above, it ust then satisfy a further hysteresis criterion in order to be accepted. The inia either side of the peak ust fall to below half the value of the peak. In the case of Figure 5.42, the left-hand iniu does not satisfy this, so this frequency point would be discarded and not used to derive a radial easureent. If a frequency point within the peak on the spectru satisfies all direction finding criteria, the Doppler shift radial current easureent is obtained for that frequency point and stored as a radial easureent. Once radial current easureents are obtained for every frequency point of a valid peak in the range spectru, these radial easureents are stored for later re-gridding. An exaple of the radial current easureents obtained for the spectru shown above for the 5th range on Day 29 at 445 h is presented below in Figure 5.43. 8 6 4 Distance fro Sabbioni () 2!2 ADCP A!4!6 Lido 2 c.s -!8!5!!5 Distance fro Sabbioni () Figure 5.43 Lido, Day 29 at 445 h - Range 5 Radial Measureents 3

Once every range spectru has been processed to obtain radial current easureents, these data are used to produce radial velocity easureents for gridpoints using inverse distance weighting techniques such as those shown in Sections 5.4 and 5.5. Radial easureents lying within 7 of each gridpoint are used to create an inverse distance weighted average radial current velocity for that gridpoint. Any velocity easureent that differs by ore than twice the standard deviation fro the ean is also discarded prior to a final ean being calculated for a gridpoint. The re-gridded radials obtained for the Lido acquisition on Day 29, 445 h are shown below in Figure 5.44. Once the radial easureents are obtained for the pre-deterined gridpoints, these data are output in the sae file forat as that produced by the PMAP2DAT analysis software. This allows all the existing secondary analysis software developed in MATLAB to be able to process the current data obtained using MATLAB DFind. 8 Radials Plot: Sabbioni & Lido Stations, Separate!2944! 6 4 Distance fro Sabbioni () 2 ADCP A!2!4 2 c.s -!6 Lido c/s = 5 on axis scales!4!2!!8!6!4!2 Distance fro Sabbioni () Figure 5.44 Lido, Day 29 at 445 h - Radial Measureents 4

ʻCurrentsGUIʼ was used to process the radial current easureents obtained using the DFind algorith for the entire deployent period. The tie-series for the ADCP Region average current obtained is shown below in Figure 5.45. Copared with the tie-series obtained using the PMAP2DAT algorith (Figure 5.9), the DFind algorith has resulted in higher peak current readings. The structure of the tie-series for the Lido station is ore representative of the tidal flow with spring and neap tides clearly present. The average current signal obtained for the ADCP region using the Sabbioni data has ore noise though. Siilarly, the coparison between 2-pass interpolated and soothed ADCP region average with the ADCP current easureents (Bin 9) shows that the use of the MATLAB DFind algorith has iproved the correlation between the two instruents, though not substantially (Figure 5.46). 5

.5!.5 2!!2 2!!2 2!!2 Tide Station! Lido 275 28 285 29 295 3 35 3 35 Day Nuber Radar (ADCP Region)! Sabbioni 275 28 285 29 295 3 35 3 35 Day Nuber Radar (ADCP Region)! Lido 275 28 285 29 295 3 35 3 35 Day Nuber Radar (ADCP Region)! Cobined Green - Average current Blue - Interpolated and Soothed average current 275 28 285 29 295 3 35 3 35 Day Nuber Figure 5.45 MATLAB DFind, ADCP Zone Average Current Tie-Series Velocity (.s - ) Velocity (.s - ) Velocity (.s - ) Height () 6

.5 2 Pass Interp. Binoial Radar vs ADCP Current 45.5 (.s - ) Radar Current!.5! Gradient =.42655 Y!Intercept =!.48837!.5!.5!!.5.5.5 ADCP Current (.s - ) Figure 5.46 MATLAB DFind Interpolated/Soothed Radar Current (ADCP Zone) vs ADCP Current (Bin 9) The MATLAB DFind algorith was developed to iprove the accuracy of the radar easureents by characterising noisy peaks in the spectra and discarding these where necessary. The DFind algorith has produced an iproved tie-series record and greater correlation with the ADCP easureents. The surface current velocity tie-series for individual transect points is too noisy to be used for a definitive ass transport calculation. 7

5.7 FFT Analysis and Accuulated Transect Profiles Although the radar surface current easureents obtained using MATLAB DFind are still too noisy to be used for calculating ass transport, alternative ethods of analysing the ADCP region and transect surface currents were investigated. 5.7. FFT Analysis This ethod analyses the transect tie-series data in the frequency doain to differentiate between low-frequency tidal driven current signals and noise. For each day of the deployent period, a Fast Fourier Transfor is created for both the ADCP signal and the ADCP region radar signal. An exaple of this for day 274 is shown below in Figure 5.47..5 Radar and ADCP Tie!series 35 FFT! Radar and ADCP Data 3 25.5 Current /s Magnitude 2 5!.5! 5!.5 274 274. 274.2 274.3 274.4 274.5 274.6 274.7 274.8 274.9 275 DAY 2 4 6 8 2 4 6 8 Frequency (Cycles/Day) Figure 5.47 FFT Analysis - Acceptable Radar Signal ( Blue - Radar, Green - ADCP) Good quality current signals fro either radar or ADCP are seen visually as sooth low-frequency sinusoidal curves, such as that shown in the first plot of Figure 5.47. These good quality current easureent signals are characterised as having strong spectral peaks at 2 cycles/day, representing the sei-diurnal tidal signal. In this analysis, good signals were identified as having greater than 35% of the area under the spectral curve within the region between and 3 cycles/day. An exaple of a poor radar signal copared with the ADCP signal is shown below for Day 279 in Figure 5.48. 8

.5 Radar and ADCP Tie!series 35 FFT! Radar and ADCP Data 3.5 25 Current /s Magnitude 2 5!.5! 5!.5 279 279. 279.2 279.3 279.4 279.5 279.6 279.7 279.8 279.9 28 DAY 2 4 6 8 2 4 6 8 Frequency (Cycles/Day) Figure 5.48 FFT Analysis - Poor Radar Signal ( Blue - Radar, Green - ADCP) In this case, the ADCP region radar signal is significantly different fro the ADCP signal. The spectral analysis shows that uch of the energy is centred around cycles/day. Data obtained on days such as this are discarded, as the spectral signal does not lie between and 3 cycles/day. The reaining good quality data are used to produce a coparison of the ADCP and the ADCP region radar data, as shown in Figure 5.49 below..5 FFT Analysed Binoial Radar current vs ADCP current FFT Analysed Binoial Radar vs ADCP Current.5 Radar Current (.s - )!.5! Gradient =.4672 Y!Intercept =!.58896!.5!.5!!.5.5.5 ADCP Current (.s - ) Figure 5.49 FFT Analysed Data vs ADCP Current 9

Copared with Figure 5.46, this plot shows a reduced nuber of outliers. There is still a significant disagreeent between the readings obtained fro each instruent though. Even when only using good quality radar data, the radar is still reporting lower current easureents than the fixed ADCP in the channel. 5.7.2 Accuulated Transect Profiles Profiles of the surface current along the channel transect were created for various bins of flood and ebb current velocity, and the state of tide. The interpolated and soothed transect profile easureents obtained in Section 5.6 were sorted into bins using the fixed ADCP velocity easureents and tide gauge data as references. The ADCP velocity ranges used for the sorting bins are listed in Table 5.. Ebb Tide Currents -. to +. /s Currents -. to -.25 /s Currents -.25 to -.5 /s Currents -.5 to -.75 /s Currents -.75 to -. /s Currents -. to -.25 /s Flood Tide Currents -. to +. /s Currents +. to +.25 /s Currents +.25 to +.5 /s Currents +.5 to +.75 /s Currents +.75 to +. /s Currents +. to +.25 /s Table 5. Transect Profile Sorting Bins A two pass iteration process was used to reove outliers beyond twice the standard deviation fro the ean. The reaining transect point easureents within each bin were plotted, together with the ean surface current for each point across the transect, as shown in Figure 5.5.

Currents!. to +.! Ebb Tide, Accuulated Transect Profile Currents!. to +.! Flood Tide, Accuulated Transect Profile /s /s!! 2 4 6 8 Currents!. to!.25! Ebb Tide, Accuulated Transect Profile 2 4 6 8 Currents +. to +.25! Flood Tide, Accuulated Transect Profile /s /s!! 2 4 6 8 Currents!.25 to!.5! Ebb Tide, Accuulated Transect Profile 2 4 6 8 Currents +.25 to +.5! Flood Tide, Accuulated Transect Profile /s /s!! 2 4 6 8 Currents!.5 to!.75! Ebb Tide, Accuulated Transect Profile 2 4 6 8 Currents +.5 to +.75! Flood Tide, Accuulated Transect Profile /s /s!! 2 4 6 8 Currents!.75 to!! Ebb Tide, Accuulated Transect Profile 2 4 6 8 Currents +.75 to +! Flood Tide, Accuulated Transect Profile /s /s!! 2 4 6 8 Currents! to!.25! Ebb Tide, Accuulated Transect Profile 2 4 6 8 Currents + to +.25! Flood Tide, Accuulated Transect Profile /s /s!! 2 4 6 8 Transect Point Nuber ( spacing) 2 4 6 8 Transect Point Nuber ( spacing) Figure 5.5 Accuulated Transect Profiles The plotted results are consistent with all the previous coparisons that have been ade between the radar and the ADCP results, with the radar consistently easuring lower current velocities. As the ADCP easured velocity increases, the variability of the radar easureents also increases. In the next section, we discover why there are discrepancies between the radar current velocities and those obtained by the ADCP.

5.8 Modified Range Resolution, Manual Direction Finding Analysis After all the previous analyses had been undertaken with the understanding that the PortMap radar had a range resolution of, the PortMap anufacturer, Helzel Messtechnik reported a fault with the prograing of the real-tie coputers within PortMap. The prograing fault was in one of the range-resolving Fast Fourier Transfor operations perfored by the CL7, whereby the incorrect half of the spectru was being used. Therefore, the desired spectru signal produced by the 52.2 MHz signal is discarded. By chance, the signals and range resolved data actually being used were in fact produced by a 5th haronic (45.42 MHz) of the priary oscillator frequency (29.84 MHz). This resulted in a change of bandwidth by a factor of 5, which in turn caused the range resolution to be 2 instead of. The power level of the 5th haronic signal was also significantly less than the 52.2 MHz signal resulting in a lower signal-to-noise ratio and further contributing to a reduction in the range of the syste. One contributing factor to the low transit power was that the 45.42 MHz frequency was well out in the edge of the transitter and antenna filters. One interesting consequence of this was that the bandwidth occupied by this very low signal was very wide, aking the 2 resolution the true spatial resolution of the syste. To copensate for the fault present in the PortMap syste, the MATLAB DFind algorith was odified to use a range resolution of 2 and a radar frequency of 45.42 MHz. Using the expected axiu range cell nuber fro Section 5. with this new range resolution of 2, we can see in Figure 5.5 the effect that this has on the achieved coverage of the channel by the radar syste. With the change in range resolution, a new high-resolution easureent grid is required for re-gridding the radial easureents. A grid with 2 spacing in both the x and y directions was created with the Sabbioni station again used as the origin. 2

28 T 22 T Sabbioni 6 T T 336 T Lido 36 T 96 T 56 T Figure 5.5 Modified Range Resolution (2 ) Estiated Maxiu Radar Range (Iage Map Courtesy of Cnes/Spot, DigitalGlobe, TerraMetrics and GoogleEarth) Using a new radar frequency of 45.42 MHz, Bragg wave spectral peaks for zero surface current will now be centred at +/-.233 Hz in the radar spectra. Although highly tie-consuing, the direction finding algorith was odified to require anual intervention to confir each Bragg peak selected for obtaining current easureents. This was done to attain the best possible current easureent results fro a select nuber of days within the six week deployent. Any peaks that visually appear too noisy were discarded by operator intervention. Data acquired by both the Sabbioni and Lido stations were analysed for a single direction looking up the channel for Days 29, 292, 293, 294 and Days 32, 33, 34, 35, 36. These days were selected because they were during spring tides and were identified as having better signal-to-noise ratio for the signals acquired. An exaple short-range radial current ap fro data acquired on DAY 29 at 232 h is shown in Figure 5.52. Note the individual current aps for each station, as a result of higher resolution and shorter range. 3