A lithological map created from multibeam backscatter data in challenging circumstances: the Lower Sea Scheldt estuary

Transcription

1 A lithological map created from multibeam backscatter data in challenging circumstances: the Lower Sea Scheldt estuary Mieke Mathys*, Marc Sas*, Frederik Roose** HYDRO12, Rotterdam, 15/11/2012 *International Marine and Dredging Consultants (IMDC), Coveliersstraat 15, B-2600 Berchem, Belgium T: **Frederik ROOSE, Belgium Flemish Ministry of Mobility and Public Works, Maritime Access Division, Tavernierkaai 3, B2000 Antwerp, Belgium

2 Introduction: goal and setting NL/B Western Scheldt Lower Sea Scheldt Sea Scheldt Rupel Upper Sea Scheldt Gent 13-Mar-13 / Hydro 12 / slide 2

3 Outline presentation General methodology Challenging circumstances How we created a lithological map Recommendations Conclusion 13-Mar-13 / Hydro 12 / slide 3

4 Methods Creation of a lithological map QTC MULTIVIEW TM Sediment classication based on amplitudes and statistical properties of backscatter images Geometrical image compensation 13-Mar-13 / Hydro 12 / slide 4

5 Methods Creation of a lithological map Automated clustering method Each cluster or acoustic class represents bottom type Supervised (ground-truthed catalogue) versus unsupervised classification (ground-truthing of acoustic classes) 13-Mar-13 / Hydro 12 / slide 5

6 Methods Multibeam campaigns 5 section maps 2 different ships (Flemish Hydrography) Veremans: largest area, deep channels Parel 2: shallow tidal flat areas Multibeam sonars of type Simrad EM 3002 dual Frequency 300 khz 10 Hz pulse rate 149 µs pulse length May-September Mar-13 / Hydro 12 / slide 6

7 Methods Van Veen Shipek Sampling campaigns Controle samples: Shortly after multbeam recordings 50 spread over each section map Calibration samples: in 6 reference zones to create catalogue for supervised classification (sand, sandy mud, muddy sand, mud, hard soil due to stiff clay, due to gravel) Taking into account tide (flood, ebb, slack time) Validation samples: After creation lithological map 13-Mar-13 / Hydro 12 / slide 7 Reineck box corer

8 Challenging circumstances: tide-dominated estuary Slack water and influence of sediment settlement Absorption and refraction depending on salinity and water depth Recordings with two vessels 13-Mar-13 / Hydro 12 / slide 8

9 Challenging circumstances Slack water and influence of sediment settlement Possible creation of concentrated mud layer Backscatter return from settling mud particles (or fluid mud) instead of underlying bottom? Comparing acoustic maps of one track during different tidal phases small variations HW 2.5h after HW 13-Mar-13 / Hydro 12 / slide 9

10 Challenging circumstances Absorption and refraction due to salt water and increasing water depth CTD measurements are not necessary in QTC MULTIVIEW software (patented compensation tables) Normalized images over the entire survey Backscatter not in db but in relative image amplitude (0-255) Geometric image compensation for water depth (and relief) still, test using 2 exterme salinity values (sound velocities) few differences in positions of footprint of outer beams 13-Mar-13 / Hydro 12 / slide 10

11 Challenging circumstances Recordings with two vessels test: same survey lines surveyed by Parel and Veremans in the same direction three times (in 8 minutes) test: reproducibility of a single transducer comparison of classification maps (difference maps) differences between sequential measurements over a single line with one transducer = same order as differences between transducers of two different ships 1.5% difference in homogeneous area, 16% difference in heterogeneous area 13-Mar-13 / Hydro 12 / slide 11

12 Creation of a lithological map Supervised classification Catalogue based on 6 reference zones Mud sand Sandy mud Muddy sand Hard bottom by gravel Hard bottom by stiff clay Best calibration catalogue (tested number of classes and comparison with samples) applied to enitre Lower Sea Scheldt MB data set 13-Mar-13 / Hydro 12 / slide 12

13 Creation of a lithological map: supervised Best calibration catalogue (15 classes): 3 classes with very small extent 4 classes contained no samples (unknown) Remaining 8 classes grouped in 3 lithological classes (based on samples): mud Hard bottom Sand + muddy sand+ sandy mud 13-Mar-13 / Hydro 12 / slide 13

14 Creation of a lithological map: unsupervised Used entire data set to create catalogue Optimum number of classes: 14 Apply catalogue to entire Lower Sea Scheldt data set 3 classes contained no samples (unknown) Remaining 11 classes grouped in 5 lithological classes (based on samples): sand+sandy mud+muddy sand 1 sand+sandy mud+muddy sand 2 sand Hard bottom Mud 70% of 460 samples lies in the corresponding lithological class 13-Mar-13 / Hydro 12 / slide 14

15 Final lithological map 13-Mar-13 / Hydro 12 / slide 15

16 Recommendations Frequency multibeam (300 khz) High resolution, but very shallow penetration Only water-bottom or water-top mud interface charcteristics (roughness, grain size), not underlying sediment volume Reason why mud and hard soil can belong to same acoustic class (similar interface characteristics: smoothness, specular surface) Lower frequency multibeam (30 khz) most likely better distinguishes between grain sizes (muddy sand, sandy mud), as it captures top sediment volume 13-Mar-13 / Hydro 12 / slide 16

17 Recommendations Sampling strategy Detailed description on board is crucial. E.g. fine mud layer at the top strongly influences acoustic signature, but will not appear from lab analysis; presence of gravel as continuous layer or scattered gives different acoustic signature, but similar lab results Supervised classification (predefined lithological classes): more intensive calibration sampling campaign is required To reduce number of outlyers in certain lithological class Unsupervised classification enables to examine purposively (across acoustic border) why a certain area belongs to a specific acoustic class, Sampling along transects across acoustic boundaries taking into account (micro-)structural elements (at interface) and not only grain size 13-Mar-13 / Hydro 12 / slide 17

18 Conclusion Study area = tide-dominated estuary Water level, currents, salinity, sediment concentration vary over tidal cycle Tests showed that moment in tide of MB recordings was not significant Variations are correctly compensated by software Preference is given to unsupervised classification In a heterogeneous environments as the Lower Sea Scheldt the entire backscatter data set should be used to create acoustic classification maps, not only applying a catalogue from a restricted calibration area (supervised classification) Unsupervised classification should be combined with a structured and detailed sampling campaign, describing more than grain size alone In case of high frequency MB, ground-thruth of acoustic classes in terms of characteristics of water-bottom interface (texture, structure) In case of lower frequency MB, characteristics of sediment volume (grain size) becomes more important 13-Mar-13 / Hydro 12 / slide 18

19 Acknowledgements Flemish Hydrography for multibeam recordings and making available survey vessels and crew Hydrology and Hydraulic engineering department of VUB for sample analyses Flanders Hydraulics Research for sample analyses 13-Mar-13 / Hydro 12 / slide 19