Source gather Receiver gather Strongest amplitude pick First arrival Traveltime picking The travel times were estimated using state-of-the art 3D seismic software (Opendtect, dGB Earth Sciences) and techniques. The data cube has been analysed to produce a cube containing the dip of the transmitted wave (via FFT). The data cube is then filtered for incoher- ent dips (dip-steering filter). Based on this new cube, the dip-cube and the coherence calculation, the strongest amplitude is picked thoroughout the dataset. The picking is the result of seed propaga- tion and cross-check (in white in Fig. 6). The result is a ‘3D’ horizon which corresponds to the maximum amplitude of the first arrival. The horizon is shifted by a constant time in order to match the ob- served first zero-crossing (Fig. 6). The resulting ‘travel time horizon’ constitutes the data used for the tomography (Fig. 7A). Figure 6. Crosshole dataset prepared for traveltime picking. 1:100 ^N ?? 5 m I II III IV 50 km N < 500 m 500–1000 m 1000–1500 m 1500–2000 m > 2000 m Thickness of chalk Ringkøbing – Danish Basin Sorgenfrei–Tornquist Zone Outer limit Fault Salt dome Late Cretaceous inversion Basement high Fyn High Stevns A Jylland Sigerslev Højerup St. Heddinge Rødvig Mande- hoved Sea cliff Village Road 1 km B Boesdal Quarry Copenhagen 5.28m (collapsed) 1 2 3 4 Figure 1. Location of the study site, modified from Stemmerik et al. (2006). A. Thickness and structural map of the Upper Cretaceous – Danian chalk in the Danish area. B. Map of the study area with the experiment site (Stevns Klint, DK). Figure 2. Field layout A. Field layout facing North. The data were collected between boreholes 3 and 4. B. Map of the field layout. A. B. Uncorrected Corrected Source gather Receiver gather Timeslice -130 130 50 100 0 -50 -100 5 m 5 m 100 ns 100 ns 5 m > 0.106 0.098 0.092 0.087 0.082 0.079 0.075 0.072 0.070 0.068 0.065 0.063 0.061 0.060 0.058 0.057 < 0.056 Discussion and Conclusions Based on the newly acquired data a high resolution cross-hole tomographic inversion has been performed between two wells (Fig. 7A). The inversion illus- trates nicely the different properties of the studied chalk succession. Three main units can be distinguished in the profile: high-velocity upper unit , lay- ered variable middle unit and low velocity unit (Fig. 7A). Comparison between core plug porosities and velocity repartitions shows a direct relationship in such water-saturated environment (Fig. 7B). The porosities are also the reflec- tion of various depositional units with the top, middle and lower units being the Stevns Klint Fm (tight bryozoan chalk), the Højerup Mb. (Bryozoan chalk with flint) and the Sigerslev Mb. (white chalk). These results show many simi- larities with the study of Nielsen et al. (2010). However, the picking time is here of c. 1h wheras the ‘traditional’ method can take up to a week (without QC). The quality control and traveltime picking method developed for this study greatly improved the efficiency and consistency in GPR cross-hole travel time picking. It is believed that using such a method is critical when doing time -lapse studies to reduce potential errors due to discrepencies between experiments and interpretations. Figure 7. Relation between velocities, lithology and porosity. A. Velocity model calculated from the traveltimes. B. Porosities from core plug samples (Nielsen et al., 2010) Distance (m) 0 1 2 3 4 5 m/ns A. Depth (m) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 2 4 6 8 10 12 14 35 40 45 50 Porosity (%) B. Data and Methods The crosshole GPR data are acquired using 100 Mhz antennae by Sensors & Software with boreholes c. 5m apart (Fig. 2): Each transmitter gather (Fig.3A) contains 57 traces and the whole dataset consists of 44 transmitter gathers themselves measured every 25 cm (Fig. 3B). A standard processing flow has been applied including: 1) time-zero calibration; 2) bandpass filtering; 3) time cut. Transmitter gathers are arranged to form a cube (Fig. 4.) Quality control is performed on the data cube to correct for acquisition and processing errors. Errors are quickly identified by looking at the receiver gather or time slice from the datacube (Fig. 4 & 5). Most common error made during data acquisition is duplicated trace. In Fig. 5 ‘a kink’ in the first arrival can be seen where the receiver position is not correct. This can also be seen as a shift in amplitude on the timeslice. Looking at the transmitter gather only does not display acquisi- tion problems of this type. Unbalanced amplitudes produced by erroneous processing are also evident (Figs. 4 & 5). Figure 3. Data coverage and apparent velocity along straight ray- path. A. One transmitter gather. B. Full data coverage. A. B. Figure 4. 3D representation of crosshole data. abscissa repre- senting the receiver positions, ordinate the source positions and vertical axis the time S o u r c e g a t h e r Receiver gather Timeslice Efficient data analysis and travel time picking methods for crosshole GPR experiments Johanna Keskinen *, Julien Moreau *, Lars Nielsen *, Lars Stemmerik **, Kresten Anderskouv *, Klaus Holliger ***, and Majken C. Looms * * Department of Geosciences and Natural Resource Management, University of Copenhagen, Denmark ** Natural History Museum of Denmark, University of Copenhagen, Copenhagen, Denmark *** Institute of Geophysics, University of Lausanne, Lausanne, Switzerland F A C U L T Y O F S C I E N C E U N I V E R S I T Y O F C O P E N H A G E N Introduction The Chalk Group forms an important reservoir for hydrocarbons and groundwater, and chalk has been studied extensively with geological and geophysical methods. High-resolution time-lapse GPR crosshole experiments are conducted to resolve fine-scale anisotropy and flow characteristics of different chalk lithologies. Time- lapse studies require processing and intepretation of large datasets in an efficient and consistent manner. To this end we have developed corresponding methods for efficient data analysis and tested it on data collected from a well-studied and known Chalk Group locality . REFERENCES Nielsen L., Looms M. C., Hansen T. M., Cordua K.S., Stemmerik L., 2010, Estimation of Chalk Hterogeneity from Stochastic Modeling Conditioned by Crosshole GPR traveltimes and Log Data in Advances in Near-Surface Seismology and Ground-penetrating radar, Geophysical Development Series, eds. Miller R. D., Bradford J. H. and Holliger K., SEG Stemmerik L., Surlyk F., Klitten K., Rasmussen, S. L., and Schovsbo N., 2006, Shallow core drilling of the Upper Creatceous Chalk at Stevns Klint, Denmark: Geological Survey of Denmark and Greenland Bulletin, 10, pp.13-16 Surlyk F., Damholt T. and Bjerager M., 2006, Stevns Klint, Denmark: Uppermost Maastrichtian chalk, Cretaceous-Tertiary boundary. and lower Danian Bryozoan mound complex: Bulletin of the Geological Society of Denmark, 54, pp. 1-48 Geological setting The experiment site is located in a former chalk quarry south of Copenhagen (Fig. 1). The geology of the area is well known from over 100 years of geological studies. The studied interval includes from base to top: the Sigerslev Member (white chalk), the Højerup Member (Bryozoan chalk, wackestones), the Fiskeler Member (K-T boundary clays) and the Stevns Klint formation (bryozoan mounds in chalk; Surlyk et al., 2006). The boreholes used for the study (Fig. 2) were cored and analysed by Nielsen et al. (2010). The top seven metres correspond to flint-rich Danian and Maastrictian chalk overlying soft bioturbated Maastrictian chalk. At the time of the experiment the groundwater table was between 2.1 and 2.8 metres below the surface. Figure 5. 2D slices extracted from the 3D data cube. Data has been corrected for dupli- cate traces and processing errors. Acknowledgments This study was financed by Maersk Oil. Opendtect was made available by dGB Earth Sciences. Python distribution was made available by Enthought. 100 ns 100 ns