The “conversion log” is a new approach to quantify seismic scattering in situ in terms of PS conversion in transmission along a vertical seismic profile (VSP): The amount of converted seismic energy is determined by slant-stacking and plotted as a function of depth, thus forming a borehole log of seismic conversion. We investigated seismic scattering of crystalline crust at the Continental Deep Drilling Site (KTB) in southern Germany where detailed knowledge exists of crustal parameters down to 9 km depth. In 1999 a deep VSP was acquired in the KTB main borehole. The experiment yielded high quality seismic data in terms of signal bandwidth, signal-to-noise ratio and stability of the source signal. The seismic data show varying levels of PS conversion along the borehole. The dip of layering and foliation is about 45° to 75° along the KTB drill hole. Under these conditions the conversion amplitudes depend only weakly on the angle between the incident seismic wave and the impedance contrast surface. The conversion log method was used to quantify energy loss by forward scattering. Field data were compared with finite-difference computations and with petrological and structural borehole information. It turned out that only 10–50% of PS forward scattering originates from conversion at lithological interfaces and structural complexity whereas 90–50% is due to velocity heterogeneity caused by fractures. The conversion log is correlated with the depth function of fracture density, and it is inversely correlated with the depth function of chlorite content, that seems to ‘heal’ the influence of cracks and fissures.
 Seismic scattering plays an important role in seismic energy propagation. The quantification of its specific portions can help to understand seismic wave attenuation and its relation to geological structure. Scattering is often analyzed in statistical terms because particular constituents, such as reflected and converted waves, cannot be separated easily in surface seismic measurements. An experimental setting that is useful for the quantification is the vertical seismic profile (VSP) because it collects the signals in one-way transmission with a higher signal-to-noise ratio than in a surface seismic setting.
 A wave type that can easily be identified in VSP data and that can be used as a scattering indicator is the PS converted wave in transmission (Figures 1b, 2b, and 3b). Conversion of waves takes place at seismic impedance changes within the crust. These contrasts can be caused, e.g., by lithological changes, complex tectonic structure, or by fractures. In this paper we focus on quantifying the conversion of compressional to shear wave energy.
 We investigated PS-conversion along the main hole of the Continental Deep Drilling Project (KTB, S Germany). The data base of our study is a new deep VSP ranging from 3.0 to 8.5 km depth [Rabbel et al., 2004]. Seismic waves were generated with dynamite charges of 0.5 to 1 kg providing stable source signals and a high signal-to-noise ratio between 15 and 240 Hz. Signals were monitored by a stationary reference geophone in 3728 m depth in the KTB pilot hole located 200 m west of the main borehole.
 The 9.1 km deep main borehole of the KTB was drilled into steeply dipping alternating stacks of biotite-gneiss and amphibolite layers (Figure 1a). The KTB geological section represents an example of strongly faulted crystalline crust deformed under compression [Emmermann and Lauterjung, 1997] (Figure 1a). The large number of borehole parameters measured during the KTB project provide a good basis for comparative studies. For this study we made use of detailed petrological and structural depth functions derived from x-ray diffractometry of drill cuttings and FMS-logging [Duyster et al., 1995; Hirschmann and Lapp, 1994].
2. Quantification of Forward Scattering
 Forward scattering can be quantified in terms of PS conversion observed in the KTB VSP data. Transmitted PS converted waves can be identified in horizontal component seismograms of a VSP as arrivals preceding the direct S wave traveling with nearly the same apparent velocity. They ‘grow out’ of the P wave first break recorded by the vertical component. The intersection of P and PS traveltime branches indicates the conversion depth. Test computations using the Zoeppritz equations showed that the transmission coefficient of PS converted waves depends only weakly on interface dip for angles between 45° and 70° such as found at the KTB [Beilecke, 2003]. Therefore, the ratio between shear and compressional wave amplitudes at the conversion depth corresponds to the local impedance contrast.
 We observed that the amplitudes of a converted phase show considerable fluctuation along its traveltime branch, probably caused by small-scale geological heterogeneity. Therefore, amplitude analysis cannot rely only on picking single amplitude values from the traces. Instead we applied a slant stacking process to derive statistically more reliable conversion amplitudes. This procedure consists of a horizontal rotation of the geophone coordinate system according to the system of anisotropy, dip filtering to remove remnants of downward traveling P waves, a move-out correction, a correction of amplitudes with respect to the P wave first break to account for spherical divergence, a residual static correction, and the stacking of 15 trace envelopes, derived from Hilbert-Transform, at each depth position [Beilecke, 2003].
 The result of these computations, termed “conversion log”, is a borehole log showing the absolute value of conversion amplitude with depth. The conversion log determined from the field data (Figure 1c) was compared with counterparts based on synthetic data computed for lithological models of the KTB site (Figures 2 and 3).
2.1. PS Forward Scattering of Field Data
 A model of the crustal structure at the KTB is shown in Figure 1a [Hirschmann, 1996]. It is based on seismic experiments and borehole data [e.g., Harjes et al., 1997; Hirschmann, 1996]. The seismogram of the NE component of the VSP showing the slower split shear wave and the resulting conversion log are shown in Figures 1b and 1c, respectively. They show that PS conversion occurs almost everywhere along the borehole. Converted wave amplitudes amount up to 20% of the incident P wave. However, strong conversions are limited to the upper part of the log (<6 km depth).
2.2. Determination of the Lithological Share
 To determine the lithologically induced share of the conversion we simulated wave propagation for two models with a finite difference (FD) modelling code of Bohlen . The models differ in geological complexity. The elastic parameters of both models represent isotropic rocks without any cracks. They were derived from laboratory measurements of KTB rock samples [Kern et al., 1991], cuttings analysis and inversion computations [Rabbel et al., 2001]. The first model (Figure 2a) is one-dimensional. It represents the lithological sequence found along the borehole with 12.5 m resolution. The layer dip of 60° corresponds to the average dip angles in situ. Using Zoeppritz equations it can be shown that PS conversion is maximum at this angle with little variation between 45° and 75° dip [Beilecke, 2003]. Therefore, the results of this model are a maximum estimate of lithologically induced forward scattering.
 The horizontal component seismic section resulting from this first model shows significant conversion; however the over all appearance differs from the complicated pattern of the field seismograms (Figure 2b). The amplitudes of the resulting conversion log (Figure 2c) are 40–80% smaller than the field data values.
 In order to determine the influence of 2D heterogeneity the first model was modified to include a varying dip along the borehole and layer offset along fault zones (Figure 3a). The model design reproduces the geometrical features of the geology at the KTB in a qualitative way (Figure 1a). Faults were assumed to be infinitely thin with no specific elastic parameters attributed. Therefore, the seismic response only depends on the model layering and tectonic complexity. The resulting seismogram (Figure 3b) shows more similarity with the field data, confirming that the second model is structurally adequate. However, the amplitude level of the PS conversion log (Figure 3c) is as low or even lower as in the 1D model. It confirms that the conversion log of the one-dimensional 60° dip model represents the maximum level of conversion that can be expected for an assumed crack-free metamorphic rock sequence of complex geometry.
3.1. Comparison of Field Data vs. Synthetic Data
 Since geophones are velocity sensors, the S/P amplitude ratio of the conversion log represents the square root of the ratio of the local energy densities of scattered PS- to incident P-waves. Besides intrinsic dissipation, PS scattering is one more component contributing to the absorption of the P-wave. A corresponding attenuation factor 1/Q could be expressed for spectral input data by the squared spectral conversion log divided by 2π. In situ, the conversion log indicates in average 8% and 3% S/P amplitude conversion for the 3–6 km and 6–8.5 km depth intervals, respectively.
 In contrast, the synthetic conversion logs of simulated crack-free lithology show average S/P amplitudes of 2%. There is no pronounced similarity between the depth trends of the modeled and field conversion logs. This indicates that the influence of mineral composition and layer geometry on PS conversion is mostly camouflaged by cracks or fractures. The complex model shows conversion values comparable to the simple 1D model. Thus, one fundamental result for the KTB type of complex metamorphic crust is that less than 50% of the PS conversion is driven by lithological contrasts and tectonic structure but most of the conversion is caused by the crack- and fracture-related heterogeneous weakening of the rocks.
3.2. Comparison of Conversion Log vs. Other Logs
 In order to verify this hypothesis we compared the conversion log to the geophysical and petrological logs of the KTB data base. We found only two logs correlating directly and inversely with the conversion log: the depth function of crack density derived from formation-microscanner (FMS) logs [Hirschmann and Lapp, 1994] (Figure 4c) and the depth function of chlorite content derived from cuttings analysis [Duyster et al., 1995] (Figure 4b). A prominent feature of the field data is the drop of the conversion log below 6 km depth. The comparison of Figures 4a and 4c shows that this shift in conversion level corresponds to the level in fracture density. Both are lowered by a factor of 2. Individual peaks, however, do not correspond, with few exceptions. Since the variation of elastic moduli is proportional to the number density of cracks [e.g., O'Connell and Budiansky, 1974], the observed correlation confirms that the major portion of PS forward scattering is related to crustal fractures. Because the cracks observed in the FMS log are much smaller than the seismic wavelength we assume that they affect scattering via the heterogeneity of effective velocity rather than directly.
 Amongst all mineral logs only chlorite correlates, however inversely, with both conversion and fracture logs (Figure 4b). Note that the relatively high level of chlorite below 6000 m depth coincides well with a low level of conversion. Chlorite is a regular component of the rock matrix the initial abundance of which may be increased by hydrothermal alteration of mafic minerals. Its increased abundance in the lower KTB section may primarily indicate a change in facies connected with less open crack porosity than in the upper section of the KTB drill hole. In addition, hydrothermal circulation in the fault system of the lower KTB section (SE1 in Figure 1a) may have contributed to a chlorite accumulation reducing pore space and PS conversion.
 (1) PS converted waves can be quantified in VSPs by slant stacking. The resulting “conversion-log” was applied to determine the PS forward scattering component of P-wave absorption. (2) At the KTB site P-waves almost continuously radiate downward travelling PS-waves the amplitudes of which vary between 3 and 10% of the incident P wave with maxima of 18%. This implies that 0.1–1% of P wave energy per cycle is continuously transferred to shear with local maxima of 3%. (3) Wavefield simulation based on the lithological log and the seismic velocities of crack-free rocks showed that only 10–50% of the conversion level can be explained by lithological and structural heterogeneity. The crack influence makes up 50–90% of the conversion. Therefore, the major portion of conversion is caused by fracture zones where seismic heterogeneity is increased compared to solid rock. (4) The KTB conversion log is proportional to the number density of cracks and inversely related to the chlorite content of the drilled formation. Because of the strong influence of the crack density on the variation of seismic velocity the conversion log is mainly a measure of open fracture porosity. The observed inverse correlation of chlorite and conversion logs may primarily indicate a change in facies. In addition, it could be evidence of a progressive healing of fracture pore space in a hydrothermal environment. This inverse influence of fractures and chlorite on conversion amplitudes raises the question whether hydrothermal alteration significantly influences the possibility of seismic imaging of faulted metamorphic rocks.
 We are grateful to R. Pechnig and H. Hoskins for valuable comments. Data acquisition was funded by DFG grant Ra496/8, ICDP grants 07-98/1 and 03/99, NSF grant EAR-9727654, GFZ Potsdam and GGA Hannover.