Time-lapse refraction seismic tomography for the detection of ground ice degradation

The ice content of the subsurface is a major factor controlling the natural hazard potential of permafrost degradation in alpine terrain. Monitoring of changes in ice content is therefore similarly important as temperature monitoring in mountain permafrost. Although electrical resistivity tomography monitoring (ERTM) proved to be a valuable tool for the observation of ice degradation, results are often ambiguous or contaminated by inversion artefacts. In theory, the sensitivity of P-wave velocity of seismic waves to phase changes between unfrozen water and ice is similar to the sensitivity of electric resistivity. Provided that the general conditions (lithology, stratigraphy, state of weathering, pore space) remain unchanged over the observation period, temporal changes in the observed travel times of repeated seismic measurements should indicate changes in the ice and water content within the pores and fractures of the subsurface material. In this paper, a time-lapse refraction seismic tomography (TLST) approach is applied as an independent method to ERTM at two test sites in the Swiss Alps. The approach was tested and validated based on a) the comparison of time-lapse seismograms and analysis of reproducibility of the seismic signal, b) the analysis of time-lapse travel time curves with respect to shifts in travel times and changes in P-wave velocities, and c) the comparison of inverted tomograms including the quantification of velocity changes. Results show a high potential of the TLST approach concerning the detection of altered subsurface conditions caused by freezing and thawing processes. For velocity changes on the order of 3000m/s even an unambiguous identification of significant ice loss is possible. Correspondence to: C. Hilbich ([email protected]) 1 Motivation Monitoring of permafrost in polar and mountainous regions becomes more and more important in the context of ongoing global warming. A key parameter concerning slope stability analyses and permafrost modelling purposes is the ice content of the subsurface and its temporal evolution (Gruber and Haeberli, 2007; Harris et al., 2009). Thermal monitoring in boreholes is a common and widespread method to observe permafrost evolution (e.g. Harris and Isaksen, 2008; PERMOS, 2009), but does only provide indirect insights into ice content changes. Many geophysical methods are more sensitive to changes in water content than to temperature variations, and in particular to phase transitions between frozen and unfrozen water. Fortier et al. (1994) showed that variations of apparent resistivity with time can be used to predict unfrozen water and ice contents of the frozen ground. Recent studies on the applicability of electrical resistivity tomography monitoring (ERTM) proved the high sensitivity of electrical resistivity tomography (ERT) to spatio-temporal changes in the subsurface ice and water contents (Hauck, 2002; Hilbich et al., 2008; Kneisel et al., 2008; Hilbich, 2009). Apart from ERTM, repeated refraction seismic measurements theoretically have a considerable potential to observe permafrost evolution, since the seismic P-wave velocity (vp) is highly sensitive to variations in the ice or water content (by changes in vp between frozen and unfrozen water of up to 2000m/s). Seismic refraction is often considered to be a valuable additional method to verify subsurface structures identified by ERT (e.g. Hauck and Vonder Mühll, 2003; Kneisel et al., 2008). It is generally capable of discriminating unfrozen and frozen sediments or massive ice, and is thus a common method to determine active layer thickness. However, P-wave velocities range from ca. 2500–4200m/s for Published by Copernicus Publications on behalf of the European Geosciences Union. 244 C. Hilbich: Time-lapse refraction seismic tomography for the detection of ground ice degradation permafrost ice and from ca. 2000–6000m/s for both frozen and unfrozen bedrock (e.g. Hauck and Kneisel, 2008). These large ranges and their wide overlap make a differentiation of stratigraphic details in bedrock and/or below the permafrost table (e.g. in rock glaciers or talus slopes) often difficult. Together with the comparatively high measurement and processing efforts, this may be a reason why refraction seismic surveys are less popular in permafrost research than e.g. ERT or GPR measurements. Nevertheless, numerous studies successfully applied refraction seismic surveys in permafrost terrain (e.g. Röthlisberger, 1972; Barsch, 1973; Harris and Cook, 1986; Vonder Mühll, 1993; Musil et al., 2002; Hauck et al., 2004; Ikeda, 2006; Hausmann et al., 2007; Maurer and Hauck, 2007). Main advantages of the method, compared to ERT surveys, are e.g. the much higher depth resolution (Lanz et al., 1998), the potential to exactly localise sharp layer boundaries, the less challenging coupling of the sensors in blocky terrain, or the applicability in terrain with electrically conductive infrastructure contaminating the resistivity signal. Provided that the general subsurface conditions (lithology, stratigraphy, state of weathering, pore space) remain unchanged over the observation period, changes in the ice and water content within the pores and fractures of the subsurface material should similarly or even better be detectable by repeated seismic measurements than for ERT. Despite the ambiguities involved in a qualitative interpretation of layers (due to overlapping velocity ranges of different materials), even comparatively small temporal changes in P-wave velocities may indicate zones with significant ice content changes due to seasonal variations or long-term climate change. According to the theoretical suitability of repeated seismic measurements to permafrost related research, a time-lapse refraction seismic tomography (TLST) approach and its potential to observe temporal changes in ice and water content in alpine permafrost will be evaluated in this paper. 2 Theory and approach Apart from a vast research related to reflection seismic monitoring of deep reservoirs in exploration geophysics (see e.g. Vesnaver et al., 2003; King, 2005), similar efforts to investigate the potential of a time-lapse refraction seismic tomography approach for the observation of shallow targets have not been reported so far. In exploration geophysics the detection of reservoir changes is preferably done using reflection seismics due to their high resolution potential at larger depths. However, reflection data rarely provide information on the very shallow parts (5–15m) of unconsolidated sediments, as reflections are usually overwhelmed by different coincidentarriving waves. They are more affected by scattering effects in highly heterogeneous material with shallow low velocity layers than seismic refractions (Lanz et al., 1998), which is the typical situation for mountain permafrost sites. According to Lanz et al. (1998) and Musil et al. (2002) refraction seismic tomography is therefore more appropriate for exploring the upper 50m of the subsurface. For deep reservoirs, a first approach in using time-lapse refraction seismics was introduced by Landrø et al. (2004) for the estimation of reservoir velocity changes. It is based on the fact, that a velocity change by only 1% will change the critical angle of refraction and thus the critical offset for refracted waves, i.e. their first appearance at the surface. For oil reservoirs located at depths of > 1000m this change in critical offset amounts to several tens of meters (Landrø et al., 2004), but for shallow targets this would be reduced to a few centimetres to decimetres, making this approach inappropriate for mountain permafrost evolution. Concerning shallow applications, no operationally applicable time-lapse refraction seismic approach was published so far, neither for permafrost, nor for other fields of interest. The presence of ice in the pore spaces of sediments can cause large increases in seismic velocity compared to the velocity when the interstitial water is unfrozen (Timur, 1968). Since ice is much stiffer than water, the wave speed is a strongly increasing function of the ice-to-water ratio. However, firm rock is much stiffer than either ice or water, therefore the wave speed is also a decreasing function of the porosity � (Zimmerman and King, 1986). In the literature several approaches to calculate the dependence of vp on ice and water content in an ice-liquid-rock matrix have been formulated, an overview is given by Carcione and Seriani (1998). Even though the different approaches to model a three-phase medium have different limitations (e.g. restriction to unconsolidated or consolidated material), all models simply relate the composite density ρ of a medium to the respective fractions � of water (�w), ice (�i ) and solid rock (�s) and their densities ρw, ρi , and ρs : ρ=�wρw +�iρi +�sρs , (1) provided that �w +�i +�s = 1. Due to this common assumption, all models presented in Carcione and Seriani (1998) show, despite their differences, similar qualitative dependencies between ice, water and the rock material and their respective P-wave velocities. According to Hauck et al. (2008), theoretically, Eq. (1) can be extended including also the air-filled pore space to account for all components of a subsurface material. Due to the markedly different velocities of air (330m/s), water (ca. 1500m/s) and ice (3500m/s), accordingly, a transition from ice to water (melting) or ice to air (melting and drainage) would result in pronounced velocity changes. Analysing the relation of vp and the gaseous (air), liquid (water), and two solid (rock/soil, ice) components of the subsurface as a function of temperature reveals that the bulk velocity of a medium is generally higher under frozen compared to that of unfrozen conditions. Thus, thawing causes a decreasing and freezing an increasing velocity, respectively. The Cryosphere, 4, 243–259, 2010 www.the-cryosphere.net/4/243/2010/ C. Hilbich: Time-lapse refraction seismic tomography for the detection of ground ice degradation 245 The velocity change depends mainly on the porosity and the initial saturation, and is thus more pronounced for unconsolidated coarse-grained sediments than for consolidated rocks (Scott et al., 1990). In a qualitative sense, these general dependencies can be used to analyse permafrost evolution via repeated seismic measurements. The principle of a repeated (time-lapse) refraction seismic approach is exemplarily illustrated in Fig. 1 for a coarse grained material with waterand/or air filled voids. From a permafrost degradation point of view, not only degradation from above (Fig. 1b) but also an overall warming of the permafrost may be detected by increasing amounts of unfrozen water or air (as a consequence of draining), as roughly indicated in Fig. 1c. Necessary conditions to reliably detect subsurface changes via repeated refraction seismic measurements include constant measurement conditions (i.e. source-receiver-geometry and signal generation) between subsequent measurements. The measured time-lapse seismic data can then be successively processed and analysed using standard methods for tomographic inversion of seismic data. In this study, a refraction seismic tomography algorithm is used that reconstructs the 2-D velocity pattern of the subsurface based on an iterative adaptation of synthetic travel times to observed travel times along calculated ray paths of the seismic P-waves (so-called SIRT algorithm, software REFLEXW, Sandmeier, 2008). A hierarchy of methods has been evaluated to detect temporal changes in the subsurface: (a) The comparison of seismograms from subsequent measurement dates (time-lapse seismograms) and analysis of reproducibility of the seismic signal. (b) The analysis of time-lapse travel time curves with respect to resolving possible shifts in travel times and changes in P-wave velocities. (c) The comparison of inverted tomograms calculated with REFLEXW including the quantification of spatiotemporal velocity changes by calculating the differences between individual tomograms (time-lapse tomography). Hereby, the 2-dimensional tomographic approach in (c) was chosen to detect lateral changes in the velocity field. For an exact localisation of vertical discontinuities (if present), a wavefront inversion would give a higher accuracy due to the enhanced smoothing effect in the tomographic approach. In the following, the potential and limitations for the application of time-lapse refraction seismic in high mountain permafrost environments will be analysed based on repeated measurements at two sites. Fig. 1. Idealised principle of time-lapse refraction seismic based on (a) a two-layer subsurface model with a coarse-blocky unfrozen overburden with air-filled voids and a saturated rock-ice-matrix underneath. The lower panels illustrate the change of travel times as a consequence of (b) a vertical shift of the refractor (e.g. the seasonally varying interface between frozen and unfrozen conditions), and (c) altered ice (and air) contents within the lower layer causing changes in seismic velocity. The corresponding travel time curves for all three scenarios are given above. 3 Site description and data sets To evaluate the TLST approach repeated refraction seismic measurements were carried out at two different permafrost test sites in the summer season of 2008. The test sites comprise a) the ventilated Lapires talus slope in the Valais, and b) the north oriented slope of the Schilthorn summit in the Bernese Alps (see Fig. 2). The Lapires site represents a permafrost site with unconsolidated sediments and is a vast concave talus slope, oriented in NE direction, which extends over more than 500m width between 2350 and 2700m altitude. Lithology consists www.the-cryosphere.net/4/243/2010/ The Cryosphere, 4, 243–259, 2010 246 C. Hilbich: Time-lapse refraction seismic tomography for the detection of ground ice degradation Table 1. Measurement details for the time-lapse refraction seismic test sites (abbreviations: TD= thaw depth, n.a. = notavailable).