Seismic Attenuation Computed from GLIMPCE Reflection Data and Comparison with Refraction Results

—Instantaneous frequency matching has been used to compute differential t* values for seismic reflection data from the Great Lakes International Multidisciplinary Program on Crustal Evolution (GLIMPCE) experiment. The differential attenuation values were converted to apparent Q ^?1 models by a fitting procedure that simultaneously solves for the interval Q ^?1 values using non-negative least squares. The bootstrap method was then used to estimate the variance in the interval Q ^?1 models. The shallow Q ^?1 structure obtained from the seismic reflection data corresponds closely with an attenuation model derived using instantaneous frequency matching on seismic refraction data along the same transect. This suggests that the effects of wave propagation and scattering on the apparent attenuation are similar for the two data sets. The Q ^?1 model from the reflection data was then compared with the structural interpretation of the reflectivity data. The highest interval Q ^?1 values (>0.01) were found near the surface, corresponding to the sedimentary rock sequence of the upper Keweenawan. Low Q ^?1 values (<0.0006) are found beneath the Midcontinent rift’s central basin. In addition to structural interpretation, seismic attenuation models derived in this way can be used to correct reflection data for dispersion, frequency and amplitude effects, and allow for improved imaging of the subsurface.     

Q-structure beneath the Tibetan Plateau from the inversion of Love- and Rayleigh-wave attenuation data

The fundamental mode Love and Rayleigh waves generated by ten earthquakes and recorded across the Tibet Plateau, at QUE, LAH, NDI, NIL, KBL, SHL, CHG, SNG and HKG are analysed. Love- and Rayleigh-wave attenuation coefficients are obtained at time periods of 5–120 s using the spectral amplitudes of these waves for 23 different paths. Love wave attenuation coefficient varies from 0.0021 km^?1, at a period of 10 s, to 0.0002 km^?1 at a period of 90 s, attaining two maxima at time periods of 10 and 115 s, and two minima at time periods of 25 and 90 s. The Rayleigh-wave attenuation coefficient also shows a similar trend. The very low value for the dissipation factor, Q_β, obtained in this study suggests high dissipation across the Tibetan paths. Backus-Gilbert inversion theory is applied to these surface wave attenuation data to obtain average Q_β^?1 models for the crust and uppermost mantle beneath the Tibetan Plateau. Independent inversion of Love- and Rayleigh-wave attenuation data shows very high attenuation at a depth of ～50–120 km ($\text{Q}{}_{\mathrm{\beta }}\text{? 10}$). The simultaneous inversion of the Love and Rayleigh wave data yields a model which includes alternating regions of high and low Q_β^?1 values. This model also shows a zone of high attenuating material at a depth of ～40–120 km. The very high inferred attenuation at a depth of ～40–120 km supports the hypothesis that the Tibetan Plateau was formed by horizontal compression, and that thickening occurred after the collision of the Indian and Eurasian plates.     

Ultrasonic measurements of Vp and Qp: relaxation spectrum of complex modulus on basalt melts

Ultrasonic compressional wave velocity V_p and quality factor Q_p have been measured in alkali basalt, olivine basalt and basic andesite melts in the frequency range of 3.4–22 MHz and in the temperature range of 1100–1400°C. Velocity and attenuation of the melts depend on frequency and temperature, showing that there are relaxation mechanisms in the melts. Complex moduli are calculated from the ultrasonic data. The results fit well a complex modulus of Arrhenius temperature dependence with log-normal Gaussian distribution in relaxation times of attenuation. The analysis yields average relaxation time, its activation energy, relaxed modulus, unrelaxed modulus and width of Gaussian distribution in relaxation times. Relaxed modulus is smaller (17.5 GPa) for basic andesite melt of high silica and high alumina contents than for the other two basalt melts (18.1–18.4 GPa). The most probable relaxation times decrease from ～ 3 × 10^?10 s for basic andesite to ～ 10^?11 s for alkali basalt at 1400°C. Activation energies of attenuation, ranging from 270 to 340 kJ mol^?1 in the three melts, are highest in basic andesite. Longitudinal viscosity values and their temperature dependences are also calculated from V_p and Q_p data. The volume viscosity values are estimated from the data using the shear viscosity values. Longitudinal, volume and shear viscosities and their activation energies are highest in the basic andesite melt of the most polymerized structure.     

Estimation of the Intrinsic Absorption and Scattering Attenuation in Northeastern Venezuela (Southeastern Caribbean) Using Coda Waves

—Northeastern Venezuela has been studied in terms of coda wave attenuation using seismograms from local earthquakes recorded by a temporary short-period seismic network. The studied area has been separated into two subregions in order to investigate lateral variations in the attenuation parameters. Coda-Q ^?1 (Q _c ^?1) has been obtained using the single-scattering theory. The contribution of the intrinsic absorption (Q _i ^?1) and scattering (Q _s ^?1) to total attenuation (Q _t ^?1) has been estimated by means of a multiple lapse time window method, based on the hypothesis of multiple isotropic scattering with uniform distribution of scatterers. Results show significant spatial variations of attenuation the estimates for intermediate depth events and for shallow events present major differences. This fact may be related to different tectonic characteristics that may be due to the presence of the Lesser Antilles subduction zone, because the intermediate depth seismic zone may be coincident with the southern continuation of the subducting slab under the arc.     

Short-period PKP phases and the anelastic mechanism of the inner core

Short-period seismograms are synthesized for PKP phases in anelastic Earth models. The synthetics were constructed using a synthetic technique valid at grazing incidence, a source-time function appropriate for deep-focus earthquakes, and an instrument response for either a short-period WWSSN or SRO seismograph. The agreement between predicted and observed amplitudes and spectral ratios requires neither a low-Q_α zone at 0.2–2 Hz nor a low or negative P-velocity gradient at the bottom of the outer core. Thin low-Q_α zones beneath the inner core boundary fit spectral ratio data that sample the upper 200 km of the inner core but fail to fit data that sample the lower inner core. Only a model having Q_α^?1?[0.003, 0.004] at 0.2–2 Hz, nearly constant with depth in the inner core, satisfies all of the spectral ratio and amplitude data. The assumption of a bulk viscosity of 10-10³ Pa s for the liquid phase of a partially molten inner core combined with the observation of low shear attenuation in the inner core at frequencies less than 0.005 Hz limit the physical parameters associated with two possible attenuation mechanisms: (1) fluid flow and viscous relaxation due to ellipsoidally shaped inclusions of melt, and (2) the solid-liquid phase transformation induced by the stress change during the passage of a seismic wave. Both mechanisms require an order of 0.1% partial melt to reproduce the observed Q_α^?1. In the outer core, the time constant of the mechanism of phase transformation is predicted to be 10⁴–10⁶ s. Confirmation of small shear attenuation in the inner core in the frequency band of seismic body waves would favor the mechanism of phase transformation.     

Seismic Q for the lithosphere of the Montenegro region (Yugoslavia): frequency,depth and time windowing effects

The availability of accelerometric data for the Montenegro earthquake of 15th April 1979 makes it possible to investigate seismic Q of the lithosphere in that region, in particular, its dependence on frequency, on the depth reached by seismic waves, and on the length of time windows in which signals are processed. Two different spectral methods, S phase energy ratio and coda envelope decay, are applied, respectively, to direct and scattered shear waves. Similar results are obtained using different portions of the recordings, i.e., coda waves for the envelope decay fit and the S wave train, with a significant duration of ～ 10 s, for the energy ratios. The same apparent Q (Q ～ 40 f, where f is the frequency expressed in Hz) that is found for other neighbouring central Mediterranean regions (e.g., Ancona, on the central Italian Adriatic coast; Valnerina, in the central Apennines; Irpinia, in the southern Apennines) is also found for the southern Yugoslavian coast, in the band 1–25 Hz up to a maximum range of ～ 120 km from the focus. This strong frequency dependence is probably connected with the type of small-scale heterogeneity and the same geological age and level of tectonic activity peculiar to all these seismotectonic areas.In order to compare the apparent Q of the whole S wave train, ～ 10 s long, with the (intrinsic) apparent Q of the single direct S wave (usually 1 s or less), the maximum entropy method is applied in the energy spectrum computation for shorter wave trains. The use of shorter time windows does not reveal any significant variation in the tendency of Q to increase linearly with frequency as the length of the time window containing the sample of the S waves decreases. This seems to indicate that scattering-dependent Q is generally inseparable from intrinsic Q in the lithosphere when estimates based on variations with distance of the seismic signal spectrum are used. While the type of linear growth with frequency does not seem to undergo any variations (it remains of the Q = qf type), the data show there are a considerable decrease in the coefficient of proportionality Q with decreasing duration of the window of S waves analysed, probably as a result of variations in seismic attenuation with depth.     

Attenuation models of the earth

Free oscillation and body wave data are used to construct average Q models for the earth. The data set includes fundamental and overtone observations of the radial, spheroidal and toroidal modes, ScS observations and amplitudes of body waves as a function of distance. The preferred model includes a low-Q zone at both the top and the bottom of the mantle. In these regions the seismic velocities are likely to be frequency dependent in the “seismic” band. Absorption in the mantle is predominantly due to losses in shear. Compressional absorption may be important in the inner core.A grain-boundary relaxation model is proposed that explains the dominance of shear over compressional dissipation, the roughly frequency independent average values for Q and the variation of Q with depth. In the high-Q regions, the lithosphere and the midmantle (200–2000 km), Q is predicted to be frequency dependent. However, the low-Q regions of the earth, where Q is roughly frequency independent, dominate the observations of attenuation.     

Seismic Wave Attenuation in the Greater Cairo Region, Egypt

In the present study, a digital waveform dataset of 216 local earthquakes recorded by the Egyptian National Seismic Network (ENSN) was used to estimate the attenuation of seismic wave energy in the greater Cairo region. The quality factor and the frequency dependence for Coda waves and S-waves were estimated and clarified. The Coda waves (Q _c) and S-waves (Q _d) quality factor were estimated by applying the single scattering model and Coda Normalization method, respectively, to bandpass-filtered seismograms of frequency bands centering at 1.5, 3, 6, 12, 18 and 24?Hz. Lapse time dependence was also studied for the area, with the Coda waves analyzed through four lapse time windows (10, 20, 30 and 40?s). The average quality factor as function of frequency is found to be Q _c?=?35?±?9f ^0.9±0.02 and Q _d?=?10?±?2f ^0.9±0.02 for Coda and S-waves, respectively. This behavior is usually correlated with the degree of tectonic complexity and the presence of heterogeneities at several scales. The variation of Q _c with frequency and lapse time shows that the lithosphere becomes more homogeneous with depth. In fact, by using the Coda Normalization method we obtained low Q _d values as expected for a heterogeneous and active zone. The intrinsic quality factor (Q _i ^?1 ) was separated from the scattering quality factor (Q _s ^?1 ) by applying the Multiple Lapse Time Domain Window Analysis (MLTWA) method under the assumption of multiple isotropic scattering with uniform distribution of scatters. The obtained results suggest that the contribution of the intrinsic attenuation (Q _i ^?1 ) prevails on the scattering attenuation (Q _s ^?1 ) at frequencies higher than 3?Hz.     

Studies of mantle structure of U.S.S.R. territory on long-range seismic profiles

Long-range seismic sounding carried out during the last few years on the territory of the U.S.S.R. has shown a basic inhomogeneity of the uppermost mantle, as well as evidence of regularities in the distribution of its seismic parameters. The following data were used: times and apparent velocities of P- and S-waves for investigation of mantle velocities, converted waves for seismic discontinuity model studies and wave attenuation for Q-factor estimation. Strong regularities were distinguished in the distribution of average seismic velocities for the uppermost mantle, in their dependence on the age and type of geostructure and on their position relative to the central part of the continent. Old platforms and the inner part of the continent are marked by velocities under the Mohorovi?i? discontinuity of more than 8.2–8.3 km s^?1, young platforms and outer parts of the continent by 8.0–8.2 km s^?1, and orogenic and rift zones by 7.8–8.0 km s^?1. The difference becomes more pronounced at a depth of about 100–200 km: for the old platform mantle velocities of 8.5–8.6 km s^?1 are typical; beneath the orogenic and rift areas, inversion zones with velocities less than 7.8 km s^?1 are observed.The converted waves show fine inhomogeneities of the crust and uppermost mantle, the presence of many discontinuities with positive and negative changes of velocity, and anisotropy of seismic waves in some of the layers. Wave attenuation allowed the determination of the Q-factor in the mantle. It varied from one region to another but a close relation between Q and P-wave velocity is the main cause of its variation.     

Dispersion and attenuation of elastic waves due to multiple scattering from cracks

Multiple scattering from cracks is considered in the two-dimensional plane-strain condition. It is assumed that identical cracks are distributed uniformly in space and that the effective waves propagate normal to the crack surfaces. Then, the apparent dispersion and attenuation are calculated as functions of frequency for three independent modes of wave propagation: SV, P and SH.The calculated results show that, in each case, the attenuation coefficient Q^?1 takes a peak value when the wavelength is nearly twice the crack width, while phase velocity has a maximum deviation from the intrinsic value at a frequency lower than the peak frequency for Q^?1.     

High-temperature anelasticity and elasticity of mantle peridotite

An apparatus has been devised which allows precise creep and relaxation measurements to be made on minerals and rocks at temperatures up to 1600°C and at very low deviatoric stresses (1 < σ < 300 bar). This paper is concerned with measurements on mantle peridotite (lherzolite) from Balmuccia (Zone of Ivrea, Italy).The reaction of the sample to a step-like increase in stress is called its “creep function”. It is shown that the creep function contains all the necessary information to derive the spectra of the quality factor Q(ω) and of Young's modulus E(ω), within the seismic range of frequencies, provided the material behaves as a linear system. This has been proven up to a strain of 5 × 10^?5.The Q^?1-spectra at 1200 and 1300°C, obtained by Fourier inversion from the creep function, show no pronounced peak in the frequency band 0.01 < tf < 1 Hz and exhibit a general tendency to decrease slightly with frequency. The creep function: ?(t) = ?_u · [1 + 3.7 · q · {(1 + 50t)^0.27 ? }], where q is related to Q, satisfactorily describes the data at high temperatures and leads to $\text{Q}{}^{\mathrm{?1}}\text{(ω, T) = 3 × 10}{}^3\text{· ω}{}^{\mathrm{<mtext>?0.27\; ·\; </mtext><mtext>exp</mtext>}}\text{(}\text{?30}\text{RT}\text{)}$E(ω) is related to Q(ω) by the material dispersion equation. Above 1100°C the unrelaxed Young's modulus decreases rapidly with temperature according to an activation energy of about 20 kcal/mole. A lowering of short period S-wave velocity by 40% and P-wave velocity by 10% occurs below the solidus. Therefore, no partial melting is required in the asthenosphere.Steady-state creep at low axial stresses (20 < σ < 100 bar), obtained from the same rock, follows the relation $\text{?}\text{?}\text{= 3 × 10}{}^7\text{· δ}{}^{1.4}\text{·}\text{exp}\text{(}\text{?125}\text{RT}\text{)}$ indicative of grain boundary diffusion or superplasticity. At higher stresses a power law $\text{?}\text{?}\text{= 45 · δ}{}^4\text{·}\text{exp}\text{(}\text{?125}\text{RT}\text{)}$ typical of dislocation creep, is found.The frequency dependence of Q and the ratio of the activation energies of Q and are indicative of so called “high-temperature background absorption”, as the dominant mechanism, and of a diffusion-controlled dislocation mobility common to both absorption and creep. From a, b, and c, relations between the effective viscosity η_f and Q of the form: $\text{log}\text{η}{}_{\mathrm{e??}}\text{=}\text{1}\text{α}\text{·}\text{log}\text{Q ? (n ? 1) ·}\text{log}\text{ω +}\text{log}\text{D}$ are derived, where α ～ 0.25, n is the power of σ, and D is a constant.     

On the Causes of Frequency-Dependent Apparent Seismological Q

Variability of the Earth’s structure makes a first-order impact on attenuation measurements which often does not receive adequate attention. Geometrical spreading (GS) can be used as a simple measure of the effects of such structure. The traditional simplified GS compensation is insufficiently accurate for attenuation measurements, and the residual GS appears as biases in both Q ₀ and η parameters in the frequency-dependent attenuation law Q(f) = Q ₀ f ^η . A new interpretation approach bypassing Q(f) and using the attenuation coefficient χ(f) = γ + πf/Q _e(f) resolves this problem by directly measuring the residual GS, denoted γ, and effective attenuation, Q _e. The approach is illustrated by re-interpreting several published datasets, including nuclear-explosion and local-earthquake codas, Pn, and synthetic 50–300-s surface waves. Some of these examples were key to establishing the Q(f) concept. In all examples considered, χ(f) shows a linear dependence on the frequency, γ ≠ 0, and Q _e can be considered frequency-independent. Short-period crustal body waves are characterized by positive γ _SP values of (0.6–2.0) × 10^?2 s^?1 interpreted as related to the downward upper-crustal reflectivity. Long-period surface waves show negative γ _LP ≈ ?1.9 × 10^?5 s^?1, which could be caused by insufficient modeling accuracy at long periods. The above γ values also provide a simple explanation for the absorption band observed within the Earth. The band is interpreted as apparent and formed by levels of Q _e ≈ 1,100 within the crust decreasing to Q _e ≈ 120 within the uppermost mantle, with frequencies of its flanks corresponding to γ _LP and γ _SP. Therefore, the observed absorption band could be purely geometrical in nature, and relaxation or scattering models may not be necessary for explaining the observed apparent Q(f). Linearity of the attenuation coefficient suggests that at all periods, the attenuation of both Rayleigh and Love waves should be principally accumulated at the sub-crustal depths (~38–100 km).     

Estimation of frequency dependent coda wave attenuation structure at the vicinity of Cairo Metropolitan Area

Estimation of seismic wave attenuation in the shallow crust in terms of coda wave Q structure previously investigated in the vicinity of Cairo Metropolitan Area was improved using seismograms of local earthquakes recorded by the Egyptian National Seismic Network. The seismic wave attenuation was measured from the time decay of coda wave amplitudes on narrow bandpass filtered seismograms based on the single scattering theory. The frequency bands of interest are from 1.5 to 18 Hz. In general, the values obtained for various events recorded at El-Fayoum and Wadi Hagul stations are very similar for all frequency bands. A regional attenuation law Q _c = 85.66 f ^0.79 was obtained.     

Comparison of Sesimic Body Wave and Coda Wave Measures of Q

—Measurements of seismic attenuation (Q ^?1) can vary considerably when made from different parts of seismograms or using different techniques, particularly at high frequencies. These discrepancies may be methodological, or may reflect earth processes. To investigate this problem, we compare body wave with coda Q ^?1 results utilizing three common techniques i) parametric fit to spectral decay, ii) coda normalization of S waves, and iii) coda amplitude decay with lapse time. Q ^?1 is measured from both body and coda waves beneath two mountain ranges and one platform, from recordings made at seismic arrays in the Caucasus and Kopet Dagh over paths ≤ 4° long. If Q is assumed frequency independent, spectral decay fits show Q _s and Q _coda near 700–800 for both mountain paths and near 2100–2200 for platform paths. Similar values are determined with the coda normalization technique. However, frequency-dependent parameterizations fit the data significantly better, with Q _s ?(1 Hz) and Q _coda?(1 Hz) near 200–300 for mountain paths and near 500–600 for platform paths. Lapse decay measurements are close to the frequency-dependent values, showing that both spectral and lapse decay methods can give similar results when Q has comparable parameterizations. Above 6 Hz, coda measurements suggest some enrichment relative to body waves, perhaps due to scattering, but intrinsic absorption appears to dominate at lower frequencies. All approaches show sharp path differences between the Eurasian platform and adjacent mountains, and all are capable of resolving spatial variations in Q.     

Constant Q and a Fractal, Stratified Earth

— Frequency-dependent measurements of the quality factor Q typically show a constant behaviour for low frequencies and a positive power-law dependence for higher frequencies. In particular, the constant Q pattern is usually explained using intrinsic attenuation models due to anelasticity with either a single or multiple superposed relaxation mechanisms — each with a particular resonance peak.¶However, in this study, I show using wave localisation theory that a constant Q may also be due to apparent attenuation due to scattering losses. Namely, this phenomenon occurs if the earth displays fractal characteristics. Moreover, if fractal characteristics exist over a limited range of scales only, even an absorption band can be created—in accordance with observations. This indicates that it may be very difficult to distinguish between intrinsic and scattering attenuation on the basis of frequency-dependent measurements of the quality factor only.     

A Summary of Attenuation Measurements from Borehole Recordings of Earthquakes: The 10 Hz Transition Problem

—Earthquake seismograms recorded by instruments in deep boreholes have low levels of background noise and wide signal bandwidth. They have been used to extend our knowledge of crustal attenuation both in the near-surface and at seismogenic depths. Site effects are of major importance to seismic hazard estimation, and the comparison of surface, shallow and deep recordings allows direct determination of the attenuation in the near-surface. All studies to date have found that Q is very low in the near-surface (～ 10 in the upper 100 m), and increases rapidly with depth. Unlike site amplification, attenuation at shallow depths exhibits little dependence on rock-type. These observations are consistent with the opening of fractures under decreasing lithostatic pressure being the principal cause of the severe near-surface attenuation. Seismograms recorded in deep boreholes are relatively unaffected by near-surface effects, and thus can be used to measure crustal attenuation to higher frequencies (≥ 100 Hz) than surface recordings. Studies using both direct and coda waves recorded at over 2 km depth find Q to be high (～ 1000) at seismogenic depths in California, increasing only weakly with frequency between 10 and 100 Hz. Intrinsic attenuation appears to be the dominant mechanism. These observations contrast with those of the rapidly increasing Q with frequency determined from surface studies in the frequency range 1 to 10 Hz. Further work is necessary to constrain the factors responsible for this apparent change in the frequency dependence of Q, but it is clearly unwise to extrapolate Q estimates made below about 10 Hz to higher frequencies.     

High-frequency P- and S-wave Attenuation in the Earth

—Investigations of the spectral characteristics of teleseismic body waves revealed that the spectral falloff rate between 1 Hz and 10 Hz is primarily controlled by anelastic attenuation along the path. In addition, the amount of high-frequency energy in teleseismic body waves is far above the level expected on the basis of Q estimates at low frequencies, thus leading to the idea of frequency dependence in Q. Q variations in the earth’s mantle can be investigated by mapping out the variations of high frequency (4 - 10 Hz) energy relative to the low frequency (1 - 3 Hz) energy in teleseismic P waves, and similar ratios at lower frequencies in teleseismic S waves. Because of the extreme sensitivity of spectral content of short-period body waves to Q variations, large uncertainties in other factors affecting spectral content can be tolerated in such studies. With the increasing number and density of broadband seismic stations recording at high sampling rates, tomographic studies of Q at high frequencies become possible.     

Incorporation of the Effects of Absorption into Kirchhoff Migration

—The effects of absorption are incorporated into Kirchhoff migration. The aim is to reconstruct the structures with true-amplitude seismic data and to increase the resolution of migrated data. A complex wave velocity is introduced into the solution of the Helmholtz equation, the starting point of Kirchhoff migration. This leads to an additional filter, the antidissipation operator, which is convolved with the wave field. The general structure of Kirchhoff migration remains unchanged. The effects of the antidissipation operator are illustrated on synthetic data. The new operator is valid for complex media with varying velocity and varying quality factor Q. Moreover there is no limitation to constant-Q, frequency-dependent Q can also be handled. The success of anelastic migration depends on how well the Q macro model is known.     

Characteristics of seismic wave attenuation in the crust and upper mantle of the Northern Caucasus

The quality factor Q as a function of frequency in an S wave range of 1–8 Hz is estimated from records of ～60 earthquakes (M _w > 3.9 and source depths of 1–60 km) obtained at the Sochi seismic station at epicentral distances of less than ～300 km. Methods of Q estimation used in the paper were developed in works by Aki, Rautian, and others; they are based on the suppression of source-related and local effects in S wave spectra with the help of coda waves measured at a fixed time from the first arrival. To compensate for directivity effects, averaging was performed over the set of events whose sources were located in a wide range of back azimuths. The geometric divergence is represented as a three-segment function: 1/R, 1, and 1/√R at epicentral distances of 1/50–1/70 to 50–70 km, 50–70 to 130–150 km, and greater than 130–150 km, respectively. The geometric divergences in this model yielded the following estimates of the quality factor: Q(f) ～ 80f ^0.9 with a base of 35–280 km and Q(f) ～ 110f ^0.8 with a base of 60–280 km. The resulting combinations of the propagation path effects (Q and the geometric divergence) can be used for predicting strong motion parameters in the Northern Caucasus.     

Shear modulus and Q of forsterite and dunite near partial melting from forced-oscillation experiments

An apparatus designed to determine the complex shear modulus of rock samples by forced torsion oscillations at high temperature and in the seismic frequency band 0.003–30 Hz is briefly described. Measurements were performed on natural dunite from Åheim (Norway) up to 1400°C and on polycrystalline forsterite up to 1500°C at 1 atm pressure. The two materials were chosen to study, by comparison, the effect of melt on the elasticity and anelasticity of mantle rocks.Between 1000 and 1200°C the absolute values of the shear modulus G are almost equal for both materials. Above 1200°C G for natural dunite decreases progressively with temperature and at 1400°C and 1 Hz reaches $\text{～}\text{1}\text{3}$ of its value at 1100°C. In contrast, G of pure forsterite depends little on temperature. For petrological reasons, supported by simultaneous measurements of the electric resistivity, there is strong evidence that the decrease of G in dunite above 1200°C is due to melt from the lower melting components of the dunite. Based on different models estimates of the melt fraction are made.At high temperature, in both materials Q^?1 is characterized by a monotonic decrease with frequency according to ω^?α, with α ≈ 0.25. An apparent activation energy of 38±5 kcal mol^?1 for forsterite and 48±8 kcal mol^?1 for dunite was found with no significant change in the regime of partial melting. From this it is concluded that Q^?1, even at partial melting, is dominated by solid state high temperature background absorption. There is no indication from these experiments for a constant-Q-band at low seismic frequencies or an increase of Q proportional to frequency as suggested by some seismologists. The present results are in good qualitative agreement with those for Young's modulus obtained previously by strain retardation experiments.