Coda Attenuation Research Articles

Frequency-dependent attenuation of P and S waves in Yunnan region

We analyzed digital seismogram data of 5 668 earthquakes that occurred in Yunnan region between July of 1999 and December of 2003. Among the 22 seismic stations, six ones, namely Baoshan, Yongsheng, Lijiang, Heqing, Yimen, and Luquan, were selected and their attenuations of P and S waves were measured by using the extended coda-normalization method. The six stations were classified into three regions according to their location, that is, Baoshan area, Yongsheng-Lijiang-Heqing (YLH) area and Luquan-Yimen (LY) area. The values of Q S −1 and Q P −1 are expressed as Q S −1 =0.008 67 f −0.86, Q P −1 =0.011 55 f −0.93, Q S −1 =0.018 24 f −0.94, Q P −1 =0.022 88 f −0.92, and Q S −1 =0.016 47 f −0.91, Q P −1 =0.028 26 f −0.97 in Baoshan, YLH, LY areas respectively. The attenuation of YLH, LY are closer to each other, however, Baoshan area is apparently lower. Comparing attenuation in the three areas with other areas of the world using the same method, it is suggested that the attenuations of P and S waves in YLH and LY areas are close to Kanto of Japan, but much higher than southeast of South Korea. The Q S −1 and Q P −1 in Baoshan area are slightly higher than southeastern South Korea. Furthermore, the results indicate that our Q S −1 in Yunnan area is close to others by analyzing the coda attenuation.

Seismic attenuation in a nonvolcanic swarm region beneath Wakayama, southwest Japan

Seismic attenuation in the Wakayama swarm area of southwest Japan is estimated by optimally fitting the theoretical S coda model to an average S coda envelope in the frequency range of 1–48 Hz. The average S coda envelope is confirmed to have a common decay curve independent of the source‐station distance. Intrinsic attenuation is found to peak at around 4 Hz, suggesting thermal diffusion in grain size or crack length domains with dimensions of the order of 0.7–1 mm, which is reasonable for crustal rock. Scattering attenuation is well approximated by the relation 0.054f−1 in this frequency range, with an inferred peak at 0.5 Hz, suggesting that the statistical model is characterized by an exponential autocorrelation function with a correlation length of about 1 km and a fractional velocity fluctuation of 21%. Scattering is the dominant cause of attenuation at frequencies below 2 Hz, while intrinsic attenuation becomes predominant above 4 Hz. Although values of coda attenuation (Qc−1) generally lie between the total and intrinsic attenuation, the differences between these three values becomes very small above 4 Hz, indicating that the coda attenuation measurements provide a reasonably good estimate of the total attenuation. The total attenuation was found to coincide well with the total apparent attenuation estimated by a linear inversion using direct S waves from swarm earthquakes, confirming the validity of the present results.

Spatial variation of coda wave attenuation in northwestern Colombia

SUMMARY One thousand seven hundred and eighty-six vertical-component, short-period observations of microearthquake codas from regional earthquakes recorded by 17 stations belonging to the National Seismological Network of Colombia were used to estimate seismic wave attenuation in Colombia. Local magnitudes range from 2.9 to 6.0 and only events occurring at hypocentral distances up to 255 km were considered for the analysis. The frequencies of interest lay between 1 and 19 Hz and the analysis was performed for each seismic station separately. Coda-wave attenuation (Q −1 ) was estimated by means of a single-scattering method whereas the separation of intrinsic absorption (Q −1 ) and scattering attenuation (Q −1 ) from total attenuation (Q −1 )w as performed using a multiple lapse time-window analysis based on the hypothesis of multiple isotropic scattering and uniform distribution of scatterers. A regionalization of the estimated Q0 (Qc at 1 Hz) values was performed and a contour map of seismic coda attenuation in Colombia is presented, where four zones with significant variations of attenuation related to different geological and tectonic characteristics can be observed. The highest attenuation is linked to the central and western regions (Q0 around 50 and 56) whereas a lower attenuation (Q0 around 69 and 67) is assigned to the northern and eastern regions. Results show that the Q −1 values are frequency dependent in the considered frequency range, and are approximated by a least-square fit to the power law Q −1 ( f ) = Q −1 ( f / f 0) −η . The exponents of the frequency dependence law ranged from η = 0.65 to 1.01 for Q −1 , η = 0.62 to 1.78 for Q −1 , η = 0.28 to 1.49 for Q −1 , and η = 0.53 to 1.67 for Q −1 . On the other hand, intrinsic absorption is found to dominate over scattering in the attenuation process for most of the stations and frequency bands analysed. Some discrepancies have been observed between the theoretical model and the observations for some frequency bands which indicate that it would be necessary to consider models for depth-dependent velocity structure and/or non-isotropic scattering patterns.

Analysis of S-Wave Attenuation in South-Central Alaska

The S -wave attenuation of the crust and upper mantle in south-central Alaska is determined from shear-wave spectral data of 60 earthquakes with magnitudes 3.0–6.3 recorded by a network of 22 strong motion stations in Anchorage. The direct S -wave attenuation ( \(Q_{{\beta}}^{-1}\) ) and coda attenuation ( \(Q_{\mathrm{c}}^{-1}\) ) were estimated by the generalized inversion and coda decay methods, respectively. The scattering and intrinsic attenuation coefficients, η s and η i , were determined by applying a multiple scattering model to the variation of integrated spectral energy with hypocentral distance for three consecutive lapse time windows (0–15, 15–30, 30–45 sec), starting from the onset of the S -wave arrival. In order to obtain the dependence of attenuation coefficients on frequency, the spectral energies over an octave bandwidth for three central frequencies, 1.0, 3.0, and 6.0 Hz, were calculated. The results show the following: (1) \(Q_{{\beta}}^{-1}\) is higher than \(Q_{\mathrm{c}}^{-1}\) for frequencies ( f ) between 0.6 and 3 Hz and less for f > 3.0 Hz; (2) for 1.0 Hz ≤ f ≤ 6.0 Hz, η i increases and η s decreases gradually with f; (3) the seismic albedo, B 0 = η s /(η s + η i ), of the medium decreases from 0.48 to 0.15 with increasing frequency for 1.0 Hz ≤ f ≤ 6.0 Hz; (4) the total attenuation coefficient, \(L_{\mathrm{e}}^{-1}={\eta}_{\mathrm{i}}+{\eta}_{\mathrm{s}}\) , is nearly frequency independent; (5) at low frequencies (∼1.0 Hz) the scattering attenuation ( \(Q_{\mathrm{s}}^{-1}=k^{-1}{\eta}_{\mathrm{s}}\) , k is the wavenumber) is comparable to the intrinsic attenuation ( \(Q_{\mathrm{i}}^{-1}=k^{-1}{\eta}_{\mathrm{i}}\) ), while for high frequencies ( f ≥ 3.0 Hz) \(Q_{\mathrm{i}}^{-1}\) dominates the attenuation process compared to \(Q_{\mathrm{s}}^{-1}\) ; and (6) at 1.0 Hz, \(Q_{\mathrm{c}}^{-1}\) lies in between \(Q_{\mathrm{i}}^{-1}\) and total attenuation, \(Q_{\mathrm{t}}^{-1}\) , but is closer to \(Q_{\mathrm{t}}^{-1}\) for frequencies higher than 1.0 Hz.

The coda attenuation of the Yao’an area in Yunnan Province

An earthquake withM=6.5 happened on January 15, 2000 in Yao’an of Yunnan Province. After the earthquake, a temporary digital network with 6 detectors around the epicenter area was set up. 402 aftershocks were located more precisely. According to coda short recording observed, the coda averaging quality factor has been acquired via Sato’s single scattering model analyses,Q c(f)=49f 0.95,f=1.5∼20.0 Hz, which has the attenuation characteristics of high structural active region.

Temporal change in coda wave attenuation observed at Volcán de Colima, México before the 1998 eruption

The last eruptive activity of Volcán de Colima (November 1998–January 1999) was preceded by the occurrence of several seismic swarms. An analysis of coda waves of local earthquakes recorded during these seismic swarms was made to estimate coda attenuation ( Q c −1) in the frequency range 2–10 Hz. We used Sato’s single scattering model [Sato, Phys. Earth Planet. Inter. 25 (1977) 27–41], beginning at twice the S-wave travel time. We found an increase in the attenuation of 34% at lower frecuencies (2 Hz) to 38% at the frecuency band around 10 Hz before the eruption. We also observed an increase of Q c −1 during the month which preceded a dome explosion on July 1998. A frequency dependence of Q c was observed with values from 76 at 2 Hz to 309 at 10 Hz. Q c values obtained at Volcán de Colima are similar to those obtained in other active areas. Temporal changes of Q c can be attributed to the change of density of the open fractures in the rocks because of the pressure generated by the pushing of magma toward the surface, which is also responsible for the inflation of the volcanic edifice.

Seismic coda attenuation after theMw= 6.2 Armenia (Colombia) earthquake of 25 January 1999

[1] Seismic wave attenuation in the central region of the Colombian Andes is studied using coda waves. Most of the events used occurred in the region on the occasion of the Armenia 25 January 1999 earthquake. The estimation of the decay rate of coda amplitudes (called coda Q−1 or Qc−1) is performed by means of the single isotropic scattering method of Sato [1977]. The attenuation parameters Qi−1 (intrinsic absorption), Qs−1 (scattering loss), and Qt−1 (total attenuation) are also estimated using the multiple lapse time window method of Hoshiba et al. [1991]. The frequencies of interest lie between 1 and 15 Hz. Results show that scattering attenuation is predominant for the frequency bands 1.5 ± 0.5 and 13.5 ± 1.5 Hz, whereas the intrinsic absorption and scattering contribute in equal shares to total attenuation for the 3 ± 1, 5 ± 1, 7.5 ± 1.5, and 10.5 ± 1.5 Hz frequencies. A comparison among the estimated attenuation parameters indicates that Qc−1 is close to total attenuation for all the studied frequency bands. On the other hand, no evidence in support of temporal changes of Qc−1 before and after the main shock has been found. Finally, the regionalization of Qc−1 values seems to correlate well with the geotectonic characteristics of the region.

Coda Q c Attenuation and Source Parameter Analysis in Friuli (NE Italy) and its Vicinity

— The digital data acquired by 16 short-period seismic stations of the Friuli-Venezia-Giulia seismic network for 56 earthquakes of magnitude 2.3–4.7 which occurred in and near NE Italy have been used to estimate the coda attenuation Q c and seismic source parameters. The entire area under study has been divided into five smaller regions, following a criterion of homogeneity in the geological characteristics and the constrains imposed by the distribution of available events. Standard IASPEI routines for coda Q c determination have been used for the analysis of attenuation in the different regions showing a marked anomaly in the values measured across the NE border between Friuli and Austria for Q 0 value. A large variation exists in the coda attenuation Q c for different regions, indicating the presence of great heterogeneities in the crust and upper mantle of the region. The mean value of Q c (f) increases from 154–203 at 1.5 Hz to 1947–2907 at 48 Hz frequency band with large standard deviation estimates.¶Using the same earthquake data, the seismic-moment, M 0, source radius, r and stress-drop, Δσ for 54 earthquakes have been estimated from P- and S-wave spectra using the Brune's seismic source model. The earthquakes with higher stress-drop (greater than 1 Kbar) occur at depths ranging from 8 to 14 km.

Attenuation of Coda Waves at the Tres Virgenes Volcanic Area, Baja California Sur, Mexico

We estimated the coda-wave attenuation quality factor Q c at the Tres Virgenes volcanic area from 26 local seismic events recorded by a portable network of seven digital stations. The quality factor Q c was estimated using the single-scattering attenuation model of Sato (1977), in the frequency range from 6 to 24 Hz. The magnitudes of the analyzed events range from 1 to 3. These events occurred at hypocentral depths ranging from 3 to 6 km and were recorded at epicentral distances of up to 30 km. Within the variation of Q c, six out of seven stations analyzed showed quite similar values at all frequencies. Data from these six stations led to an average Q c given by Q c( f ) = (50 ± 3) f (0.65±0.20) in the indicated frequency range. For the seventh station, which is located near a steam production well, anomalous low Q c values were obtained. For this particular site we found Q c( f ) = (4.3 ± 0.6) f (1.33±0.05). In this zone, we observed an abnormally strong reduction of Q c, leading to attenuation values that are three to four times larger than for the whole area of study. Results from other geophysical studies suggest the presence of a highly conductive body below this particular area. A correlation of the observed coda attenuation with a geothermal anomaly is suggested. Manuscript received 27 February 2000.

Regional features of coda attenuation of aftershocks of Wuding, Yunnan earthquake in near field

According to dispersion theory of coda of local events, using near-field data of aftershocks recorded by two mobile digital stations at the Tianxin and Yunlong after the Wuding, Yunnan M S=6.5 main-shock and sampling at different central frequencies (1.5–20 Hz), spatial distribution features of coda attenuation ratio β(f) are studied. The results show that there exists obvious non-uniformity of coda attenuation ratio β(f) given by the two stations in small region. When f<6.0 Hz the coda attenuation ratio β(f) observed on both sides of earthquake-generating fault are basically identical. Yunlong station is located on east side and Tianxin station on west side. The corresponding Q c(f) is in the range of 66–120. However, in the range of high frequency (f≥6.0 Hz) the coda in near field going through major fault encounters a strong absorption from the fracture zone. The β(f) going through the fault observed at Yunlong station is 30% lower than that in Tianxin station in which the seismic waves do not go through the fault. The reason for the difference in space distribution of coda attenuation is discussed. The results also show that factor A 0(f) of wave source is not only related to the source strength, but also to frequency f. The mean free path L of S wave obtained using 1.5–20 Hz is respectively 30 km on east side of the fault and 40 km on west side, from which it is verified that there is a stronger dispersion body when seismic waves go through the fracture zone.

コーダ波から推定されるリソスフェアのランダム不均質構造

Random inhomogeneities in the lithosphere have been investigated through the analysis of coda waves that are thought to be generated by the inhomogeneities. The power spectral density of upper crustal inhomogeneities measured directly using borehole logs obeys power law for a wide range of wavenumber, which is the characteristics of the exponential or the von Kármán autocorrelation function. The methods preferentially used for the coda analysis are phenomenological modeling based on the energy transport theory or stochastic simulation of wave propagation based on the Born approximation in random media. The parameters estimated from the methods are the scattering coefficient and coda attenuation (Qc-1) of scattering media or the fractional fluctuation of seismic wave velocity and the correlation length of random inhomogeneities. Although the frequency and lapse time dependence of Qc-1 has been widely recognized from the numerous works conducted in the last two decades, the physical meaning of Qc-1 and its lapse time dependence have not yet been clarified completely. Some recent works tried to understand it by considering layered structures with different values of scattering coefficient and intrinsic attenuation. The estimated range of fractional velocity fluctuation and correlation length is 1-8% and 0.15-1.0 [km] in the crust and 2-8% and 1-20 [km] in the upper mantle, respectively. The magnitude of fractional fluctuation shows little difference between the crust and the upper mantle, however, the correlation distance exhibits large difference that reaches one order of magnitude. In addition, the horizontal correlation length is much longer than the vertical one, forming the anisotropic inhomogeneities in the lithosphere. The recent observations of high-frequency teleseismic Pn gave an additional evidence of random inhomogeneities in the upper mantle. Thus the coda methods have great advantage to get stochastic insight into realistic image of the lithosphere.

CodaQin southwestern British Columbia, Canada

AbstractCoda decay rates of 122 locally recorded earthquakes are used to infer frequency-dependent coda Q values at 21 short-period stations in southwestern British Columbia, Canada. Using the single-scattering model, coda Q values (Qc) are derived for a range of sampling volumes corresponding to maximum lapse times of 36-59 sec and maximum sampling depths of 70-110 km. Maps of Q0 (Qc inferred at 1 Hz) for different sampling volumes show a consistent trend of increasing Q0 to the northeast away from the subduction zone and into the Coast Belt batholith, and decreasing Q0 toward the southeastern-most stations near the Vedder fault at the southern terminus of the Coast Belt. Q0 increases gradually with sampling volume at all stations. The dependence of Qc on frequencies between 2 and 16 Hz for different stations varies but on average ranges from 280 at 2 Hz to 850 at 16 Hz for sampling volumes that extend to depths of ∼90 km. An average of all data for the same sampling volume size gives a relationship of Qc = 110 f0.72 for frequencies in the 2- to 16-Hz range, which suggests significant differences in lithospheric coda attenuation properties in comparison with northwestern Washington State where Qc = 63 f0.97. Changes in coda Q correlate with the volume of brittle and fractured continental and subducting oceanic lithosphere sampled.

Intrinsic and scattering seismic wave attenuation in the Canary Islands

The Canary Islands volcanic complex is studied in terms of coda wave attenuation. A multiple lapse time window method, based on the hypothesis of multiple isotropic scattering with uniform distribution of scatterers, is applied to the available seismic data in order to obtain the intrinsic absorption (Qi−1) and the scattering attenuation (Qs−1) in the Canarian lithosphere. The analysis is performed for two hypocentral distance ranges: from 0 to 80 km and from 0 to 140 km. Results show that in both cases and for all the studied frequency bands (1–2, 2–4,6–8 and 8–10 Hz ) intrinsic absorption dominates. The low albedos found in the region indicate the low degree of heterogeneity in the Canarian lithosphere at the scale length of the studied frequencies. On the other hand the coda attenuation (Qc−1) calculated on the basis of the single‐scattering theory gives values near (Qi−1) for low frequencies and near the total attenuation (Qt−1) for high frequencies. The degree of frequency dependence of the attenuation parameters is strong in all cases. A correlation of the observed attenuation parameters with the geological evidence for a hotspot‐type archipelago and with other geophysical studies is suggested.

Attenuation of coda waves at the Cerro Prieto geothermal field, Baja California, Mexico

AbstractWe analyzed coda waves from 22 local events recorded by a temporary network of seismic stations deployed in the Cerro Prieto geothermal field of Baja California, Mexico, to estimate coda attenuation Qc in the frequency range 6 to 24 Hz. We used Sato's (1977) single scattering model for a coda window of 12.5 ± 2.5 sec beginning at twice the S-wave travel time. All events analyzed were located within the exploitation area with depths up to 6 km and source-to-receiver distances less than 8 km. At frequencies of 3, 4, and 5 Hz, rms amplitudes of coda to S waves showed no clear decay or even negative Qc values. This could be due to the inability of the theory to fit the data at low frequencies as it was suggested by Fehler et al. (1992). Within the range of variation, five of the seven stations showed similar Qc values that ranged from about 200 at 6 Hz to about 490 at 24 Hz. Assuming a power-law dependence of the form Qc (ƒ) = Q0ƒη, we estimated a value of Q0 = 111.5 ± 23 and a frequency dependence η of 0.41 ± 0.1. We found site effects at two stations that were located close to hot springs and mud pools.

Dependence of coda attenuation on frequency and lapse time in central Greece

The dependence of coda attenuationQc on frequency and lapse time was studied. Data from small local earthquakes, recorded at three stations (VMR, VSI and VFI) of the VOLNET network operating in central Greece, were used.Qc was estimated by applying the single scattering model to bandpass-filtered seismograms, over a frequency range of 1 to 12 Hz. Analysis was performed every 10 s until the end on overlapping time windows.Qc is found to depend on frequencyf in Hz according to a power law,Qc=Q0fn. ObservedQ0 ranges from 30 to 100 and the powern ranges from 0.90 to 0.70.Q0 increases andn decreases with lapse time increasing. A strong dependence ofQc on lapse time was also found. In the frequency range of 1 to 8 Hz and at a short lapse time,Qc values were found to be similar for all three stations. On the other hand, at the longest analyzed time window (50 s), the estimatedQc values show a discrepancy which is more obvious at a higher frequency band. The scattering coefficient around the central station VSI is found to range from 0.029 to 0.0041 km−1.Qc from the single scattering model andQs from the amplitude ratio of directS to coda waves for the VSI station are similar. We believe dependence of coda attenuationQc on frequency and lapse time is caused by a combination of geotectonic features and depth variation asQs.

Effect of finite thickness of scattering layer on coda Q of local earthquakes

Theories based on the assumption of an infinite medium show that the coda attenuation factor, Q c −1, of local earthquakes should be very close to the intrinsic Q i −1, whereas many observations show that Q c −1 is close to the total Q −( Q t −1). This paper shows by numerical modeling that Q c −1 depends on the thickness of the scattering layer. For source-receiver coincidence, if the thickness is greater than a half of the mean free path, Q c −1 lies between Q i −1 and Q t −1. However, when the thickness is less, Q c −1 may be higher than Q t −1. In addition, Q c −1 depends on the epicentral distance. The thinner the layer, the faster the change with distance in Q c −1. For an unbounded scattering medium Q c −1 is even lower than Q i −1 at long epicentral distances.

Stability of coda wave attenuation during the Loma Prieta, California, earthquake sequence

The Loma Prieta, California, earthquake occurred in a densely instrumented region with a history of microearthquake recording beginning more than a decade before the October 1989 mainshock. This affords an unprecedented opportunity to detect changes in seismic wave propagation in the Earth's crust associated with a major earthquake. In this study we use pairs of nearly identical earthquakes (doublets) to search for temporal changes of coda attenuation in the vicinity of the Loma Prieta earthquake. We analyze 21 earthquake doublets recorded from 1978 to 1991 that span the preseismic, coseismic, and postseismic intervals and measure the change in coda Q using a running window ratio of the doublet spectral amplitudes in three frequency bands from 2 to 15 Hz. This method provides an estimate of changes in coda Q that is insensitive to other factors that influence coda amplitudes. Our observations place an upper bound of about 5% on preseismic, coseismic, and postseismic changes of coda Q in the epicentral region of the Loma Prieta earthquake. Even at this low level, the changes are neither spatially coherent nor correlated between adjacent frequency bands. The only hint of a signal is in the preseismic data where there is a possible precursory increase in coda Q of approximately 5% in the two years before the mainshock. The stability of coda Q throughout the Loma Prieta sequence is in sharp contrast to other studies that have reported much larger precursory changes in coda Q for other earthquakes.

Coda attenuation in the area of El Chichón volcano, Chiapas, Mexico

Coda-wave attenuation is investigated using the single scattering approximation on analog records from stations of a telemetered network situated close to El Chichón volcano in southern Mexico, as well as from temporary stations deployed during 1982. The single scattering model was applied to the envelope of coda waves to characterize the coda decay rate. We obtained values of Q c which, according to the filter settings of the seismographs, and the average period measured in the traces, correspond to frequencies of about 1.3 ± 0.3 Hz. Results indicate apparently larger attenuation than observed with similar techniques in other regions of Mexico, including the Mexican Volcanic Belt. Within the limitation imposed by the data, the spatial distribution of coda Q with respect to the volcanic edifice was investigated. For paths sampling the near vicinity of the volcano a value of approximately Q c = 30 was obtained. The highest attenuation results are obtained for volumes with a slight E-W tendency, with Q c values of 28 and 25, respectively. The location of a gravity high as well as microseismic activity correlates well with this result, which is interpreted as due to the probable presence of an intrusive magma body. Lower attenuation is obtained for volumes sampling deeper than the Moho, reaching a value of Q c ≈ 87, suggesting that rocks below this boundary play a major role in controlling the overall attenuation of paths traversed by coda waves. The results also suggest a decrease in attenuation towards the south and north possibly reflecting a decrease in magmatic processes in both of these directions.

Abnormal Lg coda Q in Beijing Area

This paper presents the results of an Lg coda Q study in the Beijing area of China. The data originate from the Beijing Telemetered Network; the Q (=Q0ƒη) measurement technique is the stacked spectral ratio method (SSR). The study area, roughly 250 km in radius, is sampled repeatedly by scattered coda waves recorded at epicentral distances between 190 and 422 km. The results point to 1‐Hz Q (Q0) as low as 150 and η as high as 0.95, in a jarring incongruity with the ambient attenuation characteristics about the study region. The Lg coda Q0 is found varying inversely with η. The unequivocality with which such a relationship becomes unravelled comes about somewhat unexpectedly given the small spatial confine of the study area. If high coda attenuation (low Q0) and high η are taken to correlate with a high level of tectonic activity in a broad sense, a view generally held to be statistically accurate, then the Beijing area belongs to the same league in Eurasia, whose members include such well known “earthquake countries” as the Himalayas, the Tibetan Plateau, the Burmese Arc, and the Tianshan Mountains.

Separation of scattering attenuation and intrinsic absorption in Japan using the multiple lapse time window analysis of full seismogram envelope

Two seismic wave attenuation factors, scattering attenuation (Qs−1) and intrinsic absorption (Qi−1), can be determined from estimates of two parameters, the reciprocal of extinction length Le−1 ≡ (Qs−1 + Qi−1) ·ω/v and seismic albedo B0≡ Qs−1 / (Qs−1 + Qi−1), where ω and v are angular frequency and seismic wave velocity, respectively. These parameters are measured using the multiple lapse time window analysis method over frequency bands of 1–2, 2–4, and 4–8 Hz from 16 stations located throughout Japan. In the analysis, time‐integrated seismic wave energy from earthquakes having hypocentral distances less than 120 km and focal depths less than 40 km are compared with simulations of multiple isotropic scattering based on the assumption of spatial uniformity of Qs−1 and Qi−1. The modified coda normalization method is used to correct for sources and site effects, where the spatially non uniform distribution of the coda wave energy is considered. The Le−1 estimates do not have a distinct frequency dependence, but the Be estimates do: 0.2–0.7 for 1–2 Hz and 0.1–0.3 for 4–8 Hz. The results mean that Qs−1 is similar to Qi−1 for 1–2 Hz but is 3–5 times smaller for 4–8 Hz and that the frequency dependence of Qs−1 is larger than Qi−1. Analyses using different components of motion and different window lengths give nearly the same results. Coda attenuation (Qc−1) estimates based on the single scattering approximation are nearly equal to intrinsic absorption Qi−1 rather than to total attenuation (Qs−1 + Qi−1) for Japan.