Field Study of a Highly Active Fault Zone: the Xianshuihe Fault of Southwestern China

Field study of a highly active fault zone: The Xianshuihe fault of southwestern China

CLARENCE R. ALLEN Seismological Laboratory, California Institute of Technology, Pasadena, California 91125 LUO ZHUOLI \ QIAN HONG > Seismological Bureau of Sichuan Province, Chengdu, China WEN XUEZE ) ZHOU HUA WEI Department of Geosciences, University of Houston, Houston, Texas 77004 HUANG WEISHI Seismological Laboratory, California Institute of Technology, Pasadena, California 91125

ABSTRACT ord, repeat times estimated from slip rates, climatic environments? Added impetus was and current seismic gaps, two segments are given to the study because the Xianshuihe fault The Xianshuihe fault of western Sichuan particularly likely sites for M = 7+ earth- is the only major fault worldwide, other than the Province, China, is one of the world's most quakes in the near future: the 65-km-long San Andreas, where continuing fault creep has active faults, having produced 4 earthquakes segment between Daofu and Qianning, and been reported over significant lengths of the during this century of magnitude >1 along a the 135-km-long segment bracketing Kang- fault. 350-km length of the fault. At least 8 such ding. Continuing creep has been documented The Xianshuihe fault zone is part of a much events have occurred since 1725. In the more along some segments of the fault, and this, more extensive left-lateral fault system of at least limited 150-km-Dong segment including Lu- together with the high degree of activity and 1,400-km length, extending from southern Yun- huo and Daofu, major earthquakes in 1904, other unique attributes, makes the Xian- nan Province northwest through Sichuan into 1923, 1973, and 1981 (M = 7, IVi, 7.6, 6.9) shuihe fault one of the most promising sites in Qinghai (Fig. 1, insert) (Ding, 1984). Almost all were associated with overlapping surficial the world for earthquake prediction and segments of the system have been the locus of fault ruptures and with individual left-lateral hazard-evaluation studies. major earthquakes within the historic record displacements as large as 3.6 m. Field studies (Fig. 2), and it currently constitutes probably the indicate that this high degree of activity is INTRODUCTION most active fault system within China. It clearly typical of the fault's longer-term history. The deserves comparison with other highly active Holocene left-lateral slip rate on the north- The purpose of this study was to examine in strike-slip fault systems worldwide, such as the western segment of the fault has been 15 ± 5 the field a highly active fault zone—one of the San Andreas fault of California, the North mm/yr, decreasing to about 5 mm/yr on its most active in the world during this century. Anatolian fault of Turkey, and the Alpine fault southeastern segment, based on radiometri- During the present century, four earthquakes of of New Zealand. cally dated offset stream-channel and terrace magnitude 7 or greater have occurred along a Following recent Chinese practice, we differ- deposits and on offset glacial moraines. 350-km segment of the Xianshuihe fault, and at entiate herein between the Xianshuihe fault Physiographic features of active faulting least eight such events have occurred here since zone, extending some 350 km northwest from are fully as diagrammatic as those of Califor- 1725 (Figs. 1, 2; Table 1). Four events exceed- near Shimian to Kasu (Fig. 1), and the more nia's San Andreas fault, mainly because of ing magnitude 6.8 have occurred during this restricted Xianshuihe fault itself, which is only high-altitude preservation and the absence of century along a single 150-km segment of the one of five individual segments within the over- cultural modification on this eastern margin fault, with overlapping surface ruptures. At the all zone. These five segments (Fig. 3) are the of the Tibetan Plateau. Detailed en echelon commencement of the field work, we had sev- Moxi fault, Selaha (Kangding) fault, Zheduo- tensional and pushup features resulting from eral questions in mind. (1) Could the very high tang fault, Yalahe fault, and the restricted Xian- surface ruptures in 19*73, 1955, 1923, and degree of activity have been recognized by phys- shuihe fault, extending northwest from Laoqian- 1893 can still be recognized today, and new iographic features alone, even in the absence of ning. The nature and significance of the data have been collected bearing on the the historic record? (2) Is the high level of cur- segmentation between these different members offsets and fault-rupture lengths during these rent activity representative of the late Quater- are important questions that will be addressed in and other events. nary history, or is it merely a temporal burst of a section below. The type area of the restricted The locations and magnitudes of historic activity such as that along the North Anatolian Xianshuihe fault is the 150-km-long segment be- earthquakes suggest that the characteristic fault of Turkey since 1939? (3) Can the source tween Songlinkou Pass and Kasu, where its earthquake model may apply to the Xian- areas of recent earthquakes be related to seg- trace is relatively simple and continuous, and shuihe fault. Obvious geometric segmentation mentation of the fault? (4) In this high-altitude, where it controls the course of the Xianshui of the fault has controlled the initiation and alpine environment along the eastern margin of River and its tributaries. termination of ruptures in some events, the Tibetan Plateau, how are fault features pre- At its northwestern end, the Xianshuihe fault whereas segmentation control for others served as compared with those along other ac- overlaps the southeastern end of the Ganzi- remains obscure. Based on the historic rec- tive faults in China and elsewhere in more usual Yushu fault, with a left en echelon stepover of

Geological Society of America Bulletin, v. 103, p. 1178-1199, 29 figs., 3 tables, September 1991.

1178

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 Figure 1. Map of Xianshuihe fault in area of this study. Boxes within map indicate areas of more detailed sketch maps herein. Inset at lower left shows location of study area and regional relationship between Xianshuihe fault and some other major active faults of China and adjacent areas. All elevations are in meters.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1180 ALLEN AND OTHERS

1960 6| M^ol Figure 2. Map showing epicenters of earthquakes of magnitude (Mj) 6% and greater in the Chinese 7.2 1976 catalog (Luo, 1980), with relation to major faults in the Sichuan-Yunnan region. Box shows location of Figure 1, the area of this study. 1933

about 40 km. At its southeastern extremity, the Xianshuihe fault zone connects with the An- ninghe fault in a complex and little-studied area near Shimian. Still farther southeast, a significant en echelon offset occurs between the An- ninghe and Xiaojiang faults (Fig. 2), with the very active Zemuhe fault in the transition area. All of these individual fault zones, on the other hand, must be considered parts of the same overall left-lateral fault system. The entire system lies along the northeast boundary of a large rhomb-shaped block, or miniplate, that is bounded on the opposite southwest side by the right-lateral Red River fault system (Allen and others, 1984), and the apparent southeastward movement of this relatively stable block relative to its neighbors has been the subject of much tectonic discussion (for example, Li and Wang, 1975; Kan and others, 1977). It is admittedly somewhat surprising that two major fault systems that are almost parallel and only 200 km apart should have opposing current senses of lateral slip, although this has been explained, at least in principle, by the mechanics of the im- pingement of India into Eurasia and the conse- quent block rotations (Molnar and Tapponnier, 1975; Tapponnier and Molnar, 1976; Tapponn- ier and others, 1986).

HISTORIC EARTHQUAKES AND SEISMICITY

Regional Activity. Figure 2 shows historic large earthquakes on the Xianshuihe and associated faults of southwestern China. It should be noted that the historical record in this part of China, which is very sparsely populated (primarily by Tibetan people), is very short as compared to those of the more heavily populated areas to

TABLE 1. HISTORIC EARTHQUAKES OF M i 6.9 ON THE XIANSHUIHE FAULT ZONE the north, east, and south; the record of large events on the Ganzi-Yushu fault system to the

Date Location Magnitude Fault Rupture Maximum west is even more limited (Wen and others, length displacement (km) (m) 1985). The earliest large earthquake of record on the Xianshuihe fault zone itself is that of 1725 N. Kangding >7 Selaba? 1 1 1725, although some large earthquakes that oc- 1786 S. Kangdiog 1* Moxi-Selaha >70 ? 1816 Luhuo >n Xianshuihe 1 7 curred as early as 1515 A.D. are documented 1893 Qiaoning >7 Xianshuihe >40 ? 1904 Daofu 7 Xianshuihe ? 7 along southern extensions of the fault system in 1923 Renda 714 Xianshuihe >60 3± Yunnan (Fig. 2). The overall pattern, nonethe- 1955 Kangding n Zheduotang 27 1- 1973 Luhuo 7.6 Xianshuihe 90 3.6 less, clearly indicates that major earthquakes 1981 Daofti 6.9 Xianshuihe 44 1- have tended to occur on the major faults of the

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 Figure 3. Sketch map of Xianshuihe fault zone in Laoqianning-Kangding area. Heavy solid line indicates fault where clearly represented by Holocene scarps. Heavy dotted line indicates segments where Holocene displacements are probably continuous but are not obvious on aerial photographs owing to active river erosion or other causes. Dashed lines are major roads. Localities A-D are places where fault offsets of latest Pleistocene glacial raines have been measured (see text).

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1182 ALLEN AND OTHERS

Figure 4. Map of ground ruptures associated with the M = 7.6 Luhuo earthquake of 1973 (heavy solid line), the M=lVi "Sharato" earthquake of 1923 (dashed line), and the M= 6.9 Daofu earthquake of 1981 (dotted line). Note overlaps. Numbers show left slip (in meters) in 1973, from Tang and others (1976). 1923 data are partly from Heim (1934), and 1981 data are from Tang and others (1984b). For reference, see town locations in Figure 1.

region, and that the Xianshuihe fault zone is Xialatuo-Luhuo area and appears to have been earthquake, and it has been identified as a tem- currently the most active of these. If Figure 2 remarkably similar in both location and magni- poral seismic gap (Han and Huang, 1983; Lin were to be more homogeneous in its temporal tude to the 1973 event, although perhaps cen- and others, 1986). coverage, the Xianshuihe fault zone might well tered slightly farther southeast. Features related 1904 M = 7 Earthquake near Daofu. This appear relatively even more active. to this event have been found in trenches exca- earthquake, previously thought to be of magni- A number of investigators have discussed the vated across the Xianshuihe fault at Xialatuo tude 6, has recently been reinvestigated and as- temporal and spatial pattern of historic earth- (Huang and others, 1982). signed the higher magnitude of 7 (Huang, 1985), quakes along the Xianshuihe fault zone (for ex- 1893 M ^ 7 Earthquake near Qianning. partly on the basis of Gutenberg's unpublished ample, Huang, 1981, 1983; Tang and others, Little is known about the 1893 earthquake ex- notes filed at Pasadena. Recently re-evaluated 1984a; Qian, 1986; Lin and others, 1986; Wen, cept that it caused, major damage at Huiyuan, Milne seismograph records now indicate a mag- 1986), and herein we will only briefly summa- Jinlong, and Juluo Monasteries (Figs. 3,4), thus nitude of 6.9 (Abe, 1988). The intensity and felt rize the more important events. Unless otherwise suggesting that it was centered along the south- area of the 1904 event seem to have been similar stated, all data are from the catalogs of historic ernmost segment of the restricted Xianshuihe to those of the 1981 Daofu earthquake, and one Sichuan earthquakes (Luo, 1980; Gu, 1983). fault. Features of very recent displacement are very old resident of Goupu (Fig. 4) reported to Some of our magnitude assignments are differ- abundantly clear throughout this segment. Indis- us in 1986 that the Xianshuihe fault ruptured ent from those in the catalogs, based on more tinct right-stepping en echelon cracks could still along the same line in both 1904 and 1981. recent information available to the authors. be identified in 1986 along the fault trace about 1923 M= IVi Earthquake near Renda. The Table 1 summarizes these data. 4 km northwest of Laoqianning (Fig. 5), and by faulting associated with the 1923 earthquake 1725 Mai Earthquake North of Kang- comparison with cracks known to have been was documented by Heim (1934), although he ding. Sparse historical records of the 1725 formed farther northwest in 1923 (see below), surprisingly failed to recognize its strike-slip na- earthquake suggest that it was centered some- formation in 1893 is probable. A trench exca- ture. His 1930 photograph taken at Changcu where between Kangding and Huiyuan Monas- vated across the fault at Songlinkou Pass (Fig. 1) Pass (Fig. 6A) shows clear right-stepping en tery (Fig. 3). The most probable source is the likewise revealed offset features probably of echelon cracks and intervening pressure ridges, northern segment of the Selaha (Kangding) 1893 origin. We think that the fault ruptured and these same features are almost as visible fault. from southeast of Laoqianning to northwest of today at this locality as they were in 1930 (Fig. 1786 M = 7% Earthquake South of Kang- Songlinkou Pass, a distance of at least 40 km. It 6B). The length and spacing of the cracks sug- ding. The 1786 earthquake was clearly a very is noteworthy that this segment of the Laoqian- gests left slip of at least a meter. A few hundred large event, causing minor damage even as far ning-Kasu sector of the Xianshuihe fault is cur- meters southeast of the photo locality, a cultural away as Chengdu, more than 200 km from rently the longest-lasting one without a major berm or retaining wall crossing the fault trace at Kangding. Major destruction occurred both at Kangding and Moxi, suggesting that the source area was centered near Xuemenkan Pass (Fig. 3). An earthquake-induced landslide dammed the Dadu River (Fig. 1) for 10 days, and the Figure 5. View north- subsequent breaching killed thousands of people west of fault saddle 4 km downstream (Li and others, 1986). Wang Xin- northwest of Laoqiann- min (1986, personal commun.) has recently ing. Subtle fractures on identified detailed features of very recent dis- face of scarp beyond indi- placement which indicate that the surficial fault vidual (center) reflect left- rupture extended for at least 70 km from south lateral displacement, pre- of Moxi to northwest of Kangding. sumably associated with 1816 Ai> IVi Earthquake near Luhuo. The 1893 earthquake. 1816 earthquake was most damaging in the

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1183

Figure 6A. Photograph taken in 1930 by Albert Heim (1934), showing trace of 1923 fault rupture. View is southeast from near summit of Changcu Pass (Fig. 4). Note alternating en echelon fractures and pressure ridges, not recognized by Heim as indicative of left-lateral displacement. Also note sag pond (left).

right angles is clearly offset 3 m, although it is 1984). There can be no question, on the other much as 14 m long, at 35° to the overall fault not known whether this offset results solely from hand, but that the rupture of the 1923 event trend, and they appear to have been as much as the 1923 event. overlapped with that of the subsequent 1973 V2 m wide. Although the displacement in 1955 is Heim's map shows the rupture extending as event by at least 10 km (Fig. 4). not documented, we estimate that it must have far northwest as Xialatuo, but the fault leaves 1955 M = IVi Kangding Earthquake. The been well in excess of 1 m to produce features the valley floor (and main trail) at this point and 1955 earthquake was associated with clear this impressive. Of particular interest, however, may not have been searched for by Heim and his left-lateral surface rupture along the Zheduotang is the fault length. Even if the fault broke in local informants. Strangely, Heim concentrated fault, which is one of several members of the 1955 along its entire length of 27 km, as identi- his attention not on the principal fault within the Xianshuihe fault system in the Laoqianning- fied on aerial photographs, this is a remarkably valley, but instead on the many linear ravines on Kangding sector (Fig. 3). The fault is crossed in short rupture length for a strike-slip earthquake the adjacent mountain sides, which he thought several places by the Kangding-Lhasa highway of magnitude IVi, inasmuch as earthquakes of were eroded "earthquake cracks." We doubt his (Fig. 8), and the 1955 fractures are still easily this magnitude worldwide have been associated interpretation. Southeast of Changcu Pass, visible at the top of Zheduoshan (Cheto La) with ruptures averaging 85 km in length (Bonilla Heim's map is not clear with regard to possible Pass. Here, individual en echelon fissuresar e as and others, 1984). Seismological studies of the 1923 fault rupture, and his traverse was along the main trail on the opposite (north) side of the river from that of the fault trace in this area. We identified en echelon cracks similar to those at the Changcu Pass 20 km farther southeast near Mazi (A, Fig. 7). At the same place, the edge of a cultivated field is left laterally offset about 2 m along the line of the cracks, presumably resulting from the 1923 event. We conclude that the 1923 earthquake was associated with a maximum left slip of about 3 m, which is consistent with strike- slip earthquakes of this magnitude (Bonilla and others, 1984), and that the rupture extended at least 36 km from Xialatuo to Mazi. Because the displacement was 2 m at Mazi, however, the rupture must have extended a considerable distance still farther southeast, and it is not clear that anyone at that time looked for the fault trace either here or northwest of Xialatuo. Figure 7. Sketch map of fault-controlled drainage patterns near Mazi (Fig. 1). Note consist- Therefore, we estimate that the rupture was at ent left-lateral offsets. "X" indicates windgap along a formerly offset stream channel. At least 60 km in total length, which is again con- locality A, ±2-m offset of edge of a cultivated field presumably results from 1923 earthquake sistent with faulting associated with other earth- (M = 714), and poorly preserved en echelon cracks of the 1923 event were still visible nearby in quakes of this magnitude (Bonilla and others, 1986. See Figure 18 for air photo of this area.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1184 ALLEN AND OTHERS

nitude 7.9, recent re-evaluation by Chinese seismologists has placed the magnitude at 7.6. Many of the surface features of faulting associated with the Luhuo earthquake are almost as fresh today as they were in 1973 (Figs. 9,10,11, 12). In many places, cultural features such as field edges can be observed to be offset by amounts consistent with those observed locally in 1973 by Tang and others (1976). At the location of maximum offset near Dandu, however, the edge of a cultivated field several hundred meters southeast of the main road is clearly offset about 7.2 m, twice the 1973 offset of 3.6 m that was measured closely nearby. We can only conclude that this feature preserves the displacement of one or more earlier earthquakes, most likely that of the 1816 event, which was in other ways very similar to that of 1973. Post- 1973 creep cannot explain an offset this large. Because 1/30,000 aerial photographs of the Figure 8. View southeast along Zheduo- fault were taken only two weeks following the tang fault (arrow). Scar]) results primarily 1973 earthquake, some of the details of the fault from left-lateral displacement, most recently rupture are easily discernible. For example, the occurring here in 1955 Kaingding earthquake. fault generally broke into distinct en echelon Figure 9. View southeast of fault scarp segments that averaged a few hundred meters in near Yusi (Fig. 13) formed in association with length, and in some places the stepover between 1973 Luhuo earthquake, as viewed in 1986. 1955 earthquake confirm left slip on a north- segments was as much as Vi km (Fig. 13). At any west-trending fault (Lin and others, 1986). given point along the fault, however, there was 1973 M= 7.6 Luhuo Earthquake. Geologic typically only one fracture zone; an exception elucidated by Deng and Zhang (1984), Qian features related to the 1973 Luhuo earthquake was southeast of Dandu, where three parallel (1983), and others. In only a few places was the were remarkably well described by Tang and traces can be seen over a width of about 200 m. displacement concentrated along a simple single others (1976). The surface rupture extended Individual fracture zones themselves almost in- shear surface such as has been observed more northwest from near Renda to Kasu (Fig. 4), a variably consisted of smaller en echelon frac- often in other strike-slip earthquakes such as that distance of about 90 km, and it clearly over- tures averaging a few meters in length, separated of 1906 in California (Lawson, 1908). Probably lapped the rupture of 1923 by at least 10 km at by distinct pressure ridges at right angles, in the a significant contributing cause to this difference the southeast end. The maximum left-lateral classic manner described by Richter (1958) and was the fact that the 1973 earthquake occurred offset of 3.6 m was measured at Dandu, about 20 km northwest of the epicenter near Xuxu (Fig. 4). Vertical displacements were generally minor, but usually up on the southwest and in- variably in keeping with earlier vertical compo- nents of displacement Worldwide strike-slip earthquakes of Ms = 7.6 have typically been associated with rupture lengths of 100 km and maximum displacements of 3 to 4 m (Bonilla and others, 1984). Seismological studies of the Luhuo earthquake are remarkably consistent with the geological observations (Zhou and others, 1983a, 1983b). The source mechanism indicates almost pure left slip on a vertical fault striking N55°W, and the source-time function suggests greater slip northwest of the epicenter than to the southeast. The slip at depth extended some 15 km farther northwest than did the surface rupture, and this is consistent with the aftershock distribution. The total moment was 1.9 x 1027 dyne- cm, and the USGS magnitude (Ms) was 7.5. Figure 10. Same scarp as in Figure 9, view northeast. Note almost pure left-lateral displace- Although listed in the Chinese earthquake cata- ment. Top of ridge is offset 8 m, representing movement during previous earthquakes as well as logs (for example, Luo, 1980) as being of mag- that in 1973.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1185

quake and the previous large earthquake here in migrate with time along the length of the fault, 1816 occurred in mid-winter, thus serving to with different segments of the fault being active amplify and preserve the detailed geomorphic at different times (Tang and others, 1976; Han features of rupture along the fault. and others, 1985). Particularly striking is the re- 1981 M= 6.9 Daofu Earthquake. The most cent migration of events from the northwest end recent significant earthquake on the Xianshuihe of the fault toward the southeast, starting with fault was that of 1981 centered near Daofu, the 1967 Zhuwo earthquake, followed by the which suffered considerable damage. 1973 Luhuo earthquake, and then the 1981 (China) was 6.9, andA/s(USGS) was 6.8. Zhou Daofu earthquake (Fig. 2). The next segment in and others (1983b) determined from surface- sequence would be that from Songlinkou Pass wave inversion that the earthquake was caused southeast to near Qianning (Fig. 1), which is by pure left-lateral displacement on a vertical another reason, in addition to the aforemen- fault striking N41°W, which agrees well with tioned seismic gap, for considering this segment the local trend of the fault as observed in the particularly dangerous at the present time. Judg- field. They calculated a moment of 1.3 x 1026 ing from its length, a gap-filling earthquake dyne-cm. Peter Molnar (1990, personal com- along this fault segment would be of magnitude mun.) has more recently argued for a moment of greater than 7 (Bonilla and others, 1984). 25 6.2 x 10 dyne-cm. Surface ruptures were Characteristic Earthquakes? The fact that a much less obvious for this earthquake than for number of large earthquakes along the Xian- the 1923 and 1973 events, which is consistent shuihe fault appear to have been remarkably with its lower magnitude. Scattered fissures oc- similar to earlier events lends support to the curred in many places in the alluviated valley concept of the characteristic earthquake. This floor near Daofu, but the most consistent en hypothesis holds that a given segment of a fault Figure 11. View southeast of 1973 fault echelon cracks occurred squarely along the trace may break repeatedly with earthquakes of essen- rupture near Yusi (Fig. 13). Fissures were still of the Xianshuihe fault near Goupu (Fig. 4) and tially the same size and fault parameters, perhaps open in 1986. clearly indicated left-lateral displacement (Deng controlled by asperities and barriers on the fault and Zhang, 1984). En echelon cracks were ob- plane (Aki, 1984; Schwartz and Coppersmith, served intermittently along the fault from Song- 1984). Thus the 1816 and 1973 events along the linkou Pass northwest to Mazi, a distance of in mid-winter, and the ground was frozen to a Xianshuihe fault are perhaps characteristic —44 km (Tang and others, 1984b). depth of several tens of centimeters. Conse- earthquakes, as well as the 1904 and 1981 quently, the soil behaved in brittle fashion and in Changing Seismicity with Time. Since events. The projected earthquake on the Song- some areas tended to break into massive tilted 1700, two periods of active seismicity (1700- linkou Pass-Qianning segment would presum- blocks (Fig. 14). Indeed, some of the resulting 1816, 1893-present) have been separated by a ably be similar to that which was associated local micro-physiography along the fault trace is period of marked quiescence, and Jiang (1985) with rupture of the same segment in 1893. truly spectacular and is quite unlike that seen demonstrated that no earthquakes of M>5 oc- Current Seismicity. Figure 15 shows current along most other major strike-slip faults. It is curred during the quiescent period. During the seismicity along the Xianshuihe and nearby interesting to note that both the 1973 earth- active periods, moreover, earthquakes tended to faults. Although there is a tendency for small earthquakes to occur along or close to the major active faults, there is also considerable diffuse seismic activity throughout the region. This pattern is similar to that of southern California, where one must be careful in drawing conclusions about large earthquakes on the basis of the locations of smaller ones (Allen and others, 1965). Although almost all events exceeding magnitude 6 in both areas have occurred on the major active faults, the same is clearly not true for most of the smaller earthquakes.

GEOLOGIC ENVIRONMENT AND PREVIOUS STUDIES Jf

The Xianshuihe fault was comparatively late in being recognized and studied because of its remote location in the mountainous Tibetan region of northwestern western Sichuan (formerly Sikang) Province. The Qianning-Luhuo segment was first shown on the reconnaissance geologic map of Tan and Li (1930), in connection with regional studies of mineral deposits. Atten-

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1186 ALLEN AND OTHERS

Figure 13. Sketch map of Xianshuihe fault near Luhuo. Aligned en echelon symbols indicate locations of conspicuous en echelon cracks associated with the 1973 Luhuo earthquake (M = 7.6) that are visible on 1:30,000 aerial photographs taken two weeks after the earthquake; most of these cracks could still be readily identified in 1986. Dotted line indicates crest of broad side-hill ridge, which is associated with pronounced rift topography along the fault zone. Sag ponds are indicated by "SP". Physiographic features of recent faulting are much better preserved here than in adjacent segments where major river courses are eroding directly within the fault zone.

tion was first focused on the fault itself by Albert the "Kangting fault" by S. P. Lee (1948). Still nese geologists believe that the Xianshuihe fault Heim, while carrying out geographical explora- later, it was termed the "Luhuo-Daofu fault" by first formed as a tensile fracture during late Pa- tions in the region in 1930 (Heim, 1933, 1934). the Geological Institute of the Ministry of Geol- leozoic time and controlled the emplacement of Local residents in the Daofu-Xialatuo area ogy, and the present name was given in the late igneous plutons during the Indosinian (Late Tri- pointed out to Heim ground cracks that had 1960s by the former Seismogeological Brigade assic) and the Yanshanian (Jurassic-Cretaceous) been associated with the large earthquake there of Southwestern China. Since the time of the orogenies. Regional metamorphism took place in 1923, as well as older cracks associated with 1973 Luhuo earthquake, the Xianshuihe fault during these important and widespread diastro- earlier large earthquakes. Although Heim failed has come under intense scrutiny by the Sichuan phic episodes, and the bedrock along the fault to recognize the strike-slip character of the 1923 Seismological Bureau and by many investiga- north of Qianning is constituted primarily of and earlier displacements, his map is generally tors, especially by systematic seismogeologic quartzites, marbles, slates, and phyllites of the correct in delineating a long fault zone between mapping which commenced in the early 1980s. Late Triassic Xikang group; south of Qianning, Sharato (now Xialaltuo) and Tailing (now Lao- The bedrock of the region is made up primar- Mesozoic and Precambrian granites predomi- qianning). After extensions farther south were ily of basinal rocks of Triassic age, which rest in nate. No Jurassic, Cretaceous, or Tertiary rocks recognized, such as near Kangding by C. S. Lee turn on Paleozoic and late Precambrian rocks have been found in the region, and Quaternary and others (1938), the entire zone was named (Yang and others, 1986) (Fig. 16). Some Chi- deposits are limited to small basins (particularly along the fault), in addition to local terrace, eo- lian, and glacial deposits. Some authors have suggested that an anticlinorium involving Trias- sic and older rocks exposed north of the fault between Luhuo and Donggu has been offset left laterally about 50 km to a location south of the fault between Luhuo and Daofu (Fig. 16), but this has yet to be confirmed. Rocks along the fault zone are intensely sheared, with many tight folds and abundant mylonite and gouge. The crushed zone, which is easily erodible and thus controls the drainage pattern of the region, is about 440 m wide where exposed near Luhuo, and several tens of meters wide near Kangding (Li and others, 1985).

DESCRIPTION OF THE FAULT TRACE

General Statement Figure 14. Photograph taken immediately following 1973 earthquake of push-ups and ten- sion cracks along fault near Gelu. Thickness of blocks was partially controlled by depth of Physiographic features of recent faulting are frozen ground. View northwest. Photo by Deng Tiangang. as impressive along the Xianshuihe fault as for

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1187

any major strike-slip fault in the world, partly because of its relatively high slip rate, and partly because of superb exposures resulting from the absence of dense vegetative cover, minimal cultural modification, and the high-altitude climate. Diagrammatic features of recent displacement such as sag ponds, fault scarps, side-hill ridges, scissoring, and offset stream channels are as spectacular along some segments of the Xian- shuihe fault as are the oft-illustrated features along California's San Andreas fault in the Car- rizo Plain area (for example, Wallace, 1975; Allen, 1981,1982; Sieh and Jahns, 1984). In the paragraphs that follow, we describe some of the more obvious and significant features of the fault as traversed from northwest to southeast, from Kasu to Moxi (Fig. 1), but we make no claim for detailed or comprehensive coverage of the entire fault trace.

Kasu-Xialatuo Segment

The upper course of the Xianshui River is eroded along the Xianshuihe fault over most of this segment, as far southeast as Xuxu (Fig. 4). Near the northwest end of the fault, Kasa Lake is an impressive 2-km-long sag pond but is not on the main fault trace. Instead, it appears to be controlled by a splay fault that lies along the steep escarpment northeast of the lake and con- verges with the main break near Zhuwo. Al- though the northwest termination of the 1973 rupture was near Kasu, the Xianshuihe fault extends for another 25 km northwest, beyond which point it appears to lose continuity and fray-out into a number of branches. Figure 15. Map of instrumentally located epicenters in Xianshuihe fault region, 1973 to From Zhuwo southeast to Dandu, the princi- 1985. Immediate aftershocks of the 1973 Luhuo and 1981 Daofu earthquakes are not shown. pal recent break, and the locus of the 1973 rup-

Figure 16. Simplified geologic map of Xianshuihe fault region. Modified from Tang and others (1984).

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1188 ALLEN AND OTHERS

ture, lies near the base of the canyon wall north of the Xirashui River and is readily accessible in only a feVplaces. At Dandu, which was the site of maximum-strike-slip displacement in 1973, the fault crosses the river, and numerous good exposures of the fault are present and easily accessible^etween this point and Xuxu. Some of the best and most diagrammatic of these are near Gelu (Fig. 1), where terrace risers are laterally offset as much as 138 m, and a distinct "push-up" hill lies between the overlapping ends of two en echelon segments of the fault; this locality is illustrated and discussed in detail in a discussion of slip-rate determinations below. Between Xuxu and Xialatuo (Fig. 13), the fault leaves the floor of the river valley and cuts across high terrain for some 25 km in one of the most instructive segments anywhere along the fault. In this area, the vertical component of movement has generally been up on the northeast side, so that a distinct rift-like valley, or fault Figure 17. View southeast of blocky topography (foreground) along fault near Laohekou trough, is present along and parallel to the gentle (Fig. 13), resulting at least in part from rupture of frozen ground in 1973 (and 1816?) northeast-facing mountain slope (Fig. 17). This earthquakes. linear trough was, in fact, the locus of the main Kangding-Lhasa trail (through Laohekou and Laoluhuo) prior to the construction of the au- during the 1973 earthquake are best preserved crosses the river with an en echelon offset, and tomobile road along the river and through the (Figs. 9-12), as described above. the river then departs southward into a major more recent town of Luhuo. Numerous deeply gorge. The fault trace climbs rapidly several cut channels cross the fault nearly at right angles Xialatuo-Qiajiao Segment hundred meters to Changcu Pass, along the old in this segment, and left-laterally offset drainages Daofu-Xialatuo trail, where Albert Heim took are common (Fig. 13); one of these near Laohe- From Xialatuo southeast to Renda (Fig. 4), his noteworthy photograph following the 1923 kou has yielded carbon-14 dates critical to slip- the recent fault trace lies at the edge of the valley earthquake (Fig. 6). Fault physiography in this rate determination and is illustrated and dis- floor against the steep southern valley wall, and area is truly spectacular and is remarkably rem- cussed in detail in a later section. It is also in this fault physiography is relatively obscure. South- iniscent of that of the Carrizo Plain of Califor- area that features resulting from surface faulting east of Renda, on the other hand, the fault nia, particularly in the trench-like topography

Figure 18. Stereo aerial photographs of Mazi area. See map of Figure 7.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1189

along the active trace. The fault trace then de- scends along a fault valley, with numerous sag ponds, springs, and side-hill benches, to rejoin and cross the Xianshui River again at Qiajiao.

Qiajiao-Songlinkou Pass Segment Figure 19. Sketch map of most active trace of Numerous offset canyons are present for the Xianshuihe fault in first few kilometers southeast of Qiajiao, where Songlinkou-Longdengba the fault trace generally lies part way up the area (Fig. 1). Dashed line south valley wall and causes many benches, represents Kangding-Lha- side-hill ridges, fault trenches, and so on. Some sa highway. Four conspi- of the most diagrammatic of these drainage di- cuous sag ponds are indi- versions, with accompanying wind gaps, are cated by "SP". Numerals evident near Mazi, and the sketch map of this indicate road-maintenance area (Fig. 7) should be compared with an aerial stations. Note bend in photograph of the same area (Fig. 18). The fault near Longdengba, fault again enters the valley floor near Goupu associated locally with a (Fig. 4) and then crosses the Xianshui River for 15-m-high, northeast-fac- a final time south of Daofu. The wide crushed ing fault scarp (Fig. 21). zone of the fault is exposed in the steep canyon walls here, although it is somewhat obscured by massive landslides. The 25-km stretch of the fault between this locality and Songlinkou Pass is mostly in the high hills south of the highway and was not visited by the writers except for the southernmost segment approaching the pass (Fig. 19). This area is one of the few along the fault that is heavily forested, and even on aerial photographs, the recent fault trace is not every- where clear. river and is then in hilly terrain southeast to ments (Fig. 23), the most recent of which was Laoqianning. Many diagrammatic features of presumably in 1893. The last identifiable fea- Songlinkou Pass-Laoqianning Segment recent movement are present in this area, how- tures of recent displacement before this segment ever, because of the grassy uncultivated slopes of the fault seemingly terminates are intermittent At the summit of Songlinkou Pass lies a con- (see Fig. 5), and it is easily accessible along the scarps southwest of the Huiyuan Monastery spicuous sag pond (Fig. 20), which is surpris- old trail from Laoqianning to Daofu. Several (Fig. 3), although there is a suggestion that the ingly typical of major strike-slip faults the world low passes are clearly fault controlled. The fault 1893 rupture may have continued a few kilome- over. For example, virtually every major drain- lies along the southwest margin of the Laoqian- ters farther southeast. As discussed in a section age divide along the North Anatolian fault of ning valley, where it is locally up on the south- below, the fault clearly has a major segmenta- Turkey is marked by a sag pond at the very west side. Scarps here show multiple move- tion break at this point. summit (Allen, 1975, 1982). At Longdengba (Fig. 19), a 15-m-high, northeast-facing scarp displaces the broad, post- glacial terrace (Fig. 21), and at no other place along the restricted Xianshuihe fault is such a high vertical component observed in the Holo- cene displacement. Furthermore, the sense of vertical displacement is opposite to that most commonly seen along the fault. This scarp is probably related to the gentle but abrupt bend in the fault at Longdengba, which can be seen in Figure 19, inasmuch as the sense of vertical movement is reversed and the displacement de- creases within less than 5 km both to the northwest and southeast. Offset relationships at Longdengba are further discussed in a section below on slip-rate determinations. The Kangding-Lhasa highway closely follows the fault for about 15 km southeast from Longdengba, and classic fault scarps and sag ponds are abundant in this segment, such as il- Figure 20. View southeast from summit of Songlinkou Pass of sag pond and fault scarp (in lustrated in Figure 22. At the north edge of the shadows). Opposite side of pass (behind camera) drops off precipitously, reflecting fault- area shown in Figure 3, the fault crosses the disturbed drainage pattern.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1190 ALLEN AND OTHERS

the segment south of Taizhan, however, and the portrayal of Figure 3 is based solely on aerial- photograph examination. At the summit of the 4,200-m pass northeast of Dahaizishan, two small lakes are present that lie athwart the fault and may be sag ponds similar to that already described at Songlinkou Pass, although glacial scouring could also have been important here. These lakes are illustrated by Heim (1933), inasmuch as the former main trail between Kang- ding (Tatsienlu) and Laoqianning (Tailing) traversed this pass. The uppermost course of the major river flowing southward from the pass into Kangding is clearly fault controlled, but we see no evidence on aerial photographs that this fault continues as a neotectonic feature as far as or beyond Zhonggu. Southeast of Jinlong Monastery (Fig. 3), scattered northwest-striking fault scarps can again be seen on aerial photographs, and these coa- lesce into a major active fault line that then con- Figure 21. View southeast from near Longdengba (Fig. 19), showing sag pond and fault tinues southeast to and beyond Kangding as the scarp along Xianshuihe fault. Post-glacial terrace surface at left has been uplifted 15 m on right. Selaha (Kangding) fault. Although it is only one of several branches of the fault zone in its northern segment, southeast of Kangding it constitutes the major break of the Xianshuihe fault zone. Laoqianning-Kangding Segment termed the Yalahe fault (Fig. 3). It clearly has a The 5-km-long lake of Mugechuo (Fig. 3) is, in significant normal component of displacement, essence, a sag pond along the fault, and several If the fault trace is projected in a straight line which is consistent with its strike. A latest Pleis- places where the fault cleanly cuts latest Pleisto- southeast from its apparent termination near tocene glacial moraine is offset by the fault near cene glacial moraines (B-D, Fig. 3) are de- Laoqianning, a number of stream channels are Gedalianzi Pass (A, Fig. 3), with 22 m of south- scribed in a subsequent section. aligned in the next 10 km (Fig. 3), suggesting side-up, normal displacement and 10 m of left- The Zheduotang fault is a major branch of the that the fault is indeed continuous in this area, lateral displacement. As followed farther south, fault system in this sector, although, at least as a even if it displays only limited recent movement. the fault gradually trends more southeasterly, recent break, it does not join or merge with The most obvious recent displacement southeast and it presumably takes on an increasing com- other branches. It is crossed several times by the from Laoqianning, however, is along a fault of ponent of strike slip as it becomes parallel to the main Kangding-Lhasa highway (Fig. 8) and is distinctly different orientation that we have regional Xianshuihe trend. We have not visited easily accessible except in its northern extremity. Numerous fresh sag ponds and scarplets are present along the fault trace, and, as described above, it was the locus of significant left-lateral surface rupture during the M = IVi Kangding earthquake of 1955. Kangding-Moxi Segment Although the recent trace of the fault passes through the southern part of the city of Kang- ding itself, features of recent displacement are generally obscured here by cultural modification. After it crosses the river and starts climbing obliquely across the steep mountain front southward toward Xuemenkan Pass (Fig. 3), however, there are many features of recent faulting, perhaps resulting from the last large earthquake here in 1786. Particularly obvious are side-hill ridges, offset stream channels, and numerous small sag ponds (Fig. 24), many of which are easily accessible from the road along the river. This segment has been described in Figure 22. View north of fault scarp (beyond telephone poles) along Xianshuihe fault 1 km detail by Zhao (1985), who measured offsets of south of highway maintenance station no. 40 (Fig. 19). Fault probably ruptured here in 1893. numerous geologic features.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1191

SLIP-RATE DETERMINATION

General Statement

Determination of the long-term slip rate on the Xianshuihe fault was one of the primary objectives of this study. Slip rates have major impact on seismic-hazard evaluation (for example, Anderson and Luco, 1983; Youngs and Coppersmith, 1985), as well as on regional tectonic syntheses (for example, Molnar and Deng, 1984). Of special interest in this study was the question of whether the long-term slip rate on the Xianshuihe fault is compatible with the high seismicity observed during this century, or whether this is a temporal burst of activity that is atypical of the fault's long-term seismicity. Four areas where we found quantitative data bearing on the long-term slip rate are discussed Figure 23. View southeast along fault scarp at southwestern edge of Laoqianning valley in the sections that follow, and these results are (Fig. 3). Side-hill bench may result from 1893 earthquake, which was centered near here. then summarized in the final section.

Terrace Relations near Gelu

The writers did not examine the fault in the and traverses a bedrock spur, with a large closed Near the village of Gelu (Fig. 26), several 30-km-long segment between this area and that depression marking the recent trace. Farther terraces of the Xianshui River and its tributary, of Moxi (Fig. 1), although aerial photographs southeast, the fault cuts obliquely across the Dengdalongba Stream, are offset by the most reveal a clear en echelon stepover of about 2.5 broad Moxi terrace a few kilometers north of active trace of the Xianshuihe fault. The fault km near Yajiagen, about midway between Moxi the town, and although the surface is intensively displacement measured near point A (Fig. 26) and Xuemenkan Pass (Fig. 25). As was the case cultivated, a 3- to 5-m-high scarp (up on the by Tang and others (1976) following the 1973 at Songlinkou and Dahaizishan passes, a sag southwest) clearly marks the fault trace. Luhuo earthquake was 2.3 m horizontally and pond lies exactly at the summit of Xuemenkan Southeast of Moxi, the fault enters a very 0.5 m vertically (south side up), and abundant Pass. Features of recent faulting are again ob- mountainous region that was beyond the area of evidence of this displacement still exists 13 years vious and easily accessible near Moxi. For ex- this study. Satellite images suggest, however, later. For example, a fresh-appearing, 1-m-deep ample, immediately northwest of the Xinxin that southeast of Moxi toward Shimian, the fault gully about halfway between points C and D is hydroelectric plant (10 km northwest of Moxi), loses much of the continuity and simplicity that cleanly offset 2 ± .2 m, left laterally, and fresh the fault locally leaves the alluviated valley floor generally characterize it in the area of our study. scarplets on the face of the 5-m-high scarp at A appear 50-100 cm high. Numerous terrace risers, however, are horizontally offset significantly more than 2 m, representing earlier displacements in addition to that of 1973. For example, the horizontal offset of the 1-m-high riser at B is about 5.6 m. The most impressive offset, however, and the one that initially caught our attention on aerial photographs, is the 138-m offset of an older terrace riser from C to D; the riser is very cleanly truncated by the fault, particularly at C. This riser is about 7 m high both northeast and southwest of the fault trace, and it represents the former river bank associated with the relatively flat former flood plain of the Xianshui River on which the village of Merlu is located (Fig. 26). This terrace surface, or tread, is only about 5 m above the current flood plain of the river at E and represents the youngest widespread terrace of the Xianshui River in the Gelu area. It should be noted that it is the terrace riser that is offset 138 m, and the riser is associated in age with that of the terrace surface below the

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1192 ALLEN AND OTHERS

riser rather than with that of the surface which caps the riser,a s has been emphasized by Lensen and Vella (1971). Thus it is the age of the lower terrace surface that is now 5 m above the current river level which is relevant to the displacement of the terrace riser from C to D. At the time of our field work, we could only infer that the terrace surface was Holocene in age, but a sample obtained subsequently from about halfway between Merlu and the river, 6 cm beneath the terrace surface, has yielded a 14C age of 7,584 ± 172 yr B.P. (Guoguang Zhao, 1990, personal commun.). These relationships thus give a well- constrained Holocene slip rate at Gelu of about 18 mm/yr. Two additional geomorphic features of the Xianshuihe fault near Gelu are significant. First, the Holocene fault scarp at A (Fig. 26) is some 5 m high, and yet the fault displays virtually no vertical displacement at C, only a few hundred meters southeast along the trace. The 1973 displacement showed exactly the same pattern. Clearly the vertical component of displacement along the fault changes markedly over short dis- tances, and this varying vertical component has tended to be repeated faithfully from one earthquake to the next; such behavior is typical of dominantly strike-slip faults the world over (for example, Allen, 1981). A second geomorphic feature of interest is the oval-shaped hill immediately north of Gelu (Fig. 26), which represents a classic "push-up" (pressure ridge, rhomb- horst) between the ends of two en echelon segments of a strike-slip fault (Aydin and Nur, 1982; Mann and others, 1983; Segall and Pol- Figure 25. Map of area of en echelon offset between Selaha (Kangding) and Moxi members lard, 1980). It is also interesting that the 138-m of Xianshuihe fault zone., 15 km northwest of Moxi. For location, see Figure 1. Holocene strike-slip displacement between C and D must die out to the southeast within an almost equal distance, because the fault trace ends at the stepover; this relationship implies that the "push-up" itself must be a very active feature which is rapidly growing in elevation.

Offset Channel near Laohekou

About V2 km northwest of Laohekou (Fig. 13), a stream channel that is incised 5-10 m into alluvial-fan deposits is sharply offset by two parallel branches of the Xianshuihe fault, which it crosses at right angles (Fig. 27). Prior to incision, the stream flowed on the alluvial-fan surface with a 288-m, left-lateral diversion in crossing the fault. Incision apparently commenced when the former course was abandoned and a new shorter and steeper stream course was estab- lished across the fault. The total left-lateral offset of the new, incised channel by the two fault Figure 26. Sketch map of most active trace of Xianshuihe fault and offset terraces near Gelu branches is now 54 m. This figure, however, (Fig. 1). Letters A-E indicate localities referred to in text. Note "push-up" hill at Gelu, between probably represents the minimum tectonic offset right-stepping en echelon fault segments. since the initiation of incision, because the most

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1193

likely cause of incision was capture of the stream TABLE 2. OFFSET GLACIAL MORAINES by a beheaded channel on the downstream side Locality Approx. Approx. Horizontal Vertical of the fault which became juxtaposed with the (see Fig. 3) tot N., long. E. elev. oflset offset stream by left slip on the fault; capture would be (m) (m) (m) most likely to occur somewhat before actual jux- A. Gedalianzi Pass 30°33', 101°35' 4,120 10 22 taposition rather than after, giving rise to an ap- B. NE Changbachan 30°14", 101°43' 4,200 94 C. NW Kangding 30°«', 101°56' 3,600 60 parent right-lateral offset at that time, so that the D. SE Kangding 29°59", 101°58' 3,200 61 actual tectonic offset since the beginning of incision probably exceeds 54 m. A radiocarbon sample was obtained from or- essarily higher here than at localities farther elevation of nearly 6,000 m. At locality B, ganic sediments in the alluvial-fan deposits ex- northwest. northeast of Changbachan, underlying deposits posed in the wall of the incised gully (Fig. 27), Five kilometers southeast of Longdengba (a of an older valley glaciation are clearly identifi- and this has yielded a 14C age of 5630 ± 80 yr few hundred meters north of the sag pond in Fig. able on aerial photographs, but only a single late B.P. The sample was obtained from 2.2 m below 19), the fault cuts the edge of the post-glacial Pleistocene advance is obvious at the remaining the top of the old fan surface, however, so that terrace, which is here about 50 m high. The sites. incision actually began at a somewhat later time. left-lateral offset of the terrace edge (that is, the The latest Pleistocene glaciation in the Gong- In any event, there are thus two good reasons wall of the inner gorge) is about 160 m, and ga area has been termed the "Yumu stage" (Li why the slip rate of 9.6 mm/yr obtained by although one can only say that the eroded ter- and Feng, 1984), and it presumably corresponds dividing the Laohekou offset of 54 m by the race edge is younger than the terrace itself, this to the stage termed "Qomolangma II (Baiyu) in sample age of 5,630 yr must be a minimum rate. further implies a significant slip rate. the larger Qinghai-Xizang region of southwestern China (Zheng and Li, 1980). Only limited Terrace Relations near Longdengba Offset Glacial Moraines work on the chronology of Quaternary glaciation has been carried out in this part of China, At Longdengba, a broad terrace surface is ver- Four localities were recognized on aerial pho- but there is no reason to think that mountain tically offset 15 m by the Xianshuihe fault, and tographs where glacial moraines are offset by the glaciation in the region was not approximately the valley wall is left-laterally offset 203 m Xianshuihe fault or its major branches (A-D, synchronous with that elsewhere in the world. where it meets the terrace surface (Figs. 19,21). Fig. 3). The ages of these moraines are uncer- Detailed studies of the thick loess sequences in We were unsuccessful in obtaining radiocarbon tain, and we were unable to find materials suit- China show that Quaternary climatic fluctua- dates from materials collected within the terrace able for radiocarbon dating. Nevertheless, at tions here are synchronous with those indicated gravels, but the terrace is thought to be post- each of these localities, the moraines are rela- by deep-sea cores and hence can be correlated glacial in age because it is so widespread and so tively sharp-crested and tentatively appear to worldwide (Liu and others, 1985). In particular, similar to others throughout the region, and be- represent the latest Pleistocene ice advance, the Malan loess, which represents the latest cause it extends unmodified well up into the although at localities A and B they are not the Pleistocene cool period, correlates with Oxygen- higher valleys. This age assignment must be re- farthest downstream moraines of this advance. isotope stages 2,3, and 4. Although the latest ice garded as tentative, but if true, the offset of the That the moraines are not Holocene (Neoglacia- advance probably peaked about 18,000 yr ago, valley wall at Longdengba suggests a horizontal tion) in age is suggested by the fact that only one worldwide deglaciation, which may have oc- slip rate of 10-20 mm/yr. There is no evidence of the four glaciated valleys currently holds an curred in two stages, took place primarily be- from either the historic seismicity or the fault active glacier, and it is more than 5 km upstream tween 16,000 and 7,000 yr B.P. (Ruddiman and physiography to suggest that the slip rate is nec- from the offset moraine and at a much higher Duplessy, 1985). Three of the moraines offset by the Xian- shuihe fault were visited in the field, in order to measure the offsets, and careful photogrammetric measurements were made for the fourth locality (B), northwest of Changbachan. This latter locality was subsequently visited by Guo- guang Zhao (1990, personal commun.) who reported agreement with our photogrammetric measurements. Results are summarized in Table 2. At locality C, Zhao (1985) had previously measured a lateral offset of 76 m. The fault offsets we measured must have occurred mainly following the peak of glaciation, because the moraines had to be well stabilized before they could preserve fault displacements, Figure 27. Sketch map of offset channel 1.2 km northwest of Laohekou (Fig. 13). Stream at and because several of the offsets are well up- right formerly flowed on fan surface, now a terrace, toward upper left, with large left-lateral stream of the end moraines of this ice advance. displacement. Light dashed lines approximately portray slope of terrace surface. Subsequent to Offset moraines of latest Pleistocene age have capture by channel at upper right, 54 m of further left-lateral offset has occurred on two active been used for slip-rate determinations in several branches of the fault (heavy dotted lines), and maximum time for this offset is given by age of other parts of the world, and the ages used in the sample in terrace materials. See text. calculations include (1) 13,000 yrs for moraines

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1194 ALLEN AND OTHERS

TABLE 3. SLIP-RATE DETERMINATIONS mm/yr, with different rates for different seg- known to characterize several segments of the ments and different time windows. On the basis San Andreas fault at average rates varying from Location Fault Slip rate (mm/yr) of assumed ages of offset alluvial units and < 1 mm/yr to as much as about 30 mm/yr, and drainages, Tang and others (1984) also suggest a the highest rate effectively represents the total Gelu Xianshuihe 18 slip rate of 5-10 mm/yr for the northern seg- local long-term strain-accumulation rate (Sav- Laohekou Xianshuihe >9.6 Longdengba Xianshuihe 10-20? ment of the fault, and 3 mm/yr near Kangding. age and Burford, 1973; Thatcher, 1979). De- Gedalianzi (A) Yalahe 1 Average of B, C, D Selaha 5.5 Zhao (1985) suggests a slip rate of >6 mm/yr spite intensive searches along active faults else- for the stretch of the Selaha (Kangding) fault where in the world, fault creep was until near Kangding during the past 7,420 yr. recently recognized only at a single locality on cut by the Bocono fault in Venezuela (Schubert, It is significant that the results of seismic mo- the North Anatolian fault of Turkey, discovered 1982), (2) 19,000 ± 2,000 yr for the Wasatch ment accumulations by both Molnar and Deng in 1969 (Aytun, 1982). Thus the recent docu- fault in Utah (Swan and others, 1981), (1984) and Tang and others (1984) indicate that mentation of creep on the Xianshuihe fault is of (3) 15,000 to 20,000 yr for the Hilton Creek our long-term slip rate is compatible with the particular scientific interest. fault in California (Clark and Gillespie, 1981), high level of historic seismic activity. Therefore, (4) 10,000 to 14,000 yr for the Denali fault in this recent activity must be considered typical of Observations Alaska (Sieh, 1981), and (5) 13,000 yr for the the fault's activity during most of the Holocene Tuco fault in Peru (Yonekura and others, 1979). epoch and not a temporal burst of activity such Following the 1973 Luhuo earthquake, geo- In extensive studies of Quaternary faulting in as has been seen along the North Anatolian fault detic resurveys indicated that both horizontal south-central Tibet, Armijci and others (1986) since 1939 (Allen, 1975). A high slip rate on the and vertical displacements were continuing to assumed an age of 10,000 yr for the youngest Xianshuihe fault is also compatible with the occur within a few hundred meters of the fault glacial deposits. Everything considered, we high rate of east-west extension in southern trace at several localities. These changes were arbitrarily assume an age of 13,000 yr for the Tibet, estimated at 10 ± 6 mm/yr by Armijo particularly well documented at Xialatuo, where offset moraines shown in Figure 3. Assuming a and others (1986). vertical uplift of the southwest block averaged 13,000-yr age, average horizontal slip rates at How does a slip rate of 15 mm/yr compare 4.4 mm/yr between 1973 and 1975 but has sites A-D are 0.8, 7.2, 4.6, and 4.7 mm/yr, with those of other active faults worldwide? This subsequently decreased markedly (Ge, 1985). respectively, during Holocene time. rate would fall into the classification of "AA" Probably more tectonically significant is the hor- (Matsuda, 1975), "very high" (Slemmons, 1977), izontal motion at this site, surveyed some 26 Summary and "Class 1" (Cluff and others, 1982)—all of times since 1976 along a 133-m-long line of 9 which represent exceptionally high rates of ac- monuments. These surveys (Fig. 28) show an Our slip-rate determinations for the Xian- tivity. Indeed, with the exception of subduction almost steady, left-lateral creep of about 6 shuihe fault are summarized in Table 3. Of zones, very few faults worldwide exceed 15 mm/yr, most of which is concentrated within a these, the two that are the best constrained are mm/yr in long-term slip rate. Among these are few meters of the fault trace. Visible evidence of those from Gelu and Laohekou. The determina- California's San Andreas fault, with a maximum the continuing creep is obvious in fresh en tion at Longdengba is obviously poorly con- long-term slip rate of about 35 mm/yr (Sieh and echelon cracks in the floor of a concrete monu- strained, but it tends to confirm the observations Jahns, 1984), Alaska's Fairweather fault (58 ment placed astride the fault at Xialatuo in 1983 farther northwest. The slip rates based on the mm/yr; Plafker and others, 1978), and New in connection with a decennial commemoration offsets of glacial moraines are dependent on our Zealand's Wairarapa member of the Alpine of the 1973 event. From the measurements of estimate of the moraine ages. The very low slip fault system (34-60 mm/yr; Lensen and Vella, Figure 28, one could argue either that the hori- rate at Gedalianzi Pass (locality A) is reasonable 1971). Unlike the Xianshuihe fault, all of these zontal creep rate is relatively constant with in view of the fault orientation and high vertical faults are at plate boundaries. Those faults in time, or that a very long and gentle decay is component there. Our overall conclusion is that China that are thought to have slip rates exceed- following the 1973 earthquake, which was asso- the Xianshuihe fault zone is characterized by a ing 15 mm/yr include the Keketuohai-Ertai ciated with 10 cm of left slip at this locality slip rate of 15 ± 5 mm/yr in its northernmost fault (18 mm/yr), Puchang fault (18 mm/yr), (Tang and others, 1976). Thirteen years is ad- segment, decreasing to about 5 mm/yr in its Altyn Tagh fault (30 ± 20 mm/yr) (Molnar and mittedly a long time for post-seismic creep to southern segment. others, 1987), and the Nanhuashan fault (±18 persist at such a high rate, and there are few other earthquakes of this magnitude for which Other investigators have also assigned slip mm/yr) (Ding, 1984). Slip rates of 1-5 mm/yr data exist with which to compare. Exponential rates to the Xianshuihe fault, although generally are perhaps more typical of earthquake-pro- decay of creep following recent California on the basis of fewer data. Molnar and Deng ducing faults the world over, and very infre- earthquakes has been much faster, but the earth- (1984) used assumed moments of the 1923, quent large earthquakes have sometimes oc- quakes have also been much smaller (for exam- 1955, 1973, and 1981 earthquakes to compute curred on faults with slip rates as low as .02 ple, Smith and Wyss, 1968; Cohn and others, an approximate slip rate of 15 mm/yr for the mm/yr (Cluff and others, 1982). 1982; Louie and others, 1985). It is significant, past 80 yr; if the 1786 and 1866 earthquakes are nevertheless, that post-seismic creep in Califor- included as well, a regional deformation rate FAULT CREEP nia has occurred only on faults that also display corresponding to 10 mm/yr of slip on a continuing or episodic slip between major northwest-trending plane was calculated. (The Background events. An earthquake more comparable in "1866" event on the Ganzi-Yushu fault is now magnitude to that of the Luhuo event was the known to have occurred instead in 1854 and is Continuous or episodic surficial slip on a M = 7.5 Guatemala earthquake of 1976, where so indicated in Fig. 2.) Similarly, Tang and oth- s fault, known as fault creep, was first recognized left afterslip on the Motagua fault continued at ers (1984) used assumed moments of historic on the San Andreas fault in California (Stein- least through 1977 (Bucknam and others, 1978), earthquakes to compute an average slip rate of 9 brugge and Zacher, 1960). This behavior is now

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1195

Figure 28. Creep measurements across Xianshuihe fault at Xialatuo. Solid circles on uppermost line represent times of individual measurements; readings were generally more frequent on other lines. Inset shows geodetic arrangement of monuments. Line A'-B' is 133 m long and comprises 8 monuments. Modified from Ge (1985).

but at a very rapidly decaying rate—quite unlike Zheoduotang fault during the 1975-1984 highest creep rate measured, about 6 mm/yr at that at Xialatuo. Both in Guatemala and at Xia- period. Xialatuo, may be post-seismic creep related to latuo, alluvial thicknesses at the measurement There is a strong suggestion in California that the 1973 Luhuo earthquake, but the fact that it sites are small (±10 m at Xialatuo), suggesting the distribution of the creeping and "locked" still continues at a high rate 13 years following that whatever creep is occurring cannot be at- segments of the San Andreas fault is related to the event suggests that other processes are also tributed solely to delay in upward propagation basement rock types, with creep occurring involved. Too little is known about the spatial of slip through thick alluvium, as has been pro- primarily in areas of abundant serpentinite and distribution of creep along the fault, and its rela- posed for some California sites (for example, Franciscan basement on at least one side of the tionship to microseismicity, to draw conclusions Burford, 1972). On neither the Motagua or Xi- fault (Allen, 1968; Irwin and Barnes, 1975). We of a predictive nature, but the analogy to the anshuihe faults is there good evidence indicating are not aware of any such relationship along the Parkfield, California, situation is intriguing whether or not significant creep was occurring Xianshuihe fault, although detailed geologic (Bakun and Lindh, 1985). The repetition of in the years before the recent earthquakes. mapping has yet to be carried out in most of this seemingly "characteristic" historic earthquakes At other locations along the Xianshuihe fault, region. along the Xianshuihe fault, taken together with only sparse data of horizontal slip are available, At many of the survey sites along the Xian- spatial and temporal discontinuities in creep and for much shorter time periods than that of shuihe fault, continuing vertical displacements rates, surely indicate that this is a feature worthy the Xialatuo record (Ge, 1985). The survey site have taken place following the Luhuo earth- of further detailed studies in connection with near the northwest end of the 1973 rupture at quake, but their tectonic significance is more earthquake-prediction efforts. Zhuwo (Fig. 4), showed consistent left-lateral debatable than that of the horizontal movements slip of about 1.5 mm/yr in 12 resurveys from owing to the topographic relief along the fault FAULT SEGMENTATION mid-1980 through 1984. The next survey site trace, ground-water changes, and so on. Of par- southeast, at Dandu, showed essentially no hori- ticular interest to Chinese investigators have Significance zontal displacement during the same period. The been the anomalies in a number of the level lines site at Shawan (Fig. 4) showed questionable that took place in the months prior to the 1981 Fault segmentation has been a subject of con- right-lateral slip of about 1 mm/yr, but other Daofu earthquake (Tang and others, 1984; Ge, siderable recent interest because of the effect of nearby sites at Geluohazi and Xuxu showed no 1985) (Fig. 28). segmentation in determining the points at which significant horizontal slip. No other survey sites fault ruptures start and stop during large earth- exist along the fault except near Daofu, where Summary quakes. Thus, segmentation bears directly on the some post-seismic creep following the 1981 questions of "characteristic" earthquakes, esti- earthquake is suggested, and near Laoqianning There is no question but that significant hori- mations of maximum magnitudes, stress drops, (Fig. 3), where no significant horizontal slip has zontal creep has occurred along some segments and prediction of strong ground motions occurred during the 1979-1984 period. Sim- of the Xianshuihe fault during the 1976-1986 (Schwartz and Coppersmith, 1986). Segmenta- ilarly, no discernible slip has occurred across the time period, but not along other segments. The tion can be related to a variety of geologic fea-

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1196 ALLEN AND OTHERS

tures, all of which might be classified as zone appears to have marked the terminations of horizontal slip is about 2:1 (Table 2). The broad "barriers" or "asperities" (Aki, 1979, 1984). ruptures both on the Xianshuihe fault to the valley of Laoqianning is itself a very unusual Among such features visible to a geologist that southeast (for example, in 1973) and on the feature in this otherwise mountainous area and have been related to earthquake occurrences are Ganzi-Yushu fault to the northwest (for exam- presumably owes its existence to the dilatational (1) minor fault bends (King and Nabelek, 1985; ple, in 1854). fault jog, analogous to those represented by Bilham and Williams, 1985), (2) en echelon The very uniform curvature of the Xianshuihe many other sedimentary basins along domi- stepovers (Bakun and others, 1980; Segall and fault between Ganzi and Shimian (Fig. 2) is in- nantly strike-slip faults (Crowell, 1974; Rod- Pollard, 1980; Deng and Zhang, 1984; Sibson, triguing, suggesting block rotation about a pole gers, 1980; Sibson, 1985, 1986). Other features 1985), (3) changing relief along the fault somewhere near the Burmese border. The gentle marking this first-order segmentation point are (Schwartz and Coppersmith, 1984; Crone and curvature of the Xiaojiang fault, farther south, (1) the much higher relief to the south than to others, 1987), (4) cross structures (Schwartz and suggests rotation about the same pole but with a the north, (2) the current seismic gap along the Coppersmith, 1986; Crone and others, 1987), larger radius. This geometry is surprisingly sim- Xianshuihe fault whose southern termination is and (5) changing bedrock types and hydrologic ilar to that of members of the San Andreas fault near Laoqianning (Han and Huang, 1983), and environment (Allen, 1967; Irwin and Barnes, system in southern California, where Weldon (3) the much more distributed, net-like fault 1975). Because of the numerous relatively well and Humphreys (1986) recently pointed out the pattern to the south. Although little is known documented large earthquakes on the Xian- remarkable concentric arcs of different segments about historic large earthquakes in this area, shuihe fault, and their possible "characteristic" of the fault, again emphasizing the rotation of strike-slip rupture associated with the M 3= 7 nature, understanding of the relationship of these mini-blocks caught within the plate-boundary event of 1893 apparently terminated on the south within a few kilometers of Laoqianning. earthquakes to segmentation of the fault was one structure. Departures from the smooth arcs, not of the primary objectives of this study. More departures from linearity, represent the principal The next most obvious en echelon offset details of our segmentation study are presented asperities. along the Xianshuihe fault is that between the in Wen and others (1989). Selaha and Moxi fault segments near Yajiagen En echelon Offsets (Fig. 25), although the separation between over- Regional Fault Geometry lapping segments here is considerably less than The most obvious interruption in the conti- that at Laoqianning, and there is no clear struc- The Xianshuihe fault is only part of a much nuity of the Xianshuihe fault itself in the region tural contrast to the northwest and southeast of longer left-lateral fault system extending at least of this study, representing a first-order segmenta- the jog. This area was not visited in the field, but 1,400 km from southern Yunnan northwest tion, is the abrupt termination of active strike- photo analysis reveals the presence of a lake in through western Sichuan into southern Qinghai slip faulting near Laoqianning, together with the the jog area, and it appears to represent a typical (Fig. 1, insert). The two most obvious points of commencement farther south of normal faulting pull-apart. The M=l% earthquake of 1786 is segmentation along this entire system are (1) the of different strike (Figs. 1, 3). No detailed geo- thought by Wang (1986, personal commun.) to en echelon stepover between the Xianshuihe logic mapping has been done in this area, but it have been associated with rupture through, fault and the Ganzi-Yushu fault (Wen and oth- is clear from aerial photographs that active rather than terminating at, this discontinuity, al- ers, 1985; Wen and Bai, 1985) and (2) the fault strike-slip faulting along the restricted Xian- though reported intensities were highest in this branching between the Anninghe and Zemuhe- shuihe fault does not extend more than a few area and could have been related to the presence Xiaojiang faults near Xichang (Huang and kilometers southwest of Huiyuan Monastery, at of the asperity. Tang, 1982). Both areas have been the loci of least as a single well-defined rupture similar to En echelon offsets of still lower order are considerable historic earthquake activity (Fig. that to the northeast (Figs. 5, 24). The Yalahe common along the fault trace but do not appear 2). The dilatational nature of the stepover be- fault, as traced southeastward from the Laoqi- to have been significant in the termination or tween Ganzi and Zhuwo is demonstrated by the anning area, initially trends east as a normal initiation of fault rupture. A possible exception presence of northeast-trending normal faults, as fault before it gradually returns to the regional is the left stepover of about 1 km near Renda well as by normal-fault focal mechanisms of re- southwest strike; where a glacial moraine is cut (Fig. 4), which is at about the southern termina- cent earthquakes within the stepover (Fig. 29), by the east-trending fault at Gedalianzi Pass tion of rupture in 1973. The report by Heim and, as predicted by Sibson (1985, 1986), the (A, Fig. 3), the ratio of Holocene vertical to (1934), however, indicates that the 1923 rupture progressed through this same area, overlapping the area of subsequent 1973 rupture. Other low- order offsets are discussed by Wen and others (1989).

Fault Bends

King and Nabelek (1985) pointed out that the epicenter of the 1973 Luhuo earthquake was near a distinct bend in the fault trace near Gelu

Figure 29. Map showing fault pattern in area of en echelon overlap between Xianshuihe and Ganzi-Yushu faults (Fig. 1). Most northwest-trending faults are left slip, whereas most northeast-trending faults are normal. Focal mechanisms are shown for six of the larger earthquakes in this area since 1967. Modified from Wen and Bai (1985).

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1197

(Fig. 4). Careful tracing of the fault on Landsat tween Qianning and Yingda (Fig. 1). Indeed, based on the observations of Tang and others images indicates a bend of about 9° over a dis- the continuity of the fault in this region is so (1976), and using a slip rate of 15 mm/yr, the tance of 12 km, and we agree that this may great that, in the absence of the historical record, recurrence interval for such events would be represent an important mechanical disconti- one would not preclude a single large earth- about 67 yr. The average fault slip in 1973 at nuity. An even more abrupt bend is that at quake rupturing the entire 220-km-long seg- seismogenic depths, based on the seismic mo- Longdengba (Fig. 19), where the fault turns 10° ment. Yet this has not happened in several ment calculated by Zhou and others (1983a), in less than 2 km. The local vertical displace- hundred years, and the historic record instead however, was 3.8 m, yielding a recurrence inter- ment associated with this bend (Fig. 21) was suggests that earthquakes of shorter rupture val of about 250 yr. The actual value probably pointed out earlier, and it possibly could be as- length occur repeatedly as smaller "characteris- lies between these two extremes, because the slip sociated with the epicenter of the 1893 earth- tic" events (Schwartz and Coppersmith, 1984). rate was measured within a few hundred meters quake, which was centered in this area. This Although it is tempting to identify more obscure of the surface fault trace, and the regional strain- segment of the fault is currently a conspicuous geologic features such as alleged cross structures, accumulation rate is probably larger. Our judg- seismic gap (Han and Huang, 1983), and the small basins, and minor bends as asperities con- ment is that the historic repeat interval of about Longdengba asperity may play an important trolling these events, we hesitate to do so in the 150 yr is approximately representative of the role in initiating the next event. It represents a absence of more detailed geologic mapping and long-term behavior in the Luhuo area. promising area for instrumentation. more detailed seismographic coverage. This The seismic moment of the 1981 Daofu The gentle double curvature of the Selaha does, however, remain a very promising area of earthquake calculated by Zhou and others fault in the vicinity of Kangding (Fig. 3) is sim- field investigation. (1983b) suggests an average fault slip at seismo- ilar to situations along California's San Andreas genic depth of about 1 m, although displace- fault described by Crowell (1974), with basins SEISMIC HAZARD OF THE ments at the surface were not sufficiently large in the releasing bends and mountains in the re- XIANSHUIHE FAULT or consistent to allow systematic measurements. straining bends. In the Kangding area, the basin Again using a 15 mm/yr slip rate, repeat times of Mugechuo Lake represents a releasing bend, General Statement of 67 yr or less would be expected for earth- and trenches excavated across the fault south of quakes of M= ±1 on this fault segment, and this Kangding support the compressional nature of Numerous large earthquakes have occurred is approximately consistent with the timing of the faulting there, at the base of a 6,070-m peak. along the Xianshuihe fault within the historic those experienced in 1904 and 1981. The occurrence of these gentle bends in the record (Fig. 2; Table 1). Many of these have Both the 1973 Luhuo and 1981 Daofu Kingding region, and their possible "locking" been locally devastating, but, fortunately, popu- earthquakes were among the 26 worldwide effect on fault slip, may be related to the fact that lation density along most of the fault is very low, events used by Kanamori and Allen (1986) in this segment of the Xianshuihe fault seems to be and few large or critical structures exist. At least their study of repeat time and stress drop. These characterized by infrequent large earthquakes one village adjacent to the fault was completely two events are consistent with the overall pat- rather than more frequent moderate ones. Sim- abandoned following devastation during the tern in supporting the concept that earthquakes ilarly, the great earthquakes on the San Andreas 1893 earthquake northwest of Laoqianning with relatively short repeat times, such as these, fault in 1906 and 1857 have been related to (Heim, 1934), and entire towns and villages had have longer-than-average rupture lengths and major bends in those specific areas (Allen, to be reconstructed following the more recent lower-than-average stress drops as compared to 1968). 1973 Luhuo and 1981 Daofu earthquakes. Both earthquakes with long repeat times (for exam- Surface rupture during the 1955 Kangding the historic record and the demonstrated high ple, more than 300 yr). This conclusion could earthquake (M = 7Vi) was limited, insofar as is slip rate on the fault indicate that it must con- have an impact on the spectra used locally in known, to the gently curving Zheduotang fault tinue to be considered a very dangerous feature. earthquake-resistant design. (Fig. 3), although the entire fault segment ap- The segment of the fault from Laoqianning parently broke during this one event. As the Recurrence Intervals and Seismic Gaps northwest to Songlinkou Pass (Figs. 1, 3) last Selaha and Zheduotang branches of the Xian- broke during the A/s 7 earthquake of 1893. If shuihe fault system are, at most, only 10 km The historical record of western Sichuan is strain is accumulating on this 35-km-long seg- apart in this area, and they converge toward too short to allow confident estimates of earth- ment at 15 mm/yr, then sufficient strain has their ends, it is possible that they dip steeply quake recurrence intervals on this basis alone, now accumulated for another earthquake in this toward each other and merge at depth. This although some data are suggestive. Only the magnitude range, and the length of this segment might help to explain the unusually short surface northern part of the Xianshuihe fault has had is consistent with an earthquake of M = ±1 rupture length of 27 km in 1955 for an earth- repeated earthquakes on the same fault seg- (Bonilla and others, 1984). The current seismic quake of magnitude IVr, fault rupture at depth ments: the 1816 and 1973 earthquakes near gap, however, extends some 30 km still farther may have been more complicated, with accom- Luhuo appear very similar and suggest an inter- northwest to the area of Daofu, overlapping the panying slip on the Selaha or other branches val of about 150 yr between M = ±1Vi earth- area of rupture in 1981, and suggests an earth- that either did not reach the ground surface or quakes here, and the 1904 and 1981 events near quake of perhaps M = IVi (Han and Huang, went unnoticed in this remote, mountainous Daofu are perhaps sufficiently alike to suggest 1983). Using a somewhat different approach, if region. an interval of about 75 yr between M = ±7 we assume a fault length of 65 km, a fault width earthquakes on this fault segment. On the south- of 10 km, and 94 yr of slip accumulating at 15 Other Localities ern part of the fault near Kangding, which has mm/yr, then an earthquake of moment 3.0 x experienced even larger earthquakes, none has 1026 dyne cm is "prepared," corresponding to Other than the few examples mentioned in repeated within the historic record. about Mw = 6.9; following Kanamori and others the preceding sections, it is difficult to identify It is instructive to use the slip rates deduced in (1986), an earthquake with this corresponding asperities associated with the terminations of this study to further elucidate possible recur- fault length could not sustain significantly more ruptures during historic earthquakes on the rela- rence intervals. The average surface slip during accumulated moment and must therefore be a tively continuous restricted Xianshuihe fault be- the 1973 Luhuo earthquake was about 1 m, likely event in the near future.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 1198 ALLEN AND OTHERS

If we are correct in deducing that the slip rate and (3) general lack of concealing vegetative portunities for slip-rate determinations, for on the fault zone near Kangding is much less cover, but with sufficient grass and turf to pre- which only exploratory field work and radio- than that to the north, perhaps 5 mm/yr, then serve details of recent ruptures. Perhaps only the metric dating have been cdTHed out in the pres- recurrence intervals should be correspondingly Carrizo Plain segment of the San Andreas fault ent study. Significant creep is known worldwide longer. Assuming, after Bonilla and others in California achieves the same average excel- only on the Xianshuihe and San Andreas faults, (1984), that the average slip (30% of the maxi- lence of physiographic expression that is typical and this subject is one of particular relevance to mum) in association with a M= 7% event would of the higher-altitude segments of the Xian- earthquake-prediction efforts. The 300-yr his- be about 1.3 m, a repeat time of about 260 yr shuihe fault. tory of numerous and sometimes overlapping should characterize events such as the destruc- Two additional factors may contribute to the fault ruptures along the Xianshuihe fault is al- tive 1786 earthquake. Thus the time is again at spectacular physiographic features along the Xi- most unparalleled elsewhere in the world, and hand when a repeat of the 1786 earthquake anshuihe fault. First, many recent ground- further detailed mapping, both geologically and would come as no scientific surprise. A rupture rupturing earthquakes along the Xianshuihe seismologically, should elucidate the nature of length of 135 km, corresponding to an event of fault have occurred in mid-winter when the fault segmentation and its bearing on earth- this magnitude (Bonilla and others, 1984), ground was frozen to a depth of several centime- quake-hazard evaluation. Finally, although the would extend from the northwest end of the ters, giving rise to a blocky type of "moletrack" Xianshuihe fault is in a somewhat remote area Selaha fault southeast almost to Shimian (Fig. deformation that is truly impressive and is easily that is currently difficult to reach from the main 1), and so a very large area, would be shaken. preserved (Fig. 14). Second, the Xianshuihe population centers of China, after one reaches fault has been characterized by relatively fre- the area, accessibility to the fault is generally Summary quent earthquakes in the M = 7 to 7 Vi range, excellent because old caravan routes as well as apparently giving rise to fresher appearing phys- modern roads tend to follow the erosional In view of both the historic record and the iographic features than along faults such as the trough—the marked "rift topography"—along high measured slip rate, the entire Xianshuihe San Andreas with less frequent M = ±8 earth- the active fault trace. fault zone must be considered one of very high quakes, albeit with larger individual displace- seismic hazard. Two segments are of particular ments and with equal or higher long-term slip ACKNOWLEDGMENTS concern at this time: (1) the 65-km-long segment rates. between Daofu and Laoqianning, capable of In the Introduction, we asked the question, This study was carried out under the protocol producing a M= ±7 event, and (2) the 135-km- "Could the very high degree of activity have for scientific and technical cooperation in earth- long segment including the Selaha (Kangding) been recognized by physiographic features quake studies between the State Seismological and Moxi faults, capable of producing a larger alone, even in the .absence of the historic rec- Bureau of the People's Republic of China and event. The warning for the Daofu-Laoqianning ord?" The answer is an unqualified "yes." For the U.S. Geological Survey and the National segment is considered to be of higher confi- example, where one first sees the northern seg- Science Foundation. The field study was spon- dence, because the slip rate there is better con- ment of the Xianshuihe fault while driving from sored by the Seismological Bureau of Sichuan firmed, recent epicenters of large earthquakes Kangding to Daofu, near maintenance station Province, and we appreciate the assistance of Li have migrated toward the area from the north- no. 40 (Figs. 19,22), the scarp is so fresh and the Xinghai and Zhou Xinghe in making the multi- west, and a distinct seismic gap currently exists. sag ponds so abundant as to be fully and imme- tude of arrangements that were necessary. We On the other hand, an earthquake on the south- diately convincing of a high degree of activity. are indebted to the many geologists of the ern segment near Kangding would be of greater Even at lower altitudes in more densely vege- Sichuan Bureau who accompanied us in the consequence, owing to its larger projected size tated areas, such as near Moxi (Fig. 1), sag field and contributed markedly to the study. and closeness to population centers. Neither the ponds and fresh scarps are clearly more abun- Principal among these were Huang Shengmu, segmentation of the fault nor its history suggests dant than along, for example, the Red River Zhang Wenpu, Liu Benpei, and Yang Xinxiao. that earthquakes are likely that are significantly fault of Yunnan, with a slip rate of 2-5 mm/yr Field work was carried out with the generous larger than those that have been experienced in a somewhat similar rainfall environment help of administrative authorities in Luhuo, historically. (Allen and others, 1984). While the high- Daofu, and Kangding Counties. We especially altitude preservation along most of the Xian- appreciate the efforts of Ding Guoyu of the State COMPARISONS WITH OTHER shuihe fault certainly serves to enhance the Seismological Bureau in encouraging and sup- FAULTS WORLDWIDE impression of high activity, many of the same porting the study. American participation was kinds of offset features have been identified by supported by U.S. Geological Survey Grant No. The Xianshuihe fault is typical of major careful field studies of other regional strike-slip 14-08-0001-G1088. strike-slip faults related to intracontinental plate faults in grossly different environments, such as boundaries the world over, even though it re- in the dense bush along the Alpine fault of New REFERENCES CITED Zealand (Wellman, 1953), along the highly flects the very complex and distributed plate Abe, K., 1988, Magnitudes and origin times from Milne seismograph data: modified and cultivated terrain of the Median Earthquakes in China and California, 1898-1912, in Lee, W. H., Mey- boundary between India and Eurasia (Molnar ers, H., and Shimazaki, K., eds., Historical seismograms and earth- Tectonic Line of Japan (Okada, 1980), or in the quakes of the world: London, Academic Press, p. 37-50. and Tapponnier, 1975). Nothing in the tectonic Aki, K., 1979, Characterization of barriers on an earthquake fault: Journal of nature of the Xianshuihe fault is remarkably un- jungles engulfing the Philippine and Great Sum- Geophysical Research, v. 84, p. 6140-6148. atran faults (Allen, 1962; Untung and others, 1984, Asperities, barriers, characteristic earthquakes, and strong motion usual by worldwide standards. On the other prediction: Journal of Geophysical Research, v. 89, p. 5867-5872. 1985). Allen, C. R., 1962, Circum-Pacific faulting in the Philippines-Taiwan region: hand, it has a number of attributes that are better Journal of Geophysical Research, v. 67, p. 4795-4812. displayed and more easily studied than perhaps 1968, The tectonic environments of seismically active and inactive areas Several other attributes of the Xianshuihe along the San Andreas fault system: Stanford University Publications in anywhere else in the world. In particular, the fault make it exceptionally promising for further the Geological Sciences, v. 11, p. 70-82. 1975, Geological criteria for evaluating seismicity: Geological Society of geomorphic expression of the fault, especially at studies of fault behavior, seismic hazard evalua- America Bulletin, v. 86, p. 1041-1057. altitudes above 3,000 m, is nothing short of 1981, The modern San Andreas fault, in Ernst, W. G., ed.. The geotec- tion, and earthquake prediction. The fact that tonic development of California: Englewood Cliffs, New Jersey, dramatic, owing to the fault's (1) relatively high the fault cuts numerous glacial moraines and Prentice-Hall, p. 511-534. 1982, Comparisons between the North Anatolian fault of Turkey and slip rate, (2) absence of cultural modification, post-glacial fluvial surfaces offers excellent op- the San Andreas fault of California, in Isikara, A. M., and Vogel, A.,

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021 XIANSHUIHE FAULT, WESTERN CHINA 1199

eds., Multidisciplinary approach to earthquake prediction, Volume 2; tonic activity in southwestern China [in Chinese]: Acta Geophysica v. 95, p. 883-896. Braunschweig/Wiesbaden, West Germany, Friedr. Vieweg & Sohn, Sinica, v. 20 (2), p. 96-109. Slemmons, D. B., 1977, Faults and earthquake magnitude: U.S. Army Engineer p. 67-85. Kanamori, H., and Allen, C. R., 1986, Earthquake repeat time and average Waterways Experiment Station, Miscellaneous Paper S-73-1,166 p. Allen, C. R., St. Amand, P., Richter, C. F., and Nordquist, J. M., 1965, stress drop: American Geophysical Union Geophysical Monograph 37, Smith, S. W., and Wyss, M., 1968, Displacement on the San Andreas fault Relationship between seismicity and geologic structure in the southern p. 227-235. subsequent to the 1966 Parkfield earthquake: Seismological Society of California region: Seismological Society of America Bulletin, v. 55, Kanamori, H., Allen, C. R., and Clayton, R. W., 1986, Earthquake and seismic- America Bulletin, v. 58, p. 1955-1973. p. 753-797. iiy research using SCARLET and CEDAR: U.S. Geological Survey Steinbrugge, K. V., and Zacher, E. G., 1960, Creep on the San Andreas fault- Allen, C. R., Gillespie, A. R., Han Yuan, Sieh, K. E., Zhang Buchun, and Zhu Open-File Report 86-383, p. 345-347. Fault creep and property damage: Seismological Society of America Chengnan, 1984, Red River and associated faults, Yunnan Province, King, G., and Nabelek, J., 1985, Role of fault bends in the initiation and Bulletin, v. 50, p. 389-396. China: Quaternary geology, slip rates, and seismic hazard: Geological termination of earthquake rupture: Science, v. 228, p. 984-987. Swan, F. H„ III, Hanson, K. L., Schwartz, D. P., and Kneupfer, P. L., 1981, Society of America Bulletin, v. 95, p. 686-700. Lawson, A. C., ed., 1908, The California earthquake of April 18,1906: Carne- Study of earthquake recurrence intervals on the Wasatch fault at the Anderson, J. G., and Luco, J. E., 1983, Consequences of slip rate constraints on gie Institution of Washington Publication 87, Volume 1, p. 53-113. Little Cottonwood site, Utah: U.S. Geological Survey Open-File Report earthquake recurrence relations: Seismological Society of America Bul- Lee, C. S., Yuan, J. Q., and Guo, L. Z., 1938, Comments on the geology of east 81-450, 30 p. letin, v. 73, p. 471-496. Sikang Province [in Chinese]: Di Zhi Lun Ping, v. 5, p. 69-84. Tan, H. C., and Li, C. Y., 1930, Mineral deposits of eastern Sikang: China Armijo, R., Tapponnier, P., Merrier, J. L., and Han Tonglin, 1986, Quaternary Lee, S. P., 1948, Tectonic relation of seismicity near Kangting, east Sikang: Geological Survey Bulletin 17, p. 1-4. extension in southern Tibet: Field observations and tectonic implica- Journal of the Chinese Geophysical Society, v. 1, p. 43-50. Tang Rongchang, Wen Dehua, Deng Tiangang, and Huang Shengmu, 1976, A tions: Journal of Geophysical Research, v. 91, p. 13803-13872. Lensen, G., and Vella, P., 1971, The Waiohine River faulted terrace sequence: preliminary study of the characteristics of the ground fractures during Aydin, A., and Nur, A., 1982, Evolution of pull-apart basins and their scale Royal Society of New Zealand Bulletin 9, p. 117-119. the Luhuo M ~ 7.9 earthquake, 1973 and the origin of the earthquake independence: Tectonics, v. 1, p. 91-105. Li Jijun and Feng Zhaodong, 1984, Quaternary glaciation relics in Hengduan [in Chinese]: Acta Geophysica Sinica, v. 19 (1), p. 19-27. Aytun, A., 1982, Creep measurements in the Ismetpasa region of the North Mountains, in Li Jijun and Feng Zhaodang, Evolution of mountain Tang Rongchang, Huang Zuzhi, Qian Hong, Deng Tiangang, Jiang Lenqiang, Anatolian fault zone, in Isakara, A. M., and Vogel, A., eds., Multidis- glaciers and the Quaternary glaciation [in Chinese]: Lanzhou University Ge Peiji, Liu Shengli, Cao Yangguo, and Zhang Chenggui, 1984a, On ciplnary approach to earthquake prediction, Volume 2: Braunschweig/ Bulletin, Dec. 1984, p. 61-72. the recent tectonic activity and earthquake of the Xianshuihe fault zone, Wiesbaden, West Germany, Friedr. Vieweg & Sohn, p. 279-292. Li Ping, and Wang Liangmou, 1975, Exploration of the seismo-geological in A collection of papers of International Symposium on Continental Bakun, W. H , and Lindh, A. G„ 1985, The Parkfield, California, earthquake features of the Yunnan-West Sichuan region [in Chinese]: Scientia Geo- Seismicity and Earthquake Prediction: Beijing, China Seismological prediction experiment: Science, v. 229, p. 619-624. logic Sinica, 1975 (4), p. 308-326. Press, p. 347-363. Bakun, W. H., Stewart, R. M, Bufe, C. G„ and Marks, S. M., 1980, Implication Li Tianchi, Schuster. R. L., and Wu Jishan, 1986, Landslide dams in south- Tang Rongchang, Qian Hong, Chang Wenfu, Chang Chengqui, Cao Yangguo, of seismicity for failure of a section of the San Andreas rift: Seismologi- central China: American Society of Civil Engineers Geotechnical Spe- and Liu Shengli, 1984b, On the seismogeological setting and conditions cal Society of America Bulletin, v. 70, p. 185-201. cial Publication 3, p. 146-163. of seismogenic structures of 1981 Daofu earthquake [in Chinese]: Seis- Bilham, R., and Williams, P., 1985, Sawtooth segmentation and deformation Li Tiansao, You Zeli, Du Qifang, Wu Zixong, and Jiang Yuanming, 1985, mology and Geology, v. 6 (2), p. 33-40. processes on the southern San Andreas fault, California: Geophysical Geological characteristics and the pattern of movement on the Xian- Tapponnier, P., and Molnar, P., 1976, Slip line field theory and large-scale Research Letters, v. 12, p. 557-560. shuihe fault zone [in Chinese], in Sichuan Seismological Bureau, Pro- continental tectonics: Nature, v. 264, p. 319-324. Bonilla, M. G„ Mark, R. K., and Lienkaemper, J. J., 1984, Statistical relations ceedings of the confernece on the Xianshuihe fault zone: Beijing, China, 1977, Active faulting and tectonics in China: Journal of Geophysical among earthquake magnitude, surface rupture length, and surface fault Seismological Press, p. 1-7. Research, v. 82, p. 2905-2930. displacement: Seismological Society of America Bulletin, v. 74, Lin Banghui, Cheng Tianchang, Pu Xiaohong, Liu Wanqin, Peng Meixuan, Tapponnier, P., Peltzer, G., LeDain, L. Y., Armijo, R., and Cobbold, P., 1982, p. 2379-2411, and Zhang Weiping, 1986, Rupture processes of strong earthquakes on Propagating extrusion tectonics in Asia: New insights from simple Bucknam, R. C., Plafker, G„ and Sharp, R. V., 1978, Fault movement (after- Xianshuihe fault bell and seismic activity [in Chinese]: Acta Seismolog- experiments with plasticine: Geology, v. 10, p. 611-616. slip) following the Guatemala earthquake of February 4, 1976: ica Sinica, v. 8, p. 1-20, Tapponnier, P., Peltzer, G., and Armijo, R., 1986, On the mechanics of the Geology, v. 6, p. 170-173. Liu Tungsheng, An Zhisheng, Yuan Baoyin, and Han Jiamao, 1985, The collision between India and Asia, in Coward, M. P., and Ries, A. C., Burford, R. O., 1972, Continued slip on the Coyote Creek fault after the loess-paleosol sequence in China and climatic history: Episodes, v. 8, eds., Collision tectonics: Geological Society [London] Special Publica- Borrego Mountain earthquake: U.S. Geological Survey Professional p. 21-28. tion 19, p. 115-157. Paper 787, p. 105-111. Louie, J. N., Allen, C. R., Johnson, D. C., Haase, P. C., and Cohn, S. N„ 1985, Thatcher, W., 1979, Systematic inversion of geodetic data in central California: Clark, M. M„ and Gillespie, A. R., 1981, Record of late Quaternary faulting Fault slip in southern California: Seismological Society of America Journal of Geophysical Research, v. 84, p. 2283-2295. along the Hilton Creek fault in the Sierra Nevada, California [abs.]: Bulletin, v. 75, p. 811-833. Untung, M., Buyung, N., Kertapati, E., Undang, and Allen, C. R., 1985, Rup- Earthquake Notes, v. 52, p. 46. Luo Zhuoli, ed., 1980, Seismic history of Sichuan Province [in Chinese]: ture along the Great Sumatran fault, Indonesia, during the earthquakes Cluff, L. S., Coppersmith, K. J., and Kneupfer, P. J., 1982, Assessing degrees of Chengdu, China, People's Press, 2 volumes. of 1926 and 1943: Seismological Society of America Bulletin, v. 75, fault activity for seismic zonation: 3rd International Microzonation Mann. P., Hempton, M. R., Bradley, D. C., and Burke, K., 1983, Development p. 313-317. Conference, Proceedings, v. 1, p. 113-118. of pull-apart basins: Journal of Geology, v. 91, p. 529-554. Wallace, R. E., 1975, The San Andreas fault in the Carrizo Plain-Temblor Cohn, S. N., Allen, C. R., Gilroaii, R., and Goulty, N. R., 1982, Preearthquake Matsuda, T., 1975, Magnitude and recurrence interval of earthquakes from a Range region, California: California Division of Mines and Geology and postearthquake creep on the Imperial fault and the Brawley fault fault [in Japanese]: Zisin, v. 28, p. 269-283. Special Report 118, p. 241-250. zone: U.S. Geological Survey Professional Paper 1254, p. 161-167. Molnar, P., and Deng Qidong, 1984, Faulting associated with large earth- Weldon, R., and Humphreys, E., 1986, A kinematic model of southern Cali- Crone, A. J„ Machette, M. N., Bonilla, M. G., Lienkaemper, J. J., Pierce, K. L., quakes and the average rate of deformation in central and eastern Asia: fornia: Tectonics, v. 5, p. 33-48. Scott, W. E., and Bucknam, R. C., 1987, Surface faulting accompany- Journal of Geophysical Research, v. 89, p. 6203-6227. Wellman, H. W., 1953, Data for study of recent and late Pleistocene faulting in ing the Borah Peak earthquake and segmentation of the Lost River fault, Molnar, P., and Tapponnier, P., 1975, Cenozoic tectonics of Asia: Effects of a the South Island of New Zealand: New Zealand Journal of Science and centra] Idaho: Seismological Society of America Bulletin, v. 77, continental collision: Science, v. 189, p. 419-426. Technology. B., v. 34, p. 270-288. p. 739-770. Molnar, P., Burchfiel, B. C., Liang K'uangyi, and Zhao Ziyun, 1987, Geomor- Wen Xueze, 1986, Estimating recent activities of the Xianshuihe fault zone Crowell, J. C_, 1974, Origin of late Cenozoic basins in southern California: phic evidence for active faulting in the Altyn Tagh and northern Tibet by earthquake activity parameters: Acta Seismologica Sinica, v. 8, Society of Economic Paleontologists and Mineralogists Special Publica- and qualitative estimates of its contribution to the convergence of India p. 146-155. tion 22, p. 190-204. and Eurasia: Geology, v. 15, p. 249-253. Wen Xueze, and Bai Lianxiang, 1985, The crustal fracture pattern and seismo- Deng Qidong, and Zhang Piezhen, 1984, Research on the geometry of shear Okada, A., 1980, Quaternary faulting along the median tectonic line of genic structure of the northwestern segment of the Xianshuihe fault zone fracture zones: Journal of Geophysical Research, v. 89, p. 5699-5710. southwest Japan: Geological Society of Japan Memoir 18, p. 79 108. [in Chinese], in Sichuan Seismological Bureau, Proceedings of Confer- Ding Guoyu, 1984, Active faults of China, in A collection of papers of Interna» Plafker, G.. Hudson, T., Brum, T., and Rubin, M., 1978, Late Quaternary ence on Xianshuihe Fault Zone: Beijing, China, Seismological Press, tional Symposium on Continental Seismicity and Earthquake Predic- offsets along the Fairweather fault and crustal plate interactions in p. 33-40. tion: Beijing, China, Seismological Press, p. 225-242. southern Alaska: Canadian Journal of Earth Sciences, v. 15, Wen Xueze, Huang Shengmu, and Jiang Zaixiong, 1985, Neotectonic features Ge Peiji, 1985, Analysis of the geodetic measurements across the Xianshuihe p. 805-816. of the Ganzi-Yushu fault zone and assessment of its earthquake risk [in rupture zone fin Chinese], in Sichuan Seismological Bureau, Proceed- Qian Hong, 1983, On the mechanism of the formation of anti xi-iype ground Chinese]: Seismology and Geology, v. 7 (3), p. 23-34. ings of the Conference on the Xianshuihe Fault Zone: Beijing, China, crack zones along the Xianshuihe fault [in Chinese]: Seismology and Wen Xueze, Allen, C. R., Luo Zhuoli, Qian Hong, Zhou Huawei, and Huang Seismological Press, p. 158-165. Geology, v. 5 (3), p. 75-79. Weishi, 1989, Segmentation, geometric features, and their seismotec- Gu Gongsu, ed., 1983, Earthquake catalog of China: Beijing, China, Scientific 1986, Recent displacements along Xianshuihe fault belt and its relation tonic implications for the Holocene Xianshuihe fault zone [in Chinese}: Press. with seismic activities: Journal of Seismological Research, v. 9 (5), Acta Seismologica Sinica, v. 11, p. 362-373. Han Wei bin, and Huang Shengmu, 1983, A seismic gap on the Xianshuihe p. 601-613. Yang Zunyi, Cheng Yuqi, and Wang Hongzhen, 1986, The geology of China: fault, Sichuan Province [in Chinese]: Acta Seismologica Sinica, v. 5 (3), Richter, C. F., 1958, Elementary seismology: San Francisco, California, W. H. Oxford, England, Clarendon Press, 303 p. p. 280-286. Freeman, 768 p. Yonekura, N„ Matsuda, T., Nogami, M., and Kaizuka, S., 1979, An active Han Weibin, Xi Dengli, and Liao Jingmei, 1985, Segmentary exchange of Rodgers, D. A., 1980, Analysis of pull-apart basin development produced by en fault at the western foot of the Cordillera Bianca, Peru [in Japanese]: seismic activity on Xianshuihe fault zone [in Chinese], in Sichuan Seis- echelon strike-slip faults: International Association of Sedimentologists Journal of Geography [Tokyo], v. 88, p. 1-19. mological Bureau, Proceedings of the Conference on the Xianshuihe Special Publication 4, p. 27-41. Young, R. J., and Coppersmith, K. J., ¡985, Implications of fault slip rates and Fault Zone: Beijing, China, Seismological Press, p. 94-103. Ruddiman, W. F„ and Duplessy, J.-C., 1985, Conference on the last deglacia- earthquake recurrence models to probabilistic seismic hazard estimates: Heim, A., 1933, Minya Gongkar [in German]: Bern-Berlin, Germany, Verlag tion: Timing amd mechanism: Quaternary Research, v. 23, p. 1 17. Seismological Society of Amnerica Bulletin, v. 75, p. 939-964. Hans Huber, 244 p. Savage, J. C., and Burford, R. O., 1973, Geodetic determination of the relative Zhao Xian, 1985, The characteristics of repeated dislocations along the 1934, Earthquake region ofTaofu: Geological Society of America Bul- plate motion in California: Journal of Geophysical Research, v. 78, Kangding fault, in Sichuan Seismological Bureau, Proceedings of the letin, v. 45, p. 1035-1049. p. 832-845. Conference on the Xianshuihe Fault Zone [in Chinese]: Beijing, China, Huang Shengmu, 1981, The intermittent earthquake during the seismic attenua- Schubert, C., 1982, Neotectonics of Bocono fault, western Venezuela: Tectono- Seismological Press, p. 41-47. tion on the Xianshui River belt in Sichuan [in Chinese]: Journal of physics, v. 85, p. 205-220. Zheng Benxing, and Li Jijun, 1981, Quaternary glaciation of the Qinghai- Seismologica! Research, v. 4 (3), p. 302-307. Schwartz, D. P., and Coppersmith, K. J., 1984, Fault behavior and characteris- Xizang Plateau, in Geological and ecological studies of Qinghai-Xizang 1985, A preliminary discussion on seismic episodes of the Xianshuihe tic earthquakes: Examples from the Wasatch and San Andreas fault Plateau: Beijing, China, Science Press, p. 1631-1640. seismic zone, in Sichuan Seismological Bureau, Proceedings of the Con- zones: Journal of Geophysical Research, v. 89, p. 5681-5696. Zhou Huilan, Allen, C. R., and Kanamori, H., 1983a, Rupture complexity of ference on the Xianshuihe Fault Zone: Beijing, China, Seismological 1986, Seismic hazards: New trends in analysis using geologic data, in the 1970 Tonghai and 1973 Luhuo earthquakes, China, from P-wave Press, p. 88-93. Wallace, R. E„ ed., Active tectonics: Washington, D.C., National inversion, and relationship to surface faulting: Seismological Society of Huang Shengmu, Liu Benpei, Jiang Zaixiong, and Li Dinfu, 1982, Seismic Academy Press, p. 215-230. America Bulletin, v. 73, p. 1585-1597. faults of Xialatuo, Luhuo County, Sichuan Province [in Chinese]: Seis- Segall, P., and Pollard, D. D., 1980, Mechanics of discontinuous faults: Journal Zhou Huilan, Liu, H.-L., and Kanamori, H., 1983b, Source processes of large mology and Geology, v. 4 (1), p. 67-71. of Geophysical Research, v. 85, p. 4337-4350. earthquakes along the Xianshuihe fault in southwestern China: Seismo- Huang Zuzhi and Tang Rongchang, 1982, A preliminary study of the seismo- Sibson, R. H., 1985, Stopping of earthquake ruptures at dilatational fault jogs: logical Society of America Bulletin, v. 73, p. 537-551. geological characteristics of the Zemuhe fault [in Chinese], in The active Nature, v. 316, p. 248-251. faults of China: Beijing, China, Seismological Press, p. 251-261. 1986, Rupture interaction with fault jogs: American Geophysical Union Irwin, W. P., and Barnes, I., ] 975, Effect of geologic structure and metamor- Geophysical Monograph 37, p. 157-167. phic fluids on seismic behavior of the San Andreas fault system in Sichuan Seismological Bureau, 1984, Proceedings of the conference on the central and northern California: Geology, v. 3, p. 713-716. Xianshuihe fault: Beijing, China, Seismological Press, 254 p. Jiang Zaixiong, 1985, A study of the absence of earthquakes on the Xianshuihe Sieh, K. E., 1981, A review of geological evidence for recurrence times of large MANUSCRIPT RECEIVED BY THE SOCIETY OCTOBER 17,1988 tectonic zone from 1817 to 1892 [in Chinese]: Earthquake Research in earthquakes: American Geophysical Union, Maurice Ewing Series 4, REVISED MANUSCRIPT RECEIVED OCTOBER 15,1990 Sichuan, no. 1. p. 181-207. MANUSCRIPT ACCEPTED MARCH 5,1991 Kan Rongju, Zhang Sichang, Yan Fengtong, and Yu Linsheng, 1977, Present Sieh, K. E,, and Jahns, R. H., 1984, Holocene activity of the San Andreas fault CONTRIBUTION No. 4498, DIVISION OF GEOLOGICAL AND PLANETARY SCIENCES, tectonic stress field and its relation to the characteristics of recent tec- at Wallace Creek, California: Geological Society of America Bulletin, CALIFORNIA INSTITUTE OF TECHNOLOGY

Printed in U.S.A.

Downloaded from http://pubs.geoscienceworld.org/gsa/gsabulletin/article-pdf/103/9/1178/3381303/i0016-7606-103-9-1178.pdf by guest on 01 October 2021