Identifying the Latest Displacement and Long-Term Strong Earthquake Activity of the Haiyuan Fault Using High-Precision UAV Data, NE Tibetan Plateau

Abstract

Strong earthquake activity along fault zones can lead to the displacement of geomorphic units such as gullies and terraces while preserving earthquake event data through changes in sedimentary records near faults. The quantitative analysis of these characteristics facilitates the reconstruction of significant earthquake activity history along the fault zone. Recent advancements in acquisition technology for high-precision and high-resolution topographic data have enabled more precise identification of displacements caused by fault activity, allowing for a quantitative assessment of the characteristics of strong earthquakes on faults. The 1920 Haiyuan earthquake, which occurred on the Haiyuan fault in the northeastern Tibetan Plateau, resulted in a surface rupture zone extending nearly 240 km. Although clear traces of surface rupture have been well preserved along the fault, debate regarding the maximum displacement is ongoing. In this study, we focused on two typical offset geomorphic sites along the middle segment of the Haiyuan fault that were previously identified as having experienced the maximum displacement during the Haiyuan earthquake. High-precision geomorphologic images of the two sites were obtained through unmanned aerial vehicle (UAV) surveys, which were combined with light detection and ranging (LiDAR) data along the fault zone. Our findings revealed that the maximum horizontal displacement of the Haiyuan earthquake at the Shikaguan site was approximately 5 m, whereas, at the Tangjiapo site, it was approximately 6 m. A cumulative offset probability distribution (COPD) analysis of high-density fault displacement measurements along the ruptures indicated that the smallest offset clusters on either side of the Ganyanchi Basin were 4.5 and 5.1 m long. This analysis further indicated that the average horizontal displacements of the Haiyuan earthquake were approximately 4–6 m. Further examination of multiple gullies and geomorphic unit displacements at the Shikatougou site, along with a detailed COPD analysis of dense displacement measurements within a specified range on both sides, demonstrated that the cumulative displacement within 30 m of this section of the Haiyuan fault exhibited at least five distinct displacement clusters. These dates may represent the results of five strong earthquake events in this fault segment over the past 10,000–13,000 years. The estimated magnitude, derived from the relationship between displacement and magnitude, ranged from M_w 7.4 to 7.6, with an uneven recurrence interval of approximately 2500–3200 years.

1. Introduction

The precision of quantitative studies on active faults is dependent upon accurately identifying and measuring offset landforms, as well as constraining the timing of geomorphic surface formation or the occurrence of large earthquake events. These elements collectively characterize fault activity and the recurrence patterns of significant earthquakes [1,2,3,4,5]. Although instrumental seismic and historical literature records can extend seismic catalogs along fault zones to approximately 2000 years, they fall short of encompassing the recurrence period of large earthquakes on active faults, which can span thousands or even tens of thousands of years [5,6,7]. The advent of paleoseismology in the 1970s has allowed large earthquake chronologies to be extended via sedimentary records and faulted strata revealed by trenching, thus extending the time series of major earthquake activity along fault zones [2,5,8,9,10,11,12,13,14]. However, limitations in the trench location, scale, and quantity restrict the comprehensive acquisition of the characteristics of earthquake activity over extended time scales.

Seismic activity, particularly large earthquakes along fault zones, often results in surface rupture or deformation, which affects geomorphic or sedimentary markers crossing faults. Numerous case studies and field investigations indicate that geomorphic changes caused by major seismic events are not always recorded by obvious geomorphic markers but may manifest through subtle, long-term modifications. These changes can manifest in the subsequent evolution of landscape development, the process of morphological changes in geomorphic surfaces, changes in the water content, or the differentiation and growth of vegetation [1,5,15]. In previous studies, owing to the limitations of image resolution and the scope of field surveys, it was difficult to identify these geomorphic markers in detail [5]. In recent years, the development of high-tech remote sensing and spatial information acquisition technologies has provided unprecedented massive amounts of data and technical support for the high-precision, high-resolution, and high-density quantitative acquisition of faulted–offset landforms along fault zones. These technologies enable dense measurements of fault offsets, which, in turn, allows for the statistical separation of superimposed signals from multiple seismic events. Through measurements and statistical analysis of the displacements of different types of geomorphic units along faults, fault displacements caused by strong earthquake activity have been obtained, and the data have subsequently been used to quantitatively evaluate the characteristics and patterns of strong earthquakes in fault zones [16,17,18,19,20,21,22,23]. For example, the patterns of activity of strong earthquakes in fault zones can be defined on the basis of the measurements of displacements along a fault and analysis of the cumulative pattern of displacement (COPD), in combination with the ages of geomorphic units and paleoearthquake events exposed in trenches [19,20,21,22].

Earlier studies of segment offsets along the Haiyuan fault zone documented surface rupture associated with the 1920 M 8.5 Haiyuan earthquake. Investigations and mapping conducted in the 1990s identified surface rupture displacement at over one hundred locations along the nearly 240 km surface rupture zone [24]. Subsequent advances in technology, particularly the development of high-resolution airborne light detection and ranging (LiDAR), have enabled the acquisition of more detailed topographic data. Using this technique, 320 offset measurements were obtained along an 88 km portion of the rupture zone [21]. These kinematic records, which were obtained primarily from significant gully offsets and major geomorphic markers, may be deficient in capturing the complete kinematic history and evolution of a fault because they tend to overlook subtle geomorphic variations and lack detailed measurements across diverse geomorphic units within the same system, such as both sides of a gully and sequences of terraces. Zheng et al. [5] proposed that geomorphic features can preserve a more complete record of large earthquakes. Utilizing high-resolution topographic models derived from UAV photogrammetry, they reinterpreted fault offsets near Santang Village at the western end of the Haiyuan fault. Their analysis identified at least four surface-rupturing earthquakes, including the 1920 M 8.5 Haiyuan earthquake, whereas previous interpretations had attributed the observed fault offsets solely to the 1920 rupture. This revised interpretation is consistent with trench-exposed paleoearthquake evidence. These findings demonstrate that detailed identification of micro landforms can significantly enhance the accuracy of reconstructing earthquake event sequences along fault zones. Although the 1920 Haiyuan earthquake occurred only a century ago, varying interpretations of the scale of fault offset have emerged among researchers employing different methodologies. In fault zones with a history of significant earthquakes or complex activity patterns, it is necessary to conduct detailed investigations of geomorphic features and systematically classify the geomorphic units intersecting the fault zone. This approach aims to comprehensively obtain the accumulation of displacement associated with strong earthquake activity and analyze the characteristics of fault activity.

This study focuses on two typical faulted geomorphic sites along the Haiyuan fault that contained the surface rupture from the 1920 Haiyuan earthquake. High-resolution topographic data were acquired via a UAV at these two sites. Through high-density identification and interpretation of offset microtopography across various gullies and geomorphic units, the quantitative characteristics of the 1920 surface rupture were determined. On this basis, LiDAR-derived topographic data were used to analyze the distribution of fault offsets along a ~20 km section of the Haiyuan fault near Ganyanchi town. By integrating previous paleoearthquake trench studies and chronological data from this region and conducting detailed modeling and reconstruction of fault displacement at the two sites, the aim of this study is to establish a long-term geomorphic faulting history and a sequence of major seismic events for the central segment of the Haiyuan fault.

2. The Haiyuan Fault and the 1920 Haiyuan Earthquake

The Qilian-Haiyuan fault is the main boundary fault in the northeastern Tibetan Plateau and extends eastward from Halahu Lake in the western part of the Qilian Mountains to the Liupan Mountains, including the Halahu segment, Tuolai Mountains segment, Lenglongling segment, Jinqianghe segment, Maomao Mountains segment, Laohu Mountains segment, and easternmost segment of the Haiyuan fault segment (Figure 1) [25,26,27,28,29,30,31,32,33]. Kinematic analyses reveal that the left-lateral strike/slip rate along the Qilian–Haiyuan fault zone varies by approximately 5 mm/a; in the middle segment, the rate decreases to 1~2 mm/a near Halahu Lake in the western segment and to 1~3 mm/a near the northern Liupan Mountains in the eastern segment, where the fault zone transitions toward a thrust fault zone [25,30,33,34]. The fault’s significance in the uplift and outward expansion of the Tibetan Plateau has been extensively documented [25,31,32,33,35,36]. The M 8.5 Haiyuan earthquake occurred in 1920 on this fault, and the Haiyuan fault in the eastern section of the Qilian-Haiyuan fault zone has always been a focus of research. Previous studies have obtained many results on the fault’s activity, fault displacement distributions, slip rates, paleoearthquake activities, and seismic mechanisms along the Haiyuan fault [2,13,21,24,29,35,37,38].

Historically, trenching has been a primary method for investigating strong earthquakes within fault zones. Therefore, many trenches were excavated along the surface rupture zone of the 1920 Haiyuan earthquake [2,13,14,30,38,39,40,41]. Ran et al. [13,30,38] determined that the middle segment of the Haiyuan fault had experienced seven strong earthquake events over 10,000 years by examining the combined trenches and three-dimensional (3D) trenches and that the 1920 Haiyuan earthquake resulted in a horizontal displacement of 7 ± 0.5 m. Zhang et al. [38] proposed spatiotemporal recurrence models for strong earthquakes for different segments of the Haiyuan fault using data from 17 predecessor-excavated trenches and the method of successive limitation of paleoearthquake analysis. More than ten strong earthquake events have been recorded on the Haiyuan fault in the past ten millennia, with segmental ruptures, particularly in the middle segment near the Shikaguan site, which has experienced at least eight strong earthquake events. However, variations in the scale of trench excavation, homogeneity of exposed sedimentary strata, and differing interpretations of the surface rupture of the 1920 Haiyuan earthquake have resulted in incomplete long-term earthquake sequences [2,14]. Shao et al. [14] excavated trenches in the Ganyanchi basin segment, suggesting the presence of multiple strong earthquake events in addition to the 1920 surface rupture.

Another method for constraining strong earthquake events along fault zones is to obtain the distribution of fault displacements along faults and then relate displacement to magnitude to constrain the seismic intensity. This method assumes that varying fault displacements reflect the cumulative effects of differing numbers of strong earthquake events [17,19,20,21,22]. For the surface rupture of the Haiyuan earthquake, field observations and previous studies have revealed significant differences in displacements at different sites along the fault zone. In the areas east of Shikaguan and Shao Mayin, the displacement in the gully is 59.5 m (Figure 2a) [22,29]. In Tangjiapo, which is located west of Shikaguan, the displacement of an abandoned field embankment affected by the fault is approximately 6 m (Figure 2b). For the 1920 Haiyuan earthquake (M 8.5), 1980s field surveys identified an arc-like displacement pattern characterized by a maximum horizontal displacement of about 10 m in the central section, decreasing toward both ends [24,37,42,43,44]. Specifically, the displacement in the Shikaguan site (approximately 10 m of left-lateral displacement) was the result of the 1920 Haiyuan earthquake event (Figure 2c,d). Ren et al. [21], utilizing airborne LiDAR point cloud data from the Haiyuan fault zone, proposed that the previously identified maximum displacement in the Shikaguan site might be the cumulative effect of at least two seismic events. This conclusion is further supported by evidence of multiple seismic events revealed by paleoearthquake trenching in the Shikaguan site [45]. Previous findings have predominantly depended on large-scale, coarse measurements that capture the deformation of primarily macro-landform features. To comprehensively understand the details of surface rupture and the complexities of seismic events, it is essential to employ refined microtopographic analysis and high-precision measurements.

3. Data Acquisition and Processing

3.1. Acquisition of Fine Geomorphic Data

In recent years, the advantages of UAV photogrammetry systems have gradually emerged in geomorphic surveying. In regions with sparse vegetation, such as northwestern China, they effectively capture information on geomorphic units [22]. This study focuses on two typical faulted geomorphic sites, Tangjiapo and Shikaguan, which are situated in the middle segment of the Haiyuan fault and experienced surface rupture during the 1920 Haiyuan earthquake. For data acquisition, a DJI Matrice 300 (SZ DJI Technology Co., Ltd., Shenzhen, Guangdong Province, China) real-time kinematic (RTK) instrument equipped with L1 LiDAR was utilized. The UAV system operated at a height of 100 m above the ground, maintaining a flight speed of 10 m/s. The DJI Matrice 300 RTK is equipped with a high-precision RTK module enabling real-time differential positioning (at centimeter-level accuracy) when operating within areas covered by a network RTK signal. In remote regions where network coverage is limited, a UFO U5 base station was deployed to ensure positional stability. By connecting to this base station, the M300 UAV received real-time differential correction data, thereby achieving precise geolocation. According to specifications provided on the official DJI website (https://enterprise.dji.com/zenmuse-l1/specs, accessed on 12 October 2024), both RTK positioning methods are capable of generating centimeter-level three-dimensional coordinates of the survey area using ground control point-free SfM processing, relying on real-time differential positioning (resolving carrier-phase ambiguities through base station corrections for centimeter-level accuracy) Postprocessing of the collected imagery using DJI Terra v4.1.0 software yielded point cloud data with an average density exceeding 50 points/m². The reconstructed 3D topography of the study area achieved centimeter-level resolution, including both a digital elevation model (DEM) and a digital orthophoto map (DOM) (Figure 3 and Figure 4).

3.2. Data Processing and Geomorphic Model

Structure from motion (SfM) is a technique used for 3D reconstruction that transforms two-dimensional photographs into three-dimensional scenes by utilizing multiview images. This method involves analyzing multiple photographs captured from various angles and identifying common feature points in the overlapping sections of these images. By employing the principle of triangulation, the spatial positions of these points are calculated, allowing an accurate reconstruction of the 3D shape of the target scene [46]. When UAVs capture original point cloud data, they often include multiple echoes from vegetation and the ground, as well as noise points caused by random errors and reflections from airborne particles. To ensure that the point cloud data accurately reflect the ground surface, denoising the original point cloud and eliminating noise and floating points are necessary before the creation of the terrain model. The processed ground point cloud data are inherently discrete; by converting them into a continuous surface for a more intuitive representation, we employ the Triangulated Irregular Network (TIN) method because TINs can effectively preserve terrain discontinuities and linear features such as fault scarps and channel edges by dynamically adjusting sampling point density and layout according to terrain complexity variations. This approach minimizes data redundancy in relatively flat areas while ensuring a vector-based representation in topographically complex areas like ridges and valleys. These combined advantages make TIN particularly suitable for precise geomorphic analysis. Additionally, since point cloud data may exhibit cavities due to shooting blind spots, interpolation is performed using the surrounding point clouds to fill these gaps. An irregular triangulation network is then constructed to develop a digital elevation model of the survey areas (Figure 3 and Figure 4).

3.3. Measurement Methods for Fault Segment Offsets

The MATLAB R2023b-based graphical tool, LaDiCaoz _2v2, developed by Zielke, O et al., [47], is designed for measuring lateral fault slip by aligning offset geomorphic features (such as gullies and fan edges) on high-resolution DEMs. The tool has been enhanced to provide reproducible measurements through a process of slicing the DEM at the fault trace and then performing back-slipping of one side. This methodology achieves sub-meter precision when applied to UAV-based topographic data. Compared to manual measurements, it demonstrates enhanced accuracy. The resulting offset quantitative data facilitate quantitative analysis of fault kinematics, slip rates, and paleoearthquake dataset. The operational workflow of the software involves several sequential steps. (1) Import the DEM file of the study area to generate DEM hillshade maps and contour maps. (2) Identify the fault location and define the positions of the upstream (blue) and downstream (red) cross-sections that intersect the gully (Figure 5a). The software simultaneously displays elevation profiles for both upstream and downstream cross-sections. (3) Horizontally shift the shape of the red cross-section (step size of dx), followed by vertical movement (step size of dy), and, finally, stretching (step size of dz). The optimal displacement value is determined by minimizing the sum of the absolute elevation differences between the two sections (Figure 5b–e). (4) After the optimal displacement value is identified, a “sliding back” procedure aligns the trend lines of the gullies on both sides. This process determines the maximum and minimum displacement values and the associated measurement error (Figure 5f) [21,47,48,49,50,51].

4. Results

4.1. Displacement Measurements at the Tangjiapo Site

In the surface rupture zone formed by the 1920 Haiyuan M 8.5 earthquake, Tangjiapo was a typical site where multiple field ridges built before the earthquake were displaced by the earthquake, forming a series of en echelon tension-shear cracks with strikes of 70°~110° (Figure 2d and Figure 6a). These cracks form a left-lateral and right-step rupture model, with many extrusion bulges forming between the secondary cracks, but a few secondary cracks form tensile depressions in the left-step model (Figure 6b). The surface rupture zone at the Tangjiapo site, which is composed of multiple earthquake cracks, has angles of 30°~40° between the cracks and the fault strike, reflecting an obvious pattern of shear movement pattern of the left-lateral strike/slip fault (Figure 6c).

To measure the horizontal displacement of field ridges on both sides of secondary cracks, the LaDiCaoz displacement measurement program is employed. This program projects horizontal displacement measurements of each field ridge within the rupture zone onto the fault strike on the basis of the rupture geometry. Figure 6d shows the method used for measuring displacement for secondary cracks affecting field ridge D, which is divided into three sections. Using LaDiCaoz software, the displacement between these sections along the secondary cracks is measured, yielding values of 4.02 ± 0.35 m, 4.29 ± 0.16 m, and 5.04 ± 0.52 m and 4.52 ± 0.20 m, respectively (Figure 6d). Consequently, the displacement along the average direction of the crack between the three sections of the field ridge is calculated as 4.16 ± 0.38 m and 4.78 ± 0.56 m, respectively. On the basis of these measurements, the horizontal displacement of ridge D along the secondary crack is projected in the fault strike direction, considering the angle between the ridge and the fault strike (approximately 30° and 45° at ridge D). This projection resulted in a horizontal displacement of 6.10 ± 0.67 m along the fault strike for ridge D (Figure 6d). Similarly, the horizontal displacements along the fault strike for the other ridges are determined to be 6.39 ± 0.97 m for ridge A, 5.84 ± 0.48 m for ridge B, and 3.73 ± 0.55 m for ridge C. Given that the field ridges were constructed via local farming practices, which are typically not long-enduring, and considering that the local dry land was abandoned before the earthquake [52], it can be concluded that the observed horizontal displacement was caused only by the 1920 Haiyuan earthquake.

4.2. Displacement Measurements at the Shikaguan Site

In this study, we analyzed three gullies with varying morphological integrities and lengths, each situated within approximately 200 m along the fault strike at Shikaguan gully, to interpret geomorphic units (Figure 7a). Within a confined range, the length of the gullies correlates with their formation time. The differences in lengths and widths among the three gullies at Shikaguan indicate distinct developmental histories (Figure 2c,d and Figure 7a). This distinction helps in understanding variations in accumulated fault displacement across different gullies, as those with longer developmental histories would have undergone more prolonged interactions due to fault activity. The geomorphic characteristics on both sides of the gullies, such as texture from images, terrain height, slope variations, and vegetation distribution, can be systematically divided into four levels from youngest to oldest, namely Gully, Fan 1, Fan 2, and Fan h. Gully represents the most recent geomorphic unit of modern gullies, characterized by the water-rich and green grass area in gully A and the gully beds in other gullies. Fan 1 is the initial terrace formed by the gullies and is distributed primarily along both sides of each gully. Notably, Fan 1 in gully A is relatively wide, whereas Fan 1 in other gullies is relatively narrow. Fan 2 is found exclusively in larger-scale gullies A and B, where variations in alluvial fan width may suggest temporal differences in their formation processes. Fan h represents the highest-order geomorphic unit in the area and constitutes the earliest sloped accumulation fan, which developed extensively among the three gullies.

Using the LaDiCaoz displacement measurement program, we assessed the fault displacements of various geomorphic units and boundary lines at the Shikaguan site, with a specific focus on three gullies. Our measurements yield 12 displacement values within a 200 m range along the fault strike (Figure 7b). In gully A, the horizontal displacement of the boundary line at the upper edge of the Fan 2 on the east is approximately 19.63 ± 0.98 m (➀), whereas the lower edge of the Fan 2 exhibits a displacement of approximately 16.15 ± 1.04 m (➁). The displacement along the faults of gully A between the upstream and downstream sides is 26.44 ± 2.30 m (➂). On the west side of gully A, the boundary line of the upper edge of the Fan h units exhibits a displacement of approximately 27.15 ± 2.93 m (➃). In gully B, the eastern and western Th upper edges have horizontal displacements of 10.26 ± 0.94 m (➄) and 5.45 ± 2.04 m (➇), respectively. The centerline of gully B experiences two turns when passing through the fault, possibly indicating two separate events, with displacements of approximately 5.44 ± 0.83 m (➅) and 10.41 ± 0.81 m (➆). Gully C displays more moderate offsets, with the upper edge of the eastern Fan h displaced by 7.36 ± 1.67 m (➈) and the upper boundary of the western Fan h displaced by 10.73 ± 2.79 m (⑫). Additionally, the horizontal displacements of the centerline of gully C and the rear edge of terrace Fan 1 on the west side are approximately 4.48 ± 0.88 m (➉) and 6.36 ± 0.94 m (⑪), respectively.

5. Discussion

5.1. Maximum Co-Seismic Displacement of the 1920 Haiyuan Earthquake

Previous studies have suggested that the synchronous displacement of several gullies near the Shikaguan site was approximately 10 m, attributed to the Haiyuan earthquake event [24,37,42,43,44]. However, subsequent geological investigations have revealed a more complex history of strong earthquake activity. Trench results from the Shikaguan site indicate that sedimentary records in faulted sag ponds have documented at least two strong earthquake events over the past 13,000 years [45]. In the adjacent Gaowanzi area, multidimensional trench analyses have resealed at least seven strong earthquake events in this fault segment during the past 10,000 years [13,39,40]. Similarly, trench investigations in the Ganyanchi Basin have revealed multiple strong earthquake events within the same timeframe [10,14,30,53]. Geomorphic analyses along the fault zone suggest comparable developmental stages of landforms in Gaowanzi village, Ganyanchi town, Shikaguan, and Shaomayin village, which are located approximately 2 km east of Shikaguan gully. These findings indicate that these landforms may have formed and evolved concurrently, revealing uncertainties in previous interpretations of rupture displacement during the Haiyuan earthquake. According to patterns of the distribution of slip displacement along fault zones [1,9]), co-seismic displacement characteristics, particularly along central fault segments, exhibit relative stability with minimal variation [54]. This suggests that displacement magnitudes or cumulative slip along primary fault cores show limited divergence; even when minor structural complexities, such as stepovers or branching faults, are present, the total displacements tend to be consistent along the fault [1].

Continuous displacement measurements along the central Haiyuan fault by Ren et al. [21] have demonstrated that cumulative displacements generally cluster in multiples of approximately 5 m. At a loess slope located 100 m from the Shikaguan site, four measured gully offsets yielded displacements of 5.5 m, 10 m, 10 m, and 9.7 m. Field measurements at the Shikaguan site identified two distinct displacement values: 5.0 m and 9.9 m. The 5.0 m displacement is interpreted as a co-seismic slip from the single 1920 Haiyuan M 8.5 earthquake event, whereas the 9.9 m displacement likely represents cumulative slip from multiple large seismic events. A probability density function (PDF) is widely employed to quantify uncertainty in geomorphic offset measurements. To better constrain slip history, summing the PDFs of individual features yields a cumulative offset probability density (COPD), which manifests multimodal distributions interpreted to represent the cumulative slip from successive prehistoric earthquakes. These distributions are interpreted as follows: larger peaks reflect accumulated displacements from earlier events, while the smallest peak may represent the most recent co-seismic slip [17,19,48]. Using airborne LiDAR data from the Haiyuan fault acquired by Ren et al. [21], we focused on high-density measurements and analysis of fault displacement from Gaowanzi to the eastern part of Shaomayin village in the middle segment of the Haiyuan fault (Figure 8). The results demonstrated regularity in displacement clusters within this section, with the minimum displacement occurring at approximately 5 m. Five distinct periodic displacement clusters of 5.1 m, 9.6 m, 14.8 m, 20.9 m, and 25.3 m, along with 4.5 m, 8.4 m, 14.8 m, 19.6 m, and 25.5 m, were observed within 30 m on the east and west sides of Ganyanchi, respectively. This indicates that the latest co-seismic displacement of the Haiyuan fault in this segment is approximately 5 m, with characteristic cumulative values for each event ranging between 4 and 6 m. In this study, high-precision geomorphic surveying conducted at Tangjiapo revealed multiple field-ridge offsets ranging from 3.73 to 6.39 m (Figure 6). Additionally, a historical literature analysis by Sun et al. [52] confirmed that these offsets were exclusively associated with the 1920 Haiyuan M 8.5 earthquake. At the Shikaguan site, our measurements also revealed displacements of a minimum of 4–6 m of multiple gullies and geomorphic unit boundaries (Figure 7). On the basis of the above findings and evidence, we can infer that the co-seismic displacements of the 1920 Haiyuan M 8.5 earthquake in this segment are likely within the range of 4–6 m.

5.2. Strong Earthquake Activity in the Middle Segment of the Haiyuan Fault

Due to prolonged and intense earthquake activity, fault displacements are inevitably preserved across various geomorphic units along the fault zone. Typically, geomorphic units that formed earlier experience more seismic events, resulting in larger accumulated displacements [1,8]. Previous studies have identified different patterns of displacement accumulation—such as characteristic slip, variable slip, and nonuniform slip—depending on fault behavior [1,8]. Within a relatively confined area or a single fault segment, the magnitude and pattern of accumulated displacement tend to be relatively consistent.

In Figure 8, at least five distinct displacement clusters are evident in the Shikaguan site segment. Our measurements of several typical displacements in this area revealed that the three gullies, which vary in size, exhibited different displacement characteristics (Figure 7). We reconstructed the displacement history at the Shikaguan site (Figure 9a). Initially, gully C and geomorphic units with small displacements were adjusted to both sides of the fault to reestablish linear characteristics, setting an initial displacement of 5.12 m (Figure 9b). The retro deformation process then unfolded in several stages: restoring the upstream–downstream connectivity of gully B (achieving a cumulative offset of 10.53 m, Figure 9c), aligning the lower edge of the Fan 2 of gully A (16.15 m cumulative offset, Figure 9d), reconstructing the upper edge of the Fan 2 of gully A (19.63 m cumulative offset, Figure 9e), and finally, fully restoring the upstream–downstream connectivity of gully A (total cumulative offset of 26.79 m, Figure 9f). Consequently, it is inferred that at least five strong earthquake events have accumulated horizontal displacements of approximately 25 m in the middle section of the Haiyuan fault over the past 10,000 to 13,000 years, which aligns with paleoearthquake research findings [13,39,40,45].

On the basis of these results of displacement accumulation, we conclude that the accumulated fault displacement at this location on the Haiyuan fault is the outcome of at least five events. The horizontal displacements for these five events were 5.1 m, 5.4 m, 5.6 m, 3.5 m, and 7.2 m, respectively. We integrated surface displacement data from five distinct periods with the age estimates of paleoearthquake events to reconstruct time–displacement curves for five major earthquakes [45]. This approach enables a more intuitive understanding of the large earthquake history along the central segment of the Haiyuan fault over the past approximately 10,000–13,000 years (Figure 10). Using the empirical formula of Wells, D.L., [55] [M_w = 6.93 + 0.82 log (AD)], the magnitudes of the five events are estimated to range from M_w 7.4 to 7.6. Events of this magnitude would inevitably lead to significant seismic hazards in the densely populated Haiyuan area. Furthermore, considering the age of the geomorphic units within the 25 m of displacement and paleoearthquake events of this segment [13,29,39,40,45], these five events are estimated to have occurred over the past approximately 10,000 to 13,000 years, with a recurrence interval of approximately 2500 to 3200 years.

6. Conclusions

This study focuses on typical displacement sites in the middle segment of the Haiyuan fault, specifically the Shikaguan and Tangjiapo sites. High-precision geomorphic models were obtained at these locations using UAVs, supplemented by analyses of earlier airborne LiDAR data along the fault zones. This approach allowed for a comprehensive study of the maximum co-seismic displacement from the Haiyuan earthquake and the characteristics of strong earthquake activity along this fault segment.

The analysis of displacement at the Shikaguan and Tangjiapo sites indicates that the average co-seismic displacements resulting from the 1920 Haiyuan earthquake likely range between 4 and 6 m. Specifically, the maximum displacement is 6.4 m at Tangjiapo, and the most recent displacement is 4.5 m at the Shikaguan site. High-density displacement measurements combined with cumulative offset and paleoearthquake dating (COPD) analysis revealed five distinct clusters within a cumulative displacement range of 30 m. These clusters correspond to five paleoearthquakes with individual offsets of 5.1 m, 5.4 m, 5.6 m, 3.5 m, and 7.2 m, respectively. The estimated magnitudes of these events range from M_w 7.4 to 7.6. On the basis of previous trench dating results, these events are estimated to have occurred over the past 10,000 to 13,000 years.

This study shows that high-precision and high-resolution remote sensing data significantly enhance active tectonics research, offering insights that are unattainable through traditional methods alone. By integrating traditional research techniques with new technologies, a comprehensive sequence of strong earthquake activity along fault zones can be established, thereby reducing the uncertainty associated with paleoearthquake events determined by a single method.