Spatiotemporal Distribution of Nonseismic Landslides during the Last 22 Years in Shaanxi Province, China

Haijun Qiu, Yifei Cui, Dongdong Yang, Yanqian Pei, Sheng Hu, Shuyue Ma, Junqing Hao, Zijing LiuISPRS Int. J. Geo Inf 2019 · DOI

TL;DR

This study presents a comprehensive analysis of nonseismic landslide spatiotemporal distribution in Shaanxi Province, China, over 22 years. By compiling a detailed landslide catalog and employing statistical methods, the research reveals key temporal patterns, influencing factors, and regional differences, offering valuable insights for landslide risk assessment and engineering design.

What Problem Does This Paper Address?

Landslides pose significant threats globally, causing casualties and economic losses. Understanding their spatiotemporal distribution is crucial for effective hazard assessment, early warning systems, and risk management. However, a significant challenge lies in the difficulty of obtaining comprehensive historical landslide data, especially for nonseismic events, which often lack periodic patterns and are hard to track at a large scale. This paper addresses this gap by creating and analyzing a detailed landslide catalog for Shaanxi Province, a region known for its high landslide susceptibility due to its unique geological and geomorphological characteristics. The study aims to fill the knowledge gap regarding the temporal and spatial patterns of nonseismic landslides in this region, providing a foundation for improved risk assessment and engineering practices.

Methods

The researchers compiled a landslide catalog for Shaanxi Province from 1996 to 2017 using existing records, yearbooks, archives, and extensive fieldwork. This catalog includes information on landslide locations, types, dates, and fatalities. To analyze the data, they employed several statistical and spatial analysis techniques:

  • Power-Law Relationship: Used to fit the cumulative frequency of annual landslide numbers, daily landslide numbers, and fatality numbers. This approach helps understand the distribution of extreme events.
  • Exponential Relationship: Applied to model the cumulative frequency of time intervals between landslide events, revealing patterns in event clustering.
  • Spatial Analysis: Utilized Geographic Information System (GIS) tools, including kernel density estimation, mean center analysis, and digital elevation model (DEM) analysis, to investigate the spatial distribution of landslides and their relationship with various influencing factors.
  • Statistical Analysis: Employed methods like frequency ratio (FR) to quantify the relationship between landslide occurrences and factors such as elevation, distance to stream channels, rainfall, faults, slope gradient, land use, and geology.

Datasets

The primary dataset used in this study is a newly compiled nonseismic landslide catalog for Shaanxi Province, China, covering the period from 1996 to 2017. This catalog was constructed from:

  • Existing landslide records and archives.
  • Shaanxi Province Yearbooks of Disaster Prevention.
  • Extensive fieldwork conducted between 2015 and 2018, which included measuring and mapping 120 landslides.
  • Interpretation of remote sensing images and Google Earth imagery.

The catalog contains 695 reported landslide occurrences, classified into falls, slides, and debris flows. This inventory is noted as a minimum representation due to the difficulty in obtaining complete historical data.

Key Insights and Ideas

  • Power-Law Behavior is Pervasive: The study found that landslide occurrences, daily landslide numbers, and fatalities all exhibit power-law distributions. This suggests a self-organized criticality in landslide processes, where a few large events are balanced by many small ones.
  • Seasonal Dominance of Landslides: The research highlights a strong seasonal pattern, with the vast majority of landslides occurring between July and October, coinciding with the monsoon season and peak rainfall.
  • Regional Differences Matter: Significant differences were observed in landslide types and densities between the Loess Plateau and the Qinba Mountains, underscoring the importance of considering regional geomorphological and geological characteristics in hazard assessment.
  • Precipitation as a Primary Trigger: The study strongly links landslide frequency to rainfall, with higher annual precipitation and intense rainfall events being key triggers. Even low rainfall events can trigger landslides if preceded by prolonged periods of high rainfall, indicating the role of soil moisture.
  • Proximity to Features is Crucial: Landslides are concentrated near stream channels and faults, and their frequency increases with proximity to these features, highlighting the role of erosion and tectonic activity.

Key Results

  • Temporal Distribution:
    • Cumulative frequency of annual landslide numbers follows a power-law: CF = 23.76N₁⁻¹·³⁰ (R² = 0.97).
  • Approximately 80% of landslides and fatalities occurred between July and October.
  • Cumulative frequency of time intervals between landslide events follows an exponential function: CFt = 0.67e⁻⁰.¹¹ᵗ (R² = 0.98).
  • Cumulative frequency of daily landslide numbers also follows a power-law: CF = 0.57D⁻¹·⁶ (R² = 0.94).
  • Spatial Distribution and Influencing Factors:
    • Landslide density in the Qinba Mountains (53 landslides/10⁴ km²) is 2.5 times higher than in the Loess Plateau (21 landslides/10⁴ km²).
  • Falls are more prevalent in the Loess Plateau (55%), while debris flows are more common in the Qinba Mountains (22%).
  • Most landslides occur at elevations of 600–900 m and on slopes of 30–40°.
  • Areas with annual rainfall >700 mm account for 64% of landslides.
  • 66% of landslides occur within 5 km of a fault.
  • Cropland and grassland exhibit high landslide densities, while settlements show the highest relative frequency.
  • Fatalities:
    • Approximately 42% of collected landslides caused deaths.
  • The cumulative frequency of fatalities follows a power-law: CFf = 1.44Nf⁻¹·⁴⁰ (R² = 0.98).

Limitations and Future Work

The authors acknowledge that the landslide data collection was preferentially focused on landslides affecting infrastructure, those resulting in fatalities or damages, and those near residential areas. This means the catalog likely represents a minimum landslide number and may have limitations in terms of completeness, potentially influencing the accuracy of statistical analyses, especially for less impactful events.

Future work could involve:

  • Improving the completeness of the landslide inventory to capture a broader range of events.
  • Further refining the spatiotemporal distribution framework for nonseismic landslides on a larger scale.
  • Utilizing these findings for more accurate landslide risk assessment and management strategies.
  • Investigating the role of other less-studied factors or their interactions in triggering landslides.

Summary updated

BibTeX
@article{qiu2019spatiotemporal,
  author = {Haijun Qiu and Yifei Cui and Dongdong Yang and Yanqian Pei and Sheng Hu and Shuyue Ma and Junqing Hao and Zijing Liu},
  title = {{Spatiotemporal Distribution of Nonseismic Landslides during the Last 22 Years in Shaanxi Province, China}},
  journal = {ISPRS Int. J. Geo Inf.},
  year = {2019},
  doi = {10.3390/ijgi8110505},
}
Figure