# Identification of the Dominant Rainfall Index and Evolution of Multi-Factor Driving Mechanisms for Landslide Activity in Hong Kong (1990–2024)

## TL;DR

This study identifies the main rainfall factor and how multiple factors drive landslides in Hong Kong from 1990 to 2024.

## Contribution

A new integrated framework combining rainfall index optimization and multi-factor analysis is introduced for landslide mechanisms.

## Key findings

- The maximum 3-day cumulative rainfall index (RX3day) is the dominant rainfall indicator for landslide triggering.
- Geological and topographic factors provide a stable background for landslide spatial patterns.
- Rainfall-related interactions have increased over time, showing stronger nonlinear effects with geology and topography.

## Abstract

Revealing the spatiotemporal driving mechanisms of landslide activity is fundamental to improving long-term landslide hazard management and risk mitigation in mountainous cities. Focusing on landslide events in Hong Kong from 1990 to 2024, this study develops an integrated framework at the slope-unit scale that combines rainfall index optimization with multi-factor spatiotemporal driving analysis. First, Grey Relational Analysis (GRA) is employed to systematically evaluate the spatiotemporal associations between landslide occurrences and six commonly used rainfall indices, aiming to obtain a consistent and robust representation of rainfall triggering conditions. Subsequently, the Optimal-Parameter Geographical Detector (OPGD) model is introduced to quantitatively assess the explanatory power of individual factors—covering geological, topographic, hydro-meteorological, and human-related variables—as well as their pairwise interactions, thereby revealing the spatiotemporal evolution of landslide driving factors and their multi-factor coupling mechanisms over a 35-year period. The results indicate that the maximum 3-day cumulative rainfall index (RX3day) consistently exhibits the strongest association across different resolution parameter settings and is identified as the dominant rainfall indicator representing dynamic landslide triggering. Geological conditions and topographic factors constitute a stable background controlling the spatial heterogeneity of landslides throughout the entire study period, whereas the explanatory power of RX3day increases markedly after around 2000, gradually emerging as a primary dynamic driving factor of landslide activity. Interaction detection further demonstrates that landslide occurrence is mainly governed by nonlinear enhancement effects among multiple factors, with “geology–topography” and “rainfall–topography/geology” interactions showing the highest explanatory power, and rainfall-related interactions exhibiting continuous strengthening over time. Overall, the spatiotemporal distribution of landslides in Hong Kong is jointly controlled by long-term stable geological–topographic conditions and increasingly intensified extreme rainfall forcing.