# Research on a dynamic early warning model based on refined threshold analysis: Case study of the Tanjiawan Landslide

## TL;DR

This study develops a dynamic early warning model for landslides by analyzing rainfall and deformation data, using the Tanjiawan landslide as a case study.

## Contribution

A refined dynamic early warning model is proposed, integrating rainfall and displacement thresholds with adaptive monitoring cycles.

## Key findings

- Tanjiawan landslide deformation is closely linked to rainfall, showing 'step-like' displacement and 'lagged attenuation' in displacement rates.
- Shorter monitoring cycles improve the accuracy of displacement rate thresholds for landslide early warning.
- The model adapts monitoring cycles based on deformation stage characteristics, enabling optimized early warning.

## Abstract

After rainfall-induced landslides enter the creep deformation stage, timely mitigation is often challenging, making reasonable and effective early warning critical for reducing disaster losses. This study focuses on the Tanjiawan landslide, introducing the concept of “a single rainfall process” to characterize the rainfall process affecting landslide deformation. Based on a detailed analysis of deformation characteristics such as displacement and displacement rate under rainfall, the least squares method is used to identify the “failure inflection point” and “stable inflection point” on the “step-like” deformation curve to determine the accelerated deformation interval. This approach further establishes the antecedent rainfall threshold (Pe), current rainfall (P), and displacement rate threshold (V). Subsequently, a refined dynamic early warning model is developed by integrating the function F(V, P, Pe) with a Logistic regression model. The findings indicate: (1) The deformation of the Tanjiawan landslide is closely correlated with rainfall processes, with cumulative displacement curves exhibiting distinct “step-like” characteristics and displacement rates showing a “lagged attenuation” phenomenon. (2) Finer monitoring cycles enable more precise capture of dynamic landslide deformation, resulting in more reliable displacement rate thresholds. (3) The landslide early warning model can dynamically adjust monitoring cycles based on the evolutionary characteristics of deformation stages, achieving adaptive monitoring optimization.