Reducing false alarms in crackmeter-based early warning of small-scale slope deformation via deep learning-driven asynchronous displacement–rainfall data fusion

Ouyang, Shubing; Li, Daichao; Liu, Minjiang; Ge, Fengjian; Zheng, Xia; Li, Yuan; Wu, Sheng

doi:10.5194/nhess-26-5441-2026

Articles | Volume 26, issue 11

https://doi.org/10.5194/nhess-26-5441-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Special issue:

Early Warning Systems from Research to Operations: Status,...

https://doi.org/10.5194/nhess-26-5441-2026

© Author(s) 2026. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 26, issue 11

Research article

|

02 Nov 2026

Research article |

| 02 Nov 2026

Reducing false alarms in crackmeter-based early warning of small-scale slope deformation via deep learning-driven asynchronous displacement–rainfall data fusion

Shubing Ouyang, Daichao Li, Minjiang Liu, Fengjian Ge, Xia Zheng, Yuan Li, and Sheng Wu

Model code and software

The code of PatchDC Shubing Ouyang https://github.com/ShubingOuyangcug/PatchDC

Short summary

Crackmeter-based warning systems generate many false alarms, causing alarm fatigue. An AI model was developed to combine real-time displacement and rainfall data and distinguish genuine deformation warnings from false alarms caused by noise or other disturbances. It identified 89.38% of confirmed true warnings, and 90.32% of alerts classified as genuine were correct. Analyses of misclassified cases informed additional operational safeguards to further reduce missed warnings.