Early warning threshold of sinkhole collapse based on dynamic characteristics from groundwater monitoring: A case study of Jinshazhou of Guangzhou, China

Karst sinkhole collapse early warning based on monitoring data has long been a prominent technical problem which needs to be solved in the field of geohazard. This paper analyses the current status of sinkhole collapse monitoring and early warning research, and then presents a case study of Jinshazhou of Guangzhou on this issue. The analytical result of groundwater monitoring data shows that the data anomalies largely reflect the abrupt change of groundwater dynamic and the stage of relevant karst collapse development. On this basis, through curve fitting and residual analysis this work determines the confidence belt of sinkhole collapse early warning. Comparing the times of most frequent anomalies, occurrence of maximum and minimum outliers, and the actual times of karst collapse, it is confirmed that it is feasible to use the anomaly data analysis method to make early warning of sinkhole collapse. With shorter monitoring data intervals, capturing more anomalies, the early warning will be more accurate.