Bedrock groundwater is reported to be one of the main factors governing occurrences of deep landslides. Hence, predictions of bedrock groundwater level (BGL) changes caused by rainwater infiltration are essential for assessments of the landslide vulnerability as well as for establishments of evacuation systems. This study proposed four simple functional models which correlate antecedent precipitation indices (APIs) to BGLs. Li1 and Li2 models assume linear relationships between BGLs and APIs, and Pw1 and Pw2 models employ power functions. While Li1 and Pw1 models use an API with a single half]life time (HLT), Li2 and Pw2 models use two APIs with two different HLTs. Performances of the models were examined by using BGLs observed at8boreholes excavated at a steep hillslope underlain by weathered granitic bedrocks. Results showed that Pw2 model was the best for reproducing BGLs observed at the 5 boreholes. Pw1 and Li1 models were selected to be the best for 2 and 1 boreholes, respectively. Optimized HLTs, which were much longer than the ordinary values used for predictions of shallow landslides and debris flows, properly represented groundwater dynamics of the studied hillslope.

Key wordsFdeep landslide, bedrock groundwater, antecedent precipitation index, functional model