Comparison of Piezometric Head Relationships Using Statistical Analysis: A Case Study of Pasak Jolasit Dam

Abstract

The analysis of data from dam behavior monitoring instruments provides an understanding of the behavior occurring within the dam. Piezometers are widely used instruments to monitor seepage and internal erosion in earth dams. This study analyzes the data using three statistical methods: Simple Linear Regression (SLR), Multiple Linear Regression (MLR), and Dummy Variable Regression (DVR), to examine the relationships between piezometric head, reservoir water levels, and related factors, as well as to identify the most appropriate analytical method. The results show that SLR provides correlation coefficients (R) ranging from 0.05 to 0.70, MLR from 0.26 to 0.72, and DVR from 0.43 to 0.81. SLR is a basic and convenient method but yields the lowest correlations, and is suitable to explain the behavior of upstream piezometers. MLR provides higher correlations, especially for piezometers influenced by rainfall, but it suffers from multicollinearity. DVR, which analyzes the relationships based on temporal variations in reservoir water levels, is the most appropriate method, providing the highest correlations, reducing multicollinearity, and yielding the best AIC and BIC values. However, the suitability of each analytical method should depend on the location of the piezometers and correspond to the current behavioral characteristics.