Estimation of Rock Mass Equivalent Permeability Around Tunnel Route Using the Geostatistical Methods: A Case Study

The objective of this paper is to estimate the equivalent permeability of the rock surrounding the tailrace tunnel of the Azad Dam pumped storage power plant, using geostatistical methods. The permeability of the rock mass is a critical factor that influences the estimation of water flow rates. Since the tunnel passes through various geological units with different permeabilities, it is crucial to estimate the equivalent permeability for each unit in order to predict the water seepage from that unit into the tunnel. In order to estimate the permeability along the tunnel and underground structures, twelve exploratory boreholes were drilled, and water pressure tests were conducted. Due to the distribution of the exploratory boreholes, a study and statistical analysis are necessary to determine the permeability of the rock mass for each geological unit. Using geostatistical log kriging, multiple indicator kriging with four thresholds, and multiple indicator kriging with five thresholds, the permeability of the rock mass at the tunnel route was estimated. The results indicate that at least 40% of the rock mass has low permeability, while the remaining mass of the tunnel passes through rocks with moderate to high permeability. The accuracy of the estimated permeability values was evaluated by predicting the water inflow into the tunnel using the estimated permeability values and comparing it to the observed flow. Numerical models were generated for each geological unit to estimate the water inflow into the tunnel, based on the results of the geostatistical methods. Log kriging, multiple indicator kriging with four thresholds, and multiple indicator kriging with five thresholds were used to calculate the water inflow, resulting in 94.15, 94.15, and 127.5 L per second, respectively. The results of the modeling were compared to the observed water flow into the tunnel. Comparing the modeling results to both the statistical methods and observed values showed errors of 31.2%, 31.2%, and 6.9%, respectively. Of the three methods, the multiple indicator kriging computational method with five thresholds was found to be the most accurate, with the least amount of error and the closest approximation to the actual value. As a result, it was selected as the best method.