loading page

Comparison of travel-time and geostatistical inversion approaches for hydraulic tomography: A synthetic modeling study
  • +1
  • Huiyang Qiu,
  • Rui Hu,
  • Walter Arthur Illman
Huiyang Qiu
Hohai University

Corresponding Author:qiuhuiyang@hhu.edu.cn

Author Profile
Rui Hu
School of Earth Sciences and Engineering, Hohai University
Author Profile
University of Waterloo
Author Profile
Walter Arthur Illman
University of Waterloo
Author Profile


Hydraulic tomography (HT) has been proven to be an effective approach in mapping the heterogeneity of hydraulic parameters. The travel-time based inversion (TTI) and geostatistical inversion (GI) approaches are two of several HT methods. In particular, the GI approach is used to compute heterogeneous hydraulic conductivity (K) and specific storage (Ss) tomograms, while the TTI approach yields diffusivity (D = K/Ss) tomograms. The main objective of this paper is to evaluate these two methods through a synthetic study. Two cases are designed based on different monitoring configurations. Two independent scenarios are designed by: providing the same data sets and providing all available data for calibration, while data selection follows recommended strategies utilized by the two approaches. Then, the estimated tomograms are evaluated by visual comparison of estimated parameter distributions and assessments of model calibration and validation results. Results show that the advantages of TTI are: (1) imaging of structural features representing high D zones; (2) requirement of less data for inverse modeling; and (3) rapid computational times. In contrast, the advantages of the GI approach are: (1) the direct characterization of both K and Ss distributions; (2) better drawdown predictions; and (3) a larger estimated area. Our study suggests that the TTI approach is suitable for rapid, coarse characterization of heterogeneity that could potentially be utilized for providing hydrogeological structures for an initial model for the GI approach. The GI approach, although significantly more computationally intensive, is more robust and preferable to applications that require higher accuracy in parameter estimation.