Cessation time approach incorporating parametric and non-parametric machine-learning algorithms for recovery test data

ABSTRACT

In this study we propose a new method called the cessation time approach (CTA) for interpreting recovery tests in confined aquifers, which is based on the Theis solution. The CTA method involves selecting a residual drawdown measurement from the recovery phase and linking it to its dimensionless counterpart through simple algebraic steps. This approach is then incorporated with a regression model to estimate aquifer parameters. The performance of several parametric polynomial and non-parametric machine learning regression models was investigated using various datasets. Results show that CTA with third-order multivariable polynomials produced highly accurate parameter estimates with a normalized root mean squared error (NRMSE) within 0.5% for a field dataset. Among the machine learning algorithms tested, the radial basis function and Gaussian process regression achieved the highest accuracy with NRMSEs of 0.6%. We conclude that CTA can be a viable interpretation tool for recovery tests due to its accuracy and simplicity.

KEYWORDS:

• aquifer analysis
• machine learning
• non-parametric algorithms
• parameter estimation
• recovery test

Editor S. Archfield; Associate Editor M. Bianchi

Editor S. Archfield; Associate Editor M. Bianchi

Disclosure statement

No potential conflict of interest was reported by the authors.

Data transparency

All data used in this research can be provided on request.